linux-surface/patches/5.8/0006-surface-sam.patch

From ef00777fa69ff655ce11b70ccb321774ec82659d Mon Sep 17 00:00:00 2001
From: Maximilian Luz <luzmaximilian@gmail.com>
Date: Thu, 29 Oct 2020 22:04:38 +0100
Subject: [PATCH] PCI: Allow D3cold for hot-plug ports on Surface Books

The Microsoft Surface Book series of devices have a tablet part (so
called clipboard) that can be detached from the base of the device.
While the clipboard contains the CPU, the base can contain a discrete
GPU (dGPU). This dGPU is connected via a PCIe hot-plug port.

Currently D3cold is disallowed for all hot-plug ports. On the Surface
Book 2 and 3, this leads to increased power consumption during suspend
and when the dGPU is not used (i.e. runtime suspended). This can be
observed not only in battery drain, but also by the dGPU getting notably
warm while suspended and not in D3cold.

Testing shows that the Surface Books behave well with D3cold enabled for
hot-plug ports, alleviating the aforementioned issues. Thus white-list
D3cold for hot-plug ports on those devices.

Note: PCIe hot-plug signalling while the device is in D3cold is handled
via ACPI, out-of-band interrupts, and the surface_hotplug driver
(combined). The device will work without the surface_hotplug driver,
however, device removal/addition will only be detected on device resume.

Signed-off-by: Maximilian Luz <luzmaximilian@gmail.com>
Patchset: surface-sam
---
 drivers/pci/pci.c | 31 +++++++++++++++++++++++++++++--
 1 file changed, 29 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index c9338f914a0e..94371684783e 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -2679,6 +2679,32 @@ static const struct dmi_system_id bridge_d3_blacklist[] = {
 	{ }
 };
 
+static const struct dmi_system_id bridge_d3_hotplug_whitelist[] = {
+#ifdef CONFIG_X86
+	{
+		/*
+		 * Microsoft Surface Books have a hot-plug root port for the
+		 * discrete GPU (the device containing it can be detached form
+		 * the top-part, containing the cpu).
+		 *
+		 * If this discrete GPU is not transitioned into D3cold for
+		 * suspend, the device will become notably warm and also
+		 * consume a lot more power than desirable.
+		 *
+		 * We assume that since those devices have been confirmed
+		 * working with D3, future Surface devices will too. So let's
+		 * keep this match generic.
+		 */
+		.ident = "Microsoft Surface",
+		.matches = {
+			DMI_EXACT_MATCH(DMI_SYS_VENDOR, "Microsoft Corporation"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Surface"),
+		},
+	},
+#endif
+	{ }
+};
+
 /**
  * pci_bridge_d3_possible - Is it possible to put the bridge into D3
  * @bridge: Bridge to check
@@ -2719,10 +2745,11 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge)
 		/*
 		 * Hotplug ports handled natively by the OS were not validated
 		 * by vendors for runtime D3 at least until 2018 because there
-		 * was no OS support.
+		 * was no OS support. Explicitly whitelist systems that have
+		 * been confirmed working.
 		 */
 		if (bridge->is_hotplug_bridge)
-			return false;
+			return dmi_check_system(bridge_d3_hotplug_whitelist);
 
 		if (dmi_check_system(bridge_d3_blacklist))
 			return false;
-- 
2.29.2

From 901ceb17f1b911bb545fa41285344bb7573fb326 Mon Sep 17 00:00:00 2001
From: Maximilian Luz <luzmaximilian@gmail.com>
Date: Mon, 9 Nov 2020 14:23:00 +0100
Subject: [PATCH] PCI: Run platform power transition on initial D0 entry

On some devices and platforms, the initial platform power state is not
in sync with the power state of the PCI device.

pci_enable_device_flags() updates the state of a PCI device by reading
from the the PCI_PM_CTRL register. This may change the stored power
state of the device without running the appropriate platform power
transition.

Due to the stored power-state being changed, the later call to
pci_set_power_state(..., PCI_D0) in do_pci_enable_device() can evaluate
to a no-op if the stored state has been changed to D0 via that. This
will then prevent the appropriate platform power transition to be run,
which can on some devices and platforms lead to platform and PCI power
state being entirely different, i.e. out-of-sync. On ACPI platforms,
this can lead to power resources not being turned on, even though they
are marked as required for D0.

Specifically, on the Microsoft Surface Book 2 and 3, some ACPI power
regions that should be "on" for the D0 state (and others) are
initialized as "off" in ACPI, whereas the PCI device is in D0. As the
state is updated in pci_enable_device_flags() without ensuring that the
platform state is also updated, the power resource will never be
properly turned on. Instead, it lives in a sort of on-but-marked-as-off
zombie-state, which confuses things down the line when attempting to
transition the device into D3cold: As the resource is already marked as
off, it won't be turned off and the device does not fully enter D3cold,
causing increased power consumption during (runtime-)suspend.

By replacing pci_set_power_state() in do_pci_enable_device() with
pci_power_up(), we can force pci_platform_power_transition() to be
called, which will then check if the platform power state needs updating
and appropriate actions need to be taken.

Signed-off-by: Maximilian Luz <luzmaximilian@gmail.com>
Patchset: surface-sam
---
 drivers/pci/pci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 94371684783e..56c564c698bd 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -1664,7 +1664,7 @@ static int do_pci_enable_device(struct pci_dev *dev, int bars)
 	u16 cmd;
 	u8 pin;
 
-	err = pci_set_power_state(dev, PCI_D0);
+	err = pci_power_up(dev);
 	if (err < 0 && err != -EIO)
 		return err;
 
-- 
2.29.2

From e025027564ad4c665df448213db6041640b0eb5e Mon Sep 17 00:00:00 2001
From: Maximilian Luz <luzmaximilian@gmail.com>
Date: Sat, 31 Oct 2020 20:46:33 +0100
Subject: [PATCH] PCI: Add sysfs attribute for PCI device power state

While most PCI power-states can be queried from user-space via lspci,
this has some limits. Specifically, lspci fails to provide an accurate
value when the device is in D3cold as it has to resume the device before
it can access its power state via the configuration space, leading to it
reporting D0 or another on-state. Thus lspci can, for example, not be
used to diagnose power-consumption issues for devices that can enter
D3cold or to ensure that devices properly enter D3cold at all.

To alleviate this issue, introduce a new sysfs device attribute for the
PCI power state, showing the current power state as seen by the kernel.

Signed-off-by: Maximilian Luz <luzmaximilian@gmail.com>
Patchset: surface-sam
---
 Documentation/ABI/testing/sysfs-bus-pci |  9 +++++++++
 drivers/pci/pci-sysfs.c                 | 12 ++++++++++++
 2 files changed, 21 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-bus-pci b/Documentation/ABI/testing/sysfs-bus-pci
index 450296cc7948..881040af2611 100644
--- a/Documentation/ABI/testing/sysfs-bus-pci
+++ b/Documentation/ABI/testing/sysfs-bus-pci
@@ -360,3 +360,12 @@ Contact:	Heiner Kallweit <hkallweit1@gmail.com>
 Description:	If ASPM is supported for an endpoint, these files can be
 		used to disable or enable the individual power management
 		states. Write y/1/on to enable, n/0/off to disable.
+
+What:		/sys/bus/pci/devices/.../power_state
+Date:		November 2020
+Contact:	Linux PCI developers <linux-pci@vger.kernel.org>
+Description:
+		This file contains the current PCI power state of the device.
+		The value comes from the PCI kernel device state and can be one
+		of: "unknown", "error", "D0", D1", "D2", "D3hot", "D3cold".
+		The file is read only.
diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
index 6d78df981d41..17f186ce8e87 100644
--- a/drivers/pci/pci-sysfs.c
+++ b/drivers/pci/pci-sysfs.c
@@ -124,6 +124,17 @@ static ssize_t cpulistaffinity_show(struct device *dev,
 }
 static DEVICE_ATTR_RO(cpulistaffinity);
 
+/* PCI power state */
+static ssize_t power_state_show(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	struct pci_dev *pci_dev = to_pci_dev(dev);
+	pci_power_t state = READ_ONCE(pci_dev->current_state);
+
+	return sprintf(buf, "%s\n", pci_power_name(state));
+}
+static DEVICE_ATTR_RO(power_state);
+
 /* show resources */
 static ssize_t resource_show(struct device *dev, struct device_attribute *attr,
 			     char *buf)
@@ -581,6 +592,7 @@ static ssize_t driver_override_show(struct device *dev,
 static DEVICE_ATTR_RW(driver_override);
 
 static struct attribute *pci_dev_attrs[] = {
+	&dev_attr_power_state.attr,
 	&dev_attr_resource.attr,
 	&dev_attr_vendor.attr,
 	&dev_attr_device.attr,
-- 
2.29.2

From f87195e108325927ad44baf3ed464d3c878c85c0 Mon Sep 17 00:00:00 2001
From: Maximilian Luz <luzmaximilian@gmail.com>
Date: Mon, 17 Aug 2020 01:23:20 +0200
Subject: [PATCH] misc: surface_sam: Add file2alias support for Surface SAM
 devices

Implement file2alias support for Surface System Aggregator Module (SSAM)
devices. This allows modules to be auto-loaded for specific devices via
their respective module alias.

Patchset: surface-sam
---
 include/linux/mod_devicetable.h   | 17 +++++++++++++++++
 scripts/mod/devicetable-offsets.c |  7 +++++++
 scripts/mod/file2alias.c          | 21 +++++++++++++++++++++
 3 files changed, 45 insertions(+)

diff --git a/include/linux/mod_devicetable.h b/include/linux/mod_devicetable.h
index e14cbe444afc..a35f82fb36da 100644
--- a/include/linux/mod_devicetable.h
+++ b/include/linux/mod_devicetable.h
@@ -836,4 +836,21 @@ struct mhi_device_id {
 	kernel_ulong_t driver_data;
 };
 
+/* Surface System Aggregator Module */
+
+#define SSAM_MATCH_CHANNEL	0x1
+#define SSAM_MATCH_INSTANCE	0x2
+#define SSAM_MATCH_FUNCTION	0x4
+
+struct ssam_device_id {
+	__u8 match_flags;
+
+	__u8 category;
+	__u8 channel;
+	__u8 instance;
+	__u8 function;
+
+	kernel_ulong_t driver_data;
+};
+
 #endif /* LINUX_MOD_DEVICETABLE_H */
diff --git a/scripts/mod/devicetable-offsets.c b/scripts/mod/devicetable-offsets.c
index 010be8ba2116..35b9899b0b0f 100644
--- a/scripts/mod/devicetable-offsets.c
+++ b/scripts/mod/devicetable-offsets.c
@@ -241,5 +241,12 @@ int main(void)
 	DEVID(mhi_device_id);
 	DEVID_FIELD(mhi_device_id, chan);
 
+	DEVID(ssam_device_id);
+	DEVID_FIELD(ssam_device_id, match_flags);
+	DEVID_FIELD(ssam_device_id, category);
+	DEVID_FIELD(ssam_device_id, channel);
+	DEVID_FIELD(ssam_device_id, instance);
+	DEVID_FIELD(ssam_device_id, function);
+
 	return 0;
 }
diff --git a/scripts/mod/file2alias.c b/scripts/mod/file2alias.c
index 9599e2a3f1e6..079672e0747a 100644
--- a/scripts/mod/file2alias.c
+++ b/scripts/mod/file2alias.c
@@ -1364,6 +1364,26 @@ static int do_mhi_entry(const char *filename, void *symval, char *alias)
 	return 1;
 }
 
+/* Looks like: ssam:cNtNiNfN
+ *
+ * N is exactly 2 digits, where each is an upper-case hex digit.
+ */
+static int do_ssam_entry(const char *filename, void *symval, char *alias)
+{
+	DEF_FIELD(symval, ssam_device_id, match_flags);
+	DEF_FIELD(symval, ssam_device_id, category);
+	DEF_FIELD(symval, ssam_device_id, channel);
+	DEF_FIELD(symval, ssam_device_id, instance);
+	DEF_FIELD(symval, ssam_device_id, function);
+
+	sprintf(alias, "ssam:c%02X", category);
+	ADD(alias, "t", match_flags & SSAM_MATCH_CHANNEL, channel);
+	ADD(alias, "i", match_flags & SSAM_MATCH_INSTANCE, instance);
+	ADD(alias, "f", match_flags & SSAM_MATCH_FUNCTION, function);
+
+	return 1;
+}
+
 /* Does namelen bytes of name exactly match the symbol? */
 static bool sym_is(const char *name, unsigned namelen, const char *symbol)
 {
@@ -1438,6 +1458,7 @@ static const struct devtable devtable[] = {
 	{"tee", SIZE_tee_client_device_id, do_tee_entry},
 	{"wmi", SIZE_wmi_device_id, do_wmi_entry},
 	{"mhi", SIZE_mhi_device_id, do_mhi_entry},
+	{"ssam", SIZE_ssam_device_id, do_ssam_entry},
 };
 
 /* Create MODULE_ALIAS() statements.
-- 
2.29.2

From 7debd8e51563a4742e06c56857891a13227211b5 Mon Sep 17 00:00:00 2001
From: Maximilian Luz <luzmaximilian@gmail.com>
Date: Mon, 17 Aug 2020 01:44:30 +0200
Subject: [PATCH] misc: Add support for Surface System Aggregator Module

Add support for the Surface System Aggregator Module (SSAM), an embedded
controller that can be found on 5th and later generation Microsoft
Surface devices. The responsibilities of this EC vary from device to
device. It provides battery information on all 5th and later generation
devices, temperature sensor and cooling capability access, functionality
for clipboard detaching on the Surface Books (2 and 3), as well as
HID-over-SSAM input devices, including keyboard on the Surface Laptop 1
and 2, and keyboard as well as touchpad input on the Surface Laptop 3
and Surface Book 3.

Patchset: surface-sam
---
 Documentation/driver-api/index.rst            |    1 +
 .../surface_aggregator/client-api.rst         |   38 +
 .../driver-api/surface_aggregator/client.rst  |  394 +++
 .../surface_aggregator/clients/cdev.rst       |   85 +
 .../surface_aggregator/clients/dtx.rst        |  712 +++++
 .../surface_aggregator/clients/index.rst      |   22 +
 .../surface_aggregator/clients/san.rst        |   44 +
 .../driver-api/surface_aggregator/index.rst   |   21 +
 .../surface_aggregator/internal-api.rst       |   67 +
 .../surface_aggregator/internal.rst           |   50 +
 .../surface_aggregator/overview.rst           |   76 +
 .../driver-api/surface_aggregator/ssh.rst     |  343 +++
 drivers/misc/Kconfig                          |    1 +
 drivers/misc/Makefile                         |    1 +
 drivers/misc/surface_aggregator/Kconfig       |   67 +
 drivers/misc/surface_aggregator/Makefile      |   18 +
 drivers/misc/surface_aggregator/bus.c         |  424 +++
 drivers/misc/surface_aggregator/bus.h         |   27 +
 .../misc/surface_aggregator/clients/Kconfig   |  151 +
 .../misc/surface_aggregator/clients/Makefile  |   11 +
 .../clients/surface_acpi_notify.c             |  884 ++++++
 .../clients/surface_aggregator_cdev.c         |  299 ++
 .../clients/surface_aggregator_registry.c     |  656 +++++
 .../clients/surface_battery.c                 | 1192 ++++++++
 .../surface_aggregator/clients/surface_dtx.c  | 1279 +++++++++
 .../surface_aggregator/clients/surface_hid.c  |  922 ++++++
 .../clients/surface_hotplug.c                 |  269 ++
 .../clients/surface_perfmode.c                |  122 +
 drivers/misc/surface_aggregator/controller.c  | 2557 +++++++++++++++++
 drivers/misc/surface_aggregator/controller.h  |  288 ++
 drivers/misc/surface_aggregator/core.c        |  831 ++++++
 drivers/misc/surface_aggregator/ssh_msgb.h    |  201 ++
 .../surface_aggregator/ssh_packet_layer.c     | 2009 +++++++++++++
 .../surface_aggregator/ssh_packet_layer.h     |  175 ++
 drivers/misc/surface_aggregator/ssh_parser.c  |  229 ++
 drivers/misc/surface_aggregator/ssh_parser.h  |  157 +
 .../surface_aggregator/ssh_request_layer.c    | 1254 ++++++++
 .../surface_aggregator/ssh_request_layer.h    |  142 +
 drivers/misc/surface_aggregator/trace.h       |  625 ++++
 include/linux/mod_devicetable.h               |    5 +-
 include/linux/surface_acpi_notify.h           |   39 +
 include/linux/surface_aggregator/controller.h |  832 ++++++
 include/linux/surface_aggregator/device.h     |  430 +++
 include/linux/surface_aggregator/serial_hub.h |  659 +++++
 include/uapi/linux/surface_aggregator/cdev.h  |   58 +
 include/uapi/linux/surface_aggregator/dtx.h   |  150 +
 scripts/mod/devicetable-offsets.c             |    3 +-
 scripts/mod/file2alias.c                      |   10 +-
 48 files changed, 18823 insertions(+), 7 deletions(-)
 create mode 100644 Documentation/driver-api/surface_aggregator/client-api.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/client.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/clients/cdev.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/clients/dtx.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/clients/index.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/clients/san.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/index.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/internal-api.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/internal.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/overview.rst
 create mode 100644 Documentation/driver-api/surface_aggregator/ssh.rst
 create mode 100644 drivers/misc/surface_aggregator/Kconfig
 create mode 100644 drivers/misc/surface_aggregator/Makefile
 create mode 100644 drivers/misc/surface_aggregator/bus.c
 create mode 100644 drivers/misc/surface_aggregator/bus.h
 create mode 100644 drivers/misc/surface_aggregator/clients/Kconfig
 create mode 100644 drivers/misc/surface_aggregator/clients/Makefile
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_acpi_notify.c
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_aggregator_cdev.c
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_aggregator_registry.c
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_battery.c
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_dtx.c
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_hid.c
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_hotplug.c
 create mode 100644 drivers/misc/surface_aggregator/clients/surface_perfmode.c
 create mode 100644 drivers/misc/surface_aggregator/controller.c
 create mode 100644 drivers/misc/surface_aggregator/controller.h
 create mode 100644 drivers/misc/surface_aggregator/core.c
 create mode 100644 drivers/misc/surface_aggregator/ssh_msgb.h
 create mode 100644 drivers/misc/surface_aggregator/ssh_packet_layer.c
 create mode 100644 drivers/misc/surface_aggregator/ssh_packet_layer.h
 create mode 100644 drivers/misc/surface_aggregator/ssh_parser.c
 create mode 100644 drivers/misc/surface_aggregator/ssh_parser.h
 create mode 100644 drivers/misc/surface_aggregator/ssh_request_layer.c
 create mode 100644 drivers/misc/surface_aggregator/ssh_request_layer.h
 create mode 100644 drivers/misc/surface_aggregator/trace.h
 create mode 100644 include/linux/surface_acpi_notify.h
 create mode 100644 include/linux/surface_aggregator/controller.h
 create mode 100644 include/linux/surface_aggregator/device.h
 create mode 100644 include/linux/surface_aggregator/serial_hub.h
 create mode 100644 include/uapi/linux/surface_aggregator/cdev.h
 create mode 100644 include/uapi/linux/surface_aggregator/dtx.h

diff --git a/Documentation/driver-api/index.rst b/Documentation/driver-api/index.rst
index 6567187e7687..e36363f0972b 100644
--- a/Documentation/driver-api/index.rst
+++ b/Documentation/driver-api/index.rst
@@ -99,6 +99,7 @@ available subsections can be seen below.
    serial/index
    sm501
    smsc_ece1099
+   surface_aggregator/index
    switchtec
    sync_file
    vfio-mediated-device
diff --git a/Documentation/driver-api/surface_aggregator/client-api.rst b/Documentation/driver-api/surface_aggregator/client-api.rst
new file mode 100644
index 000000000000..a1117d57036a
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/client-api.rst
@@ -0,0 +1,38 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+===============================
+Client Driver API Documentation
+===============================
+
+.. contents::
+    :depth: 2
+
+
+Serial Hub Communication
+========================
+
+.. kernel-doc:: include/linux/surface_aggregator/serial_hub.h
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_packet_layer.c
+    :export:
+
+
+Controller and Core Interface
+=============================
+
+.. kernel-doc:: include/linux/surface_aggregator/controller.h
+
+.. kernel-doc:: drivers/misc/surface_aggregator/controller.c
+    :export:
+
+.. kernel-doc:: drivers/misc/surface_aggregator/core.c
+    :export:
+
+
+Client Bus and Client Device API
+================================
+
+.. kernel-doc:: include/linux/surface_aggregator/device.h
+
+.. kernel-doc:: drivers/misc/surface_aggregator/bus.c
+    :export:
diff --git a/Documentation/driver-api/surface_aggregator/client.rst b/Documentation/driver-api/surface_aggregator/client.rst
new file mode 100644
index 000000000000..f1275bf53837
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/client.rst
@@ -0,0 +1,394 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+.. |ssam_controller| replace:: :c:type:`struct ssam_controller <ssam_controller>`
+.. |ssam_device| replace:: :c:type:`struct ssam_device <ssam_device>`
+.. |ssam_device_driver| replace:: :c:type:`struct ssam_device_driver <ssam_device_driver>`
+.. |ssam_client_bind| replace:: :c:func:`ssam_client_bind`
+.. |ssam_client_link| replace:: :c:func:`ssam_client_link`
+.. |ssam_get_controller| replace:: :c:func:`ssam_get_controller`
+.. |ssam_controller_get| replace:: :c:func:`ssam_controller_get`
+.. |ssam_controller_put| replace:: :c:func:`ssam_controller_put`
+.. |ssam_device_alloc| replace:: :c:func:`ssam_device_alloc`
+.. |ssam_device_add| replace:: :c:func:`ssam_device_add`
+.. |ssam_device_remove| replace:: :c:func:`ssam_device_remove`
+.. |ssam_device_driver_register| replace:: :c:func:`ssam_device_driver_register`
+.. |ssam_device_driver_unregister| replace:: :c:func:`ssam_device_driver_unregister`
+.. |module_ssam_device_driver| replace:: :c:func:`module_ssam_device_driver`
+.. |SSAM_DEVICE| replace:: :c:func:`SSAM_DEVICE`
+.. |ssam_notifier_register| replace:: :c:func:`ssam_notifier_register`
+.. |ssam_notifier_unregister| replace:: :c:func:`ssam_notifier_unregister`
+.. |ssam_request_sync| replace:: :c:func:`ssam_request_sync`
+.. |ssam_event_mask| replace:: :c:type:`enum ssam_event_mask <ssam_event_mask>`
+
+
+======================
+Writing Client Drivers
+======================
+
+For the API documentation, refer to:
+
+.. toctree::
+   :maxdepth: 2
+
+   client-api
+
+
+Overview
+========
+
+Client drivers can be set up in two main ways, depending on how the
+corresponding device is made available to the system. We specifically
+differentiate between devices that are presented to the system via one of
+the conventional ways, e.g. as platform devices via ACPI, and devices that
+are non-discoverable and instead need to be explicitly provided by some
+other mechanism, as discussed further below.
+
+
+Non-SSAM Client Drivers
+=======================
+
+All communication with the SAM EC is handled via the |ssam_controller|
+representing that EC to the kernel. Drivers targetting a non-SSAM device
+(and thus not being a |ssam_device_driver|) need to explicitly establish a
+connection/relation to that controller. This can be done via the
+|ssam_client_bind| function. Said function returns a reference to the SSAM
+controller, but, more importantly, also establishes a device link between
+client device and controller (this can also be done separate via
+|ssam_client_link|). It is important to do this, as it, first, guarantees
+that the returned controller is valid for use in the client driver for as
+long as this driver is bound to its device, i.e. that the driver gets
+unbound before the controller ever becomes invalid, and, second, as it
+ensures correct suspend/resume ordering. This setup should be done in the
+driver's probe function, and may be used to defer probing in case the SSAM
+subsystem is not ready yet, for example:
+
+.. code-block:: c
+
+   static int client_driver_probe(struct platform_device *pdev)
+   {
+           struct ssam_controller *ctrl;
+           int status;
+
+           status = ssam_client_bind(&pdev->dev, &ctrl);
+           if (status)
+                   return status == -ENXIO ? -EPROBE_DEFER : status;
+
+           // ...
+
+           return 0;
+   }
+
+The controller may be separately obtained via |ssam_get_controller| and its
+lifetime be guaranteed via |ssam_controller_get| and |ssam_controller_put|.
+Note that none of these functions, however, guarantee that the controller
+will not be shut down or suspended. These functions essentially only operate
+on the reference, i.e. only guarantee a bare minimum of accessibility
+without any guarantees at all on practical operability.
+
+
+Adding SSAM Devices
+===================
+
+If a device does not already exist/is not already provided via conventional
+means, it should be provided as |ssam_device| via the SSAM client device
+hub. New devices can be added to this hub by entering their UID into the
+corresponding registry. SSAM devices can also be manually allocated via
+|ssam_device_alloc|, subsequently to which they have to be added via
+|ssam_device_add| and eventually removed via |ssam_device_remove|. By
+default, the parent of the device is set to the controller device provided
+for allocation, however this may be changed before the device is added. Note
+that, when changing the parent device, care must be taken to ensure that the
+controller lifetime and suspend/resume ordering guarantees, in the default
+setup provided through the parent-child relation, are preserved. If
+necessary, by use of |ssam_client_link| as is done for non-SSAM client
+drivers and described in more detail above.
+
+A client device must always be removed by the party which added the
+respective device before the controller shuts down. Such removal can be
+guaranteed by linking the driver providing the SSAM device to the controller
+via |ssam_client_link|, causing it to unbind before the controller driver
+unbinds. Client devices registered with the controller as parent are
+automatically removed when the controller shuts down, but this should not be
+relied upon, especially as this does not extend to client devices with a
+different parent.
+
+
+SSAM Client Drivers
+===================
+
+SSAM client device drivers are, in essence, no different than other device
+driver types. They are represented via |ssam_device_driver| and bind to a
+|ssam_device| via its UID (:c:type:`struct ssam_device.uid <ssam_device>`)
+member and the match table
+(:c:type:`struct ssam_device_driver.match_table <ssam_device_driver>`),
+which should be set when declaring the driver struct instance. Refer to the
+|SSAM_DEVICE| macro documentation for more details on how to define members
+of the driver's match table.
+
+The UID for SSAM client devices consists of a ``domain``, a ``category``,
+a ``target``, an ``instance``, and a ``function``. The ``domain`` is used
+differentiate between physical SAM devices
+(:c:type:`SSAM_DOMAIN_SERIALHUB <ssam_device_domain>`), i.e. devices that can
+be accessed via the Surface Serial Hub, and virtual ones
+(:c:type:`SSAM_DOMAIN_VIRTUAL <ssam_device_domain>`), such as client-device
+hubs, that have no real representation on the SAM EC and are solely used on
+the kernel/driver-side. For physical devices, ``category`` represents the
+target category, ``target`` the target ID, and ``instace`` the instance ID
+used to access the physical SAM device. In addition, ``function`` references
+a specific device functionality, but has no meaning to the SAM EC. The
+(default) name of a client device is generated based on its UID.
+
+A driver instance can be registered via |ssam_device_driver_register| and
+unregistered via |ssam_device_driver_unregister|. For convenience, the
+|module_ssam_device_driver| macro may be used to define module init- and
+exit-functions registering the driver.
+
+The controller associated with a SSAM client device can be found in its
+:c:type:`struct ssam_device.ctrl <ssam_device>` member. This reference is
+guaranteed to be valid for at least as long as the client driver is bound,
+but should also be valid for as long as the client device exists. Note,
+however, that access outside of the bound client driver must ensure that the
+controller device is not suspended while making any requests or
+(un)registering event notifiers (and thus should generally be avoided). This
+is guaranteed when the controller is accessed from inside the bound client
+driver.
+
+
+Making Synchronous Requests
+===========================
+
+Synchronous requests are (currently) the main form of host-initiated
+communication with the EC. There are a couple of ways to define and execute
+such requests, however, most of them boil down to something similar as shown
+in the example below. This example defines a write-read request, meaning
+that the caller provides an argument to the SAM EC and receives a response.
+The caller needs to know the (maximum) length of the response payload and
+provide a buffer for it.
+
+Care must be taken to ensure that any command payload data passed to the SAM
+EC is provided in little-endian format and, similarly, any response payload
+data received from it is converted from little-endian to host endianness.
+
+.. code-block:: c
+
+   int perform_request(struct ssam_controller *ctrl, u32 arg, u32 *ret)
+   {
+           struct ssam_request rqst;
+           struct ssam_response resp;
+           int status;
+
+           /* Convert request argument to little-endian. */
+           __le32 arg_le = cpu_to_le32(arg);
+           __le32 ret_le = cpu_to_le32(0);
+
+           /*
+            * Initialize request specification. Replace this with your values.
+            * The rqst.payload field may be NULL if rqst.length is zero,
+            * indicating that the request does not have any argument.
+            *
+            * Note: The request parameters used here are not valid, i.e.
+            *       they do not correspond to an actual SAM/EC request.
+            */
+           rqst.target_category = SSAM_SSH_TC_SAM;
+           rqst.target_id = 0x01;
+           rqst.command_id = 0x02;
+           rqst.instance_id = 0x03;
+           rqst.flags = SSAM_REQUEST_HAS_RESPONSE;
+           rqst.length = sizeof(arg_le);
+           rqst.payload = (u8 *)&arg_le;
+
+           /* Initialize request response. */
+           resp.capacity = sizeof(ret_le);
+           resp.length = 0;
+           resp.pointer = (u8 *)&ret_le;
+
+           /*
+            * Perform actual request. The response pointer may be null in case
+            * the request does not have any response. This must be consistent
+            * with the SSAM_REQUEST_HAS_RESPONSE flag set in the specification
+            * above.
+            */
+           status = ssam_request_sync(ctrl, &rqst, &resp);
+           if (status)
+               return status;
+
+           /*
+            * Alternatively use
+            *
+            *   ssam_request_sync_onstack(ctrl, &rqst, &resp, sizeof(arg_le));
+            *
+            * to perform the request, allocating the message buffer directly
+            * on the stack as opposed to via kzalloc().
+            */
+
+           /*
+            * Convert request response back to native format. Note that in the
+            * error case, this value is not touched.
+            */
+           *ret = le32_to_cpu(ret_le);
+
+           return status;
+   }
+
+Note that |ssam_request_sync| in its essence is a wrapper over lower-level
+request primitives, which may also be used to perform requests. Refer to its
+implementation and documentation for more details.
+
+An arguably more user-friendly way of defining such functions is by using
+one of the generator macros, for example via:
+
+.. code-block:: c
+
+   SSAM_DEFINE_SYNC_REQUEST_W(__ssam_tmp_perf_mode_set, __le32, {
+           .target_category = SSAM_SSH_TC_TMP,
+           .target_id       = 0x01,
+           .command_id      = 0x03,
+           .instance_id     = 0x00,
+   });
+
+This example defines a function
+
+.. code-block:: c
+
+   int __ssam_tmp_perf_mode_set(struct ssam_controller *ctrl, const __le32 *arg);
+
+executing the specified request, with the controller passed in when calling
+said function. In this example, the argument is provided via the ``arg``
+pointer. Note that the generated function allocates the message buffer on
+the stack. Thus, if the argument provided via the request is large, these
+kinds of macros should be avoided. Also note that, in contrast to the
+previous non-macro example, this function does not do any endianness
+conversion, which has to be handled by the caller. Apart from those
+differences the function generated by the macro is similar to the one
+provided in the non-macro example above.
+
+The full list of such function-generating macros is
+
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_N` for requests without return value and
+  without argument.
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_R` for requests with return value but no
+  argument.
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_W` for requests without return value but
+  with argument.
+
+Refer to their respective documentation for more details. For each one of
+these macros, a special variant is provided, which targets request types
+applicable to multiple instances of the same device type:
+
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_MD_N`
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_MD_R`
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_MD_W`
+
+The difference of those macros to the previously mentioned versions is, that
+the device target and instance IDs are not fixed for the generated function,
+but instead have to be provided by the caller of said function.
+
+Additionally, variants for direct use with client devices, i.e.
+|ssam_device|, are also provided. These can, for example, be used as
+follows:
+
+.. code-block:: c
+
+   SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_sta, __le32, {
+           .target_category = SSAM_SSH_TC_BAT,
+           .command_id      = 0x01,
+   });
+
+This invocation of the macro defines a function
+
+.. code-block:: c
+
+   int ssam_bat_get_sta(struct ssam_device *sdev, __le32 *ret);
+
+executing the specified request, using the device IDs and controller given
+in the client device. The full list of such macros for client devices is:
+
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_CL_N`
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_CL_R`
+- :c:func:`SSAM_DEFINE_SYNC_REQUEST_CL_W`
+
+
+Handling Events
+===============
+
+To receive events from the SAM EC, an event notifier must be registered for
+the desired event via |ssam_notifier_register|. The notifier must be
+unregistered via |ssam_notifier_unregister| once it is not required any
+more.
+
+Event notifiers are registered by providing (at minimum) a callback to call
+in case an event has been received, the registry specifying how the event
+should be enabled, an event ID specifying for which target category and,
+optionally and depending on the registry used, for which instance ID events
+should be enabled, and finally, flags describing how the EC will send these
+events. Additionally, a priority for the respective notifier may be
+specified, which determines its order in relation to any other notifier
+registered for the same target category.
+
+By default, event notifiers will receive all events for the specific target
+category, regardless of the instance ID specified when registering the
+notifier. The core may be instructed to only call a notifier if the target
+ID or instance ID (or both) of the event match the ones implied by the
+notifier IDs (in case of target ID, the target ID of the registry), by
+providing an event mask (see |ssam_event_mask|).
+
+In general, the target ID of the registry is also the target ID of the
+enabled event (with the notable exception being keyboard input events on the
+Surface Laptop 1 and 2, which are enabled via a registry with target ID 1,
+but provide events with target ID 2).
+
+A full example for registering an event notifier and handling received
+events is provided below:
+
+.. code-block:: c
+
+   u32 notifier_callback(struct ssam_event_notifier *nf,
+                         const struct ssam_event *event)
+   {
+           int status = ...
+
+           /* Handle the event here ... */
+
+           /* Convert return value and indicate that we handled the event. */
+           return ssam_notifier_from_errno(status) | SSAM_NOTIF_HANDLED;
+   }
+
+   int setup_notifier(struct ssam_device *sdev,
+                      struct ssam_event_notifier *nf)
+   {
+           /* Set priority wrt. other handlers of same target category. */
+           nf->base.priority = 1;
+
+           /* Set event/notifier callback. */
+           nf->base.fn = notifier_callback;
+
+           /* Specify event registry, i.e. how events get enabled/disabled. */
+           nf->event.reg = SSAM_EVENT_REGISTRY_KIP;
+
+           /* Specify which event to enable/disable */
+           nf->event.id.target_category = sdev->uid.category;
+           nf->event.id.instance = sdev->uid.instance;
+
+           /*
+            * Specify for which events the notifier callback gets executed.
+            * This essentially tells the core if it can skip notifiers that
+            * don't have target or instance IDs matching those of the event.
+            */
+           nf->event.mask = SSAM_EVENT_MASK_STRICT;
+
+           /* Specify event flags. */
+           nf->event.flags = SSAM_EVENT_SEQUENCED;
+
+           return ssam_notifier_register(sdev->ctrl, nf);
+   }
+
+Multiple event notifiers can be registered for the same event. The event
+handler core takes care of enabling and disabling events when notifiers are
+registered and unregistered, by keeping track of how many notifiers for a
+specific event (combination of registry, event target category, and event
+instance ID) are currently registered. This means that a specific event will
+be enabled when the first notifier for it is being registered and disabled
+when the last notifier for it is being unregistered. Note that the event
+flags are therefore only used on the first registered notifier, however, one
+should take care that notifiers for a specific event are always registered
+with the same flag and it is considered a bug to do otherwise.
diff --git a/Documentation/driver-api/surface_aggregator/clients/cdev.rst b/Documentation/driver-api/surface_aggregator/clients/cdev.rst
new file mode 100644
index 000000000000..720c9c03aa0e
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/clients/cdev.rst
@@ -0,0 +1,85 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+.. |u8| replace:: :c:type:`u8 <u8>`
+.. |u16| replace:: :c:type:`u16 <u16>`
+.. |ssam_cdev_request| replace:: :c:type:`struct ssam_cdev_request <ssam_cdev_request>`
+.. |ssam_request_flags| replace:: :c:type:`enum ssam_request_flags <ssam_request_flags>`
+
+==============================
+User-Space EC Interface (cdev)
+==============================
+
+The ``surface_aggregator_cdev`` module provides a misc-device for the SSAM
+controller to allow for a (more or less) direct connection from user-space to
+the SAM EC. It is intended to be used for development and debugging, and
+therefore should not be used or relied upon in any other way. Note that this
+module is not loaded automatically, but instead must be loaded manually.
+
+The provided interface is accessible through the ``/dev/surface/aggregator``
+device-file. All functionality of this interface is provided via IOCTLs.
+These IOCTLs and their respective input/output parameter structs are defined in
+``include/uapi/linux/surface_aggregator/cdev.h``.
+
+
+Controller IOCTLs
+=================
+
+The following IOCTLs are provided:
+
+.. flat-table:: Controller IOCTLs
+   :widths: 1 1 1 1 4
+   :header-rows: 1
+
+   * - Type
+     - Number
+     - Direction
+     - Name
+     - Description
+
+   * - ``0xA5``
+     - ``1``
+     - ``WR``
+     - ``REQUEST``
+     - Perform synchronous SAM request.
+
+
+``REQUEST``
+-----------
+
+Defined as ``_IOWR(0xA5, 1, struct ssam_cdev_request)``.
+
+Executes a synchronous SAM request. The request specification is passed in
+as argument of type |ssam_cdev_request|, which is then written to/modified
+by the IOCTL to return status and result of the request.
+
+Request payload data must be allocated separately and is passed in via the
+``payload.data`` and ``payload.length`` members. If a response is required,
+the response buffer must be allocated by the caller and passed in via the
+``response.data`` member. The ``response.length`` member must be set to the
+capacity of this buffer, or if no response is required, zero. Upon
+completion of the request, the call will write the response to the response
+buffer (if its capacity allows it) and overwrite the length field with the
+actual size of the response, in bytes.
+
+Additionally, if the request has a response, this should be indicated via
+the request flags, as is done with in-kernel requests. Request flags can be
+set via the ``flags`` member and the values correspond to the values found
+in |ssam_request_flags|.
+
+Finally, the status of the request itself is returned in the ``status``
+member (a negative value indicating failure). Note that failure indication
+of the IOCTL is separated from failure indication of the request: The IOCTL
+returns a negative status code if anything failed during setup of the
+request (``-EFAULT``) or if the provided argument or any of its fields are
+invalid (``-EINVAL``). In this case, the status value of the request
+argument may be set, providing more detail on what went wrong (e.g.
+``-ENOMEM`` for out-of-memory), but this value may also be zero. The IOCTL
+will return with a zero status code in case the request has been set up,
+submitted, and completed (i.e. handed back to user-space) successfully from
+inside the IOCTL, but the request ``status`` member may still be negative in
+case the actual execution of the request failed after it has been submitted.
+
+A full definition of the argument struct is provided below:
+
+.. kernel-doc:: include/uapi/linux/surface_aggregator/cdev.h
+   :functions: ssam_cdev_request
diff --git a/Documentation/driver-api/surface_aggregator/clients/dtx.rst b/Documentation/driver-api/surface_aggregator/clients/dtx.rst
new file mode 100644
index 000000000000..e974c2b04e9f
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/clients/dtx.rst
@@ -0,0 +1,712 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+.. |__u16| replace:: :c:type:`__u16 <__u16>`
+.. |sdtx_event| replace:: :c:type:`struct sdtx_event <sdtx_event>`
+.. |sdtx_event_code| replace:: :c:type:`enum sdtx_event_code <sdtx_event_code>`
+.. |sdtx_base_info| replace:: :c:type:`struct sdtx_base_info <sdtx_base_info>`
+.. |sdtx_device_mode| replace:: :c:type:`struct sdtx_device_mode <sdtx_device_mode>`
+
+======================================================
+User-Space DTX (Clipboard Detachment System) Interface
+======================================================
+
+The ``surface_dtx`` driver is responsible for proper clipboard detachment
+and re-attachment handling. To this end, it provides the ``/dev/surface/dtx``
+device file, through which it can interface with a user-space daemon. This
+daemon is then ultimately responsible for determining and taking necessary
+actions, such as unmounting devices attached to the base,
+unloading/reloading the graphics-driver, user-notifications, etc.
+
+There are two basic communication principles used in this driver: Commands
+(in other parts of the documentation also referred to as requests) and
+events. Commands are sent to the EC and may have a different implications in
+different contexts. Events are sent by the EC upon some internal state
+change. Commands are always driver-initiated, whereas events are always
+initiated by the EC.
+
+.. contents::
+
+Nomenclature
+============
+
+* **Clipboard:**
+  The detachable upper part of the Surface Book, housing the screen and CPU.
+
+* **Base:**
+  The lower part of the Surface Book from which the clipboard can be
+  detached, optionally (model dependent) housing the discrete GPU (dGPU).
+
+* **Latch:**
+  The mechanism keeping the clipboard attached to the base in normal
+  operation and allowing it to be detached when requested.
+
+* **Silently ignored commands:**
+  The command is accepted by the EC as a valid command and acknowledged
+  (following the standard communication protocol), but the EC does not act
+  upon it, i.e. ignores it.e upper part of the
+
+
+Detachment Process
+==================
+
+Warning: This part of the documentation is based on reverse engineering and
+testing and thus may contain errors or be incomplete.
+
+Latch States
+------------
+
+The latch mechanism has two major states: *open* and *closed*. In the
+*closed* state (default), the clipboard is secured to the base, whereas in
+the *open* state, the clipboard can be removed by a user.
+
+The latch can additionally be locked and, correspondingly, unlocked, which
+can influence the detachment procedure. Specifically, this locking mechanism
+is intended to prevent the the dGPU, positioned in the base of the device,
+from being hot-unplugged while in use. More details can be found in the
+documentation for the detachment procedure below. By default, the latch is
+unlocked.
+
+Detachment Procedure
+--------------------
+
+Note that the detachment process is governed fully by the EC. The
+``surface_dtx`` driver only relays events from the EC to user-space and
+commands from user-space to the EC, i.e. it does not influence this process.
+
+The detachment process is started with the user pressing the *detach* button
+on the base of the device or executing the ``SDTX_IOCTL_LATCH_REQUEST`` IOCTL.
+Following that:
+
+1. The EC turns on the indicator led on the detach-button, sends a
+   *detach-request* event (``SDTX_EVENT_REQUEST``), and awaits further
+   instructions/commands. In case the latch is unlocked, the led will flash
+   green. If the latch has been locked, the led will be solid red
+
+2. The event is, via the ``surface_dtx`` driver, relayed to user-space, where
+   an appropriate user-space daemon can handle it and send instructions back
+   to the EC via IOCTLs provided by this driver.
+
+3. The EC waits for instructions from user-space and acts according to them.
+   If the EC does not receive any instructions in a given period, it will
+   time out and continue as follows:
+
+   - If the latch is unlocked, the EC will open the latch and the clipboard
+     can be detached from the base. This is the exact behavior as without
+     this driver or any user-space daemon. See the ``SDTX_IOCTL_LATCH_CONFIRM``
+     description below for more details on the follow-up behavior of the EC.
+
+   - If the latch is locked, the EC will *not* open the latch, meaning the
+     clipboard cannot be detached from the base. Furthermore, the EC sends
+     an cancel event (``SDTX_EVENT_CANCEL``) detailing this with the cancel
+     reason ``SDTX_DETACH_TIMEDOUT`` (see :ref:`events` for details).
+
+Valid responses by a user-space daemon to a detachment request event are:
+
+- Execute ``SDTX_IOCTL_LATCH_REQUEST``. This will immediately abort the
+  detachment process. Furthermore, the EC will send a detach-request event,
+  similar to the user pressing the detach-button to cancel said process (see
+  below).
+
+- Execute ``SDTX_IOCTL_LATCH_CONFIRM``. This will cause the EC to open the
+  latch, after which the user can separate clipboard and base.
+
+  As this changes the latch state, a *latch-status* event
+  (``SDTX_EVENT_LATCH_STATUS``) will be sent once the latch has been opened
+  successfully. If the EC fails to open the latch, e.g. due to hardware
+  error or low battery, a latch-cancel event (``SDTX_EVENT_CANCEL``) will be
+  sent with the cancel reason indicating the specific failure.
+
+  If the latch is currently locked, the latch will automatically be
+  unlocked before it is opened.
+
+- Execute ``SDTX_IOCTL_LATCH_HEARTBEAT``. This will reset the internal timeout.
+  No other actions will be performed, i.e. the detachment process will neither
+  be completed nor canceled, and the EC will still be waiting for further
+  responses.
+
+- Execute ``SDTX_IOCTL_LATCH_CANCEL``. This will abort the detachment process,
+  similar to ``SDTX_IOCTL_LATCH_REQUEST``, described above, or the button
+  press, described below. A *generic request* event (``SDTX_EVENT_REQUEST``)
+  is send in response to this. In contrast to those, however, this command
+  does not trigger a new detachment process if none is currently in
+  progress.
+
+- Do nothing. The detachment process eventually times out as described in
+  point 3.
+
+See :ref:`ioctls` for more details on these responses.
+
+It is important to note that, if the user presses the detach button at any
+point when a detachment operation is in progress (i.e. after the the EC has
+sent the initial *detach-request* event (``SDTX_EVENT_REQUEST``) and before
+it received the corresponding response concluding the process), the
+detachment process is canceled on the EC-level and an identical event is
+being sent. Thus a *detach-request* event, by itself, does not signal the
+start of the detachment process.
+
+The detachment process may further be canceled by the EC due to hardware
+failures or a low clipboard battery. This is done via a cancel event
+(``SDTX_EVENT_CANCEL``) with the corresponding cancel reason.
+
+
+User-Space Interface Documentation
+==================================
+
+Error Codes and Status Values
+-----------------------------
+
+Error and status codes are divided into different categories, which can be
+used to determine if the status code is an error, and, if it is, the
+severity and type of that error. The current categories are:
+
+.. flat-table:: Overview of Status/Error Categories.
+   :widths: 2 1 3
+   :header-rows: 1
+
+   * - Name
+     - Value
+     - Short Description
+
+   * - ``STATUS``
+     - ``0x0000``
+     - Non-error status codes.
+
+   * - ``RUNTIME_ERROR``
+     - ``0x1000``
+     - Non-critical runtime errors.
+
+   * - ``HARDWARE_ERROR``
+     - ``0x2000``
+     - Critical hardware failures.
+
+   * - ``UNKNOWN``
+     - ``0xF000``
+     - Unknown error codes.
+
+Other categories are reserved for future use. The ``SDTX_CATEGORY()`` macro
+can be used to determine the category of any status value. The
+``SDTX_SUCCESS()`` macro can be used to check if the status value is a
+success value (``SDTX_CATEGORY_STATUS``) or if it indicates a failure.
+
+Unknown status or error codes sent by the EC are assigned to the ``UNKNOWN``
+category by the driver and may be implemented via their own code in the
+future.
+
+Currently used error codes are:
+
+.. flat-table:: Overview of Error Codes.
+   :widths: 2 1 1 3
+   :header-rows: 1
+
+   * - Name
+     - Category
+     - Value
+     - Short Description
+
+   * - ``SDTX_DETACH_NOT_FEASIBLE``
+     - ``RUNTIME``
+     - ``0x1001``
+     - Detachment not feasible due to low clipboard battery.
+
+   * - ``SDTX_DETACH_TIMEDOUT``
+     - ``RUNTIME``
+     - ``0x1002``
+     - Detachment process timed out while the latch was locked.
+
+   * - ``SDTX_ERR_FAILED_TO_OPEN``
+     - ``HARDWARE``
+     - ``0x2001``
+     - Failed to open latch.
+
+   * - ``SDTX_ERR_FAILED_TO_REMAIN_OPEN``
+     - ``HARDWARE``
+     - ``0x2002``
+     - Failed to keep latch open.
+
+   * - ``SDTX_ERR_FAILED_TO_CLOSE``
+     - ``HARDWARE``
+     - ``0x2003``
+     - Failed to close latch.
+
+Other error codes are reserved for future use. Non-error status codes may
+overlap and are generally only unique within their use-case:
+
+.. flat-table:: Latch Status Codes.
+   :widths: 2 1 1 3
+   :header-rows: 1
+
+   * - Name
+     - Category
+     - Value
+     - Short Description
+
+   * - ``SDTX_LATCH_CLOSED``
+     - ``STATUS``
+     - ``0x0000``
+     - Latch is closed/has been closed.
+
+   * - ``SDTX_LATCH_OPENED``
+     - ``STATUS``
+     - ``0x0001``
+     - Latch is open/has been opened.
+
+.. flat-table:: Base State Codes.
+   :widths: 2 1 1 3
+   :header-rows: 1
+
+   * - Name
+     - Category
+     - Value
+     - Short Description
+
+   * - ``SDTX_BASE_DETACHED``
+     - ``STATUS``
+     - ``0x0000``
+     - Base has been detached/is not present.
+
+   * - ``SDTX_BASE_ATTACHED``
+     - ``STATUS``
+     - ``0x0001``
+     - Base has been attached/is present.
+
+Again, other codes are reserved for future use.
+
+.. _events:
+
+Events
+------
+
+Events can be received by reading from the device file. They are disabled by
+default and have to be enabled by executing ``SDTX_IOCTL_EVENTS_ENABLE``
+first. All events follow the layout prescribed by |sdtx_event|. Specific
+event types can be identified by their event code, described in
+|sdtx_event_code|. Note that other event codes are reserved for future use,
+thus an event parser must be able to handle any unknown/unsupported event
+types gracefully, by relying on the payload length given in the event header.
+
+Currently provided event types are:
+
+.. flat-table:: Overview of DTX events.
+   :widths: 2 1 1 3
+   :header-rows: 1
+
+   * - Name
+     - Code
+     - Payload
+     - Short Description
+
+   * - ``SDTX_EVENT_REQUEST``
+     - ``1``
+     - ``0`` bytes
+     - Detachment process initiated/aborted.
+
+   * - ``SDTX_EVENT_CANCEL``
+     - ``2``
+     - ``2`` bytes
+     - EC canceled detachment process.
+
+   * - ``SDTX_EVENT_BASE_CONNECTION``
+     - ``3``
+     - ``4`` bytes
+     - Base connection state changed.
+
+   * - ``SDTX_EVENT_LATCH_STATUS``
+     - ``4``
+     - ``2`` bytes
+     - Latch status changed.
+
+   * - ``SDTX_EVENT_DEVICE_MODE``
+     - ``5``
+     - ``2`` bytes
+     - Device mode changed.
+
+Individual events in more detail:
+
+``SDTX_EVENT_REQUEST``
+^^^^^^^^^^^^^^^^^^^^^^
+
+Sent when a detachment process is started or, if in progress, aborted by the
+user, either via a detach button press or a detach request
+(``SDTX_IOCTL_LATCH_REQUEST``) being sent from user-space.
+
+Does not have any payload.
+
+``SDTX_EVENT_CANCEL``
+^^^^^^^^^^^^^^^^^^^^^
+
+Sent when a detachment process is canceled by the EC due to unfulfilled
+preconditions (e.g. clipboard battery too low to detach) or hardware
+failure. The reason for cancellation is given in the event payload detailed
+below and can be one of
+
+* ``SDTX_DETACH_TIMEDOUT``: Detachment timed out while the latch was locked.
+  The latch has neither been opened nor unlocked.
+
+* ``SDTX_DETACH_NOT_FEASIBLE``: Detachment not feasible due to low clipboard
+  battery.
+
+* ``SDTX_ERR_FAILED_TO_OPEN``: Could not open the latch (hardware failure).
+
+* ``SDTX_ERR_FAILED_TO_REMAIN_OPEN``: Could not keep the latch open (hardware
+  failure).
+
+* ``SDTX_ERR_FAILED_TO_CLOSE``: Could not close the latch (hardware failure).
+
+Other error codes in this context are reserved for future use.
+
+These codes can be classified via the ``SDTX_CATEGORY()`` macro to discern
+between critical hardware errors (``SDTX_CATEGORY_HARDWARE_ERROR``) or
+runtime errors (``SDTX_CATEGORY_RUNTIME_ERROR``), the latter of which may
+happen during normal operation if certain preconditions for detachment are
+not given.
+
+.. flat-table:: Detachment Cancel Event Payload
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Field
+     - Type
+     - Description
+
+   * - ``reason``
+     - |__u16|
+     - Reason for cancellation.
+
+``SDTX_EVENT_BASE_CONNECTION``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Sent when the base connection state has changed, i.e. when the base has been
+attached, detached, or detachment has become infeasible due to low clipboard
+battery. The new state and, if a base is connected, ID of the base is
+provided as payload of type |sdtx_base_info| with its layout presented
+below:
+
+.. flat-table:: Base-Connection-Change Event Payload
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Field
+     - Type
+     - Description
+
+   * - ``state``
+     - |__u16|
+     - Base connection state.
+
+   * - ``base_id``
+     - |__u16|
+     - Type of base connected (zero if none).
+
+Possible values for ``state`` are:
+
+* ``SDTX_BASE_DETACHED``,
+* ``SDTX_BASE_ATTACHED``, and
+* ``SDTX_DETACH_NOT_FEASIBLE``.
+
+Other values are reserved for future use.
+
+``SDTX_EVENT_LATCH_STATUS``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Sent when the latch status has changed, i.e. when the latch has been opened,
+closed, or an error occurred. The current status is provided as payload:
+
+.. flat-table:: Latch-Status-Change Event Payload
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Field
+     - Type
+     - Description
+
+   * - ``status``
+     - |__u16|
+     - Latch status.
+
+Possible values for ``status`` are:
+
+* ``SDTX_LATCH_CLOSED``,
+* ``SDTX_LATCH_OPENED``,
+* ``SDTX_ERR_FAILED_TO_OPEN``,
+* ``SDTX_ERR_FAILED_TO_REMAIN_OPEN``, and
+* ``SDTX_ERR_FAILED_TO_CLOSE``.
+
+Other values are reserved for future use.
+
+``SDTX_EVENT_DEVICE_MODE``
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Sent when the device mode has changed. The new device mode is provided as
+payload:
+
+.. flat-table:: Device-Mode-Change Event Payload
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Field
+     - Type
+     - Description
+
+   * - ``mode``
+     - |__u16|
+     - Device operation mode.
+
+Possible values for ``mode`` are:
+
+* ``SDTX_DEVICE_MODE_TABLET``,
+* ``SDTX_DEVICE_MODE_LAPTOP``, and
+* ``SDTX_DEVICE_MODE_STUDIO``.
+
+Other values are reserved for future use.
+
+.. _ioctls:
+
+IOCTLs
+------
+
+The following IOCTLs are provided:
+
+.. flat-table:: Overview of DTX IOCTLs
+   :widths: 1 1 1 1 4
+   :header-rows: 1
+
+   * - Type
+     - Number
+     - Direction
+     - Name
+     - Description
+
+   * - ``0xA5``
+     - ``0x21``
+     - ``-``
+     - ``EVENTS_ENABLE``
+     - Enable events for the current file descriptor.
+
+   * - ``0xA5``
+     - ``0x22``
+     - ``-``
+     - ``EVENTS_DISABLE``
+     - Disable events for the current file descriptor.
+
+   * - ``0xA5``
+     - ``0x23``
+     - ``-``
+     - ``LATCH_LOCK``
+     - Lock the latch.
+
+   * - ``0xA5``
+     - ``0x24``
+     - ``-``
+     - ``LATCH_UNLOCK``
+     - Unlock the latch.
+
+   * - ``0xA5``
+     - ``0x25``
+     - ``-``
+     - ``LATCH_REQUEST``
+     - Request clipboard detachment.
+
+   * - ``0xA5``
+     - ``0x26``
+     - ``-``
+     - ``LATCH_CONFIRM``
+     - Confirm clipboard detachment request.
+
+   * - ``0xA5``
+     - ``0x27``
+     - ``-``
+     - ``LATCH_HEARTBEAT``
+     - Send heartbeat signal to EC.
+
+   * - ``0xA5``
+     - ``0x28``
+     - ``-``
+     - ``LATCH_CANCEL``
+     - Cancel detachment process.
+
+   * - ``0xA5``
+     - ``0x29``
+     - ``R``
+     - ``GET_BASE_INFO``
+     - Get current base/connection information.
+
+   * - ``0xA5``
+     - ``0x2A``
+     - ``R``
+     - ``GET_DEVICE_MODE``
+     - Get current device operation mode.
+
+   * - ``0xA5``
+     - ``0x2B``
+     - ``R``
+     - ``GET_LATCH_STATUS``
+     - Get current device latch status.
+
+``SDTX_IOCTL_EVENTS_ENABLE``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x22)``.
+
+Enable events for the current file descriptor. Events can be obtained by
+reading from the device, if enabled. Events are disabled by default.
+
+``SDTX_IOCTL_EVENTS_DISABLE``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x22)``.
+
+Disable events for the current file descriptor. Events can be obtained by
+reading from the device, if enabled. Events are disabled by default.
+
+``SDTX_IOCTL_LATCH_LOCK``
+^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x23)``.
+
+Locks the latch, causing the detachment procedure to abort without opening
+the latch on timeout. The latch is unlocked by default. This command will be
+silently ignored if the latch is already locked.
+
+``SDTX_IOCTL_LATCH_UNLOCK``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x24)``.
+
+Unlocks the latch, causing the detachment procedure to open the latch on
+timeout. The latch is unlocked by default. This command will not open the
+latch when sent during an ongoing detachment process. It will be silently
+ignored if the latch is already unlocked.
+
+``SDTX_IOCTL_LATCH_REQUEST``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x25)``.
+
+Generic latch request. Behavior depends on the context: If no
+detachment-process is active, detachment is requested. Otherwise the
+currently active detachment-process will be aborted.
+
+If a detachment process is canceled by this operation, a generic detachment
+request event (``SDTX_EVENT_REQUEST``) will be sent.
+
+This essentially behaves the same as a detachment button press.
+
+``SDTX_IOCTL_LATCH_CONFIRM``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x26)``.
+
+Acknowledges and confirms a latch request. If sent during an ongoing
+detachment process, this command causes the latch to be opened immediately.
+The latch will also be opened if it has been locked. In this case, the latch
+lock is reset to the unlocked state.
+
+This command will be silently ignored if there is currently no detachment
+procedure in progress.
+
+``SDTX_IOCTL_LATCH_HEARTBEAT``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x27)``.
+
+Sends a heartbeat, essentially resetting the detachment timeout. This
+command can be used to keep the detachment process alive while work required
+for the detachment to succeed is still in progress.
+
+This command will be silently ignored if there is currently no detachment
+procedure in progress.
+
+``SDTX_IOCTL_LATCH_CANCEL``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IO(0xA5, 0x28)``.
+
+Cancels detachment in progress (if any). If a detachment process is canceled
+by this operation, a generic detachment request event
+(``SDTX_EVENT_REQUEST``) will be sent.
+
+This command will be silently ignored if there is currently no detachment
+procedure in progress.
+
+``SDTX_IOCTL_GET_BASE_INFO``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IOR(0xA5, 0x29, struct sdtx_base_info)``.
+
+Get the current base connection state (i.e. attached/detached) and the type
+of the base connected to the clipboard. This is command essentially provides
+a way to query the information provided by the base connection change event
+(``SDTX_EVENT_BASE_CONNECTION``).
+
+Possible values for ``struct sdtx_base_info.state`` are:
+
+* ``SDTX_BASE_DETACHED``,
+* ``SDTX_BASE_ATTACHED``, and
+* ``SDTX_DETACH_NOT_FEASIBLE``.
+
+Other values are reserved for future use.
+
+``SDTX_IOCTL_GET_DEVICE_MODE``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IOR(0xA5, 0x2A, __u16)``.
+
+Returns the device operation mode, indicating if and how the base is
+attached to the clipboard. This is command essentially provides a way to
+query the information provided by the device mode change event
+(``SDTX_EVENT_DEVICE_MODE``).
+
+Returned values are:
+
+* ``SDTX_DEVICE_MODE_LAPTOP``
+* ``SDTX_DEVICE_MODE_TABLET``
+* ``SDTX_DEVICE_MODE_STUDIO``
+
+See |sdtx_device_mode| for details. Other values are reserved for future
+use.
+
+
+``SDTX_IOCTL_GET_LATCH_STATUS``
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Defined as ``_IOR(0xA5, 0x2B, __u16)``.
+
+Get the current latch status or (presumably) the last error encountered when
+trying to open/close the latch. This is command essentially provides a way
+to query the information provided by the latch status change event
+(``SDTX_EVENT_LATCH_STATUS``).
+
+Returned values are:
+
+* ``SDTX_LATCH_CLOSED``,
+* ``SDTX_LATCH_OPENED``,
+* ``SDTX_ERR_FAILED_TO_OPEN``,
+* ``SDTX_ERR_FAILED_TO_REMAIN_OPEN``, and
+* ``SDTX_ERR_FAILED_TO_CLOSE``.
+
+Other values are reserved for future use.
+
+A Note on Base IDs
+------------------
+
+Base types/IDs provided via ``SDTX_EVENT_BASE_CONNECTION`` or
+``SDTX_IOCTL_GET_BASE_INFO`` are directly forwarded from from the EC in the
+lower byte of the combined |__u16| value, with the driver storing the EC
+type from which this ID comes in the high byte (without this, base IDs over
+different types of ECs may be overlapping).
+
+The ``SDTX_DEVICE_TYPE()`` macro can be used to determine the EC device
+type. This can be one of
+
+* ``SDTX_DEVICE_TYPE_HID``, for Surface Aggregator Module over HID, and
+
+* ``SDTX_DEVICE_TYPE_SSH``, for Surface Aggregator Module over Surface Serial
+  Hub.
+
+Note that currently only the ``SSH`` type EC is supported, however ``HID``
+type is reserved for future use.
+
+Structures and Enums
+--------------------
+
+.. kernel-doc:: include/uapi/linux/surface_aggregator/dtx.h
diff --git a/Documentation/driver-api/surface_aggregator/clients/index.rst b/Documentation/driver-api/surface_aggregator/clients/index.rst
new file mode 100644
index 000000000000..98ea9946b8a2
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/clients/index.rst
@@ -0,0 +1,22 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+===========================
+Client Driver Documentation
+===========================
+
+This is the documentation for client drivers themselves. Refer to
+:doc:`../client` for documentation on how to write client drivers.
+
+.. toctree::
+   :maxdepth: 1
+
+   cdev
+   dtx
+   san
+
+.. only::  subproject and html
+
+   Indices
+   =======
+
+   * :ref:`genindex`
diff --git a/Documentation/driver-api/surface_aggregator/clients/san.rst b/Documentation/driver-api/surface_aggregator/clients/san.rst
new file mode 100644
index 000000000000..1bf830ad367d
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/clients/san.rst
@@ -0,0 +1,44 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+.. |san_client_link| replace:: :c:func:`san_client_link`
+.. |san_dgpu_notifier_register| replace:: :c:func:`san_dgpu_notifier_register`
+.. |san_dgpu_notifier_unregister| replace:: :c:func:`san_dgpu_notifier_unregister`
+
+===================
+Surface ACPI Notify
+===================
+
+The Surface ACPI Notify (SAN) device provides the bridge between ACPI and
+SAM controller. Specifically, ACPI code can execute requests and handle
+battery and thermal events via this interface. In addition to this, events
+relating to the discrete GPU (dGPU) of the Surface Book 2 can be sent from
+ACPI code (note: the Surface Book 3 uses a different method for this). The
+only currently known event sent via this interface is a dGPU power-on
+notification. While this driver handles the former part internally, it only
+relays the dGPU events to any other driver interested via its public API and
+does not handle them.
+
+The public interface of this driver is split into two parts: Client
+registration and notifier-block registration.
+
+A client to the SAN interface can be linked as consumer to the SAN device
+via |san_client_link|. This can be used to ensure that the a client
+receiving dGPU events does not miss any events due to the SAN interface not
+being set up as this forces the client driver to unbind once the SAN driver
+is unbound.
+
+Notifier-blocks can be registered by any device for as long as the module is
+loaded, regardless of being linked as client or not. Registration is done
+with |san_dgpu_notifier_register|. If the notifier is not needed any more, it
+should be unregistered via |san_dgpu_notifier_unregister|.
+
+Consult the API documentation below for more details.
+
+
+API Documentation
+=================
+
+.. kernel-doc:: include/linux/surface_acpi_notify.h
+
+.. kernel-doc:: drivers/misc/surface_aggregator/clients/surface_acpi_notify.c
+    :export:
diff --git a/Documentation/driver-api/surface_aggregator/index.rst b/Documentation/driver-api/surface_aggregator/index.rst
new file mode 100644
index 000000000000..9fa70eedca59
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/index.rst
@@ -0,0 +1,21 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+=======================================
+Surface System Aggregator Module (SSAM)
+=======================================
+
+.. toctree::
+   :maxdepth: 2
+
+   overview
+   ssh
+   client
+   internal
+   clients/index
+
+.. only::  subproject and html
+
+   Indices
+   =======
+
+   * :ref:`genindex`
diff --git a/Documentation/driver-api/surface_aggregator/internal-api.rst b/Documentation/driver-api/surface_aggregator/internal-api.rst
new file mode 100644
index 000000000000..db6a70119f49
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/internal-api.rst
@@ -0,0 +1,67 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+==========================
+Internal API Documentation
+==========================
+
+.. contents::
+    :depth: 2
+
+
+Packet Transport Layer
+======================
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_parser.h
+    :internal:
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_parser.c
+    :internal:
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_msgb.h
+    :internal:
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_packet_layer.h
+    :internal:
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_packet_layer.c
+    :internal:
+
+
+Request Transport Layer
+=======================
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_request_layer.h
+    :internal:
+
+.. kernel-doc:: drivers/misc/surface_aggregator/ssh_request_layer.c
+    :internal:
+
+
+Controller
+==========
+
+.. kernel-doc:: drivers/misc/surface_aggregator/controller.h
+    :internal:
+
+.. kernel-doc:: drivers/misc/surface_aggregator/controller.c
+    :internal:
+
+
+Client Device Bus
+=================
+
+.. kernel-doc:: drivers/misc/surface_aggregator/bus.c
+    :internal:
+
+
+Core
+====
+
+.. kernel-doc:: drivers/misc/surface_aggregator/core.c
+    :internal:
+
+
+Trace Helpers
+=============
+
+.. kernel-doc:: drivers/misc/surface_aggregator/trace.h
diff --git a/Documentation/driver-api/surface_aggregator/internal.rst b/Documentation/driver-api/surface_aggregator/internal.rst
new file mode 100644
index 000000000000..6c020b87ad62
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/internal.rst
@@ -0,0 +1,50 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+=====================
+Core Driver Internals
+=====================
+
+For the API documentation, refer to:
+
+.. toctree::
+   :maxdepth: 2
+
+   internal-api
+
+
+Overview
+========
+
+The SSAM core implementation is structured in layers, somewhat following the
+SSH protocol structure:
+
+Lower-level packet transport is implemented in the *packet transport layer
+(PTL)*, directly building on top of the serial device (serdev)
+infrastructure of the kernel. As the name indicates, this layer deals with
+the packet transport logic and handles things like packet validation, packet
+acknowledgement (ACKing), packet (retransmission) timeouts, and relaying
+packet payloads to higher-level layers.
+
+Above this sits the *request transport layer (RTL)*. This layer is centered
+around command-type packet payloads, i.e. requests (sent from host to EC),
+responses of the EC to those requests, and events (sent from EC to host).
+It, specifically, distinguishes events from request responses, matches
+responses to their corresponding requests, and implements request timeouts.
+
+The *controller* layer is building on top of this and essentially decides
+how request responses and, especially, events are dealt with. It provides an
+event notifier system, handles event activation/deactivation, provides a
+workqueue for event and asynchronous request completion, and also manages
+the message counters required for building command messages (``SEQ``,
+``RQID``). This layer basically provides a fundamental interface to the SAM
+EC for use in other kernel drivers.
+
+While the controller layer already provides an interface for other kernel
+drivers, the client *bus* extends this interface to provide support for
+native SSAM devices, i.e. devices that are not defined in ACPI and not
+implemented as platform devices, via :c:type:`struct ssam_device <ssam_device>`
+and :c:type:`struct ssam_device_driver <ssam_device_driver>`. This aims to
+simplify management of client devices and client drivers.
+
+Refer to :doc:`client` for documentation regarding the client device/driver
+API and interface options for other kernel drivers.
diff --git a/Documentation/driver-api/surface_aggregator/overview.rst b/Documentation/driver-api/surface_aggregator/overview.rst
new file mode 100644
index 000000000000..b4ef7e25748c
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/overview.rst
@@ -0,0 +1,76 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+========
+Overview
+========
+
+The Surface/System Aggregator Module (SAM, SSAM) is an (arguably *the*)
+embedded controller (EC) on Microsoft Surface devices. It has been originally
+introduced on 4th generation devices (Surface Pro 4, Surface Book 1), but
+its responsibilities and feature-set have since been expanded significantly
+with the following generations.
+
+
+Features and Integration
+========================
+
+Not much is currently known about SAM on 4th generation devices (Surface Pro
+4, Surface Book 1), due to the use of a different communication interface
+between host and EC (as detailed below). On 5th (Surface Pro 2017, Surface
+Book 2, Surface Laptop 1) and later generation devices, SAM is responsible
+for providing battery information (both current status and static values,
+such as maximum capacity etc.), as well as an assortment of temperature
+sensors (e.g. skin temperature) and cooling/performance-mode setting to the
+host. On the Surface Book 2, specifically, it additionally provides an
+interface for properly handling clipboard detachment (i.e. separating the
+display part from the keyboard part of the device), on the Surface Laptop 1
+and 2 it is required for keyboard HID input. This HID subsystem has been
+restructured for 7th generation devices and on those, specifically Surface
+Laptop 3 and Surface Book 3, is responsible for all major HID input (i.e.
+keyboard and touchpad).
+
+While the features have not changed much on a coarse level since the 5th
+generation, internal interfaces have undergone some rather large changes. On
+5th and 6th generation devices, both battery and temperature information is
+exposed to ACPI via a shim driver (referred to as Surface ACPI Notify, or
+SAN), translating ACPI generic serial bus write-/read-accesses to SAM
+requests. On 7th generation devices, this additional layer is gone and these
+devices require a driver hooking directly into the SAM interface. Equally,
+on newer generations, less devices are declared in ACPI, making them a bit
+harder to discover and requiring us to hard-code a sort of device registry.
+Due to this, a SSAM bus and subsystem with client devices
+(:c:type:`struct ssam_device <ssam_device>`) has been implemented.
+
+
+Communication
+=============
+
+The type of communication interface between host and EC depends on the
+generation of the Surface device. On 4th generation devices, host and EC
+communicate via HID, specifically using a HID-over-I2C device, whereas on
+5th and later generations, communication takes place via a USART serial
+device. In accordance to the drivers found on other operating systems, we
+refer to the serial device and its driver as Surface Serial Hub (SSH) and
+when needed to differentiate between both types of SAM as SAM-over-SSH, in
+contrast to SAM-over-HID for the former variant.
+
+Currently, this subsystem only supports SAM-over-SSH. The SSH communication
+interface is described in more detail below. The HID interface has not been
+reverse engineered yet and it is, at the moment, unclear how many (and
+which) concepts of the SSH interface detailed below can be transferred to
+it.
+
+Surface Serial Hub
+------------------
+
+As already elaborated above, the Surface Serial Hub (SSH) is the
+communication interface for SAM on 5th- and all later-generation Surface
+devices. On the highest level, communication can be separated into two main
+types: Requests, messages sent from host to EC that may trigger a direct
+response from the EC (explicitly associated with the request), and events
+(sometimes also referred to as notifications), sent from EC to host without
+being a direct response to a previous request. We may also refer to requests
+without response as commands. In general, events need to be enabled via one
+of multiple dedicated commands before they are sent by the EC.
+
+See :doc:`ssh` for a more technical protocol documentation.
diff --git a/Documentation/driver-api/surface_aggregator/ssh.rst b/Documentation/driver-api/surface_aggregator/ssh.rst
new file mode 100644
index 000000000000..0b68228010e9
--- /dev/null
+++ b/Documentation/driver-api/surface_aggregator/ssh.rst
@@ -0,0 +1,343 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+.. |u8| replace:: :c:type:`u8 <u8>`
+.. |u16| replace:: :c:type:`u16 <u16>`
+.. |TYPE| replace:: ``TYPE``
+.. |LEN| replace:: ``LEN``
+.. |SEQ| replace:: ``SEQ``
+.. |SYN| replace:: ``SYN``
+.. |NAK| replace:: ``NAK``
+.. |ACK| replace:: ``ACK``
+.. |DATA| replace:: ``DATA``
+.. |DATA_SEQ| replace:: ``DATA_SEQ``
+.. |DATA_NSQ| replace:: ``DATA_NSQ``
+.. |TC| replace:: ``TC``
+.. |TID| replace:: ``TID``
+.. |IID| replace:: ``IID``
+.. |RQID| replace:: ``RQID``
+.. |CID| replace:: ``CID``
+
+===========================
+Surface Serial Hub Protocol
+===========================
+
+The Surface Serial Hub (SSH) is the central communication interface for the
+embedded Surface Aggregator Module controller (SAM or EC) on newer Surface
+generations. We will refer to this protocol and interface as SAM-over-SSH,
+as opposed to SAM-over-HID for the older generations.
+
+On Surface devices with SAM-over-SSH, SAM is connected to the host via UART
+and defined in ACPI as device with ID ``MSHW0084``. On these devices,
+significant functionality is provided via SAM, including access to battery
+and power information and events, thermal read-outs and events, and many
+more. For Surface Laptops, keyboard input is handled via HID directed
+through SAM, on the Surface Laptop 3 and Surface Book 3 this also includes
+touchpad input.
+
+Note that the standard disclaimer for this subsystem also applies to this
+document: All of this has been reverse-engineered and may thus be erroneous
+and/or incomplete.
+
+All CRCs used in the following are two-byte ``crc_ccitt_false(0xffff, ...)``.
+All multi-byte values are little-endian, there is no implicit padding between
+values.
+
+
+SSH Packet Protocol: Definitions
+================================
+
+The fundamental communication unit of the SSH protocol is a frame
+(:c:type:`struct ssh_frame <ssh_frame>`). A frame consists of the following
+fields, packed together and in order:
+
+.. flat-table:: SSH Frame
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Field
+     - Type
+     - Description
+
+   * - |TYPE|
+     - |u8|
+     - Type identifier of the frame.
+
+   * - |LEN|
+     - |u16|
+     - Length of the payload associated with the frame.
+
+   * - |SEQ|
+     - |u8|
+     - Sequence ID (see explanation below).
+
+Each frame structure is followed by a CRC over this structure. The CRC over
+the frame structure (|TYPE|, |LEN|, and |SEQ| fields) is placed directly
+after the frame structure and before the payload. The payload is followed by
+its own CRC (over all payload bytes). If the payload is not present (i.e.
+the frame has ``LEN=0``), the CRC of the payload is still present and will
+evaluate to ``0xffff``. The |LEN| field does not include any of the CRCs, it
+equals the number of bytes inbetween the CRC of the frame and the CRC of the
+payload.
+
+Additionally, the following fixed two-byte sequences are used:
+
+.. flat-table:: SSH Byte Sequences
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Name
+     - Value
+     - Description
+
+   * - |SYN|
+     - ``[0xAA, 0x55]``
+     - Synchronization bytes.
+
+A message consists of |SYN|, followed by the frame (|TYPE|, |LEN|, |SEQ| and
+CRC) and, if specified in the frame (i.e. ``LEN > 0``), payload bytes,
+followed finally, regardless if the payload is present, the payload CRC. The
+messages corresponding to an exchange are, in part, identified by having the
+same sequence ID (|SEQ|), stored inside the frame (more on this in the next
+section). The sequence ID is a wrapping counter.
+
+A frame can have the following types
+(:c:type:`enum ssh_frame_type <ssh_frame_type>`):
+
+.. flat-table:: SSH Frame Types
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Name
+     - Value
+     - Short Description
+
+   * - |NAK|
+     - ``0x04``
+     - Sent on error in previously received message.
+
+   * - |ACK|
+     - ``0x40``
+     - Sent to acknowledge receival of |DATA| frame.
+
+   * - |DATA_SEQ|
+     - ``0x80``
+     - Sent to transfer data. Sequenced.
+
+   * - |DATA_NSQ|
+     - ``0x00``
+     - Same as |DATA_SEQ|, but does not need to be ACKed.
+
+Both |NAK|- and |ACK|-type frames are used to control flow of messages and
+thus do not carry a payload. |DATA_SEQ|- and |DATA_NSQ|-type frames on the
+other hand must carry a payload. The flow sequence and interaction of
+different frame types will be described in more depth in the next section.
+
+
+SSH Packet Protocol: Flow Sequence
+==================================
+
+Each exchange begins with |SYN|, followed by a |DATA_SEQ|- or
+|DATA_NSQ|-type frame, followed by its CRC, payload, and payload CRC. In
+case of a |DATA_NSQ|-type frame, the exchange is then finished. In case of a
+|DATA_SEQ|-type frame, the receiving party has to acknowledge receival of
+the frame by responding with a message containing an |ACK|-type frame with
+the same sequence ID of the |DATA| frame. In other words, the sequence ID of
+the |ACK| frame specifies the |DATA| frame to be acknowledged. In case of an
+error, e.g. an invalid CRC, the receiving party responds with a message
+containing an |NAK|-type frame. As the sequence ID of the previous data
+frame, for which an error is indicated via the |NAK| frame, cannot be relied
+upon, the sequence ID of the |NAK| frame should not be used and is set to
+zero. After receival of an |NAK| frame, the sending party should re-send all
+outstanding (non-ACKed) messages.
+
+Sequence IDs are not synchronized between the two parties, meaning that they
+are managed independently for each party. Identifying the messages
+corresponding to a single exchange thus relies on the sequence ID as well as
+the type of the message, and the context. Specifically, the sequence ID is
+used to associate an ``ACK`` with its ``DATA_SEQ``-type frame, but not
+``DATA_SEQ``- or ``DATA_NSQ``-type frames with other ``DATA``- type frames.
+
+An example exchange might look like this:
+
+::
+
+    tx: -- SYN FRAME(D) CRC(F) PAYLOAD CRC(P) -----------------------------
+    rx: ------------------------------------- SYN FRAME(A) CRC(F) CRC(P) --
+
+where both frames have the same sequence ID (``SEQ``). Here, ``FRAME(D)``
+indicates a |DATA_SEQ|-type frame, ``FRAME(A)`` an ``ACK``-type frame,
+``CRC(F)`` the CRC over the previous frame, ``CRC(P)`` the CRC over the
+previous payload. In case of an error, the exchange would look like this:
+
+::
+
+    tx: -- SYN FRAME(D) CRC(F) PAYLOAD CRC(P) -----------------------------
+    rx: ------------------------------------- SYN FRAME(N) CRC(F) CRC(P) --
+
+upon which the sender should re-send the message. ``FRAME(N)`` indicates an
+|NAK|-type frame. Note that the sequence ID of the |NAK|-type frame is fixed
+to zero. For |DATA_NSQ|-type frames, both exchanges are the same:
+
+::
+
+    tx: -- SYN FRAME(DATA_NSQ) CRC(F) PAYLOAD CRC(P) ----------------------
+    rx: -------------------------------------------------------------------
+
+Here, an error can be detected, but not corrected or indicated to the
+sending party. These exchanges are symmetric, i.e. switching rx and tx
+results again in a valid exchange. Currently, no longer exchanges are known.
+
+
+Commands: Requests, Responses, and Events
+=========================================
+
+Commands are sent as payload inside a data frame. Currently, this is the
+only known payload type of |DATA| frames, with a payload-type value of
+``0x80`` (:c:type:`SSH_PLD_TYPE_CMD <ssh_payload_type>`).
+
+The command-type payload (:c:type:`struct ssh_command <ssh_command>`)
+consists of an eight-byte command structure, followed by optional and
+variable length command data. The length of this optional data is derived
+from the frame payload length given in the corresponding frame, i.e. it is
+``frame.len - sizeof(struct ssh_command)``. The command struct contains the
+following fields, packed together and in order:
+
+.. flat-table:: SSH Command
+   :widths: 1 1 4
+   :header-rows: 1
+
+   * - Field
+     - Type
+     - Description
+
+   * - |TYPE|
+     - |u8|
+     - Type of the payload. For commands always ``0x80``.
+
+   * - |TC|
+     - |u8|
+     - Target category.
+
+   * - |TID| (out)
+     - |u8|
+     - Target ID for outgoing (host to EC) commands.
+
+   * - |TID| (in)
+     - |u8|
+     - Target ID for incoming (EC to host) commands.
+
+   * - |IID|
+     - |u8|
+     - Instance ID.
+
+   * - |RQID|
+     - |u16|
+     - Request ID.
+
+   * - |CID|
+     - |u8|
+     - Command ID.
+
+The command struct and data, in general, does not contain any failure
+detection mechanism (e.g. CRCs), this is solely done on the frame level.
+
+Command-type payloads are used by the host to send commands and requests to
+the EC as well as by the EC to send responses and events back to the host.
+We differentiate between requests (sent by the host), responses (sent by the
+EC in response to a request), and events (sent by the EC without a
+preceeding request).
+
+Commands and events are uniquely identified by their target category
+(``TC``) and command ID (``CID``). The target category specifies a general
+category for the command (e.g. system in general, vs. battery and ac, vs.
+temperature, and so on), while the command ID specifies the command inside
+that category. Only the combination of |TC| + |CID| is unique. Additionally,
+commands have an instance ID (``IID``), which is used to differentiate
+between different sub-devices. For example ``TC=3`` ``CID=1`` is a
+request to get the temperature on a thermal sensor, where |IID| specifies
+the respective sensor. If the instance ID is not used, it should be set to
+zero. If instance IDs are used, they, in general, start with a value of one,
+whereas zero may be used for instance independent queries, if applicable. A
+response to a request should have the same target category, command ID, and
+instance ID as the corresponding request.
+
+Responses are matched to their corresponding request via the request ID
+(``RQID``) field. This is a 16 bit wrapping counter similar to the sequence
+ID on the frames. Note that the sequence ID of the frames for a
+request-response pair does not match. Only the request ID has to match.
+Frame-protocol wise these are two separate exchanges, and may even be
+separated, e.g. by an event being sent after the request but before the
+response. Not all commands produce a response, and this is not detectable by
+|TC| + |CID|. It is the responsibility of the issuing party to wait for a
+response (or signal this to the communication framework, as is done in
+SAN/ACPI via the ``SNC`` flag).
+
+Events are identified by unique and reserved request IDs. These IDs should
+not be used by the host when sending a new request. They are used on the
+host to, first, detect events and, second, match them with a registered
+event handler. Request IDs for events are chosen by the host and directed to
+the EC when setting up and enabling an event source (via the
+enable-event-source request). The EC then uses the specified request ID for
+events sent from the respective source. Note that an event should still be
+identified by its target category, command ID, and, if applicable, instance
+ID, as a single event source can send multiple different event types. In
+general, however, a single target category should map to a single reserved
+event request ID.
+
+Furthermore, requests, responses, and events have an associated target ID
+(``TID``). This target ID is split into output (host to EC) and input (EC to
+host) fields, with the respecting other field (e.g. output field on incoming
+messages) set to zero. Two ``TID`` values are known: Primary (``0x01``) and
+secondary (``0x02``). In general, the response to a request should have the
+same ``TID`` value, however, the field (output vs. input) should be used in
+accordance to the direction in which the response is sent (i.e. on the input
+field, as responses are generally sent from the EC to the host).
+
+Note that, even though requests and events should be uniquely identifiable
+by target category and command ID alone, the EC may require specific
+priority and instance ID values to accept a command. A command that is
+accepted for ``TID=1``, for example, may not be accepted for ``TID=2``
+and vice versa.
+
+
+Limitations and Observations
+============================
+
+The protocol can, in theory, handle up to ``U8_MAX`` frames in parallel,
+with up to ``U16_MAX`` pending requests (neglecting request IDs reserved for
+events). In practice, however, this is more limited. From our testing
+(altough via a python and thus a user-space program), it seems that the EC
+can handle up to four requests (mostly) reliably in parallel at a certain
+time. With five or more requests in parallel, consistent discarding of
+commands (ACKed frame but no command response) has been observed. For five
+simultaneous commands, this reproducibly resulted in one command being
+dropped and four commands being handled.
+
+However, it has also been noted that, even with three requests in parallel,
+occasional frame drops happen. Apart from this, with a limit of three
+pending requests, no dropped commands (i.e. command being dropped but frame
+carrying command being ACKed) have been observed. In any case, frames (and
+possibly also commands) should be re-sent by the host if a certain timeout
+is exceeded. This is done by the EC for frames with a timeout of one second,
+up to two re-tries (i.e. three transmissions in total). The limit of
+re-tries also applies to received NAKs, and, in a worst case scenario, can
+lead to entire messages being dropped.
+
+While this also seems to work fine for pending data frames as long as no
+transmission failures occur, implementation and handling of these seems to
+depend on the assumption that there is only one non-acknowledged data frame.
+In particular, the detection of repeated frames relies on the last sequence
+number. This means that, if a frame that has been successfully received by
+the EC is sent again, e.g. due to the host not receiving an |ACK|, the EC
+will only detect this if it has the sequence ID of the last frame received
+by the EC. As an example: Sending two frames with ``SEQ=0`` and ``SEQ=1``
+followed by a repetition of ``SEQ=0`` will not detect the second ``SEQ=0``
+frame as such, and thus execute the command in this frame each time it has
+been received, i.e. twice in this example. Sending ``SEQ=0``, ``SEQ=1`` and
+then repeating ``SEQ=1`` will detect the second ``SEQ=1`` as repetition of
+the first one and ignore it, thus executing the contained command only once.
+
+In conclusion, this suggests a limit of at most one pending un-ACKed frame
+(per party, effectively leading to synchronous communication regarding
+frames) and at most three pending commands. The limit to synchronous frame
+transfers seems to be consistent with behavior observed on Windows.
diff --git a/drivers/misc/Kconfig b/drivers/misc/Kconfig
index be901ffc66fe..88f4231d0ed1 100644
--- a/drivers/misc/Kconfig
+++ b/drivers/misc/Kconfig
@@ -472,5 +472,6 @@ source "drivers/misc/ocxl/Kconfig"
 source "drivers/misc/cardreader/Kconfig"
 source "drivers/misc/habanalabs/Kconfig"
 source "drivers/misc/uacce/Kconfig"
+source "drivers/misc/surface_aggregator/Kconfig"
 source "drivers/misc/ipts/Kconfig"
 endmenu
diff --git a/drivers/misc/Makefile b/drivers/misc/Makefile
index f97938d777e1..b0caee1dd3b7 100644
--- a/drivers/misc/Makefile
+++ b/drivers/misc/Makefile
@@ -58,3 +58,4 @@ obj-$(CONFIG_HABANA_AI)		+= habanalabs/
 obj-$(CONFIG_UACCE)		+= uacce/
 obj-$(CONFIG_XILINX_SDFEC)	+= xilinx_sdfec.o
 obj-$(CONFIG_MISC_IPTS)		+= ipts/
+obj-$(CONFIG_SURFACE_AGGREGATOR)	+= surface_aggregator/
diff --git a/drivers/misc/surface_aggregator/Kconfig b/drivers/misc/surface_aggregator/Kconfig
new file mode 100644
index 000000000000..47dd8fdffac3
--- /dev/null
+++ b/drivers/misc/surface_aggregator/Kconfig
@@ -0,0 +1,67 @@
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+
+menuconfig SURFACE_AGGREGATOR
+	tristate "Microsoft Surface System Aggregator Module Subsystem and Drivers"
+	depends on SERIAL_DEV_BUS
+	depends on ACPI
+	select CRC_CCITT
+	help
+	  The Surface System Aggregator Module (Surface SAM or SSAM) is an
+	  embedded controller (EC) found on 5th- and later-generation Microsoft
+	  Surface devices (i.e. Surface Pro 5, Surface Book 2, Surface Laptop,
+	  and newer, with exception of Surface Go series devices).
+
+	  Depending on the device in question, this EC provides varying
+	  functionality, including:
+	  - EC access from ACPI via Surface ACPI Notify (5th- and 6th-generation)
+	  - battery status information (all devices)
+	  - thermal sensor access (all devices)
+	  - performance mode / cooling mode control (all devices)
+	  - clipboard detachment system control (Surface Book 2 and 3)
+	  - HID / keyboard input (Surface Laptops, Surface Book 3)
+
+	  This option controls whether the Surface SAM subsystem core will be
+	  built. This includes a driver for the Surface Serial Hub (SSH), which
+	  is the device responsible for the communication with the EC, and a
+	  basic kernel interface exposing the EC functionality to other client
+	  drivers, i.e. allowing them to make requests to the EC and receive
+	  events from it. Selecting this option alone will not provide any
+	  client drivers and therefore no functionality beyond the in-kernel
+	  interface. Said functionality is the repsonsibility of the respective
+	  client drivers.
+
+	  Note: While 4th-generation Surface devices also make use of a SAM EC,
+	  due to a difference in the communication interface of the controller,
+	  only 5th and later generations are currently supported. Specifically,
+	  devices using SAM-over-SSH are supported, whereas devices using
+	  SAM-over-HID, which is used on the 4th generation, are currently not
+	  supported.
+
+config SURFACE_AGGREGATOR_BUS
+	bool "Surface System Aggregator Module Bus"
+	depends on SURFACE_AGGREGATOR
+	default y
+	help
+	  Expands the Surface System Aggregator Module (SSAM) core driver by
+	  providing a dedicated bus and client-device type.
+
+	  This bus and device type are intended to provide and simplify support
+	  for non-platform and non-ACPI SSAM devices, i.e. SSAM devices that are
+	  not auto-detectable via the conventional means (e.g. ACPI).
+
+config SURFACE_AGGREGATOR_ERROR_INJECTION
+	bool "Surface System Aggregator Module Error Injection Capabilities"
+	depends on SURFACE_AGGREGATOR
+	depends on FUNCTION_ERROR_INJECTION
+	help
+	  Provides error-injection capabilities for the Surface System
+	  Aggregator Module subsystem and Surface Serial Hub driver.
+
+	  Specifically, exports error injection hooks to be used with the
+	  kernel's function error injection capabilities to simulate underlying
+	  transport and communication problems, such as invalid data sent to or
+	  received from the EC, dropped data, and communication timeouts.
+	  Intended for development and debugging.
+
+source "drivers/misc/surface_aggregator/clients/Kconfig"
diff --git a/drivers/misc/surface_aggregator/Makefile b/drivers/misc/surface_aggregator/Makefile
new file mode 100644
index 000000000000..b48ffc37ab52
--- /dev/null
+++ b/drivers/misc/surface_aggregator/Makefile
@@ -0,0 +1,18 @@
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+
+# For include/trace/define_trace.h to include trace.h
+CFLAGS_core.o = -I$(src)
+
+obj-$(CONFIG_SURFACE_AGGREGATOR) += surface_aggregator.o
+obj-$(CONFIG_SURFACE_AGGREGATOR) += clients/
+
+surface_aggregator-objs := core.o
+surface_aggregator-objs += ssh_parser.o
+surface_aggregator-objs += ssh_packet_layer.o
+surface_aggregator-objs += ssh_request_layer.o
+surface_aggregator-objs += controller.o
+
+ifeq ($(CONFIG_SURFACE_AGGREGATOR_BUS),y)
+surface_aggregator-objs += bus.o
+endif
diff --git a/drivers/misc/surface_aggregator/bus.c b/drivers/misc/surface_aggregator/bus.c
new file mode 100644
index 000000000000..5e734bbd18cd
--- /dev/null
+++ b/drivers/misc/surface_aggregator/bus.c
@@ -0,0 +1,424 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface System Aggregator Module bus and device integration.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/device.h>
+#include <linux/slab.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/device.h>
+
+#include "bus.h"
+#include "controller.h"
+
+
+static ssize_t modalias_show(struct device *dev, struct device_attribute *attr,
+			     char *buf)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+
+	return snprintf(buf, PAGE_SIZE - 1, "ssam:d%02Xc%02Xt%02Xi%02xf%02X\n",
+			sdev->uid.domain, sdev->uid.category, sdev->uid.target,
+			sdev->uid.instance, sdev->uid.function);
+}
+static DEVICE_ATTR_RO(modalias);
+
+static struct attribute *ssam_device_attrs[] = {
+	&dev_attr_modalias.attr,
+	NULL,
+};
+ATTRIBUTE_GROUPS(ssam_device);
+
+static int ssam_device_uevent(struct device *dev, struct kobj_uevent_env *env)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+
+	return add_uevent_var(env, "MODALIAS=ssam:d%02Xc%02Xt%02Xi%02xf%02X",
+			      sdev->uid.domain, sdev->uid.category,
+			      sdev->uid.target, sdev->uid.instance,
+			      sdev->uid.function);
+}
+
+static void ssam_device_release(struct device *dev)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+
+	ssam_controller_put(sdev->ctrl);
+	kfree(sdev);
+}
+
+const struct device_type ssam_device_type = {
+	.name    = "surface_aggregator_device",
+	.groups  = ssam_device_groups,
+	.uevent  = ssam_device_uevent,
+	.release = ssam_device_release,
+};
+EXPORT_SYMBOL_GPL(ssam_device_type);
+
+
+/**
+ * ssam_device_alloc() - Allocate and initialize a SSAM client device.
+ * @ctrl: The controller under which the device should be added.
+ * @uid:  The UID of the device to be added.
+ *
+ * Allocates and initializes a new client device. The parent of the device
+ * will be set to the controller device and the name will be set based on the
+ * UID. Note that the device still has to be added via ssam_device_add().
+ * Refer to that function for more details.
+ *
+ * Return: Returns the newly allocated and initialized SSAM client device, or
+ * %NULL if it could not be allocated.
+ */
+struct ssam_device *ssam_device_alloc(struct ssam_controller *ctrl,
+				      struct ssam_device_uid uid)
+{
+	struct ssam_device *sdev;
+
+	sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
+	if (!sdev)
+		return NULL;
+
+	device_initialize(&sdev->dev);
+	sdev->dev.bus = &ssam_bus_type;
+	sdev->dev.type = &ssam_device_type;
+	sdev->dev.parent = ssam_controller_device(ctrl);
+	sdev->ctrl = ssam_controller_get(ctrl);
+	sdev->uid = uid;
+
+	dev_set_name(&sdev->dev, "%02x:%02x:%02x:%02x:%02x",
+		     sdev->uid.domain, sdev->uid.category, sdev->uid.target,
+		     sdev->uid.instance, sdev->uid.function);
+
+	return sdev;
+}
+EXPORT_SYMBOL_GPL(ssam_device_alloc);
+
+/**
+ * ssam_device_add() - Add a SSAM client device.
+ * @sdev: The SSAM client device to be added.
+ *
+ * Added client devices must be guaranteed to always have a valid and active
+ * controller. Thus, this function will fail with %-ENXIO if the controller of
+ * the device has not been initialized yet, has been suspended, or has been
+ * shut down.
+ *
+ * The caller of this function should ensure that the corresponding call to
+ * ssam_device_remove() is issued before the controller is shut down. If the
+ * added device is a direct child of the controller device (default), it will
+ * be automatically removed when the controller is shut down.
+ *
+ * By default, the controller device will become the parent of the newly
+ * created client device. The parent may be changed before ssam_device_add is
+ * called, but care must be taken that a) the correct suspend/resume ordering
+ * is guaranteed and b) the client device does not oultive the controller,
+ * i.e. that the device is removed before the controller is being shut down.
+ * In case these guarantees have to be manually enforced, please refer to the
+ * ssam_client_link() and ssam_client_bind() functions, which are intended to
+ * set up device-links for this purpose.
+ *
+ * Return: Returns zero on success, a negative error code on failure.
+ */
+int ssam_device_add(struct ssam_device *sdev)
+{
+	int status;
+
+	/*
+	 * Ensure that we can only add new devices to a controller if it has
+	 * been started and is not going away soon. This works in combination
+	 * with ssam_controller_remove_clients to ensure driver presence for the
+	 * controller device, i.e. it ensures that the controller (sdev->ctrl)
+	 * is always valid and can be used for requests as long as the client
+	 * device we add here is registered as child under it. This essentially
+	 * guarantees that the client driver can always expect the preconditions
+	 * for functions like ssam_request_sync (controller has to be started
+	 * and is not suspended) to hold and thus does not have to check for
+	 * them.
+	 *
+	 * Note that for this to work, the controller has to be a parent device.
+	 * If it is not a direct parent, care has to be taken that the device is
+	 * removed via ssam_device_remove(), as device_unregister does not
+	 * remove child devices recursively.
+	 */
+	ssam_controller_statelock(sdev->ctrl);
+
+	if (sdev->ctrl->state != SSAM_CONTROLLER_STARTED) {
+		ssam_controller_stateunlock(sdev->ctrl);
+		return -ENXIO;
+	}
+
+	status = device_add(&sdev->dev);
+
+	ssam_controller_stateunlock(sdev->ctrl);
+	return status;
+}
+EXPORT_SYMBOL_GPL(ssam_device_add);
+
+/**
+ * ssam_device_remove() - Remove a SSAM client device.
+ * @sdev: The device to remove.
+ *
+ * Removes and unregisters the provided SSAM client device.
+ */
+void ssam_device_remove(struct ssam_device *sdev)
+{
+	device_unregister(&sdev->dev);
+}
+EXPORT_SYMBOL_GPL(ssam_device_remove);
+
+
+/**
+ * ssam_device_id_compatible() - Check if a device ID matches a UID.
+ * @id:  The device ID as potential match.
+ * @uid: The device UID matching against.
+ *
+ * Check if the given ID is a match for the given UID, i.e. if a device with
+ * the provided UID is compatible to the given ID following the match rules
+ * described in its &ssam_device_id.match_flags member.
+ *
+ * Return: Returns %true iff the given UID is compatible to the match rule
+ * described by the given ID, %false otherwise.
+ */
+static bool ssam_device_id_compatible(const struct ssam_device_id *id,
+				      struct ssam_device_uid uid)
+{
+	if (id->domain != uid.domain || id->category != uid.category)
+		return false;
+
+	if ((id->match_flags & SSAM_MATCH_TARGET) && id->target != uid.target)
+		return false;
+
+	if ((id->match_flags & SSAM_MATCH_INSTANCE) && id->instance != uid.instance)
+		return false;
+
+	if ((id->match_flags & SSAM_MATCH_FUNCTION) && id->function != uid.function)
+		return false;
+
+	return true;
+}
+
+/**
+ * ssam_device_id_is_null() - Check if a device ID is null.
+ * @id: The device ID to check.
+ *
+ * Check if a given device ID is null, i.e. all zeros. Used to check for the
+ * end of ``MODULE_DEVICE_TABLE(ssam, ...)`` or similar lists.
+ *
+ * Return: Returns %true if the given ID represents a null ID, %false
+ * otherwise.
+ */
+static bool ssam_device_id_is_null(const struct ssam_device_id *id)
+{
+	return id->match_flags == 0
+		&& id->domain == 0
+		&& id->category == 0
+		&& id->target == 0
+		&& id->instance == 0
+		&& id->function == 0
+		&& id->driver_data == 0;
+}
+
+/**
+ * ssam_device_id_match() - Find the matching ID table entry for the given UID.
+ * @table: The table to search in.
+ * @uid:   The UID to matched against the individual table entries.
+ *
+ * Find the first match for the provided device UID in the provided ID table
+ * and return it. Returns %NULL if no match could be found.
+ */
+const struct ssam_device_id *ssam_device_id_match(
+		const struct ssam_device_id *table,
+		const struct ssam_device_uid uid)
+{
+	const struct ssam_device_id *id;
+
+	for (id = table; !ssam_device_id_is_null(id); ++id)
+		if (ssam_device_id_compatible(id, uid))
+			return id;
+
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(ssam_device_id_match);
+
+/**
+ * ssam_device_get_match() - Find and return the ID matching the device in the
+ * ID table of the bound driver.
+ * @dev: The device for which to get the matching ID table entry.
+ *
+ * Find the fist match for the UID of the device in the ID table of the
+ * currently bound driver and return it. Returns %NULL if the device does not
+ * have a driver bound to it, the driver does not have match_table (i.e. it is
+ * %NULL), or there is no match in the driver's match_table.
+ *
+ * This function essentially calls ssam_device_id_match() with the ID table of
+ * the bound device driver and the UID of the device.
+ *
+ * Return: Returns the first match for the UID of the device in the device
+ * driver's match table, or %NULL if no such match could be found.
+ */
+const struct ssam_device_id *ssam_device_get_match(
+		const struct ssam_device *dev)
+{
+	const struct ssam_device_driver *sdrv;
+
+	sdrv = to_ssam_device_driver(dev->dev.driver);
+	if (!sdrv)
+		return NULL;
+
+	if (!sdrv->match_table)
+		return NULL;
+
+	return ssam_device_id_match(sdrv->match_table, dev->uid);
+}
+EXPORT_SYMBOL_GPL(ssam_device_get_match);
+
+/**
+ * ssam_device_get_match_data() - Find the ID matching the device in hte
+ * ID table of the bound driver and return its ``driver_data`` member.
+ * @dev: The device for which to get the match data.
+ *
+ * Find the fist match for the UID of the device in the ID table of the
+ * corresponding driver and return its driver_data. Returns %NULL if the
+ * device does not have a driver bound to it, the driver does not have
+ * match_table (i.e. it is %NULL), there is no match in the driver's
+ * match_table, or the match does not have any driver_data.
+ *
+ * This function essentially calls ssam_device_get_match() and, if any match
+ * could be found, returns its ``struct ssam_device_id.driver_data`` member.
+ *
+ * Return: Returns the driver data associated with the first match for the UID
+ * of the device in the device driver's match table, or %NULL if no such match
+ * could be found.
+ */
+const void *ssam_device_get_match_data(const struct ssam_device *dev)
+{
+	const struct ssam_device_id *id;
+
+	id = ssam_device_get_match(dev);
+	if (!id)
+		return NULL;
+
+	return (const void *)id->driver_data;
+}
+EXPORT_SYMBOL_GPL(ssam_device_get_match_data);
+
+
+static int ssam_bus_match(struct device *dev, struct device_driver *drv)
+{
+	struct ssam_device_driver *sdrv = to_ssam_device_driver(drv);
+	struct ssam_device *sdev = to_ssam_device(dev);
+
+	if (!is_ssam_device(dev))
+		return 0;
+
+	return !!ssam_device_id_match(sdrv->match_table, sdev->uid);
+}
+
+static int ssam_bus_probe(struct device *dev)
+{
+	return to_ssam_device_driver(dev->driver)
+		->probe(to_ssam_device(dev));
+}
+
+static int ssam_bus_remove(struct device *dev)
+{
+	struct ssam_device_driver *sdrv = to_ssam_device_driver(dev->driver);
+
+	if (sdrv->remove)
+		sdrv->remove(to_ssam_device(dev));
+
+	return 0;
+}
+
+struct bus_type ssam_bus_type = {
+	.name   = "surface_aggregator",
+	.match  = ssam_bus_match,
+	.probe  = ssam_bus_probe,
+	.remove = ssam_bus_remove,
+};
+EXPORT_SYMBOL_GPL(ssam_bus_type);
+
+
+/**
+ * __ssam_device_driver_register() - Register a SSAM client device driver.
+ * @sdrv:  The driver to register.
+ * @owner: The module owning the provided driver.
+ *
+ * Please refer to the ssam_device_driver_register() macro for the normal way
+ * to register a driver from inside its owning module.
+ */
+int __ssam_device_driver_register(struct ssam_device_driver *sdrv,
+				  struct module *owner)
+{
+	sdrv->driver.owner = owner;
+	sdrv->driver.bus = &ssam_bus_type;
+
+	/* force drivers to async probe so I/O is possible in probe */
+	sdrv->driver.probe_type = PROBE_PREFER_ASYNCHRONOUS;
+
+	return driver_register(&sdrv->driver);
+}
+EXPORT_SYMBOL_GPL(__ssam_device_driver_register);
+
+/**
+ * ssam_device_driver_unregister - Unregister a SSAM device driver.
+ * @sdrv: The driver to unregister.
+ */
+void ssam_device_driver_unregister(struct ssam_device_driver *sdrv)
+{
+	driver_unregister(&sdrv->driver);
+}
+EXPORT_SYMBOL_GPL(ssam_device_driver_unregister);
+
+
+static int ssam_remove_device(struct device *dev, void *_data)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+
+	if (is_ssam_device(dev))
+		ssam_device_remove(sdev);
+
+	return 0;
+}
+
+/**
+ * ssam_controller_remove_clients() - Remove SSAM client devices registered as
+ * direct children under the given controller.
+ * @ctrl: The controller to remove all direct clients for.
+ *
+ * Remove all SSAM client devices registered as direct children under the
+ * given controller. Note that this only accounts for direct children ot the
+ * controller device. This does not take care of any client devices where the
+ * parent device has been manually set before calling ssam_device_add. Refer
+ * to ssam_device_add()/ssam_device_remove() for more details on those cases.
+ *
+ * To avoid new devices being added in parallel to this call, the main
+ * controller lock (not statelock) must be held during this (and if
+ * necessary, any subsequent deinitialization) call.
+ */
+void ssam_controller_remove_clients(struct ssam_controller *ctrl)
+{
+	struct device *dev;
+
+	dev = ssam_controller_device(ctrl);
+	device_for_each_child_reverse(dev, NULL, ssam_remove_device);
+}
+
+
+/**
+ * ssam_bus_register() - Register and set-up the SSAM client device bus.
+ */
+int ssam_bus_register(void)
+{
+	return bus_register(&ssam_bus_type);
+}
+
+/**
+ * ssam_bus_unregister() - Unregister the SSAM client device bus.
+ */
+void ssam_bus_unregister(void)
+{
+	return bus_unregister(&ssam_bus_type);
+}
diff --git a/drivers/misc/surface_aggregator/bus.h b/drivers/misc/surface_aggregator/bus.h
new file mode 100644
index 000000000000..7712baaed6a5
--- /dev/null
+++ b/drivers/misc/surface_aggregator/bus.h
@@ -0,0 +1,27 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Surface System Aggregator Module bus and device integration.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _SURFACE_AGGREGATOR_BUS_H
+#define _SURFACE_AGGREGATOR_BUS_H
+
+#include <linux/surface_aggregator/controller.h>
+
+#ifdef CONFIG_SURFACE_AGGREGATOR_BUS
+
+void ssam_controller_remove_clients(struct ssam_controller *ctrl);
+
+int ssam_bus_register(void);
+void ssam_bus_unregister(void);
+
+#else /* CONFIG_SURFACE_AGGREGATOR_BUS */
+
+static inline void ssam_controller_remove_clients(struct ssam_controller *ctrl) {}
+static inline int ssam_bus_register(void) { return 0; }
+static inline void ssam_bus_unregister(void) {}
+
+#endif /* CONFIG_SURFACE_AGGREGATOR_BUS */
+#endif /* _SURFACE_AGGREGATOR_BUS_H */
diff --git a/drivers/misc/surface_aggregator/clients/Kconfig b/drivers/misc/surface_aggregator/clients/Kconfig
new file mode 100644
index 000000000000..3c438cb3f0ca
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/Kconfig
@@ -0,0 +1,151 @@
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+
+config SURFACE_AGGREGATOR_CDEV
+	tristate "Surface System Aggregator Module User-Space Interface"
+	depends on SURFACE_AGGREGATOR
+	help
+	  Provides a misc-device interface to the Surface System Aggregator
+	  Module (SSAM) controller.
+
+	  This option provides a module (called surface_aggregator_cdev), that,
+	  when loaded, will add a client device (and its respective driver) to
+	  the SSAM controller. Said client device manages a misc-device
+	  interface (/dev/surface/aggregator), which can be used by user-space
+	  tools to directly communicate with the SSAM EC by sending requests and
+	  receiving the correspondign responses.
+
+	  The provided interface is intended for debugging and development only,
+	  and should not be used otherwise.
+
+config SURFACE_AGGREGATOR_REGISTRY
+	tristate "Surface System Aggregator Module Device Registry"
+	depends on SURFACE_AGGREGATOR_BUS
+	default m
+	help
+	  Device-registry and device-hubs for Surface System Aggregator Module
+	  (SSAM) devices.
+
+	  Provides a module and driver which act as device-registry for SSAM
+	  client devices that cannot be detected automatically, e.g. via ACPI.
+	  Such devices are instead provided via this registry and attached via
+	  device hubs, also provided in this module.
+
+	  Devices provided via this registry are:
+	  - performance / cooling mode device (all generations)
+	  - battery/AC devices (7th generation)
+	  - HID input devices (7th generation)
+
+	  Note that this module only provides the respective client devices.
+	  Drivers for these devices still need to be selected via the other
+	  options.
+
+config SURFACE_ACPI_NOTIFY
+	tristate "Surface ACPI Notify Driver"
+	depends on SURFACE_AGGREGATOR
+	default m
+	help
+	  Surface ACPI Notify (SAN) driver for Microsoft Surface devices.
+
+	  This driver provides support for the ACPI interface (called SAN) of
+	  the Surface System Aggregator Module (SSAM) EC. This interface is used
+	  on 5th- and 6th-generation Microsoft Surface devices (including
+	  Surface Pro 5 and 6, Surface Book 2, Surface Laptops 1 and 2, and in
+	  reduced functionality on the Surface Laptop 3) to execute SSAM
+	  requests directly from ACPI code, as well as receive SSAM events and
+	  turn them into ACPI notifications. It essentially acts as a
+	  translation layer between the SSAM controller and ACPI.
+
+	  Specifically, this driver may be needed for battery status reporting,
+	  thermal sensor access, and real-time clock information, depending on
+	  the Surface device in question.
+
+config SURFACE_BATTERY
+	tristate "Surface Battery Driver"
+	depends on SURFACE_AGGREGATOR_BUS
+	select POWER_SUPPLY
+	default m
+	help
+	  Driver for battery and AC-adapter devices connected/managed via the
+	  Surface System Aggregator Module (SSAM) EC.
+
+	  This driver provides battery-/AC-information and -status support for
+	  Surface devices where said data is not exposed via the standard ACPI
+	  devices. On those models (7th-generation, i.e. Surface pro 7, Surface
+	  Laptop 3, and Surface Book 3), battery-/AC-status and -information is
+	  instead handled directly via SSAM client devices.
+
+config SURFACE_DTX
+	tristate "Surface Detachment System Driver"
+	depends on SURFACE_AGGREGATOR
+	depends on INPUT
+	default m
+	help
+	  Driver for the Surface Book clipboard detachment system (DTX).
+
+	  On the Surface Book series devices, the display part containing the
+	  CPU (called the clipboard) can be detached from the base (containing a
+	  battery, the keyboard, and, optionally, a discrete GPU) by (if
+	  necessary) unlocking and opening the latch connecting both parts.
+
+	  This driver provides a user-space interface that can influence the
+	  behavior of this process, which includes the option to abort it in
+	  case the base is still in use or speed it up in case it is not.
+
+	  Note that this module can be built without support for the Surface
+	  Aggregator Bus (i.e. CONFIG_SURFACE_AGGREGATOR_BUS=n). In that case,
+	  some devices, specifically the Surface Book 3, will not be supported.
+
+config SURFACE_HID
+	tristate "Surface HID Transport Driver"
+	depends on SURFACE_AGGREGATOR_BUS
+	depends on HID
+	default m
+	help
+	  Transport driver for HID devices connected via the Surface System
+	  Aggregator Module (SSAM).
+
+	  This driver provides support for HID input devices (e.g. touchpad and
+	  keyboard) connected via SSAM. It is required for keyboard input on the
+	  Surface Laptop 1 and 2, as well as keyboard and touchpad input on the
+	  Surface Laptop 3 and Surface Book 3.
+
+	  Note that this module can be built without support for the Surface
+	  Aggregator Bus (i.e. CONFIG_SURFACE_AGGREGATOR_BUS=n). In that case,
+	  some devices, specifically the Surface Book 3 and Surface Laptop 3,
+	  will not be supported.
+
+config SURFACE_HOTPLUG
+	tristate "Surface Hot-Plug System Driver"
+	depends on SURFACE_AGGREGATOR
+	depends on SURFACE_ACPI_NOTIFY
+	depends on GPIO_SYSFS
+	default m
+	help
+	  Driver for the Surface discrete GPU (dGPU) hot-plug system.
+
+	  This driver manages the dGPU power on the Surface Books, including
+	  when hot-plugging it by detaching the clipboard (display part
+	  containing the CPU) from the base (containing the keyboard and dGPU)
+	  of the device when it is running. This driver also provides a
+	  user-space interface via which the dGPU power-state (on/off) can be
+	  set, allowing users to turn off the dGPU in order to reduce power
+	  consumption.
+
+config SURFACE_PERFMODE
+	tristate "Surface Performance-Mode Driver"
+	depends on SURFACE_AGGREGATOR_BUS
+	depends on SYSFS
+	default m
+	help
+	  Driver for the performance-/cooling-mode interface of Microsoft
+	  Surface devices.
+
+	  Microsoft Surface devices using the Surface System Aggregator Module
+	  (SSAM) can be switched between different performance modes. This,
+	  depending on the device, can influence their cooling behavior and may
+	  influence power limits, allowing users to choose between performance
+	  and higher power-draw, or lower power-draw and more silent operation.
+
+	  This driver provides a user-space interface (via sysfs) for
+	  controlling said mode via the corresponding client device.
diff --git a/drivers/misc/surface_aggregator/clients/Makefile b/drivers/misc/surface_aggregator/clients/Makefile
new file mode 100644
index 000000000000..7320922ba755
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/Makefile
@@ -0,0 +1,11 @@
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+
+obj-$(CONFIG_SURFACE_AGGREGATOR_CDEV)		+= surface_aggregator_cdev.o
+obj-$(CONFIG_SURFACE_AGGREGATOR_REGISTRY)	+= surface_aggregator_registry.o
+obj-$(CONFIG_SURFACE_ACPI_NOTIFY)		+= surface_acpi_notify.o
+obj-$(CONFIG_SURFACE_BATTERY)			+= surface_battery.o
+obj-$(CONFIG_SURFACE_DTX)			+= surface_dtx.o
+obj-$(CONFIG_SURFACE_HID)			+= surface_hid.o
+obj-$(CONFIG_SURFACE_HOTPLUG)			+= surface_hotplug.o
+obj-$(CONFIG_SURFACE_PERFMODE) 			+= surface_perfmode.o
diff --git a/drivers/misc/surface_aggregator/clients/surface_acpi_notify.c b/drivers/misc/surface_aggregator/clients/surface_acpi_notify.c
new file mode 100644
index 000000000000..7a5b8f280036
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_acpi_notify.c
@@ -0,0 +1,884 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Driver for the Surface ACPI Notify (SAN) interface/shim.
+ *
+ * Translates communication from ACPI to Surface System Aggregator Module
+ * (SSAM/SAM) requests and back, specifically SAM-over-SSH. Translates SSAM
+ * events back to ACPI notifications. Allows handling of discrete GPU
+ * notifications sent from ACPI via the SAN interface by providing them to any
+ * registered external driver.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <asm/unaligned.h>
+#include <linux/acpi.h>
+#include <linux/delay.h>
+#include <linux/jiffies.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/notifier.h>
+#include <linux/platform_device.h>
+#include <linux/rwsem.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_acpi_notify.h>
+
+
+struct san_data {
+	struct device *dev;
+	struct ssam_controller *ctrl;
+
+	struct acpi_connection_info info;
+
+	struct ssam_event_notifier nf_bat;
+	struct ssam_event_notifier nf_tmp;
+};
+
+#define to_san_data(ptr, member) \
+	container_of(ptr, struct san_data, member)
+
+
+/* -- dGPU notifier interface. ---------------------------------------------- */
+
+struct san_rqsg_if {
+	struct rw_semaphore lock;
+	struct device *dev;
+	struct blocking_notifier_head nh;
+};
+
+static struct san_rqsg_if san_rqsg_if = {
+	.lock = __RWSEM_INITIALIZER(san_rqsg_if.lock),
+	.dev = NULL,
+	.nh = BLOCKING_NOTIFIER_INIT(san_rqsg_if.nh),
+};
+
+static int san_set_rqsg_interface_device(struct device *dev)
+{
+	int status = 0;
+
+	down_write(&san_rqsg_if.lock);
+	if (!san_rqsg_if.dev && dev)
+		san_rqsg_if.dev = dev;
+	else
+		status = -EBUSY;
+	up_write(&san_rqsg_if.lock);
+
+	return status;
+}
+
+/**
+ * san_client_link() - Link client as consumer to SAN device.
+ * @client: The client to link.
+ *
+ * Sets up a device link between the provided client device as consumer and
+ * the SAN device as provider. This function can be used to ensure that the
+ * SAN interface has been set up and will be set up for as long as the driver
+ * of the client device is bound. This guarantees that, during that time, all
+ * dGPU events will be received by any registered notifier.
+ *
+ * The link will be automatically removed once the client device's driver is
+ * unbound.
+ *
+ * Return: Returns zero on succes, %-ENXIO if the SAN interface has not been
+ * set up yet, and %-ENOMEM if device link creation failed.
+ */
+int san_client_link(struct device *client)
+{
+	const u32 flags = DL_FLAG_PM_RUNTIME | DL_FLAG_AUTOREMOVE_CONSUMER;
+	struct device_link *link;
+
+	down_read(&san_rqsg_if.lock);
+
+	if (!san_rqsg_if.dev) {
+		up_read(&san_rqsg_if.lock);
+		return -ENXIO;
+	}
+
+	link = device_link_add(client, san_rqsg_if.dev, flags);
+	if (!link) {
+		up_read(&san_rqsg_if.lock);
+		return -ENOMEM;
+	}
+
+	if (READ_ONCE(link->status) == DL_STATE_SUPPLIER_UNBIND) {
+		up_read(&san_rqsg_if.lock);
+		return -ENXIO;
+	}
+
+	up_read(&san_rqsg_if.lock);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(san_client_link);
+
+/**
+ * san_dgpu_notifier_register() - Register a SAN dGPU notifier.
+ * @nb: The notifier-block to register.
+ *
+ * Registers a SAN dGPU notifier, receiving any new SAN dGPU events sent from
+ * ACPI. The registered notifier will be called with &struct san_dgpu_event
+ * as notifier data and the command ID of that event as notifier action.
+ */
+int san_dgpu_notifier_register(struct notifier_block *nb)
+{
+	return blocking_notifier_chain_register(&san_rqsg_if.nh, nb);
+}
+EXPORT_SYMBOL_GPL(san_dgpu_notifier_register);
+
+/**
+ * san_dgpu_notifier_unregister() - Unregister a SAN dGPU notifier.
+ * @nb: The notifier-block to unregister.
+ */
+int san_dgpu_notifier_unregister(struct notifier_block *nb)
+{
+	return blocking_notifier_chain_unregister(&san_rqsg_if.nh, nb);
+}
+EXPORT_SYMBOL_GPL(san_dgpu_notifier_unregister);
+
+static int san_dgpu_notifier_call(struct san_dgpu_event *evt)
+{
+	int ret;
+
+	ret = blocking_notifier_call_chain(&san_rqsg_if.nh, evt->command, evt);
+	return notifier_to_errno(ret);
+}
+
+
+/* -- ACPI _DSM event relay. ------------------------------------------------ */
+
+#define SAN_DSM_REVISION	0
+
+static const guid_t SAN_DSM_UUID =
+	GUID_INIT(0x93b666c5, 0x70c6, 0x469f, 0xa2, 0x15, 0x3d,
+		  0x48, 0x7c, 0x91, 0xab, 0x3c);
+
+enum san_dsm_event_fn {
+	SAN_DSM_EVENT_FN_BAT1_STAT = 0x03,
+	SAN_DSM_EVENT_FN_BAT1_INFO = 0x04,
+	SAN_DSM_EVENT_FN_ADP1_STAT = 0x05,
+	SAN_DSM_EVENT_FN_ADP1_INFO = 0x06,
+	SAN_DSM_EVENT_FN_BAT2_STAT = 0x07,
+	SAN_DSM_EVENT_FN_BAT2_INFO = 0x08,
+	SAN_DSM_EVENT_FN_THERMAL   = 0x09,
+	SAN_DSM_EVENT_FN_DPTF      = 0x0a,
+};
+
+enum sam_event_cid_bat {
+	SAM_EVENT_CID_BAT_BIX  = 0x15,
+	SAM_EVENT_CID_BAT_BST  = 0x16,
+	SAM_EVENT_CID_BAT_ADP  = 0x17,
+	SAM_EVENT_CID_BAT_PROT = 0x18,
+	SAM_EVENT_CID_BAT_DPTF = 0x4f,
+};
+
+enum sam_event_cid_tmp {
+	SAM_EVENT_CID_TMP_TRIP = 0x0b,
+};
+
+struct san_event_work {
+	struct delayed_work work;
+	struct device *dev;
+	struct ssam_event event;	// must be last
+};
+
+static int san_acpi_notify_event(struct device *dev, u64 func,
+				 union acpi_object *param)
+{
+	acpi_handle san = ACPI_HANDLE(dev);
+	union acpi_object *obj;
+	int status = 0;
+
+	if (!acpi_check_dsm(san, &SAN_DSM_UUID, SAN_DSM_REVISION, 1 << func))
+		return 0;
+
+	dev_dbg(dev, "notify event 0x%02llx\n", func);
+
+	obj = acpi_evaluate_dsm_typed(san, &SAN_DSM_UUID, SAN_DSM_REVISION,
+				      func, param, ACPI_TYPE_BUFFER);
+	if (!obj)
+		return -EFAULT;
+
+	if (obj->buffer.length != 1 || obj->buffer.pointer[0] != 0) {
+		dev_err(dev, "got unexpected result from _DSM\n");
+		status = -EPROTO;
+	}
+
+	ACPI_FREE(obj);
+	return status;
+}
+
+static int san_evt_bat_adp(struct device *dev, const struct ssam_event *event)
+{
+	int status;
+
+	status = san_acpi_notify_event(dev, SAN_DSM_EVENT_FN_ADP1_STAT, NULL);
+	if (status)
+		return status;
+
+	/*
+	 * Enusre that the battery states get updated correctly.
+	 * When the battery is fully charged and an adapter is plugged in, it
+	 * sometimes is not updated correctly, instead showing it as charging.
+	 * Explicitly trigger battery updates to fix this.
+	 */
+
+	status = san_acpi_notify_event(dev, SAN_DSM_EVENT_FN_BAT1_STAT, NULL);
+	if (status)
+		return status;
+
+	return san_acpi_notify_event(dev, SAN_DSM_EVENT_FN_BAT2_STAT, NULL);
+}
+
+static int san_evt_bat_bix(struct device *dev, const struct ssam_event *event)
+{
+	enum san_dsm_event_fn fn;
+
+	if (event->instance_id == 0x02)
+		fn = SAN_DSM_EVENT_FN_BAT2_INFO;
+	else
+		fn = SAN_DSM_EVENT_FN_BAT1_INFO;
+
+	return san_acpi_notify_event(dev, fn, NULL);
+}
+
+static int san_evt_bat_bst(struct device *dev, const struct ssam_event *event)
+{
+	enum san_dsm_event_fn fn;
+
+	if (event->instance_id == 0x02)
+		fn = SAN_DSM_EVENT_FN_BAT2_STAT;
+	else
+		fn = SAN_DSM_EVENT_FN_BAT1_STAT;
+
+	return san_acpi_notify_event(dev, fn, NULL);
+}
+
+static int san_evt_bat_dptf(struct device *dev, const struct ssam_event *event)
+{
+	union acpi_object payload;
+
+	/*
+	 * The Surface ACPI expects a buffer and not a package. It specifically
+	 * checks for ObjectType (Arg3) == 0x03. This will cause a warning in
+	 * acpica/nsarguments.c, but that warning can be safely ignored.
+	 */
+	payload.type = ACPI_TYPE_BUFFER;
+	payload.buffer.length = event->length;
+	payload.buffer.pointer = (u8 *)&event->data[0];
+
+	return san_acpi_notify_event(dev, SAN_DSM_EVENT_FN_DPTF, &payload);
+}
+
+static unsigned long san_evt_bat_delay(u8 cid)
+{
+	switch (cid) {
+	case SAM_EVENT_CID_BAT_ADP:
+		/*
+		 * Wait for battery state to update before signalling adapter
+		 * change.
+		 */
+		return msecs_to_jiffies(5000);
+
+	case SAM_EVENT_CID_BAT_BST:
+		/* Ensure we do not miss anything important due to caching. */
+		return msecs_to_jiffies(2000);
+
+	default:
+		return 0;
+	}
+}
+
+static bool san_evt_bat(const struct ssam_event *event, struct device *dev)
+{
+	int status;
+
+	switch (event->command_id) {
+	case SAM_EVENT_CID_BAT_BIX:
+		status = san_evt_bat_bix(dev, event);
+		break;
+
+	case SAM_EVENT_CID_BAT_BST:
+		status = san_evt_bat_bst(dev, event);
+		break;
+
+	case SAM_EVENT_CID_BAT_ADP:
+		status = san_evt_bat_adp(dev, event);
+		break;
+
+	case SAM_EVENT_CID_BAT_PROT:
+		/*
+		 * TODO: Implement support for battery protection status change
+		 *       event.
+		 */
+		return true;
+
+	case SAM_EVENT_CID_BAT_DPTF:
+		status = san_evt_bat_dptf(dev, event);
+		break;
+
+	default:
+		return false;
+	}
+
+	if (status)
+		dev_err(dev, "error handling power event (cid = %x)\n",
+			event->command_id);
+
+	return true;
+}
+
+static void san_evt_bat_workfn(struct work_struct *work)
+{
+	struct san_event_work *ev;
+
+	ev = container_of(work, struct san_event_work, work.work);
+	san_evt_bat(&ev->event, ev->dev);
+	kfree(ev);
+}
+
+static u32 san_evt_bat_nf(struct ssam_event_notifier *nf,
+			  const struct ssam_event *event)
+{
+	struct san_data *d = to_san_data(nf, nf_bat);
+	struct san_event_work *work;
+	unsigned long delay = san_evt_bat_delay(event->command_id);
+
+	if (delay == 0)
+		return san_evt_bat(event, d->dev) ? SSAM_NOTIF_HANDLED : 0;
+
+	work = kzalloc(sizeof(*work) + event->length, GFP_KERNEL);
+	if (!work)
+		return ssam_notifier_from_errno(-ENOMEM);
+
+	INIT_DELAYED_WORK(&work->work, san_evt_bat_workfn);
+	work->dev = d->dev;
+
+	memcpy(&work->event, event, sizeof(struct ssam_event) + event->length);
+
+	schedule_delayed_work(&work->work, delay);
+	return SSAM_NOTIF_HANDLED;
+}
+
+static int san_evt_tmp_trip(struct device *dev, const struct ssam_event *event)
+{
+	union acpi_object param;
+
+	/*
+	 * The Surface ACPI expects an integer and not a package. This will
+	 * cause a warning in acpica/nsarguments.c, but that warning can be
+	 * safely ignored.
+	 */
+	param.type = ACPI_TYPE_INTEGER;
+	param.integer.value = event->instance_id;
+
+	return san_acpi_notify_event(dev, SAN_DSM_EVENT_FN_THERMAL, &param);
+}
+
+static bool san_evt_tmp(const struct ssam_event *event, struct device *dev)
+{
+	int status;
+
+	switch (event->command_id) {
+	case SAM_EVENT_CID_TMP_TRIP:
+		status = san_evt_tmp_trip(dev, event);
+		break;
+
+	default:
+		return false;
+	}
+
+	if (status) {
+		dev_err(dev, "error handling thermal event (cid = %x)\n",
+			event->command_id);
+	}
+
+	return true;
+}
+
+static u32 san_evt_tmp_nf(struct ssam_event_notifier *nf,
+			  const struct ssam_event *event)
+{
+	struct san_data *d = to_san_data(nf, nf_tmp);
+
+	return san_evt_tmp(event, d->dev) ? SSAM_NOTIF_HANDLED : 0;
+}
+
+
+/* -- ACPI GSB OperationRegion handler -------------------------------------- */
+
+struct gsb_data_in {
+	u8 cv;
+} __packed;
+
+struct gsb_data_rqsx {
+	u8 cv;				// command value (san_gsb_request_cv)
+	u8 tc;				// target category
+	u8 tid;				// target ID
+	u8 iid;				// instance ID
+	u8 snc;				// expect-response-flag?
+	u8 cid;				// command ID
+	u16 cdl;			// payload length
+	u8 pld[];			// payload
+} __packed;
+
+struct gsb_data_etwl {
+	u8 cv;				// command value (should be 0x02)
+	u8 etw3;			// unknown
+	u8 etw4;			// unknown
+	u8 msg[];			// error message (ASCIIZ)
+} __packed;
+
+struct gsb_data_out {
+	u8 status;			// _SSH communication status
+	u8 len;				// _SSH payload length
+	u8 pld[];			// _SSH payload
+} __packed;
+
+union gsb_buffer_data {
+	struct gsb_data_in   in;	// common input
+	struct gsb_data_rqsx rqsx;	// RQSX input
+	struct gsb_data_etwl etwl;	// ETWL input
+	struct gsb_data_out  out;	// output
+};
+
+struct gsb_buffer {
+	u8 status;			// GSB AttribRawProcess status
+	u8 len;				// GSB AttribRawProcess length
+	union gsb_buffer_data data;
+} __packed;
+
+#define SAN_GSB_MAX_RQSX_PAYLOAD  (U8_MAX - 2 - sizeof(struct gsb_data_rqsx))
+#define SAN_GSB_MAX_RESPONSE	  (U8_MAX - 2 - sizeof(struct gsb_data_out))
+
+#define SAN_GSB_COMMAND		0
+
+enum san_gsb_request_cv {
+	SAN_GSB_REQUEST_CV_RQST = 0x01,
+	SAN_GSB_REQUEST_CV_ETWL = 0x02,
+	SAN_GSB_REQUEST_CV_RQSG = 0x03,
+};
+
+#define SAN_REQUEST_NUM_TRIES	5
+
+static acpi_status san_etwl(struct san_data *d, struct gsb_buffer *b)
+{
+	struct gsb_data_etwl *etwl = &b->data.etwl;
+
+	if (b->len < sizeof(struct gsb_data_etwl)) {
+		dev_err(d->dev, "invalid ETWL package (len = %d)\n", b->len);
+		return AE_OK;
+	}
+
+	dev_err(d->dev, "ETWL(0x%02x, 0x%02x): %.*s\n", etwl->etw3, etwl->etw4,
+		(unsigned int)(b->len - sizeof(struct gsb_data_etwl)),
+		(char *)etwl->msg);
+
+	// indicate success
+	b->status = 0x00;
+	b->len = 0x00;
+
+	return AE_OK;
+}
+
+static struct gsb_data_rqsx *san_validate_rqsx(struct device *dev,
+		const char *type, struct gsb_buffer *b)
+{
+	struct gsb_data_rqsx *rqsx = &b->data.rqsx;
+
+	if (b->len < sizeof(struct gsb_data_rqsx)) {
+		dev_err(dev, "invalid %s package (len = %d)\n", type, b->len);
+		return NULL;
+	}
+
+	if (get_unaligned(&rqsx->cdl) != b->len - sizeof(struct gsb_data_rqsx)) {
+		dev_err(dev, "bogus %s package (len = %d, cdl = %d)\n",
+			type, b->len, get_unaligned(&rqsx->cdl));
+		return NULL;
+	}
+
+	if (get_unaligned(&rqsx->cdl) > SAN_GSB_MAX_RQSX_PAYLOAD) {
+		dev_err(dev, "payload for %s package too large (cdl = %d)\n",
+			type, get_unaligned(&rqsx->cdl));
+		return NULL;
+	}
+
+	return rqsx;
+}
+
+static void gsb_rqsx_response_error(struct gsb_buffer *gsb, int status)
+{
+	gsb->status = 0x00;
+	gsb->len = 0x02;
+	gsb->data.out.status = (u8)(-status);
+	gsb->data.out.len = 0x00;
+}
+
+static void gsb_rqsx_response_success(struct gsb_buffer *gsb, u8 *ptr, size_t len)
+{
+	gsb->status = 0x00;
+	gsb->len = len + 2;
+	gsb->data.out.status = 0x00;
+	gsb->data.out.len = len;
+
+	if (len)
+		memcpy(&gsb->data.out.pld[0], ptr, len);
+}
+
+static acpi_status san_rqst_fixup_suspended(struct san_data *d,
+					    struct ssam_request *rqst,
+					    struct gsb_buffer *gsb)
+{
+	if (rqst->target_category == SSAM_SSH_TC_BAS && rqst->command_id == 0x0D) {
+		u8 base_state = 1;
+
+		/* Base state quirk:
+		 * The base state may be queried from ACPI when the EC is still
+		 * suspended. In this case it will return '-EPERM'. This query
+		 * will only be triggered from the ACPI lid GPE interrupt, thus
+		 * we are either in laptop or studio mode (base status 0x01 or
+		 * 0x02). Furthermore, we will only get here if the device (and
+		 * EC) have been suspended.
+		 *
+		 * We now assume that the device is in laptop mode (0x01). This
+		 * has the drawback that it will wake the device when unfolding
+		 * it in studio mode, but it also allows us to avoid actively
+		 * waiting for the EC to wake up, which may incur a notable
+		 * delay.
+		 */
+
+		dev_dbg(d->dev, "rqst: fixup: base-state quirk\n");
+
+		gsb_rqsx_response_success(gsb, &base_state, sizeof(base_state));
+		return AE_OK;
+	}
+
+	gsb_rqsx_response_error(gsb, -ENXIO);
+	return AE_OK;
+}
+
+static acpi_status san_rqst(struct san_data *d, struct gsb_buffer *buffer)
+{
+	u8 rspbuf[SAN_GSB_MAX_RESPONSE];
+	struct gsb_data_rqsx *gsb_rqst;
+	struct ssam_request rqst;
+	struct ssam_response rsp;
+	int status = 0;
+
+	gsb_rqst = san_validate_rqsx(d->dev, "RQST", buffer);
+	if (!gsb_rqst)
+		return AE_OK;
+
+	rqst.target_category = gsb_rqst->tc;
+	rqst.target_id = gsb_rqst->tid;
+	rqst.command_id = gsb_rqst->cid;
+	rqst.instance_id = gsb_rqst->iid;
+	rqst.flags = gsb_rqst->snc ? SSAM_REQUEST_HAS_RESPONSE : 0;
+	rqst.length = get_unaligned(&gsb_rqst->cdl);
+	rqst.payload = &gsb_rqst->pld[0];
+
+	rsp.capacity = ARRAY_SIZE(rspbuf);
+	rsp.length = 0;
+	rsp.pointer = &rspbuf[0];
+
+	// handle suspended device
+	if (d->dev->power.is_suspended) {
+		dev_warn(d->dev, "rqst: device is suspended, not executing\n");
+		return san_rqst_fixup_suspended(d, &rqst, buffer);
+	}
+
+	status = __ssam_retry(ssam_request_sync_onstack, SAN_REQUEST_NUM_TRIES,
+			      d->ctrl, &rqst, &rsp, SAN_GSB_MAX_RQSX_PAYLOAD);
+
+	if (!status) {
+		gsb_rqsx_response_success(buffer, rsp.pointer, rsp.length);
+	} else {
+		dev_err(d->dev, "rqst: failed with error %d\n", status);
+		gsb_rqsx_response_error(buffer, status);
+	}
+
+	return AE_OK;
+}
+
+static acpi_status san_rqsg(struct san_data *d, struct gsb_buffer *buffer)
+{
+	struct gsb_data_rqsx *gsb_rqsg;
+	struct san_dgpu_event evt;
+	int status;
+
+	gsb_rqsg = san_validate_rqsx(d->dev, "RQSG", buffer);
+	if (!gsb_rqsg)
+		return AE_OK;
+
+	evt.category = gsb_rqsg->tc;
+	evt.target = gsb_rqsg->tid;
+	evt.command = gsb_rqsg->cid;
+	evt.instance = gsb_rqsg->iid;
+	evt.length = get_unaligned(&gsb_rqsg->cdl);
+	evt.payload = &gsb_rqsg->pld[0];
+
+	status = san_dgpu_notifier_call(&evt);
+	if (!status) {
+		gsb_rqsx_response_success(buffer, NULL, 0);
+	} else {
+		dev_err(d->dev, "rqsg: failed with error %d\n", status);
+		gsb_rqsx_response_error(buffer, status);
+	}
+
+	return AE_OK;
+}
+
+static acpi_status san_opreg_handler(u32 function,
+		acpi_physical_address command, u32 bits, u64 *value64,
+		void *opreg_context, void *region_context)
+{
+	struct san_data *d = to_san_data(opreg_context, info);
+	struct gsb_buffer *buffer = (struct gsb_buffer *)value64;
+	int accessor_type = (function & 0xFFFF0000) >> 16;
+
+	if (command != SAN_GSB_COMMAND) {
+		dev_warn(d->dev, "unsupported command: 0x%02llx\n", command);
+		return AE_OK;
+	}
+
+	if (accessor_type != ACPI_GSB_ACCESS_ATTRIB_RAW_PROCESS) {
+		dev_err(d->dev, "invalid access type: 0x%02x\n", accessor_type);
+		return AE_OK;
+	}
+
+	// buffer must have at least contain the command-value
+	if (buffer->len == 0) {
+		dev_err(d->dev, "request-package too small\n");
+		return AE_OK;
+	}
+
+	switch (buffer->data.in.cv) {
+	case SAN_GSB_REQUEST_CV_RQST:
+		return san_rqst(d, buffer);
+
+	case SAN_GSB_REQUEST_CV_ETWL:
+		return san_etwl(d, buffer);
+
+	case SAN_GSB_REQUEST_CV_RQSG:
+		return san_rqsg(d, buffer);
+
+	default:
+		dev_warn(d->dev, "unsupported SAN0 request (cv: 0x%02x)\n",
+			 buffer->data.in.cv);
+		return AE_OK;
+	}
+}
+
+
+/* -- Driver setup. --------------------------------------------------------- */
+
+static int san_events_register(struct platform_device *pdev)
+{
+	struct san_data *d = platform_get_drvdata(pdev);
+	int status;
+
+	d->nf_bat.base.priority = 1;
+	d->nf_bat.base.fn = san_evt_bat_nf;
+	d->nf_bat.event.reg = SSAM_EVENT_REGISTRY_SAM;
+	d->nf_bat.event.id.target_category = SSAM_SSH_TC_BAT;
+	d->nf_bat.event.id.instance = 0;
+	d->nf_bat.event.mask = SSAM_EVENT_MASK_TARGET;
+	d->nf_bat.event.flags = SSAM_EVENT_SEQUENCED;
+
+	d->nf_tmp.base.priority = 1;
+	d->nf_tmp.base.fn = san_evt_tmp_nf;
+	d->nf_tmp.event.reg = SSAM_EVENT_REGISTRY_SAM;
+	d->nf_tmp.event.id.target_category = SSAM_SSH_TC_TMP;
+	d->nf_tmp.event.id.instance = 0;
+	d->nf_tmp.event.mask = SSAM_EVENT_MASK_TARGET;
+	d->nf_tmp.event.flags = SSAM_EVENT_SEQUENCED;
+
+	status = ssam_notifier_register(d->ctrl, &d->nf_bat);
+	if (status)
+		return status;
+
+	status = ssam_notifier_register(d->ctrl, &d->nf_tmp);
+	if (status)
+		ssam_notifier_unregister(d->ctrl, &d->nf_bat);
+
+	return status;
+}
+
+static void san_events_unregister(struct platform_device *pdev)
+{
+	struct san_data *d = platform_get_drvdata(pdev);
+
+	ssam_notifier_unregister(d->ctrl, &d->nf_bat);
+	ssam_notifier_unregister(d->ctrl, &d->nf_tmp);
+}
+
+#define san_consumer_printk(level, dev, handle, fmt, ...)			\
+do {										\
+	char *path = "<error getting consumer path>";				\
+	struct acpi_buffer buffer = {						\
+		.length = ACPI_ALLOCATE_BUFFER,					\
+		.pointer = NULL,						\
+	};									\
+										\
+	if (ACPI_SUCCESS(acpi_get_name(handle, ACPI_FULL_PATHNAME, &buffer)))	\
+		path = buffer.pointer;						\
+										\
+	dev_##level(dev, "[%s]: " fmt, path, ##__VA_ARGS__);			\
+	kfree(buffer.pointer);							\
+} while (0)
+
+#define san_consumer_dbg(dev, handle, fmt, ...) \
+	san_consumer_printk(dbg, dev, handle, fmt, ##__VA_ARGS__)
+
+#define san_consumer_warn(dev, handle, fmt, ...) \
+	san_consumer_printk(warn, dev, handle, fmt, ##__VA_ARGS__)
+
+static bool is_san_consumer(struct platform_device *pdev, acpi_handle handle)
+{
+	struct acpi_handle_list dep_devices;
+	acpi_handle supplier = ACPI_HANDLE(&pdev->dev);
+	acpi_status status;
+	int i;
+
+	if (!acpi_has_method(handle, "_DEP"))
+		return false;
+
+	status = acpi_evaluate_reference(handle, "_DEP", NULL, &dep_devices);
+	if (ACPI_FAILURE(status)) {
+		san_consumer_dbg(&pdev->dev, handle, "failed to evaluate _DEP\n");
+		return false;
+	}
+
+	for (i = 0; i < dep_devices.count; i++) {
+		if (dep_devices.handles[i] == supplier)
+			return true;
+	}
+
+	return false;
+}
+
+static acpi_status san_consumer_setup(acpi_handle handle, u32 lvl,
+				      void *context, void **rv)
+{
+	const u32 flags = DL_FLAG_PM_RUNTIME | DL_FLAG_AUTOREMOVE_SUPPLIER;
+	struct platform_device *pdev = context;
+	struct acpi_device *adev;
+	struct device_link *link;
+
+	if (!is_san_consumer(pdev, handle))
+		return AE_OK;
+
+	// ignore ACPI devices that are not present
+	if (acpi_bus_get_device(handle, &adev) != 0)
+		return AE_OK;
+
+	san_consumer_dbg(&pdev->dev, handle, "creating device link\n");
+
+	// try to set up device links, ignore but log errors
+	link = device_link_add(&adev->dev, &pdev->dev, flags);
+	if (!link) {
+		san_consumer_warn(&pdev->dev, handle,
+				  "failed to create device link\n");
+		return AE_OK;
+	}
+
+	return AE_OK;
+}
+
+static int san_consumer_links_setup(struct platform_device *pdev)
+{
+	acpi_status status;
+
+	status = acpi_walk_namespace(ACPI_TYPE_DEVICE, ACPI_ROOT_OBJECT,
+				     ACPI_UINT32_MAX, san_consumer_setup, NULL,
+				     pdev, NULL);
+
+	return status ? -EFAULT : 0;
+}
+
+static int san_probe(struct platform_device *pdev)
+{
+	acpi_handle san = ACPI_HANDLE(&pdev->dev);
+	struct ssam_controller *ctrl;
+	struct san_data *data;
+	acpi_status astatus;
+	int status;
+
+	status = ssam_client_bind(&pdev->dev, &ctrl);
+	if (status)
+		return status == -ENXIO ? -EPROBE_DEFER : status;
+
+	status = san_consumer_links_setup(pdev);
+	if (status)
+		return status;
+
+	data = devm_kzalloc(&pdev->dev, sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	data->dev = &pdev->dev;
+	data->ctrl = ctrl;
+
+	platform_set_drvdata(pdev, data);
+
+	astatus = acpi_install_address_space_handler(san, ACPI_ADR_SPACE_GSBUS,
+			&san_opreg_handler, NULL, &data->info);
+	if (ACPI_FAILURE(astatus))
+		return -ENXIO;
+
+	status = san_events_register(pdev);
+	if (status)
+		goto err_enable_events;
+
+	status = san_set_rqsg_interface_device(&pdev->dev);
+	if (status)
+		goto err_install_dev;
+
+	acpi_walk_dep_device_list(san);
+	return 0;
+
+err_install_dev:
+	san_events_unregister(pdev);
+err_enable_events:
+	acpi_remove_address_space_handler(san, ACPI_ADR_SPACE_GSBUS,
+					  &san_opreg_handler);
+	return status;
+}
+
+static int san_remove(struct platform_device *pdev)
+{
+	acpi_handle san = ACPI_HANDLE(&pdev->dev);
+
+	san_set_rqsg_interface_device(NULL);
+	acpi_remove_address_space_handler(san, ACPI_ADR_SPACE_GSBUS,
+					  &san_opreg_handler);
+	san_events_unregister(pdev);
+
+	/*
+	 * We have unregistered our event sources. Now we need to ensure that
+	 * all delayed works they may have spawned are run to completion.
+	 */
+	flush_scheduled_work();
+
+	return 0;
+}
+
+static const struct acpi_device_id san_match[] = {
+	{ "MSHW0091" },
+	{ },
+};
+MODULE_DEVICE_TABLE(acpi, san_match);
+
+static struct platform_driver surface_acpi_notify = {
+	.probe = san_probe,
+	.remove = san_remove,
+	.driver = {
+		.name = "surface_acpi_notify",
+		.acpi_match_table = san_match,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+module_platform_driver(surface_acpi_notify);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Surface ACPI Notify driver for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/clients/surface_aggregator_cdev.c b/drivers/misc/surface_aggregator/clients/surface_aggregator_cdev.c
new file mode 100644
index 000000000000..f5e81cd67357
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_aggregator_cdev.c
@@ -0,0 +1,299 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Provides user-space access to the SSAM EC via the /dev/surface/aggregator
+ * misc device. Intended for debugging and development.
+ *
+ * Copyright (C) 2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/fs.h>
+#include <linux/kernel.h>
+#include <linux/kref.h>
+#include <linux/miscdevice.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/rwsem.h>
+#include <linux/slab.h>
+#include <linux/uaccess.h>
+
+#include <linux/surface_aggregator/cdev.h>
+#include <linux/surface_aggregator/controller.h>
+
+#define SSAM_CDEV_DEVICE_NAME	"surface_aggregator_cdev"
+
+struct ssam_cdev {
+	struct kref kref;
+	struct rw_semaphore lock;
+	struct ssam_controller *ctrl;
+	struct miscdevice mdev;
+};
+
+static void __ssam_cdev_release(struct kref *kref)
+{
+	kfree(container_of(kref, struct ssam_cdev, kref));
+}
+
+static struct ssam_cdev *ssam_cdev_get(struct ssam_cdev *cdev)
+{
+	if (cdev)
+		kref_get(&cdev->kref);
+
+	return cdev;
+}
+
+static void ssam_cdev_put(struct ssam_cdev *cdev)
+{
+	if (cdev)
+		kref_put(&cdev->kref, __ssam_cdev_release);
+}
+
+static int ssam_cdev_device_open(struct inode *inode, struct file *filp)
+{
+	struct miscdevice *mdev = filp->private_data;
+	struct ssam_cdev *cdev = container_of(mdev, struct ssam_cdev, mdev);
+
+	filp->private_data = ssam_cdev_get(cdev);
+	return stream_open(inode, filp);
+}
+
+static int ssam_cdev_device_release(struct inode *inode, struct file *filp)
+{
+	ssam_cdev_put(filp->private_data);
+	return 0;
+}
+
+static long ssam_cdev_request(struct ssam_cdev *cdev, unsigned long arg)
+{
+	struct ssam_cdev_request __user *r;
+	struct ssam_cdev_request rqst;
+	struct ssam_request spec;
+	struct ssam_response rsp;
+	const void __user *plddata;
+	void __user *rspdata;
+	int status = 0, ret = 0, tmp;
+
+	r = (struct ssam_cdev_request __user *)arg;
+	ret = copy_struct_from_user(&rqst, sizeof(rqst), r, sizeof(*r));
+	if (ret)
+		goto out;
+
+	plddata = u64_to_user_ptr(rqst.payload.data);
+	rspdata = u64_to_user_ptr(rqst.response.data);
+
+	// setup basic request fields
+	spec.target_category = rqst.target_category;
+	spec.target_id = rqst.target_id;
+	spec.command_id = rqst.command_id;
+	spec.instance_id = rqst.instance_id;
+	spec.flags = rqst.flags;
+	spec.length = rqst.payload.length;
+	spec.payload = NULL;
+
+	rsp.capacity = rqst.response.length;
+	rsp.length = 0;
+	rsp.pointer = NULL;
+
+	// get request payload from user-space
+	if (spec.length) {
+		if (!plddata) {
+			ret = -EINVAL;
+			goto out;
+		}
+
+		spec.payload = kzalloc(spec.length, GFP_KERNEL);
+		if (!spec.payload) {
+			status = -ENOMEM;
+			ret = -EFAULT;
+			goto out;
+		}
+
+		if (copy_from_user((void *)spec.payload, plddata, spec.length)) {
+			ret = -EFAULT;
+			goto out;
+		}
+	}
+
+	// allocate response buffer
+	if (rsp.capacity) {
+		if (!rspdata) {
+			ret = -EINVAL;
+			goto out;
+		}
+
+		rsp.pointer = kzalloc(rsp.capacity, GFP_KERNEL);
+		if (!rsp.pointer) {
+			status = -ENOMEM;
+			ret = -EFAULT;
+			goto out;
+		}
+	}
+
+	// perform request
+	status = ssam_request_sync(cdev->ctrl, &spec, &rsp);
+	if (status)
+		goto out;
+
+	// copy response to user-space
+	if (rsp.length && copy_to_user(rspdata, rsp.pointer, rsp.length))
+		ret = -EFAULT;
+
+out:
+	// always try to set response-length and status
+	tmp = put_user(rsp.length, &r->response.length);
+	if (tmp)
+		ret = tmp;
+
+	tmp = put_user(status, &r->status);
+	if (tmp)
+		ret = tmp;
+
+	// cleanup
+	kfree(spec.payload);
+	kfree(rsp.pointer);
+
+	return ret;
+}
+
+static long __ssam_cdev_device_ioctl(struct ssam_cdev *cdev, unsigned int cmd,
+				     unsigned long arg)
+{
+	switch (cmd) {
+	case SSAM_CDEV_REQUEST:
+		return ssam_cdev_request(cdev, arg);
+
+	default:
+		return -ENOTTY;
+	}
+}
+
+static long ssam_cdev_device_ioctl(struct file *file, unsigned int cmd,
+				   unsigned long arg)
+{
+	struct ssam_cdev *cdev = file->private_data;
+	long status;
+
+	// ensure that controller is valid for as long as we need it
+	if (down_read_killable(&cdev->lock))
+		return -ERESTARTSYS;
+
+	if (!cdev->ctrl) {
+		up_read(&cdev->lock);
+		return -ENODEV;
+	}
+
+	status = __ssam_cdev_device_ioctl(cdev, cmd, arg);
+
+	up_read(&cdev->lock);
+	return status;
+}
+
+static const struct file_operations ssam_controller_fops = {
+	.owner          = THIS_MODULE,
+	.open           = ssam_cdev_device_open,
+	.release        = ssam_cdev_device_release,
+	.unlocked_ioctl = ssam_cdev_device_ioctl,
+	.compat_ioctl   = ssam_cdev_device_ioctl,
+	.llseek         = noop_llseek,
+};
+
+static int ssam_dbg_device_probe(struct platform_device *pdev)
+{
+	struct ssam_controller *ctrl;
+	struct ssam_cdev *cdev;
+	int status;
+
+	status = ssam_client_bind(&pdev->dev, &ctrl);
+	if (status)
+		return status == -ENXIO ? -EPROBE_DEFER : status;
+
+	cdev = kzalloc(sizeof(*cdev), GFP_KERNEL);
+	if (!cdev)
+		return -ENOMEM;
+
+	kref_init(&cdev->kref);
+	init_rwsem(&cdev->lock);
+	cdev->ctrl = ctrl;
+
+	cdev->mdev.parent   = &pdev->dev;
+	cdev->mdev.minor    = MISC_DYNAMIC_MINOR;
+	cdev->mdev.name     = "surface_aggregator";
+	cdev->mdev.nodename = "surface/aggregator";
+	cdev->mdev.fops     = &ssam_controller_fops;
+
+	status = misc_register(&cdev->mdev);
+	if (status) {
+		kfree(cdev);
+		return status;
+	}
+
+	platform_set_drvdata(pdev, cdev);
+	return 0;
+}
+
+static int ssam_dbg_device_remove(struct platform_device *pdev)
+{
+	struct ssam_cdev *cdev = platform_get_drvdata(pdev);
+
+	misc_deregister(&cdev->mdev);
+
+	/*
+	 * The controller is only guaranteed to be valid for as long as the
+	 * driver is bound. Remove controller so that any lingering open files
+	 * cannot access it any more after we're gone.
+	 */
+	down_write(&cdev->lock);
+	cdev->ctrl = NULL;
+	up_write(&cdev->lock);
+
+	ssam_cdev_put(cdev);
+	return 0;
+}
+
+static struct platform_device *ssam_cdev_device;
+
+static struct platform_driver ssam_cdev_driver = {
+	.probe = ssam_dbg_device_probe,
+	.remove = ssam_dbg_device_remove,
+	.driver = {
+		.name = SSAM_CDEV_DEVICE_NAME,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+static int __init ssam_debug_init(void)
+{
+	int status;
+
+	ssam_cdev_device = platform_device_alloc(SSAM_CDEV_DEVICE_NAME,
+						 PLATFORM_DEVID_NONE);
+	if (!ssam_cdev_device)
+		return -ENOMEM;
+
+	status = platform_device_add(ssam_cdev_device);
+	if (status)
+		goto err_device;
+
+	status = platform_driver_register(&ssam_cdev_driver);
+	if (status)
+		goto err_driver;
+
+	return 0;
+
+err_driver:
+	platform_device_del(ssam_cdev_device);
+err_device:
+	platform_device_put(ssam_cdev_device);
+	return status;
+}
+module_init(ssam_debug_init);
+
+static void __exit ssam_debug_exit(void)
+{
+	platform_driver_unregister(&ssam_cdev_driver);
+	platform_device_unregister(ssam_cdev_device);
+}
+module_exit(ssam_debug_exit);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("User-space interface for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/clients/surface_aggregator_registry.c b/drivers/misc/surface_aggregator/clients/surface_aggregator_registry.c
new file mode 100644
index 000000000000..77ef266b88f3
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_aggregator_registry.c
@@ -0,0 +1,656 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface System Aggregator Module (SSAM) client device registry.
+ *
+ * Registry for non-platform/non-ACPI SSAM client devices, i.e. devices that
+ * cannot be auto-detected. Provides device-hubs for these devices.
+ *
+ * Copyright (C) 2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/acpi.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/notifier.h>
+#include <linux/platform_device.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/device.h>
+
+
+/* -- Device registry. ------------------------------------------------------ */
+
+static const struct software_node ssam_node_root = {
+	.name = "ssam_platform_hub",
+};
+
+static const struct software_node ssam_node_hub_main = {
+	.name = "ssam:00:00:01:00:00",
+	.parent = &ssam_node_root,
+};
+
+static const struct software_node ssam_node_hub_base = {
+	.name = "ssam:00:00:02:00:00",
+	.parent = &ssam_node_root,
+};
+
+static const struct software_node ssam_node_bat_ac = {
+	.name = "ssam:01:02:01:01:01",
+	.parent = &ssam_node_hub_main,
+};
+
+static const struct software_node ssam_node_bat_main = {
+	.name = "ssam:01:02:01:01:00",
+	.parent = &ssam_node_hub_main,
+};
+
+static const struct software_node ssam_node_bat_sb3base = {
+	.name = "ssam:01:02:02:01:00",
+	.parent = &ssam_node_hub_base,
+};
+
+static const struct software_node ssam_node_tmp_perf = {
+	.name = "ssam:01:03:01:00:01",
+	.parent = &ssam_node_hub_main,
+};
+
+static const struct software_node ssam_node_bas_dtx = {
+	.name = "ssam:01:11:01:00:00",
+	.parent = &ssam_node_hub_main,
+};
+
+static const struct software_node ssam_node_hid_main_keyboard = {
+	.name = "ssam:01:15:02:01:00",
+	.parent = &ssam_node_hub_main,
+};
+
+static const struct software_node ssam_node_hid_main_touchpad = {
+	.name = "ssam:01:15:02:03:00",
+	.parent = &ssam_node_hub_main,
+};
+
+static const struct software_node ssam_node_hid_main_iid5 = {
+	.name = "ssam:01:15:02:05:00",
+	.parent = &ssam_node_hub_main,
+};
+
+static const struct software_node ssam_node_hid_base_keyboard = {
+	.name = "ssam:01:15:02:01:00",
+	.parent = &ssam_node_hub_base,
+};
+
+static const struct software_node ssam_node_hid_base_touchpad = {
+	.name = "ssam:01:15:02:03:00",
+	.parent = &ssam_node_hub_base,
+};
+
+static const struct software_node ssam_node_hid_base_iid5 = {
+	.name = "ssam:01:15:02:05:00",
+	.parent = &ssam_node_hub_base,
+};
+
+static const struct software_node ssam_node_hid_base_iid6 = {
+	.name = "ssam:01:15:02:06:00",
+	.parent = &ssam_node_hub_base,
+};
+
+
+static const struct software_node *ssam_node_group_sb2[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_tmp_perf,
+	NULL,
+};
+
+static const struct software_node *ssam_node_group_sb3[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_hub_base,
+	&ssam_node_tmp_perf,
+	&ssam_node_bat_ac,
+	&ssam_node_bat_main,
+	&ssam_node_bat_sb3base,
+	&ssam_node_hid_base_keyboard,
+	&ssam_node_hid_base_touchpad,
+	&ssam_node_hid_base_iid5,
+	&ssam_node_hid_base_iid6,
+	&ssam_node_bas_dtx,
+	NULL,
+};
+
+static const struct software_node *ssam_node_group_sl1[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_tmp_perf,
+	NULL,
+};
+
+static const struct software_node *ssam_node_group_sl2[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_tmp_perf,
+	NULL,
+};
+
+static const struct software_node *ssam_node_group_sl3[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_tmp_perf,
+	&ssam_node_bat_ac,
+	&ssam_node_bat_main,
+	&ssam_node_hid_main_keyboard,
+	&ssam_node_hid_main_touchpad,
+	&ssam_node_hid_main_iid5,
+	NULL,
+};
+
+static const struct software_node *ssam_node_group_sp5[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_tmp_perf,
+	NULL,
+};
+
+static const struct software_node *ssam_node_group_sp6[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_tmp_perf,
+	NULL,
+};
+
+static const struct software_node *ssam_node_group_sp7[] = {
+	&ssam_node_root,
+	&ssam_node_hub_main,
+	&ssam_node_tmp_perf,
+	&ssam_node_bat_ac,
+	&ssam_node_bat_main,
+	NULL,
+};
+
+
+/* -- Device registry helper functions. ------------------------------------- */
+
+static int ssam_uid_from_string(const char *str, struct ssam_device_uid *uid)
+{
+	u8 d, tc, tid, iid, fn;
+	int n;
+
+	n = sscanf(str, "ssam:%hhx:%hhx:%hhx:%hhx:%hhx", &d, &tc, &tid, &iid, &fn);
+	if (n != 5)
+		return -EINVAL;
+
+	uid->domain = d;
+	uid->category = tc;
+	uid->target = tid;
+	uid->instance = iid;
+	uid->function = fn;
+
+	return 0;
+}
+
+static int ssam_hub_remove_devices_fn(struct device *dev, void *data)
+{
+	if (!is_ssam_device(dev))
+		return 0;
+
+	ssam_device_remove(to_ssam_device(dev));
+	return 0;
+}
+
+static void ssam_hub_remove_devices(struct device *parent)
+{
+	device_for_each_child_reverse(parent, NULL, ssam_hub_remove_devices_fn);
+}
+
+static int ssam_hub_add_device(struct device *parent,
+			       struct ssam_controller *ctrl,
+			       struct fwnode_handle *node)
+{
+	struct ssam_device_uid uid;
+	struct ssam_device *sdev;
+	int status;
+
+	status = ssam_uid_from_string(fwnode_get_name(node), &uid);
+	if (status)
+		return -ENODEV;
+
+	sdev = ssam_device_alloc(ctrl, uid);
+	if (!sdev)
+		return -ENOMEM;
+
+	sdev->dev.parent = parent;
+	sdev->dev.fwnode = node;
+
+	status = ssam_device_add(sdev);
+	if (status)
+		ssam_device_put(sdev);
+
+	return status;
+}
+
+static int ssam_hub_add_devices(struct device *parent,
+				struct ssam_controller *ctrl,
+				struct fwnode_handle *node)
+{
+	struct fwnode_handle *child;
+	int status;
+
+	fwnode_for_each_child_node(node, child) {
+		status = ssam_hub_add_device(parent, ctrl, child);
+		if (status && status != -ENODEV)
+			goto err;
+	}
+
+	return 0;
+err:
+	ssam_hub_remove_devices(parent);
+	return status;
+}
+
+
+/* -- SSAM main-hub driver. ------------------------------------------------- */
+
+static int ssam_hub_probe(struct ssam_device *sdev)
+{
+	struct fwnode_handle *node = dev_fwnode(&sdev->dev);
+
+	if (!node)
+		return -ENODEV;
+
+	return ssam_hub_add_devices(&sdev->dev, sdev->ctrl, node);
+}
+
+static void ssam_hub_remove(struct ssam_device *sdev)
+{
+	ssam_hub_remove_devices(&sdev->dev);
+}
+
+static const struct ssam_device_id ssam_hub_match[] = {
+	{ SSAM_VDEV(HUB, 0x01, 0x00, 0x00) },
+	{ },
+};
+
+static struct ssam_device_driver ssam_hub_driver = {
+	.probe = ssam_hub_probe,
+	.remove = ssam_hub_remove,
+	.match_table = ssam_hub_match,
+	.driver = {
+		.name = "surface_aggregator_device_hub",
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+
+/* -- SSAM base-hub driver. ------------------------------------------------- */
+
+enum ssam_base_hub_state {
+	SSAM_BASE_HUB_UNINITIALIZED,
+	SSAM_BASE_HUB_CONNECTED,
+	SSAM_BASE_HUB_DISCONNECTED,
+};
+
+struct ssam_base_hub {
+	struct ssam_device *sdev;
+
+	struct mutex lock;
+	enum ssam_base_hub_state state;
+
+	struct ssam_event_notifier notif;
+};
+
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_bas_query_opmode, u8, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x0d,
+	.instance_id     = 0x00,
+});
+
+#define SSAM_BAS_OPMODE_TABLET		0x00
+#define SSAM_EVENT_BAS_CID_CONNECTION	0x0c
+
+static int ssam_base_hub_query_state(struct ssam_device *sdev,
+				     enum ssam_base_hub_state *state)
+{
+	u8 opmode;
+	int status;
+
+	status = ssam_retry(ssam_bas_query_opmode, sdev->ctrl, &opmode);
+	if (status < 0) {
+		dev_err(&sdev->dev, "failed to query base state: %d\n", status);
+		return status;
+	}
+
+	if (opmode != SSAM_BAS_OPMODE_TABLET)
+		*state = SSAM_BASE_HUB_CONNECTED;
+	else
+		*state = SSAM_BASE_HUB_DISCONNECTED;
+
+	return 0;
+}
+
+
+static ssize_t ssam_base_hub_state_show(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+	struct ssam_base_hub *hub = ssam_device_get_drvdata(sdev);
+	bool connected;
+
+	mutex_lock(&hub->lock);
+	connected = hub->state == SSAM_BASE_HUB_CONNECTED;
+	mutex_unlock(&hub->lock);
+
+	return snprintf(buf, PAGE_SIZE - 1, "%d\n", connected);
+}
+
+static struct device_attribute ssam_base_hub_attr_state =
+	__ATTR(state, 0444, ssam_base_hub_state_show, NULL);
+
+static struct attribute *ssam_base_hub_attrs[] = {
+	&ssam_base_hub_attr_state.attr,
+	NULL,
+};
+
+const struct attribute_group ssam_base_hub_group = {
+	.attrs = ssam_base_hub_attrs,
+};
+
+
+static int ssam_base_hub_update(struct ssam_device *sdev,
+				enum ssam_base_hub_state new)
+{
+	struct ssam_base_hub *hub = ssam_device_get_drvdata(sdev);
+	struct fwnode_handle *node = dev_fwnode(&sdev->dev);
+	int status = 0;
+
+	mutex_lock(&hub->lock);
+	if (hub->state == new) {
+		mutex_unlock(&hub->lock);
+		return 0;
+	}
+	hub->state = new;
+
+	if (hub->state == SSAM_BASE_HUB_CONNECTED)
+		status = ssam_hub_add_devices(&sdev->dev, sdev->ctrl, node);
+
+	if (hub->state != SSAM_BASE_HUB_CONNECTED || status)
+		ssam_hub_remove_devices(&sdev->dev);
+
+	mutex_unlock(&hub->lock);
+
+	if (status) {
+		dev_err(&sdev->dev, "failed to update base-hub devices: %d\n",
+			status);
+	}
+
+	return status;
+}
+
+static u32 ssam_base_hub_notif(struct ssam_event_notifier *nf,
+			       const struct ssam_event *event)
+{
+	struct ssam_base_hub *hub;
+	struct ssam_device *sdev;
+	enum ssam_base_hub_state new;
+
+	hub = container_of(nf, struct ssam_base_hub, notif);
+	sdev = hub->sdev;
+
+	if (event->command_id != SSAM_EVENT_BAS_CID_CONNECTION)
+		return 0;
+
+	if (event->length < 1) {
+		dev_err(&sdev->dev, "unexpected payload size: %u\n",
+			event->length);
+		return 0;
+	}
+
+	if (event->data[0])
+		new = SSAM_BASE_HUB_CONNECTED;
+	else
+		new = SSAM_BASE_HUB_DISCONNECTED;
+
+	ssam_base_hub_update(sdev, new);
+
+	/*
+	 * Do not return SSAM_NOTIF_HANDLED: The event should be picked up and
+	 * consumed by the detachment system driver. We're just a (more or less)
+	 * silent observer.
+	 */
+	return 0;
+}
+
+static int __maybe_unused ssam_base_hub_resume(struct device *dev)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+	enum ssam_base_hub_state state;
+	int status;
+
+	status = ssam_base_hub_query_state(sdev, &state);
+	if (status)
+		return status;
+
+	return ssam_base_hub_update(sdev, state);
+}
+static SIMPLE_DEV_PM_OPS(ssam_base_hub_pm_ops, NULL, ssam_base_hub_resume);
+
+static int ssam_base_hub_probe(struct ssam_device *sdev)
+{
+	enum ssam_base_hub_state state;
+	struct ssam_base_hub *hub;
+	int status;
+
+	hub = devm_kzalloc(&sdev->dev, sizeof(*hub), GFP_KERNEL);
+	if (!hub)
+		return -ENOMEM;
+
+	mutex_init(&hub->lock);
+
+	hub->sdev = sdev;
+	hub->state = SSAM_BASE_HUB_UNINITIALIZED;
+
+	hub->notif.base.priority = 1000;  // this notifier should run first
+	hub->notif.base.fn = ssam_base_hub_notif;
+	hub->notif.event.reg = SSAM_EVENT_REGISTRY_SAM;
+	hub->notif.event.id.target_category = SSAM_SSH_TC_BAS,
+	hub->notif.event.id.instance = 0,
+	hub->notif.event.mask = SSAM_EVENT_MASK_NONE;
+	hub->notif.event.flags = SSAM_EVENT_SEQUENCED;
+
+	status = ssam_notifier_register(sdev->ctrl, &hub->notif);
+	if (status)
+		return status;
+
+	ssam_device_set_drvdata(sdev, hub);
+
+	status = ssam_base_hub_query_state(sdev, &state);
+	if (status) {
+		ssam_notifier_unregister(sdev->ctrl, &hub->notif);
+		return status;
+	}
+
+	status = ssam_base_hub_update(sdev, state);
+	if (status) {
+		ssam_notifier_unregister(sdev->ctrl, &hub->notif);
+		return status;
+	}
+
+	status = sysfs_create_group(&sdev->dev.kobj, &ssam_base_hub_group);
+	if (status) {
+		ssam_notifier_unregister(sdev->ctrl, &hub->notif);
+		ssam_hub_remove_devices(&sdev->dev);
+	}
+
+	return status;
+}
+
+static void ssam_base_hub_remove(struct ssam_device *sdev)
+{
+	struct ssam_base_hub *hub = ssam_device_get_drvdata(sdev);
+
+	sysfs_remove_group(&sdev->dev.kobj, &ssam_base_hub_group);
+
+	ssam_notifier_unregister(sdev->ctrl, &hub->notif);
+	ssam_hub_remove_devices(&sdev->dev);
+
+	mutex_destroy(&hub->lock);
+}
+
+static const struct ssam_device_id ssam_base_hub_match[] = {
+	{ SSAM_VDEV(HUB, 0x02, 0x00, 0x00) },
+	{ },
+};
+
+static struct ssam_device_driver ssam_base_hub_driver = {
+	.probe = ssam_base_hub_probe,
+	.remove = ssam_base_hub_remove,
+	.match_table = ssam_base_hub_match,
+	.driver = {
+		.name = "surface_aggregator_base_hub",
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+		.pm = &ssam_base_hub_pm_ops,
+	},
+};
+
+
+/* -- SSAM platform/meta-hub driver. ---------------------------------------- */
+
+static const struct acpi_device_id ssam_platform_hub_match[] = {
+	/* Surface Pro 4, 5, and 6 */
+	{ "MSHW0081", (unsigned long)ssam_node_group_sp5 },
+
+	/* Surface Pro 6 (OMBR >= 0x10) */
+	{ "MSHW0111", (unsigned long)ssam_node_group_sp6 },
+
+	/* Surface Pro 7 */
+	{ "MSHW0116", (unsigned long)ssam_node_group_sp7 },
+
+	/* Surface Book 2 */
+	{ "MSHW0107", (unsigned long)ssam_node_group_sb2 },
+
+	/* Surface Book 3 */
+	{ "MSHW0117", (unsigned long)ssam_node_group_sb3 },
+
+	/* Surface Laptop 1 */
+	{ "MSHW0086", (unsigned long)ssam_node_group_sl1 },
+
+	/* Surface Laptop 2 */
+	{ "MSHW0112", (unsigned long)ssam_node_group_sl2 },
+
+	/* Surface Laptop 3 (13", Intel) */
+	{ "MSHW0114", (unsigned long)ssam_node_group_sl3 },
+
+	/* Surface Laptop 3 (15", AMD) */
+	{ "MSHW0110", (unsigned long)ssam_node_group_sl3 },
+
+	{ },
+};
+MODULE_DEVICE_TABLE(acpi, ssam_platform_hub_match);
+
+static int ssam_platform_hub_probe(struct platform_device *pdev)
+{
+	const struct software_node **nodes;
+	struct ssam_controller *ctrl;
+	struct fwnode_handle *root;
+	int status;
+
+	nodes = (const struct software_node **)acpi_device_get_match_data(&pdev->dev);
+	if (!nodes)
+		return -ENODEV;
+
+	/*
+	 * As we're adding the SSAM client devices as children under this device
+	 * and not the SSAM controller, we need to add a device link to the
+	 * controller to ensure that we remove all of our devices before the
+	 * controller is removed. This also guarantees proper ordering for
+	 * suspend/resume of the devices on this hub.
+	 */
+	status = ssam_client_bind(&pdev->dev, &ctrl);
+	if (status)
+		return status == -ENXIO ? -EPROBE_DEFER : status;
+
+	status = software_node_register_node_group(nodes);
+	if (status)
+		return status;
+
+	root = software_node_fwnode(&ssam_node_root);
+	if (!root)
+		return -ENOENT;
+
+	set_secondary_fwnode(&pdev->dev, root);
+
+	status = ssam_hub_add_devices(&pdev->dev, ctrl, root);
+	if (status) {
+		software_node_unregister_node_group(nodes);
+		return status;
+	}
+
+	platform_set_drvdata(pdev, nodes);
+	return 0;
+}
+
+static int ssam_platform_hub_remove(struct platform_device *pdev)
+{
+	const struct software_node **nodes = platform_get_drvdata(pdev);
+
+	ssam_hub_remove_devices(&pdev->dev);
+	set_secondary_fwnode(&pdev->dev, NULL);
+	software_node_unregister_node_group(nodes);
+	return 0;
+}
+
+static struct platform_driver ssam_platform_hub_driver = {
+	.probe = ssam_platform_hub_probe,
+	.remove = ssam_platform_hub_remove,
+	.driver = {
+		.name = "surface_aggregator_platform_hub",
+		.acpi_match_table = ssam_platform_hub_match,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+
+/* -- Module initialization. ------------------------------------------------ */
+
+static int __init ssam_device_hub_init(void)
+{
+	int status;
+
+	status = platform_driver_register(&ssam_platform_hub_driver);
+	if (status)
+		goto err_platform;
+
+	status = ssam_device_driver_register(&ssam_hub_driver);
+	if (status)
+		goto err_main;
+
+	status = ssam_device_driver_register(&ssam_base_hub_driver);
+	if (status)
+		goto err_base;
+
+	return 0;
+
+err_base:
+	ssam_device_driver_unregister(&ssam_hub_driver);
+err_main:
+	platform_driver_unregister(&ssam_platform_hub_driver);
+err_platform:
+	return status;
+}
+module_init(ssam_device_hub_init);
+
+static void __exit ssam_device_hub_exit(void)
+{
+	ssam_device_driver_unregister(&ssam_base_hub_driver);
+	ssam_device_driver_unregister(&ssam_hub_driver);
+	platform_driver_unregister(&ssam_platform_hub_driver);
+}
+module_exit(ssam_device_hub_exit);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Device-registry for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/clients/surface_battery.c b/drivers/misc/surface_aggregator/clients/surface_battery.c
new file mode 100644
index 000000000000..7ed9ba3f98d7
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_battery.c
@@ -0,0 +1,1192 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface battery and AC device driver.
+ *
+ * Provides support for battery and AC devices connected via the Surface
+ * System Aggregator Module.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <asm/unaligned.h>
+#include <linux/delay.h>
+#include <linux/jiffies.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/power_supply.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+
+#include <linux/surface_aggregator/device.h>
+
+#define SPWR_AC_BAT_UPDATE_DELAY	msecs_to_jiffies(5000)
+
+
+/* -- Module parameters. ---------------------------------------------------- */
+
+static unsigned int cache_time = 1000;
+module_param(cache_time, uint, 0644);
+MODULE_PARM_DESC(cache_time, "battery state chaching time in milliseconds [default: 1000]");
+
+
+/* -- SAM Interface. -------------------------------------------------------- */
+
+enum sam_event_cid_bat {
+	SAM_EVENT_CID_BAT_BIX         = 0x15,
+	SAM_EVENT_CID_BAT_BST         = 0x16,
+	SAM_EVENT_CID_BAT_ADP         = 0x17,
+	SAM_EVENT_CID_BAT_PROT        = 0x18,
+	SAM_EVENT_CID_BAT_DPTF        = 0x53,
+};
+
+enum sam_battery_sta {
+	SAM_BATTERY_STA_OK            = 0x0f,
+	SAM_BATTERY_STA_PRESENT	      = 0x10,
+};
+
+enum sam_battery_state {
+	SAM_BATTERY_STATE_DISCHARGING = BIT(0),
+	SAM_BATTERY_STATE_CHARGING    = BIT(1),
+	SAM_BATTERY_STATE_CRITICAL    = BIT(2),
+};
+
+enum sam_battery_power_unit {
+	SAM_BATTERY_POWER_UNIT_mW     = 0,
+	SAM_BATTERY_POWER_UNIT_mA     = 1,
+};
+
+/* Equivalent to data returned in ACPI _BIX method, revision 0 */
+struct spwr_bix {
+	u8  revision;
+	__le32 power_unit;
+	__le32 design_cap;
+	__le32 last_full_charge_cap;
+	__le32 technology;
+	__le32 design_voltage;
+	__le32 design_cap_warn;
+	__le32 design_cap_low;
+	__le32 cycle_count;
+	__le32 measurement_accuracy;
+	__le32 max_sampling_time;
+	__le32 min_sampling_time;
+	__le32 max_avg_interval;
+	__le32 min_avg_interval;
+	__le32 bat_cap_granularity_1;
+	__le32 bat_cap_granularity_2;
+	__u8 model[21];
+	__u8 serial[11];
+	__u8 type[5];
+	__u8 oem_info[21];
+} __packed;
+
+static_assert(sizeof(struct spwr_bix) == 119);
+
+#define SPWR_BIX_REVISION		0
+
+/* Equivalent to data returned in ACPI _BST method */
+struct spwr_bst {
+	__le32 state;
+	__le32 present_rate;
+	__le32 remaining_cap;
+	__le32 present_voltage;
+} __packed;
+
+static_assert(sizeof(struct spwr_bst) == 16);
+
+#define SPWR_BATTERY_VALUE_UNKNOWN	0xffffffff
+
+/* Get battery status (_STA) */
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_sta, __le32, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x01,
+});
+
+/* Get battery static information (_BIX) */
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_bix, struct spwr_bix, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x02,
+});
+
+/* Get battery dynamic information (_BST) */
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_bst, struct spwr_bst, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x03,
+});
+
+/* Set battery trip point (_BTP) */
+static SSAM_DEFINE_SYNC_REQUEST_CL_W(ssam_bat_set_btp, __le32, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x04,
+});
+
+/* Get platform power soruce for battery (DPTF PSRC) */
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_psrc, __le32, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x0d,
+});
+
+/*
+ * The following requests are currently unused. They are nevertheless included
+ * for documentation of the SAM interface.
+ */
+
+/* Get maximum platform power for battery (DPTF PMAX) */
+__always_unused
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_pmax, __le32, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x0b,
+});
+
+/* Get adapter rating (DPTF ARTG) */
+__always_unused
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_artg, __le32, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x0f,
+});
+
+/* Unknown (DPTF PSOC) */
+__always_unused
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_bat_get_psoc, __le32, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x0c,
+});
+
+/* Unknown (DPTF CHGI/ INT3403 SPPC) */
+__always_unused
+static SSAM_DEFINE_SYNC_REQUEST_CL_W(ssam_bat_set_chgi, __le32, {
+	.target_category = SSAM_SSH_TC_BAT,
+	.command_id      = 0x0e,
+});
+
+
+/* -- Common power-subsystem interface. ------------------------------------- */
+
+struct spwr_psy_properties {
+	const char *name;
+	struct ssam_event_registry registry;
+};
+
+struct spwr_battery_device {
+	struct ssam_device *sdev;
+
+	char name[32];
+	struct power_supply *psy;
+	struct power_supply_desc psy_desc;
+
+	struct delayed_work update_work;
+
+	struct ssam_event_notifier notif;
+
+	struct mutex lock;
+	unsigned long timestamp;
+
+	__le32 sta;
+	struct spwr_bix bix;
+	struct spwr_bst bst;
+	u32 alarm;
+};
+
+struct spwr_ac_device {
+	struct ssam_device *sdev;
+
+	char name[32];
+	struct power_supply *psy;
+	struct power_supply_desc psy_desc;
+
+	struct ssam_event_notifier notif;
+
+	struct mutex lock;
+
+	__le32 state;
+};
+
+static enum power_supply_property spwr_ac_props[] = {
+	POWER_SUPPLY_PROP_ONLINE,
+};
+
+static enum power_supply_property spwr_battery_props_chg[] = {
+	POWER_SUPPLY_PROP_STATUS,
+	POWER_SUPPLY_PROP_PRESENT,
+	POWER_SUPPLY_PROP_TECHNOLOGY,
+	POWER_SUPPLY_PROP_CYCLE_COUNT,
+	POWER_SUPPLY_PROP_VOLTAGE_MIN_DESIGN,
+	POWER_SUPPLY_PROP_VOLTAGE_NOW,
+	POWER_SUPPLY_PROP_CURRENT_NOW,
+	POWER_SUPPLY_PROP_CHARGE_FULL_DESIGN,
+	POWER_SUPPLY_PROP_CHARGE_FULL,
+	POWER_SUPPLY_PROP_CHARGE_NOW,
+	POWER_SUPPLY_PROP_CAPACITY,
+	POWER_SUPPLY_PROP_CAPACITY_LEVEL,
+	POWER_SUPPLY_PROP_MODEL_NAME,
+	POWER_SUPPLY_PROP_MANUFACTURER,
+	POWER_SUPPLY_PROP_SERIAL_NUMBER,
+};
+
+static enum power_supply_property spwr_battery_props_eng[] = {
+	POWER_SUPPLY_PROP_STATUS,
+	POWER_SUPPLY_PROP_PRESENT,
+	POWER_SUPPLY_PROP_TECHNOLOGY,
+	POWER_SUPPLY_PROP_CYCLE_COUNT,
+	POWER_SUPPLY_PROP_VOLTAGE_MIN_DESIGN,
+	POWER_SUPPLY_PROP_VOLTAGE_NOW,
+	POWER_SUPPLY_PROP_POWER_NOW,
+	POWER_SUPPLY_PROP_ENERGY_FULL_DESIGN,
+	POWER_SUPPLY_PROP_ENERGY_FULL,
+	POWER_SUPPLY_PROP_ENERGY_NOW,
+	POWER_SUPPLY_PROP_CAPACITY,
+	POWER_SUPPLY_PROP_CAPACITY_LEVEL,
+	POWER_SUPPLY_PROP_MODEL_NAME,
+	POWER_SUPPLY_PROP_MANUFACTURER,
+	POWER_SUPPLY_PROP_SERIAL_NUMBER,
+};
+
+static bool spwr_battery_present(struct spwr_battery_device *bat)
+{
+	return le32_to_cpu(bat->sta) & SAM_BATTERY_STA_PRESENT;
+}
+
+static int spwr_battery_load_sta(struct spwr_battery_device *bat)
+{
+	return ssam_retry(ssam_bat_get_sta, bat->sdev, &bat->sta);
+}
+
+static int spwr_battery_load_bix(struct spwr_battery_device *bat)
+{
+	int status;
+
+	if (!spwr_battery_present(bat))
+		return 0;
+
+	status = ssam_retry(ssam_bat_get_bix, bat->sdev, &bat->bix);
+
+	// enforce NULL terminated strings in case anything goes wrong...
+	bat->bix.model[ARRAY_SIZE(bat->bix.model) - 1] = 0;
+	bat->bix.serial[ARRAY_SIZE(bat->bix.serial) - 1] = 0;
+	bat->bix.type[ARRAY_SIZE(bat->bix.type) - 1] = 0;
+	bat->bix.oem_info[ARRAY_SIZE(bat->bix.oem_info) - 1] = 0;
+
+	return status;
+}
+
+static int spwr_battery_load_bst(struct spwr_battery_device *bat)
+{
+	if (!spwr_battery_present(bat))
+		return 0;
+
+	return ssam_retry(ssam_bat_get_bst, bat->sdev, &bat->bst);
+}
+
+static int spwr_battery_set_alarm_unlocked(struct spwr_battery_device *bat,
+					   u32 value)
+{
+	__le32 value_le = cpu_to_le32(value);
+
+	bat->alarm = value;
+	return ssam_retry(ssam_bat_set_btp, bat->sdev, &value_le);
+}
+
+static int spwr_battery_set_alarm(struct spwr_battery_device *bat, u32 value)
+{
+	int status;
+
+	mutex_lock(&bat->lock);
+	status = spwr_battery_set_alarm_unlocked(bat, value);
+	mutex_unlock(&bat->lock);
+
+	return status;
+}
+
+static int spwr_battery_update_bst_unlocked(struct spwr_battery_device *bat,
+					    bool cached)
+{
+	unsigned long cache_deadline;
+	int status;
+
+	cache_deadline = bat->timestamp + msecs_to_jiffies(cache_time);
+	if (cached && bat->timestamp && time_is_after_jiffies(cache_deadline))
+		return 0;
+
+	status = spwr_battery_load_sta(bat);
+	if (status)
+		return status;
+
+	status = spwr_battery_load_bst(bat);
+	if (status)
+		return status;
+
+	bat->timestamp = jiffies;
+	return 0;
+}
+
+static int spwr_battery_update_bst(struct spwr_battery_device *bat, bool cached)
+{
+	int status;
+
+	mutex_lock(&bat->lock);
+	status = spwr_battery_update_bst_unlocked(bat, cached);
+	mutex_unlock(&bat->lock);
+
+	return status;
+}
+
+static int spwr_battery_update_bix_unlocked(struct spwr_battery_device *bat)
+{
+	int status;
+
+	status = spwr_battery_load_sta(bat);
+	if (status)
+		return status;
+
+	status = spwr_battery_load_bix(bat);
+	if (status)
+		return status;
+
+	status = spwr_battery_load_bst(bat);
+	if (status)
+		return status;
+
+	if (bat->bix.revision != SPWR_BIX_REVISION) {
+		dev_warn(&bat->sdev->dev, "unsupported battery revision: %u\n",
+			 bat->bix.revision);
+	}
+
+	bat->timestamp = jiffies;
+	return 0;
+}
+
+static int spwr_ac_update_unlocked(struct spwr_ac_device *ac)
+{
+	int status;
+	u32 old = ac->state;
+
+	status = ssam_retry(ssam_bat_get_psrc, ac->sdev, &ac->state);
+	if (status < 0)
+		return status;
+
+	return old != ac->state;
+}
+
+static int spwr_ac_update(struct spwr_ac_device *ac)
+{
+	int status;
+
+	mutex_lock(&ac->lock);
+	status = spwr_ac_update_unlocked(ac);
+	mutex_unlock(&ac->lock);
+
+	return status;
+}
+
+static u32 sprw_battery_get_full_cap_safe(struct spwr_battery_device *bat)
+{
+	u32 full_cap = get_unaligned_le32(&bat->bix.last_full_charge_cap);
+
+	if (full_cap == 0 || full_cap == SPWR_BATTERY_VALUE_UNKNOWN)
+		full_cap = get_unaligned_le32(&bat->bix.design_cap);
+
+	return full_cap;
+}
+
+static bool spwr_battery_is_full(struct spwr_battery_device *bat)
+{
+	u32 state = get_unaligned_le32(&bat->bst.state);
+	u32 full_cap = sprw_battery_get_full_cap_safe(bat);
+	u32 remaining_cap = get_unaligned_le32(&bat->bst.remaining_cap);
+
+	return full_cap != SPWR_BATTERY_VALUE_UNKNOWN && full_cap != 0
+		&& remaining_cap != SPWR_BATTERY_VALUE_UNKNOWN
+		&& remaining_cap >= full_cap
+		&& state == 0;
+}
+
+static int spwr_battery_recheck_full(struct spwr_battery_device *bat)
+{
+	bool present;
+	u32 unit;
+	int status;
+
+	mutex_lock(&bat->lock);
+	unit = get_unaligned_le32(&bat->bix.power_unit);
+	present = spwr_battery_present(bat);
+
+	status = spwr_battery_update_bix_unlocked(bat);
+	if (status)
+		goto out;
+
+	// if battery has been attached, (re-)initialize alarm
+	if (!present && spwr_battery_present(bat)) {
+		u32 cap_warn = get_unaligned_le32(&bat->bix.design_cap_warn);
+
+		status = spwr_battery_set_alarm_unlocked(bat, cap_warn);
+		if (status)
+			goto out;
+	}
+
+	/*
+	 * Warn if the unit has changed. This is something we genuinely don't
+	 * expect to happen, so make this a big warning. If it does, we'll
+	 * need to add support for it.
+	 */
+	WARN_ON(unit != get_unaligned_le32(&bat->bix.power_unit));
+
+out:
+	mutex_unlock(&bat->lock);
+
+	if (!status)
+		power_supply_changed(bat->psy);
+
+	return status;
+}
+
+static int spwr_battery_recheck_status(struct spwr_battery_device *bat)
+{
+	int status;
+
+	status = spwr_battery_update_bst(bat, false);
+	if (!status)
+		power_supply_changed(bat->psy);
+
+	return status;
+}
+
+static int spwr_battery_recheck_adapter(struct spwr_battery_device *bat)
+{
+	u32 full_cap = sprw_battery_get_full_cap_safe(bat);
+	u32 remaining_cap = get_unaligned_le32(&bat->bst.remaining_cap);
+
+	if (full_cap == 0 || full_cap == SPWR_BATTERY_VALUE_UNKNOWN)
+		return 0;
+
+	if (remaining_cap == SPWR_BATTERY_VALUE_UNKNOWN)
+		return 0;
+
+	/*
+	 * Handle battery update quirk:
+	 * When the battery is fully charged and the adapter is plugged in or
+	 * removed, the EC does not send a separate event for the state
+	 * (charging/discharging) change. Furthermore it may take some time until
+	 * the state is updated on the battery. Schedule an update to solve this.
+	 */
+
+	if (remaining_cap >= full_cap)
+		schedule_delayed_work(&bat->update_work, SPWR_AC_BAT_UPDATE_DELAY);
+
+	return 0;
+}
+
+static int spwr_ac_recheck(struct spwr_ac_device *ac)
+{
+	int status;
+
+	status = spwr_ac_update(ac);
+	if (status > 0)
+		power_supply_changed(ac->psy);
+
+	return status >= 0 ? 0 : status;
+}
+
+static u32 spwr_notify_bat(struct ssam_event_notifier *nf,
+			   const struct ssam_event *event)
+{
+	struct spwr_battery_device *bat;
+	int status;
+
+	bat = container_of(nf, struct spwr_battery_device, notif);
+
+	dev_dbg(&bat->sdev->dev, "power event (cid = 0x%02x, iid = %d, tid = %d)\n",
+		event->command_id, event->instance_id, event->target_id);
+
+	// handled here, needs to be handled for all targets/instances
+	if (event->command_id == SAM_EVENT_CID_BAT_ADP) {
+		status = spwr_battery_recheck_adapter(bat);
+		return ssam_notifier_from_errno(status) | SSAM_NOTIF_HANDLED;
+	}
+
+	if (bat->sdev->uid.target != event->target_id)
+		return 0;
+
+	if (bat->sdev->uid.instance != event->instance_id)
+		return 0;
+
+	switch (event->command_id) {
+	case SAM_EVENT_CID_BAT_BIX:
+		status = spwr_battery_recheck_full(bat);
+		break;
+
+	case SAM_EVENT_CID_BAT_BST:
+		status = spwr_battery_recheck_status(bat);
+		break;
+
+	case SAM_EVENT_CID_BAT_PROT:
+		/*
+		 * TODO: Implement support for battery protection status change
+		 *       event.
+		 */
+		status = 0;
+		break;
+
+	case SAM_EVENT_CID_BAT_DPTF:
+		/*
+		 * TODO: Implement support for DPTF event.
+		 */
+		status = 0;
+		break;
+
+	default:
+		return 0;
+	}
+
+	return ssam_notifier_from_errno(status) | SSAM_NOTIF_HANDLED;
+}
+
+static u32 spwr_notify_ac(struct ssam_event_notifier *nf,
+			  const struct ssam_event *event)
+{
+	struct spwr_ac_device *ac;
+	int status;
+
+	ac = container_of(nf, struct spwr_ac_device, notif);
+
+	dev_dbg(&ac->sdev->dev, "power event (cid = 0x%02x, iid = %d, tid = %d)\n",
+		event->command_id, event->instance_id, event->target_id);
+
+	/*
+	 * Allow events of all targets/instances here. Global adapter status
+	 * seems to be handled via target=1 and instance=1, but events are
+	 * reported on all targets/instances in use.
+	 *
+	 * While it should be enough to just listen on 1/1, listen everywhere to
+	 * make sure we don't miss anything.
+	 */
+
+	switch (event->command_id) {
+	case SAM_EVENT_CID_BAT_ADP:
+		status = spwr_ac_recheck(ac);
+		return ssam_notifier_from_errno(status) | SSAM_NOTIF_HANDLED;
+
+	default:
+		return 0;
+	}
+}
+
+static void spwr_battery_update_bst_workfn(struct work_struct *work)
+{
+	struct delayed_work *dwork = to_delayed_work(work);
+	struct spwr_battery_device *bat;
+	int status;
+
+	bat = container_of(dwork, struct spwr_battery_device, update_work);
+
+	status = spwr_battery_update_bst(bat, false);
+	if (!status)
+		power_supply_changed(bat->psy);
+
+	if (status) {
+		dev_err(&bat->sdev->dev, "failed to update battery state: %d\n",
+			status);
+	}
+}
+
+static int spwr_battery_prop_status(struct spwr_battery_device *bat)
+{
+	u32 state = get_unaligned_le32(&bat->bst.state);
+	u32 present_rate = get_unaligned_le32(&bat->bst.present_rate);
+
+	if (state & SAM_BATTERY_STATE_DISCHARGING)
+		return POWER_SUPPLY_STATUS_DISCHARGING;
+
+	if (state & SAM_BATTERY_STATE_CHARGING)
+		return POWER_SUPPLY_STATUS_CHARGING;
+
+	if (spwr_battery_is_full(bat))
+		return POWER_SUPPLY_STATUS_FULL;
+
+	if (present_rate == 0)
+		return POWER_SUPPLY_STATUS_NOT_CHARGING;
+
+	return POWER_SUPPLY_STATUS_UNKNOWN;
+}
+
+static int spwr_battery_prop_technology(struct spwr_battery_device *bat)
+{
+	if (!strcasecmp("NiCd", bat->bix.type))
+		return POWER_SUPPLY_TECHNOLOGY_NiCd;
+
+	if (!strcasecmp("NiMH", bat->bix.type))
+		return POWER_SUPPLY_TECHNOLOGY_NiMH;
+
+	if (!strcasecmp("LION", bat->bix.type))
+		return POWER_SUPPLY_TECHNOLOGY_LION;
+
+	if (!strncasecmp("LI-ION", bat->bix.type, 6))
+		return POWER_SUPPLY_TECHNOLOGY_LION;
+
+	if (!strcasecmp("LiP", bat->bix.type))
+		return POWER_SUPPLY_TECHNOLOGY_LIPO;
+
+	return POWER_SUPPLY_TECHNOLOGY_UNKNOWN;
+}
+
+static int spwr_battery_prop_capacity(struct spwr_battery_device *bat)
+{
+	u32 full_cap = sprw_battery_get_full_cap_safe(bat);
+	u32 remaining_cap = get_unaligned_le32(&bat->bst.remaining_cap);
+
+	if (full_cap == 0 || full_cap == SPWR_BATTERY_VALUE_UNKNOWN)
+		return -ENODEV;
+
+	if (remaining_cap == SPWR_BATTERY_VALUE_UNKNOWN)
+		return -ENODEV;
+
+	return remaining_cap * 100 / full_cap;
+}
+
+static int spwr_battery_prop_capacity_level(struct spwr_battery_device *bat)
+{
+	u32 state = get_unaligned_le32(&bat->bst.state);
+	u32 remaining_cap = get_unaligned_le32(&bat->bst.remaining_cap);
+
+	if (state & SAM_BATTERY_STATE_CRITICAL)
+		return POWER_SUPPLY_CAPACITY_LEVEL_CRITICAL;
+
+	if (spwr_battery_is_full(bat))
+		return POWER_SUPPLY_CAPACITY_LEVEL_FULL;
+
+	if (remaining_cap <= bat->alarm)
+		return POWER_SUPPLY_CAPACITY_LEVEL_LOW;
+
+	return POWER_SUPPLY_CAPACITY_LEVEL_NORMAL;
+}
+
+static int spwr_ac_get_property(struct power_supply *psy,
+				enum power_supply_property psp,
+				union power_supply_propval *val)
+{
+	struct spwr_ac_device *ac = power_supply_get_drvdata(psy);
+	int status;
+
+	mutex_lock(&ac->lock);
+
+	status = spwr_ac_update_unlocked(ac);
+	if (status)
+		goto out;
+
+	switch (psp) {
+	case POWER_SUPPLY_PROP_ONLINE:
+		val->intval = !!le32_to_cpu(ac->state);
+		break;
+
+	default:
+		status = -EINVAL;
+		goto out;
+	}
+
+out:
+	mutex_unlock(&ac->lock);
+	return status;
+}
+
+static int spwr_battery_get_property(struct power_supply *psy,
+				     enum power_supply_property psp,
+				     union power_supply_propval *val)
+{
+	struct spwr_battery_device *bat = power_supply_get_drvdata(psy);
+	u32 value;
+	int status;
+
+	mutex_lock(&bat->lock);
+
+	status = spwr_battery_update_bst_unlocked(bat, true);
+	if (status)
+		goto out;
+
+	// abort if battery is not present
+	if (!spwr_battery_present(bat) && psp != POWER_SUPPLY_PROP_PRESENT) {
+		status = -ENODEV;
+		goto out;
+	}
+
+	switch (psp) {
+	case POWER_SUPPLY_PROP_STATUS:
+		val->intval = spwr_battery_prop_status(bat);
+		break;
+
+	case POWER_SUPPLY_PROP_PRESENT:
+		val->intval = spwr_battery_present(bat);
+		break;
+
+	case POWER_SUPPLY_PROP_TECHNOLOGY:
+		val->intval = spwr_battery_prop_technology(bat);
+		break;
+
+	case POWER_SUPPLY_PROP_CYCLE_COUNT:
+		value = get_unaligned_le32(&bat->bix.cycle_count);
+		if (value != SPWR_BATTERY_VALUE_UNKNOWN)
+			val->intval = value;
+		else
+			status = -ENODEV;
+		break;
+
+	case POWER_SUPPLY_PROP_VOLTAGE_MIN_DESIGN:
+		value = get_unaligned_le32(&bat->bix.design_voltage);
+		if (value != SPWR_BATTERY_VALUE_UNKNOWN)
+			val->intval = value * 1000;
+		else
+			status = -ENODEV;
+		break;
+
+	case POWER_SUPPLY_PROP_VOLTAGE_NOW:
+		value = get_unaligned_le32(&bat->bst.present_voltage);
+		if (value != SPWR_BATTERY_VALUE_UNKNOWN)
+			val->intval = value * 1000;
+		else
+			status = -ENODEV;
+		break;
+
+	case POWER_SUPPLY_PROP_CURRENT_NOW:
+	case POWER_SUPPLY_PROP_POWER_NOW:
+		value = get_unaligned_le32(&bat->bst.present_rate);
+		if (value != SPWR_BATTERY_VALUE_UNKNOWN)
+			val->intval = value * 1000;
+		else
+			status = -ENODEV;
+		break;
+
+	case POWER_SUPPLY_PROP_CHARGE_FULL_DESIGN:
+	case POWER_SUPPLY_PROP_ENERGY_FULL_DESIGN:
+		value = get_unaligned_le32(&bat->bix.design_cap);
+		if (value != SPWR_BATTERY_VALUE_UNKNOWN)
+			val->intval = value * 1000;
+		else
+			status = -ENODEV;
+		break;
+
+	case POWER_SUPPLY_PROP_CHARGE_FULL:
+	case POWER_SUPPLY_PROP_ENERGY_FULL:
+		value = get_unaligned_le32(&bat->bix.last_full_charge_cap);
+		if (value != SPWR_BATTERY_VALUE_UNKNOWN)
+			val->intval = value * 1000;
+		else
+			status = -ENODEV;
+		break;
+
+	case POWER_SUPPLY_PROP_CHARGE_NOW:
+	case POWER_SUPPLY_PROP_ENERGY_NOW:
+		value = get_unaligned_le32(&bat->bst.remaining_cap);
+		if (value != SPWR_BATTERY_VALUE_UNKNOWN)
+			val->intval = value * 1000;
+		else
+			status = -ENODEV;
+		break;
+
+	case POWER_SUPPLY_PROP_CAPACITY:
+		val->intval = spwr_battery_prop_capacity(bat);
+		break;
+
+	case POWER_SUPPLY_PROP_CAPACITY_LEVEL:
+		val->intval = spwr_battery_prop_capacity_level(bat);
+		break;
+
+	case POWER_SUPPLY_PROP_MODEL_NAME:
+		val->strval = bat->bix.model;
+		break;
+
+	case POWER_SUPPLY_PROP_MANUFACTURER:
+		val->strval = bat->bix.oem_info;
+		break;
+
+	case POWER_SUPPLY_PROP_SERIAL_NUMBER:
+		val->strval = bat->bix.serial;
+		break;
+
+	default:
+		status = -EINVAL;
+		break;
+	}
+
+out:
+	mutex_unlock(&bat->lock);
+	return status;
+}
+
+
+static ssize_t spwr_battery_alarm_show(struct device *dev,
+				       struct device_attribute *attr,
+				       char *buf)
+{
+	struct power_supply *psy = dev_get_drvdata(dev);
+	struct spwr_battery_device *bat = power_supply_get_drvdata(psy);
+
+	return sprintf(buf, "%d\n", bat->alarm * 1000);
+}
+
+static ssize_t spwr_battery_alarm_store(struct device *dev,
+					struct device_attribute *attr,
+					const char *buf, size_t count)
+{
+	struct power_supply *psy = dev_get_drvdata(dev);
+	struct spwr_battery_device *bat = power_supply_get_drvdata(psy);
+	unsigned long value;
+	int status;
+
+	status = kstrtoul(buf, 0, &value);
+	if (status)
+		return status;
+
+	if (!spwr_battery_present(bat))
+		return -ENODEV;
+
+	status = spwr_battery_set_alarm(bat, value / 1000);
+	if (status)
+		return status;
+
+	return count;
+}
+
+static const struct device_attribute alarm_attr = {
+	.attr = {.name = "alarm", .mode = 0644},
+	.show = spwr_battery_alarm_show,
+	.store = spwr_battery_alarm_store,
+};
+
+
+static void spwr_ac_init(struct spwr_ac_device *ac,
+			    struct ssam_device *sdev,
+			    struct ssam_event_registry registry,
+			    const char *name)
+{
+	mutex_init(&ac->lock);
+	strncpy(ac->name, name, ARRAY_SIZE(ac->name) - 1);
+
+	ac->sdev = sdev;
+
+	ac->notif.base.priority = 1;
+	ac->notif.base.fn = spwr_notify_ac;
+	ac->notif.event.reg = registry;
+	ac->notif.event.id.target_category = sdev->uid.category;
+	ac->notif.event.id.instance = 0;
+	ac->notif.event.mask = SSAM_EVENT_MASK_NONE;
+	ac->notif.event.flags = SSAM_EVENT_SEQUENCED;
+
+	ac->psy_desc.name = ac->name;
+	ac->psy_desc.type = POWER_SUPPLY_TYPE_MAINS;
+	ac->psy_desc.properties = spwr_ac_props;
+	ac->psy_desc.num_properties = ARRAY_SIZE(spwr_ac_props);
+	ac->psy_desc.get_property = spwr_ac_get_property;
+}
+
+static void spwr_ac_destroy(struct spwr_ac_device *ac)
+{
+	mutex_destroy(&ac->lock);
+}
+
+static int spwr_ac_register(struct spwr_ac_device *ac)
+{
+	struct power_supply_config psy_cfg = {};
+	__le32 sta;
+	int status;
+
+	// make sure the device is there and functioning properly
+	status = ssam_retry(ssam_bat_get_sta, ac->sdev, &sta);
+	if (status)
+		return status;
+
+	if ((le32_to_cpu(sta) & SAM_BATTERY_STA_OK) != SAM_BATTERY_STA_OK)
+		return -ENODEV;
+
+	psy_cfg.drv_data = ac;
+	ac->psy = power_supply_register(&ac->sdev->dev, &ac->psy_desc, &psy_cfg);
+	if (IS_ERR(ac->psy))
+		return PTR_ERR(ac->psy);
+
+	status = ssam_notifier_register(ac->sdev->ctrl, &ac->notif);
+	if (status)
+		power_supply_unregister(ac->psy);
+
+	return status;
+}
+
+static int spwr_ac_unregister(struct spwr_ac_device *ac)
+{
+	ssam_notifier_unregister(ac->sdev->ctrl, &ac->notif);
+	power_supply_unregister(ac->psy);
+	return 0;
+}
+
+static void spwr_battery_init(struct spwr_battery_device *bat,
+			      struct ssam_device *sdev,
+			      struct ssam_event_registry registry,
+			      const char *name)
+{
+	mutex_init(&bat->lock);
+	strncpy(bat->name, name, ARRAY_SIZE(bat->name) - 1);
+
+	bat->sdev = sdev;
+
+	bat->notif.base.priority = 1;
+	bat->notif.base.fn = spwr_notify_bat;
+	bat->notif.event.reg = registry;
+	bat->notif.event.id.target_category = sdev->uid.category;
+	bat->notif.event.id.instance = 0;
+	bat->notif.event.mask = SSAM_EVENT_MASK_NONE;
+	bat->notif.event.flags = SSAM_EVENT_SEQUENCED;
+
+	bat->psy_desc.name = bat->name;
+	bat->psy_desc.type = POWER_SUPPLY_TYPE_BATTERY;
+	bat->psy_desc.get_property = spwr_battery_get_property;
+
+	INIT_DELAYED_WORK(&bat->update_work, spwr_battery_update_bst_workfn);
+}
+
+static void spwr_battery_destroy(struct spwr_battery_device *bat)
+{
+	mutex_destroy(&bat->lock);
+}
+
+static int spwr_battery_register(struct spwr_battery_device *bat)
+{
+	struct power_supply_config psy_cfg = {};
+	__le32 sta;
+	int status;
+
+	// make sure the device is there and functioning properly
+	status = ssam_retry(ssam_bat_get_sta, bat->sdev, &sta);
+	if (status)
+		return status;
+
+	if ((le32_to_cpu(sta) & SAM_BATTERY_STA_OK) != SAM_BATTERY_STA_OK)
+		return -ENODEV;
+
+	status = spwr_battery_update_bix_unlocked(bat);
+	if (status)
+		return status;
+
+	if (spwr_battery_present(bat)) {
+		u32 cap_warn = get_unaligned_le32(&bat->bix.design_cap_warn);
+
+		status = spwr_battery_set_alarm_unlocked(bat, cap_warn);
+		if (status)
+			return status;
+	}
+
+	switch (get_unaligned_le32(&bat->bix.power_unit)) {
+	case SAM_BATTERY_POWER_UNIT_mW:
+		bat->psy_desc.properties = spwr_battery_props_eng;
+		bat->psy_desc.num_properties = ARRAY_SIZE(spwr_battery_props_eng);
+		break;
+
+	case SAM_BATTERY_POWER_UNIT_mA:
+		bat->psy_desc.properties = spwr_battery_props_chg;
+		bat->psy_desc.num_properties = ARRAY_SIZE(spwr_battery_props_chg);
+		break;
+
+	default:
+		dev_err(&bat->sdev->dev, "unsupported battery power unit: %u\n",
+			get_unaligned_le32(&bat->bix.power_unit));
+		return -ENOTSUPP;
+	}
+
+	psy_cfg.drv_data = bat;
+	bat->psy = power_supply_register(&bat->sdev->dev, &bat->psy_desc, &psy_cfg);
+	if (IS_ERR(bat->psy))
+		return PTR_ERR(bat->psy);
+
+	status = ssam_notifier_register(bat->sdev->ctrl, &bat->notif);
+	if (status)
+		goto err_notif;
+
+	status = device_create_file(&bat->psy->dev, &alarm_attr);
+	if (status)
+		goto err_file;
+
+	return 0;
+
+err_file:
+	ssam_notifier_unregister(bat->sdev->ctrl, &bat->notif);
+err_notif:
+	power_supply_unregister(bat->psy);
+	return status;
+}
+
+static void spwr_battery_unregister(struct spwr_battery_device *bat)
+{
+	ssam_notifier_unregister(bat->sdev->ctrl, &bat->notif);
+	cancel_delayed_work_sync(&bat->update_work);
+	device_remove_file(&bat->psy->dev, &alarm_attr);
+	power_supply_unregister(bat->psy);
+}
+
+
+/* -- Power management. ----------------------------------------------------- */
+
+#ifdef CONFIG_PM_SLEEP
+
+static int surface_battery_resume(struct device *dev)
+{
+	return spwr_battery_recheck_full(dev_get_drvdata(dev));
+}
+
+static int surface_ac_resume(struct device *dev)
+{
+	return spwr_ac_recheck(dev_get_drvdata(dev));
+}
+
+#else /* CONFIG_PM_SLEEP */
+
+#define surface_battery_resume	NULL
+#define surface_ac_resume	NULL
+
+#endif /* CONFIG_PM_SLEEP */
+
+SIMPLE_DEV_PM_OPS(surface_battery_pm_ops, NULL, surface_battery_resume);
+SIMPLE_DEV_PM_OPS(surface_ac_pm_ops, NULL, surface_ac_resume);
+
+
+/* -- Battery driver. ------------------------------------------------------- */
+
+static int surface_battery_probe(struct ssam_device *sdev)
+{
+	const struct spwr_psy_properties *p;
+	struct spwr_battery_device *bat;
+	int status;
+
+	p = ssam_device_get_match_data(sdev);
+	if (!p)
+		return -ENODEV;
+
+	bat = devm_kzalloc(&sdev->dev, sizeof(*bat), GFP_KERNEL);
+	if (!bat)
+		return -ENOMEM;
+
+	spwr_battery_init(bat, sdev, p->registry, p->name);
+	ssam_device_set_drvdata(sdev, bat);
+
+	status = spwr_battery_register(bat);
+	if (status)
+		spwr_battery_destroy(bat);
+
+	return status;
+}
+
+static void surface_battery_remove(struct ssam_device *sdev)
+{
+	struct spwr_battery_device *bat = ssam_device_get_drvdata(sdev);
+
+	spwr_battery_unregister(bat);
+	spwr_battery_destroy(bat);
+}
+
+static const struct spwr_psy_properties spwr_psy_props_bat1 = {
+	.name = "BAT1",
+	.registry = SSAM_EVENT_REGISTRY_SAM,
+};
+
+static const struct spwr_psy_properties spwr_psy_props_bat2_sb3 = {
+	.name = "BAT2",
+	.registry = SSAM_EVENT_REGISTRY_KIP,
+};
+
+static const struct ssam_device_id surface_battery_match[] = {
+	{ SSAM_SDEV(BAT, 0x01, 0x01, 0x00), (unsigned long)&spwr_psy_props_bat1     },
+	{ SSAM_SDEV(BAT, 0x02, 0x01, 0x00), (unsigned long)&spwr_psy_props_bat2_sb3 },
+	{ },
+};
+MODULE_DEVICE_TABLE(ssam, surface_battery_match);
+
+static struct ssam_device_driver surface_battery_driver = {
+	.probe = surface_battery_probe,
+	.remove = surface_battery_remove,
+	.match_table = surface_battery_match,
+	.driver = {
+		.name = "surface_battery",
+		.pm = &surface_battery_pm_ops,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+
+/* -- AC driver. ------------------------------------------------------------ */
+
+static int surface_ac_probe(struct ssam_device *sdev)
+{
+	const struct spwr_psy_properties *p;
+	struct spwr_ac_device *ac;
+	int status;
+
+	p = ssam_device_get_match_data(sdev);
+	if (!p)
+		return -ENODEV;
+
+	ac = devm_kzalloc(&sdev->dev, sizeof(*ac), GFP_KERNEL);
+	if (!ac)
+		return -ENOMEM;
+
+	spwr_ac_init(ac, sdev, p->registry, p->name);
+	ssam_device_set_drvdata(sdev, ac);
+
+	status = spwr_ac_register(ac);
+	if (status)
+		spwr_ac_destroy(ac);
+
+	return status;
+}
+
+static void surface_ac_remove(struct ssam_device *sdev)
+{
+	struct spwr_ac_device *ac = ssam_device_get_drvdata(sdev);
+
+	spwr_ac_unregister(ac);
+	spwr_ac_destroy(ac);
+}
+
+static const struct spwr_psy_properties spwr_psy_props_adp1 = {
+	.name = "ADP1",
+	.registry = SSAM_EVENT_REGISTRY_SAM,
+};
+
+static const struct ssam_device_id surface_ac_match[] = {
+	{ SSAM_SDEV(BAT, 0x01, 0x01, 0x01), (unsigned long)&spwr_psy_props_adp1 },
+	{ },
+};
+MODULE_DEVICE_TABLE(ssam, surface_ac_match);
+
+static struct ssam_device_driver surface_ac_driver = {
+	.probe = surface_ac_probe,
+	.remove = surface_ac_remove,
+	.match_table = surface_ac_match,
+	.driver = {
+		.name = "surface_ac",
+		.pm = &surface_ac_pm_ops,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+
+/* -- Module setup. --------------------------------------------------------- */
+
+static int __init surface_battery_init(void)
+{
+	int status;
+
+	status = ssam_device_driver_register(&surface_battery_driver);
+	if (status)
+		return status;
+
+	status = ssam_device_driver_register(&surface_ac_driver);
+	if (status)
+		ssam_device_driver_unregister(&surface_battery_driver);
+
+	return status;
+}
+module_init(surface_battery_init);
+
+static void __exit surface_battery_exit(void)
+{
+	ssam_device_driver_unregister(&surface_battery_driver);
+	ssam_device_driver_unregister(&surface_ac_driver);
+}
+module_exit(surface_battery_exit);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Battery/AC driver for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/clients/surface_dtx.c b/drivers/misc/surface_aggregator/clients/surface_dtx.c
new file mode 100644
index 000000000000..346b54848b8f
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_dtx.c
@@ -0,0 +1,1279 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface Book (gen. 2 and later) detachment system (DTX) driver.
+ *
+ * Provides a user-space interface to properly handle clipboard/tablet
+ * (containing screen and processor) detachment from the base of the device
+ * (containing the keyboard and optionally a discrete GPU). Allows to
+ * acknowledge (to speed things up), abort (e.g. in case the dGPU is stil in
+ * use), or request detachment via user-space.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/acpi.h>
+#include <linux/delay.h>
+#include <linux/fs.h>
+#include <linux/input.h>
+#include <linux/ioctl.h>
+#include <linux/kernel.h>
+#include <linux/kfifo.h>
+#include <linux/kref.h>
+#include <linux/miscdevice.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/poll.h>
+#include <linux/rwsem.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/device.h>
+#include <linux/surface_aggregator/dtx.h>
+
+
+/* -- SSAM interface. ------------------------------------------------------- */
+
+enum sam_event_cid_bas {
+	SAM_EVENT_CID_DTX_CONNECTION			= 0x0c,
+	SAM_EVENT_CID_DTX_REQUEST			= 0x0e,
+	SAM_EVENT_CID_DTX_CANCEL			= 0x0f,
+	SAM_EVENT_CID_DTX_LATCH_STATUS			= 0x11,
+};
+
+enum ssam_bas_base_state {
+	SSAM_BAS_BASE_STATE_DETACH_SUCCESS		= 0x00,
+	SSAM_BAS_BASE_STATE_ATTACHED			= 0x01,
+	SSAM_BAS_BASE_STATE_NOT_FEASIBLE		= 0x02,
+};
+
+enum ssam_bas_latch_status {
+	SSAM_BAS_LATCH_STATUS_CLOSED			= 0x00,
+	SSAM_BAS_LATCH_STATUS_OPENED			= 0x01,
+	SSAM_BAS_LATCH_STATUS_FAILED_TO_OPEN		= 0x02,
+	SSAM_BAS_LATCH_STATUS_FAILED_TO_REMAIN_OPEN	= 0x03,
+	SSAM_BAS_LATCH_STATUS_FAILED_TO_CLOSE		= 0x04,
+};
+
+enum ssam_bas_cancel_reason {
+	SSAM_BAS_CANCEL_REASON_NOT_FEASIBLE		= 0x00,  // low battery
+	SSAM_BAS_CANCEL_REASON_TIMEOUT			= 0x02,
+	SSAM_BAS_CANCEL_REASON_FAILED_TO_OPEN		= 0x03,
+	SSAM_BAS_CANCEL_REASON_FAILED_TO_REMAIN_OPEN	= 0x04,
+	SSAM_BAS_CANCEL_REASON_FAILED_TO_CLOSE		= 0x05,
+};
+
+struct ssam_bas_base_info {
+	u8 state;
+	u8 base_id;
+} __packed;
+
+static_assert(sizeof(struct ssam_bas_base_info) == 2);
+
+static SSAM_DEFINE_SYNC_REQUEST_N(ssam_bas_latch_lock, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x06,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_N(ssam_bas_latch_unlock, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x07,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_N(ssam_bas_latch_request, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x08,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_N(ssam_bas_latch_confirm, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x09,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_N(ssam_bas_latch_heartbeat, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x0a,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_N(ssam_bas_latch_cancel, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x0b,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_bas_get_base, struct ssam_bas_base_info, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x0c,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_bas_get_device_mode, u8, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x0d,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_bas_get_latch_status, u8, {
+	.target_category = SSAM_SSH_TC_BAS,
+	.target_id       = 0x01,
+	.command_id      = 0x11,
+	.instance_id     = 0x00,
+});
+
+
+/* -- Main structures. ------------------------------------------------------ */
+
+enum sdtx_device_state {
+	SDTX_DEVICE_SHUTDOWN_BIT    = BIT(0),
+	SDTX_DEVICE_DIRTY_BASE_BIT  = BIT(1),
+	SDTX_DEVICE_DIRTY_MODE_BIT  = BIT(2),
+	SDTX_DEVICE_DIRTY_LATCH_BIT = BIT(3),
+};
+
+struct sdtx_device {
+	struct kref kref;
+	struct rw_semaphore lock;
+
+	struct device *dev;
+	struct ssam_controller *ctrl;
+	unsigned long flags;
+
+	struct miscdevice mdev;
+	wait_queue_head_t waitq;
+	struct mutex write_lock;
+	struct rw_semaphore client_lock;
+	struct list_head client_list;
+
+	struct delayed_work state_work;
+	struct {
+		struct ssam_bas_base_info base;
+		u8 device_mode;
+		u8 latch_status;
+	} state;
+
+	struct delayed_work mode_work;
+	struct input_dev *mode_switch;
+
+	struct ssam_event_notifier notif;
+};
+
+enum sdtx_client_state {
+	SDTX_CLIENT_EVENTS_ENABLED_BIT = BIT(0),
+};
+
+struct sdtx_client {
+	struct sdtx_device *ddev;
+	struct list_head node;
+	unsigned long flags;
+
+	struct fasync_struct *fasync;
+
+	struct mutex read_lock;
+	DECLARE_KFIFO(buffer, u8, 512);
+};
+
+static void __sdtx_device_release(struct kref *kref)
+{
+	struct sdtx_device *ddev = container_of(kref, struct sdtx_device, kref);
+
+	mutex_destroy(&ddev->write_lock);
+	kfree(ddev);
+}
+
+static struct sdtx_device *sdtx_device_get(struct sdtx_device *ddev)
+{
+	if (ddev)
+		kref_get(&ddev->kref);
+
+	return ddev;
+}
+
+static void sdtx_device_put(struct sdtx_device *ddev)
+{
+	if (ddev)
+		kref_put(&ddev->kref, __sdtx_device_release);
+}
+
+
+/* -- Firmware value translations. ------------------------------------------ */
+
+static u16 sdtx_translate_base_state(struct sdtx_device *ddev, u8 state)
+{
+	switch (state) {
+	case SSAM_BAS_BASE_STATE_ATTACHED:
+		return SDTX_BASE_ATTACHED;
+
+	case SSAM_BAS_BASE_STATE_DETACH_SUCCESS:
+		return SDTX_BASE_DETACHED;
+
+	case SSAM_BAS_BASE_STATE_NOT_FEASIBLE:
+		return SDTX_DETACH_NOT_FEASIBLE;
+
+	default:
+		dev_err(ddev->dev, "unknown base state: 0x%02x\n", state);
+		return SDTX_UNKNOWN(state);
+	}
+}
+
+static u16 sdtx_translate_latch_status(struct sdtx_device *ddev, u8 status)
+{
+	switch (status) {
+	case SSAM_BAS_LATCH_STATUS_CLOSED:
+		return SDTX_LATCH_CLOSED;
+
+	case SSAM_BAS_LATCH_STATUS_OPENED:
+		return SDTX_LATCH_OPENED;
+
+	case SSAM_BAS_LATCH_STATUS_FAILED_TO_OPEN:
+		return SDTX_ERR_FAILED_TO_OPEN;
+
+	case SSAM_BAS_LATCH_STATUS_FAILED_TO_REMAIN_OPEN:
+		return SDTX_ERR_FAILED_TO_REMAIN_OPEN;
+
+	case SSAM_BAS_LATCH_STATUS_FAILED_TO_CLOSE:
+		return SDTX_ERR_FAILED_TO_CLOSE;
+
+	default:
+		dev_err(ddev->dev, "unknown latch status: 0x%02x\n", status);
+		return SDTX_UNKNOWN(status);
+	}
+}
+
+static u16 sdtx_translate_cancel_reason(struct sdtx_device *ddev, u8 reason)
+{
+	switch (reason) {
+	case SSAM_BAS_CANCEL_REASON_NOT_FEASIBLE:
+		return SDTX_DETACH_NOT_FEASIBLE;
+
+	case SSAM_BAS_CANCEL_REASON_TIMEOUT:
+		return SDTX_DETACH_TIMEDOUT;
+
+	case SSAM_BAS_CANCEL_REASON_FAILED_TO_OPEN:
+		return SDTX_ERR_FAILED_TO_OPEN;
+
+	case SSAM_BAS_CANCEL_REASON_FAILED_TO_REMAIN_OPEN:
+		return SDTX_ERR_FAILED_TO_REMAIN_OPEN;
+
+	case SSAM_BAS_CANCEL_REASON_FAILED_TO_CLOSE:
+		return SDTX_ERR_FAILED_TO_CLOSE;
+
+	default:
+		dev_err(ddev->dev, "unknown cancel reason: 0x%02x\n", reason);
+		return SDTX_UNKNOWN(reason);
+	}
+}
+
+
+/* -- IOCTLs. --------------------------------------------------------------- */
+
+static int sdtx_ioctl_get_base_info(struct sdtx_device *ddev,
+				    struct sdtx_base_info __user *buf)
+{
+	struct ssam_bas_base_info raw;
+	struct sdtx_base_info info;
+	int status;
+
+	status = ssam_retry(ssam_bas_get_base, ddev->ctrl, &raw);
+	if (status < 0)
+		return status;
+
+	info.state = sdtx_translate_base_state(ddev, raw.state);
+	info.base_id = SDTX_BASE_TYPE_SSH(raw.base_id);
+
+	if (copy_to_user(buf, &info, sizeof(info)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int sdtx_ioctl_get_device_mode(struct sdtx_device *ddev, u16 __user *buf)
+{
+	u8 mode;
+	int status;
+
+	status = ssam_retry(ssam_bas_get_device_mode, ddev->ctrl, &mode);
+	if (status < 0)
+		return status;
+
+	return put_user(mode, buf);
+}
+
+static int sdtx_ioctl_get_latch_status(struct sdtx_device *ddev, u16 __user *buf)
+{
+	u8 latch;
+	int status;
+
+	status = ssam_retry(ssam_bas_get_latch_status, ddev->ctrl, &latch);
+	if (status < 0)
+		return status;
+
+	return put_user(sdtx_translate_latch_status(ddev, latch), buf);
+}
+
+static long __surface_dtx_ioctl(struct sdtx_client *client, unsigned int cmd,
+				unsigned long arg)
+{
+	struct sdtx_device *ddev = client->ddev;
+
+	switch (cmd) {
+	case SDTX_IOCTL_EVENTS_ENABLE:
+		set_bit(SDTX_CLIENT_EVENTS_ENABLED_BIT, &client->flags);
+		return 0;
+
+	case SDTX_IOCTL_EVENTS_DISABLE:
+		clear_bit(SDTX_CLIENT_EVENTS_ENABLED_BIT, &client->flags);
+		return 0;
+
+	case SDTX_IOCTL_LATCH_LOCK:
+		return ssam_retry(ssam_bas_latch_lock, ddev->ctrl);
+
+	case SDTX_IOCTL_LATCH_UNLOCK:
+		return ssam_retry(ssam_bas_latch_unlock, ddev->ctrl);
+
+	case SDTX_IOCTL_LATCH_REQUEST:
+		return ssam_retry(ssam_bas_latch_request, ddev->ctrl);
+
+	case SDTX_IOCTL_LATCH_CONFIRM:
+		return ssam_retry(ssam_bas_latch_confirm, ddev->ctrl);
+
+	case SDTX_IOCTL_LATCH_HEARTBEAT:
+		return ssam_retry(ssam_bas_latch_heartbeat, ddev->ctrl);
+
+	case SDTX_IOCTL_LATCH_CANCEL:
+		return ssam_retry(ssam_bas_latch_cancel, ddev->ctrl);
+
+	case SDTX_IOCTL_GET_BASE_INFO:
+		return sdtx_ioctl_get_base_info(ddev,
+				(struct sdtx_base_info __user *)arg);
+
+	case SDTX_IOCTL_GET_DEVICE_MODE:
+		return sdtx_ioctl_get_device_mode(ddev, (u16 __user *)arg);
+
+	case SDTX_IOCTL_GET_LATCH_STATUS:
+		return sdtx_ioctl_get_latch_status(ddev, (u16 __user *)arg);
+
+	default:
+		return -EINVAL;
+	}
+}
+
+static long surface_dtx_ioctl(struct file *file, unsigned int cmd,
+			      unsigned long arg)
+{
+	struct sdtx_client *client = file->private_data;
+	long status;
+
+	if (down_read_killable(&client->ddev->lock))
+		return -ERESTARTSYS;
+
+	if (test_bit(SDTX_DEVICE_SHUTDOWN_BIT, &client->ddev->flags)) {
+		up_read(&client->ddev->lock);
+		return -ENODEV;
+	}
+
+	status = __surface_dtx_ioctl(client, cmd, arg);
+
+	up_read(&client->ddev->lock);
+	return status;
+}
+
+
+/* -- File operations. ------------------------------------------------------ */
+
+static int surface_dtx_open(struct inode *inode, struct file *file)
+{
+	struct sdtx_device *ddev;
+	struct sdtx_client *client;
+
+	ddev = container_of(file->private_data, struct sdtx_device, mdev);
+
+	// initialize client
+	client = kzalloc(sizeof(*client), GFP_KERNEL);
+	if (!client)
+		return -ENOMEM;
+
+	client->ddev = sdtx_device_get(ddev);
+
+	INIT_LIST_HEAD(&client->node);
+
+	mutex_init(&client->read_lock);
+	INIT_KFIFO(client->buffer);
+
+	file->private_data = client;
+
+	// attach client
+	down_write(&ddev->client_lock);
+
+	// do not add a new client if the device has been shut down
+	if (test_bit(SDTX_DEVICE_SHUTDOWN_BIT, &ddev->flags)) {
+		up_write(&ddev->client_lock);
+		sdtx_device_put(client->ddev);
+		kfree(client);
+		return -ENODEV;
+	}
+
+	list_add_tail(&client->node, &ddev->client_list);
+	up_write(&ddev->client_lock);
+
+	stream_open(inode, file);
+	return 0;
+}
+
+static int surface_dtx_release(struct inode *inode, struct file *file)
+{
+	struct sdtx_client *client = file->private_data;
+
+	// detach client
+	down_write(&client->ddev->client_lock);
+	list_del(&client->node);
+	up_write(&client->ddev->client_lock);
+
+	sdtx_device_put(client->ddev);
+	mutex_destroy(&client->read_lock);
+	kfree(client);
+
+	return 0;
+}
+
+static ssize_t surface_dtx_read(struct file *file, char __user *buf,
+				size_t count, loff_t *offs)
+{
+	struct sdtx_client *client = file->private_data;
+	struct sdtx_device *ddev = client->ddev;
+	unsigned int copied;
+	int status = 0;
+
+	if (down_read_killable(&ddev->lock))
+		return -ERESTARTSYS;
+
+	// make sure we're not shut down
+	if (test_bit(SDTX_DEVICE_SHUTDOWN_BIT, &ddev->flags)) {
+		up_read(&ddev->lock);
+		return -ENODEV;
+	}
+
+	do {
+		// check availability, wait if necessary
+		if (kfifo_is_empty(&client->buffer)) {
+			up_read(&ddev->lock);
+
+			if (file->f_flags & O_NONBLOCK)
+				return -EAGAIN;
+
+			status = wait_event_interruptible(ddev->waitq,
+					!kfifo_is_empty(&client->buffer)
+					|| test_bit(SDTX_DEVICE_SHUTDOWN_BIT,
+						    &ddev->flags));
+			if (status < 0)
+				return status;
+
+			if (down_read_killable(&client->ddev->lock))
+				return -ERESTARTSYS;
+
+			// need to check that we're not shut down again
+			if (test_bit(SDTX_DEVICE_SHUTDOWN_BIT, &ddev->flags)) {
+				up_read(&ddev->lock);
+				return -ENODEV;
+			}
+		}
+
+		// try to read from fifo
+		if (mutex_lock_interruptible(&client->read_lock)) {
+			up_read(&ddev->lock);
+			return -ERESTARTSYS;
+		}
+
+		status = kfifo_to_user(&client->buffer, buf, count, &copied);
+		mutex_unlock(&client->read_lock);
+
+		if (status < 0) {
+			up_read(&ddev->lock);
+			return status;
+		}
+
+		// we might not have gotten anything, check this here
+		if (copied == 0 && (file->f_flags & O_NONBLOCK)) {
+			up_read(&ddev->lock);
+			return -EAGAIN;
+		}
+	} while (copied == 0);
+
+	up_read(&ddev->lock);
+	return copied;
+}
+
+static __poll_t surface_dtx_poll(struct file *file, struct poll_table_struct *pt)
+{
+	struct sdtx_client *client = file->private_data;
+	__poll_t events = 0;
+
+	if (down_read_killable(&client->ddev->lock))
+		return -ERESTARTSYS;
+
+	if (test_bit(SDTX_DEVICE_SHUTDOWN_BIT, &client->ddev->flags)) {
+		up_read(&client->ddev->lock);
+		return EPOLLHUP | EPOLLERR;
+	}
+
+	poll_wait(file, &client->ddev->waitq, pt);
+
+	if (!kfifo_is_empty(&client->buffer))
+		events |= EPOLLIN | EPOLLRDNORM;
+
+	up_read(&client->ddev->lock);
+	return events;
+}
+
+static int surface_dtx_fasync(int fd, struct file *file, int on)
+{
+	struct sdtx_client *client = file->private_data;
+
+	return fasync_helper(fd, file, on, &client->fasync);
+}
+
+static const struct file_operations surface_dtx_fops = {
+	.owner          = THIS_MODULE,
+	.open           = surface_dtx_open,
+	.release        = surface_dtx_release,
+	.read           = surface_dtx_read,
+	.poll           = surface_dtx_poll,
+	.fasync         = surface_dtx_fasync,
+	.unlocked_ioctl = surface_dtx_ioctl,
+	.compat_ioctl   = surface_dtx_ioctl,
+	.llseek         = no_llseek,
+};
+
+
+/* -- Event handling/forwarding. -------------------------------------------- */
+
+/*
+ * The device operation mode is not immediately updated on the EC when the
+ * base has been connected, i.e. querying the device mode inside the
+ * connection event callback yields an outdated value. Thus, we can only
+ * determine the new tablet-mode switch and device mode values after some
+ * time.
+ *
+ * These delays have been chosen by experimenting. We first delay on connect
+ * events, then check and validate the device mode against the base state and
+ * if invalid delay again by the "recheck" delay.
+ */
+#define SDTX_DEVICE_MODE_DELAY_CONNECT	msecs_to_jiffies(100)
+#define SDTX_DEVICE_MODE_DELAY_RECHECK	msecs_to_jiffies(100)
+
+static void sdtx_update_device_mode(struct sdtx_device *ddev, unsigned long delay);
+
+
+struct sdtx_status_event {
+	struct sdtx_event e;
+	__u16 v;
+} __packed;
+
+struct sdtx_base_info_event {
+	struct sdtx_event e;
+	struct sdtx_base_info v;
+} __packed;
+
+union sdtx_generic_event {
+	struct sdtx_event common;
+	struct sdtx_status_event status;
+	struct sdtx_base_info_event base;
+};
+
+/* Must be executed with ddev->write_lock held. */
+static void sdtx_push_event(struct sdtx_device *ddev, struct sdtx_event *evt)
+{
+	const size_t len = sizeof(struct sdtx_event) + evt->length;
+	struct sdtx_client *client;
+
+	down_read(&ddev->client_lock);
+	list_for_each_entry(client, &ddev->client_list, node) {
+		if (!test_bit(SDTX_CLIENT_EVENTS_ENABLED_BIT, &client->flags))
+			continue;
+
+		if (likely(kfifo_avail(&client->buffer) >= len))
+			kfifo_in(&client->buffer, (const u8 *)evt, len);
+		else
+			dev_warn(ddev->dev, "event buffer overrun\n");
+
+		kill_fasync(&client->fasync, SIGIO, POLL_IN);
+	}
+	up_read(&ddev->client_lock);
+
+	wake_up_interruptible(&ddev->waitq);
+}
+
+static u32 sdtx_notifier(struct ssam_event_notifier *nf,
+			 const struct ssam_event *in)
+{
+	struct sdtx_device *ddev = container_of(nf, struct sdtx_device, notif);
+	union sdtx_generic_event event;
+	size_t len;
+
+	// validate event payload length
+	switch (in->command_id) {
+	case SAM_EVENT_CID_DTX_CONNECTION:
+		len = 2;
+		break;
+
+	case SAM_EVENT_CID_DTX_REQUEST:
+		len = 0;
+		break;
+
+	case SAM_EVENT_CID_DTX_CANCEL:
+		len = 1;
+		break;
+
+	case SAM_EVENT_CID_DTX_LATCH_STATUS:
+		len = 1;
+		break;
+
+	default:
+		return 0;
+	};
+
+	if (in->length != len) {
+		dev_err(ddev->dev, "unexpected payload size for event 0x%02x: "
+			"got %u, expected %zu", in->command_id, in->length, len);
+		return 0;
+	}
+
+	mutex_lock(&ddev->write_lock);
+
+	// translate event
+	switch (in->command_id) {
+	case SAM_EVENT_CID_DTX_CONNECTION:
+		clear_bit(SDTX_DEVICE_DIRTY_BASE_BIT, &ddev->flags);
+
+		// if state has not changed: do not send new event
+		if (ddev->state.base.state == in->data[0]
+		    && ddev->state.base.base_id == in->data[1])
+			goto out;
+
+		ddev->state.base.state = in->data[0];
+		ddev->state.base.base_id = in->data[1];
+
+		event.base.e.length = sizeof(struct sdtx_base_info);
+		event.base.e.code = SDTX_EVENT_BASE_CONNECTION;
+		event.base.v.state = sdtx_translate_base_state(ddev, in->data[0]);
+		event.base.v.base_id = SDTX_BASE_TYPE_SSH(in->data[1]);
+		break;
+
+	case SAM_EVENT_CID_DTX_REQUEST:
+		event.common.code = SDTX_EVENT_REQUEST;
+		event.common.length = 0;
+		break;
+
+	case SAM_EVENT_CID_DTX_CANCEL:
+		event.status.e.length = sizeof(u16);
+		event.status.e.code = SDTX_EVENT_CANCEL;
+		event.status.v = sdtx_translate_cancel_reason(ddev, in->data[0]);
+		break;
+
+	case SAM_EVENT_CID_DTX_LATCH_STATUS:
+		clear_bit(SDTX_DEVICE_DIRTY_LATCH_BIT, &ddev->flags);
+
+		// if state has not changed: do not send new event
+		if (ddev->state.latch_status == in->data[0])
+			goto out;
+
+		ddev->state.latch_status = in->data[0];
+
+		event.status.e.length = sizeof(u16);
+		event.status.e.code = SDTX_EVENT_LATCH_STATUS;
+		event.status.v = sdtx_translate_latch_status(ddev, in->data[0]);
+		break;
+	}
+
+	sdtx_push_event(ddev, &event.common);
+
+	// update device mode on base connection change
+	if (in->command_id == SAM_EVENT_CID_DTX_CONNECTION) {
+		unsigned long delay;
+
+		delay = in->data[0] ? SDTX_DEVICE_MODE_DELAY_CONNECT : 0;
+		sdtx_update_device_mode(ddev, delay);
+	}
+
+out:
+	mutex_unlock(&ddev->write_lock);
+	return SSAM_NOTIF_HANDLED;
+}
+
+
+/* -- State update functions. ----------------------------------------------- */
+
+static bool sdtx_device_mode_invalid(u8 mode, u8 base_state)
+{
+	return ((base_state == SSAM_BAS_BASE_STATE_ATTACHED)
+			&& (mode == SDTX_DEVICE_MODE_TABLET))
+		|| ((base_state == SSAM_BAS_BASE_STATE_DETACH_SUCCESS)
+			&& (mode != SDTX_DEVICE_MODE_TABLET));
+}
+
+static void sdtx_device_mode_workfn(struct work_struct *work)
+{
+	struct sdtx_device *ddev;
+	struct sdtx_status_event event;
+	struct ssam_bas_base_info base;
+	int status, tablet;
+	u8 mode;
+
+	ddev = container_of(work, struct sdtx_device, mode_work.work);
+
+	// get operation mode
+	status = ssam_retry(ssam_bas_get_device_mode, ddev->ctrl, &mode);
+	if (status) {
+		dev_err(ddev->dev, "failed to get device mode: %d\n", status);
+		return;
+	}
+
+	// get base info
+	status = ssam_retry(ssam_bas_get_base, ddev->ctrl, &base);
+	if (status) {
+		dev_err(ddev->dev, "failed to get base info: %d\n", status);
+		return;
+	}
+
+	/*
+	 * In some cases (specifically when attaching the base), the device
+	 * mode isn't updated right away. Thus we check if the device mode
+	 * makes sense for the given base state and try again later if it
+	 * doesn't.
+	 */
+	if (sdtx_device_mode_invalid(mode, base.state)) {
+		dev_dbg(ddev->dev, "device mode is invalid, trying again\n");
+		sdtx_update_device_mode(ddev, SDTX_DEVICE_MODE_DELAY_RECHECK);
+		return;
+	}
+
+	mutex_lock(&ddev->write_lock);
+	clear_bit(SDTX_DEVICE_DIRTY_MODE_BIT, &ddev->flags);
+
+	// avoid sending duplicate device-mode events
+	if (ddev->state.device_mode == mode) {
+		mutex_unlock(&ddev->write_lock);
+		return;
+	}
+
+	ddev->state.device_mode = mode;
+
+	event.e.length = sizeof(u16);
+	event.e.code = SDTX_EVENT_DEVICE_MODE;
+	event.v = mode;
+
+	sdtx_push_event(ddev, &event.e);
+
+	// send SW_TABLET_MODE event
+	tablet = mode != SDTX_DEVICE_MODE_LAPTOP;
+	input_report_switch(ddev->mode_switch, SW_TABLET_MODE, tablet);
+	input_sync(ddev->mode_switch);
+
+	mutex_unlock(&ddev->write_lock);
+}
+
+static void sdtx_update_device_mode(struct sdtx_device *ddev, unsigned long delay)
+{
+	schedule_delayed_work(&ddev->mode_work, delay);
+}
+
+
+static void __sdtx_device_state_update_base(struct sdtx_device *ddev,
+					    struct ssam_bas_base_info info)
+{
+	struct sdtx_base_info_event event;
+
+	// prevent duplicate events
+	if (ddev->state.base.state == info.state
+	    && ddev->state.base.base_id == info.base_id)
+		return;
+
+	ddev->state.base = info;
+
+	event.e.length = sizeof(struct sdtx_base_info);
+	event.e.code = SDTX_EVENT_BASE_CONNECTION;
+	event.v.state = sdtx_translate_base_state(ddev, info.state);
+	event.v.base_id = SDTX_BASE_TYPE_SSH(info.base_id);
+
+	sdtx_push_event(ddev, &event.e);
+}
+
+static void __sdtx_device_state_update_mode(struct sdtx_device *ddev, u8 mode)
+{
+	struct sdtx_status_event event;
+	int tablet;
+
+	/*
+	 * Note: This function must be called after updating the base state
+	 * via __sdtx_device_state_update_base(), as we rely on the updated
+	 * base state value in the validity check below.
+	 */
+
+	if (sdtx_device_mode_invalid(mode, ddev->state.base.state)) {
+		dev_dbg(ddev->dev, "device mode is invalid, trying again\n");
+		sdtx_update_device_mode(ddev, SDTX_DEVICE_MODE_DELAY_RECHECK);
+		return;
+	}
+
+	// prevent duplicate events
+	if (ddev->state.device_mode == mode)
+		return;
+
+	ddev->state.device_mode = mode;
+
+	// send event
+	event.e.length = sizeof(u16);
+	event.e.code = SDTX_EVENT_DEVICE_MODE;
+	event.v = mode;
+
+	sdtx_push_event(ddev, &event.e);
+
+	// send SW_TABLET_MODE event
+	tablet = mode != SDTX_DEVICE_MODE_LAPTOP;
+	input_report_switch(ddev->mode_switch, SW_TABLET_MODE, tablet);
+	input_sync(ddev->mode_switch);
+}
+
+static void __sdtx_device_state_update_latch(struct sdtx_device *ddev, u8 status)
+{
+	struct sdtx_status_event event;
+
+	// prevent duplicate events
+	if (ddev->state.latch_status == status)
+		return;
+
+	ddev->state.latch_status = status;
+
+	event.e.length = sizeof(struct sdtx_base_info);
+	event.e.code = SDTX_EVENT_BASE_CONNECTION;
+	event.v = sdtx_translate_latch_status(ddev, status);
+
+	sdtx_push_event(ddev, &event.e);
+}
+
+static void sdtx_device_state_workfn(struct work_struct *work)
+{
+	struct sdtx_device *ddev;
+	struct ssam_bas_base_info base;
+	u8 mode, latch;
+	int status;
+
+	ddev = container_of(work, struct sdtx_device, state_work.work);
+
+	// mark everyting as dirty
+	set_bit(SDTX_DEVICE_DIRTY_BASE_BIT, &ddev->flags);
+	set_bit(SDTX_DEVICE_DIRTY_MODE_BIT, &ddev->flags);
+	set_bit(SDTX_DEVICE_DIRTY_LATCH_BIT, &ddev->flags);
+
+	/*
+	 * Ensure that the state gets marked as dirty before continuing to
+	 * query it. Necessary to ensure that clear_bit() calls in
+	 * sdtx_notifier() and sdtx_device_mode_workfn() actually clear these
+	 * bits if an event is received while updating the state here.
+	 */
+	smp_mb__after_atomic();
+
+	status = ssam_retry(ssam_bas_get_base, ddev->ctrl, &base);
+	if (status) {
+		dev_err(ddev->dev, "failed to get base state: %d\n", status);
+		return;
+	}
+
+	status = ssam_retry(ssam_bas_get_device_mode, ddev->ctrl, &mode);
+	if (status) {
+		dev_err(ddev->dev, "failed to get device mode: %d\n", status);
+		return;
+	}
+
+	status = ssam_retry(ssam_bas_get_latch_status, ddev->ctrl, &latch);
+	if (status) {
+		dev_err(ddev->dev, "failed to get latch status: %d\n", status);
+		return;
+	}
+
+	mutex_lock(&ddev->write_lock);
+
+	/*
+	 * If the respective dirty-bit has been cleared, an event has been
+	 * received, updating this state. The queried state may thus be out of
+	 * date. At this point, we can safely assume that the state provided
+	 * by the event is either up to date, or we're about to receive
+	 * another event updating it.
+	 */
+
+	if (test_and_clear_bit(SDTX_DEVICE_DIRTY_BASE_BIT, &ddev->flags))
+		__sdtx_device_state_update_base(ddev, base);
+
+	if (test_and_clear_bit(SDTX_DEVICE_DIRTY_MODE_BIT, &ddev->flags))
+		__sdtx_device_state_update_mode(ddev, mode);
+
+	if (test_and_clear_bit(SDTX_DEVICE_DIRTY_LATCH_BIT, &ddev->flags))
+		__sdtx_device_state_update_latch(ddev, latch);
+
+	mutex_unlock(&ddev->write_lock);
+}
+
+static void sdtx_update_device_state(struct sdtx_device *ddev, unsigned long delay)
+{
+	schedule_delayed_work(&ddev->state_work, delay);
+}
+
+
+/* -- Common device initialization. ----------------------------------------- */
+
+static int sdtx_device_init(struct sdtx_device *ddev, struct device *dev,
+			    struct ssam_controller *ctrl)
+{
+	int status, tablet_mode;
+
+	// basic initialization
+	kref_init(&ddev->kref);
+	ddev->dev = dev;
+	ddev->ctrl = ctrl;
+
+	ddev->mdev.minor = MISC_DYNAMIC_MINOR;
+	ddev->mdev.name = "surface_dtx";
+	ddev->mdev.nodename = "surface/dtx";
+	ddev->mdev.fops = &surface_dtx_fops;
+
+	ddev->notif.base.priority = 1;
+	ddev->notif.base.fn = sdtx_notifier;
+	ddev->notif.event.reg = SSAM_EVENT_REGISTRY_SAM;
+	ddev->notif.event.id.target_category = SSAM_SSH_TC_BAS;
+	ddev->notif.event.id.instance = 0;
+	ddev->notif.event.mask = SSAM_EVENT_MASK_NONE;
+	ddev->notif.event.flags = SSAM_EVENT_SEQUENCED;
+
+	init_waitqueue_head(&ddev->waitq);
+	mutex_init(&ddev->write_lock);
+	init_rwsem(&ddev->client_lock);
+	INIT_LIST_HEAD(&ddev->client_list);
+
+	INIT_DELAYED_WORK(&ddev->mode_work, sdtx_device_mode_workfn);
+	INIT_DELAYED_WORK(&ddev->state_work, sdtx_device_state_workfn);
+
+	/*
+	 * Get current device state. We want to guarantee that events are only
+	 * sent when state actually changes. Thus we cannot use special
+	 * "uninitialized" values, as that would cause problems when manually
+	 * querying the state in surface_dtx_pm_complete(). I.e. we would not
+	 * be able to detect state changes there if no change event has been
+	 * received between driver initialization and first device suspension.
+	 *
+	 * Note that we also need to do this before registring the event
+	 * notifier, as that may access the state values.
+	 */
+	status = ssam_retry(ssam_bas_get_base, ddev->ctrl, &ddev->state.base);
+	if (status)
+		return status;
+
+	status = ssam_retry(ssam_bas_get_device_mode, ddev->ctrl,
+			    &ddev->state.device_mode);
+	if (status)
+		return status;
+
+	status = ssam_retry(ssam_bas_get_latch_status, ddev->ctrl,
+			    &ddev->state.latch_status);
+	if (status)
+		return status;
+
+	// set up tablet mode switch
+	ddev->mode_switch = input_allocate_device();
+	if (!ddev->mode_switch)
+		return -ENOMEM;
+
+	ddev->mode_switch->name = "Microsoft Surface DTX Device Mode Switch";
+	ddev->mode_switch->phys = "ssam/01:11:01:00:00/input0";
+	ddev->mode_switch->id.bustype = BUS_HOST;
+	ddev->mode_switch->dev.parent = ddev->dev;
+
+	tablet_mode = (ddev->state.device_mode != SDTX_DEVICE_MODE_LAPTOP);
+	input_set_capability(ddev->mode_switch, EV_SW, SW_TABLET_MODE);
+	input_report_switch(ddev->mode_switch, SW_TABLET_MODE, tablet_mode);
+
+	status = input_register_device(ddev->mode_switch);
+	if (status) {
+		input_free_device(ddev->mode_switch);
+		return status;
+	}
+
+	// set up event notifier
+	status = ssam_notifier_register(ddev->ctrl, &ddev->notif);
+	if (status)
+		goto err_notif;
+
+	// register miscdevice
+	status = misc_register(&ddev->mdev);
+	if (status)
+		goto err_mdev;
+
+	/*
+	 * Update device state in case it has changed between getting the
+	 * initial mode and registring the event notifier.
+	 */
+	sdtx_update_device_state(ddev, 0);
+	return 0;
+
+err_notif:
+	ssam_notifier_unregister(ddev->ctrl, &ddev->notif);
+	cancel_delayed_work_sync(&ddev->mode_work);
+err_mdev:
+	input_unregister_device(ddev->mode_switch);
+	return status;
+}
+
+static struct sdtx_device *sdtx_device_setup(struct device *dev,
+					     struct ssam_controller *ctrl)
+{
+	struct sdtx_device *ddev;
+	int status;
+
+	ddev = kzalloc(sizeof(*ddev), GFP_KERNEL);
+	if (!ddev)
+		return ERR_PTR(-ENOMEM);
+
+	status = sdtx_device_init(ddev, dev, ctrl);
+	if (status) {
+		kfree(ddev);
+		return ERR_PTR(status);
+	}
+
+	return ddev;
+}
+
+static void sdtx_device_destroy(struct sdtx_device *ddev)
+{
+	struct sdtx_client *client;
+
+	// disable notifiers, prevent new events from arriving
+	ssam_notifier_unregister(ddev->ctrl, &ddev->notif);
+
+	// stop mode_work, prevent access to mode_switch
+	cancel_delayed_work_sync(&ddev->mode_work);
+
+	// stop state_work
+	cancel_delayed_work_sync(&ddev->state_work);
+
+	// with mode_work canceled, we can unregister the mode_switch
+	input_unregister_device(ddev->mode_switch);
+
+	/*
+	 * Mark device as shut-down. Prevent new clients from being added and
+	 * new operations from being executed.
+	 */
+	set_bit(SDTX_DEVICE_SHUTDOWN_BIT, &ddev->flags);
+
+	// wake up async clients
+	down_write(&ddev->client_lock);
+	list_for_each_entry(client, &ddev->client_list, node) {
+		kill_fasync(&client->fasync, SIGIO, POLL_HUP);
+	}
+	up_write(&ddev->client_lock);
+
+	// wake up blocking clients
+	wake_up_interruptible(&ddev->waitq);
+
+	/*
+	 * Wait for clients to finish their current operation. After this, the
+	 * controller and device references are guaranteed to be no longer in
+	 * use.
+	 */
+	down_write(&ddev->lock);
+	ddev->dev = NULL;
+	ddev->ctrl = NULL;
+	up_write(&ddev->lock);
+
+	// finally remove the misc-device
+	misc_deregister(&ddev->mdev);
+
+	/*
+	 * We're now guaranteed that sdtx_device_open() won't be called any
+	 * more, so we can now drop out reference.
+	 */
+	sdtx_device_put(ddev);
+}
+
+
+/* -- PM ops. --------------------------------------------------------------- */
+
+#ifdef CONFIG_PM_SLEEP
+
+static void surface_dtx_pm_complete(struct device *dev)
+{
+	struct sdtx_device *ddev = dev_get_drvdata(dev);
+
+	/*
+	 * Normally, the EC will store events while suspended (i.e. in
+	 * display-off state) and release them when resumed (i.e. transitioned
+	 * to display-on state). During hibernation, however, the EC will be
+	 * shut down and does not store events. Furthermore, events might be
+	 * dropped during prolonged suspension (it is scurrently unknown how
+	 * big this event buffer is and how it behaves on overruns).
+	 *
+	 * To prevent any problems, we update the device state here. We do
+	 * this delayed to ensure that any events sent by the EC directly
+	 * after resuming will be handled first. The delay below has been
+	 * chosen (experimentally), so that there should be ample time for
+	 * these events to be handled, before we check and, if necessary,
+	 * update the state.
+	 */
+	sdtx_update_device_state(ddev, msecs_to_jiffies(1000));
+}
+
+static const struct dev_pm_ops surface_dtx_pm_ops = {
+	.complete = surface_dtx_pm_complete,
+};
+
+#else /* CONFIG_PM_SLEEP */
+
+static const struct dev_pm_ops surface_dtx_pm_ops = {};
+
+#endif /* CONFIG_PM_SLEEP */
+
+
+/* -- Platform driver. ------------------------------------------------------ */
+
+static int surface_dtx_platform_probe(struct platform_device *pdev)
+{
+	struct ssam_controller *ctrl;
+	struct sdtx_device *ddev;
+	int status;
+
+	// link to EC
+	status = ssam_client_bind(&pdev->dev, &ctrl);
+	if (status)
+		return status == -ENXIO ? -EPROBE_DEFER : status;
+
+	ddev = sdtx_device_setup(&pdev->dev, ctrl);
+	if (IS_ERR(ddev))
+		return PTR_ERR(ddev);
+
+	platform_set_drvdata(pdev, ddev);
+	return 0;
+}
+
+static int surface_dtx_platform_remove(struct platform_device *pdev)
+{
+	sdtx_device_destroy(platform_get_drvdata(pdev));
+	return 0;
+}
+
+static const struct acpi_device_id surface_dtx_acpi_match[] = {
+	{ "MSHW0133", 0 },
+	{ },
+};
+MODULE_DEVICE_TABLE(acpi, surface_dtx_acpi_match);
+
+static struct platform_driver surface_dtx_platform_driver = {
+	.probe = surface_dtx_platform_probe,
+	.remove = surface_dtx_platform_remove,
+	.driver = {
+		.name = "surface_dtx_pltf",
+		.acpi_match_table = surface_dtx_acpi_match,
+		.pm = &surface_dtx_pm_ops,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+
+/* -- SSAM device driver. --------------------------------------------------- */
+
+#ifdef CONFIG_SURFACE_AGGREGATOR_BUS
+
+static int surface_dtx_ssam_probe(struct ssam_device *sdev)
+{
+	struct sdtx_device *ddev;
+
+	ddev = sdtx_device_setup(&sdev->dev, sdev->ctrl);
+	if (IS_ERR(ddev))
+		return PTR_ERR(ddev);
+
+	ssam_device_set_drvdata(sdev, ddev);
+	return 0;
+}
+
+static void surface_dtx_ssam_remove(struct ssam_device *sdev)
+{
+	sdtx_device_destroy(ssam_device_get_drvdata(sdev));
+}
+
+static const struct ssam_device_id surface_dtx_ssam_match[] = {
+	{ SSAM_SDEV(BAS, 0x01, 0x00, 0x00) },
+	{ },
+};
+MODULE_DEVICE_TABLE(ssam, surface_dtx_ssam_match);
+
+static struct ssam_device_driver surface_dtx_ssam_driver = {
+	.probe = surface_dtx_ssam_probe,
+	.remove = surface_dtx_ssam_remove,
+	.match_table = surface_dtx_ssam_match,
+	.driver = {
+		.name = "surface_dtx",
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+static int ssam_dtx_driver_register(void)
+{
+	return ssam_device_driver_register(&surface_dtx_ssam_driver);
+}
+
+static void ssam_dtx_driver_unregister(void)
+{
+	ssam_device_driver_unregister(&surface_dtx_ssam_driver);
+}
+
+#else /* CONFIG_SURFACE_AGGREGATOR_BUS */
+
+static int ssam_dtx_driver_register(void)
+{
+	return 0;
+}
+
+static void ssam_dtx_driver_unregister(void)
+{
+}
+
+#endif /* CONFIG_SURFACE_AGGREGATOR_BUS */
+
+
+/* -- Module setup. --------------------------------------------------------- */
+
+static int __init surface_dtx_init(void)
+{
+	int status;
+
+	status = ssam_dtx_driver_register();
+	if (status)
+		return status;
+
+	status = platform_driver_register(&surface_dtx_platform_driver);
+	if (status)
+		ssam_dtx_driver_unregister();
+
+	return status;
+}
+module_init(surface_dtx_init);
+
+static void __exit surface_dtx_exit(void)
+{
+	platform_driver_unregister(&surface_dtx_platform_driver);
+	ssam_dtx_driver_unregister();
+}
+module_exit(surface_dtx_exit);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Detachment-system driver for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/clients/surface_hid.c b/drivers/misc/surface_aggregator/clients/surface_hid.c
new file mode 100644
index 000000000000..033abc9aa95e
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_hid.c
@@ -0,0 +1,922 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface System Aggregator Module (SSAM) HID device driver.
+ *
+ * Provides support for HID input devices connected via the Surface System
+ * Aggregator Module.
+ *
+ * Copyright (C) 2019-2020 Blaž Hrastnik <blaz@mxxn.io>,
+ *                         Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <asm/unaligned.h>
+#include <linux/acpi.h>
+#include <linux/hid.h>
+#include <linux/input.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/types.h>
+#include <linux/usb/ch9.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/device.h>
+
+
+enum surface_hid_descriptor_entry {
+	SURFACE_HID_DESC_HID    = 0,
+	SURFACE_HID_DESC_REPORT = 1,
+	SURFACE_HID_DESC_ATTRS  = 2,
+};
+
+struct surface_hid_descriptor {
+	__u8 desc_len;			// = 9
+	__u8 desc_type;			// = HID_DT_HID
+	__le16 hid_version;
+	__u8 country_code;
+	__u8 num_descriptors;		// = 1
+
+	__u8 report_desc_type;		// = HID_DT_REPORT
+	__le16 report_desc_len;
+} __packed;
+
+static_assert(sizeof(struct surface_hid_descriptor) == 9);
+
+struct surface_hid_attributes {
+	__le32 length;
+	__le16 vendor;
+	__le16 product;
+	__le16 version;
+	__u8 _unknown[22];
+} __packed;
+
+static_assert(sizeof(struct surface_hid_attributes) == 32);
+
+struct surface_hid_device;
+
+struct surface_hid_device_ops {
+	int (*get_descriptor)(struct surface_hid_device *shid, u8 entry,
+			      u8 *buf, size_t len);
+	int (*output_report)(struct surface_hid_device *shid, u8 report_id,
+			     u8 *data, size_t len);
+	int (*get_feature_report)(struct surface_hid_device *shid, u8 report_id,
+				  u8 *data, size_t len);
+	int (*set_feature_report)(struct surface_hid_device *shid, u8 report_id,
+				  u8 *data, size_t len);
+};
+
+struct surface_hid_device {
+	struct device *dev;
+	struct ssam_controller *ctrl;
+	struct ssam_device_uid uid;
+
+	struct surface_hid_descriptor hid_desc;
+	struct surface_hid_attributes attrs;
+
+	struct ssam_event_notifier notif;
+	struct hid_device *hid;
+
+	struct surface_hid_device_ops ops;
+};
+
+
+/* -- SAM interface (HID). -------------------------------------------------- */
+
+#ifdef CONFIG_SURFACE_AGGREGATOR_BUS
+
+struct surface_hid_buffer_slice {
+	__u8 entry;
+	__le32 offset;
+	__le32 length;
+	__u8 end;
+	__u8 data[];
+} __packed;
+
+static_assert(sizeof(struct surface_hid_buffer_slice) == 10);
+
+enum surface_hid_cid {
+	SURFACE_HID_CID_OUTPUT_REPORT      = 0x01,
+	SURFACE_HID_CID_GET_FEATURE_REPORT = 0x02,
+	SURFACE_HID_CID_SET_FEATURE_REPORT = 0x03,
+	SURFACE_HID_CID_GET_DESCRIPTOR     = 0x04,
+};
+
+static int ssam_hid_get_descriptor(struct surface_hid_device *shid, u8 entry,
+				   u8 *buf, size_t len)
+{
+	u8 buffer[sizeof(struct surface_hid_buffer_slice) + 0x76];
+	struct surface_hid_buffer_slice *slice;
+	struct ssam_request rqst;
+	struct ssam_response rsp;
+	u32 buffer_len, offset, length;
+	int status;
+
+	/*
+	 * Note: The 0x76 above has been chosen because that's what's used by
+	 * the Windows driver. Together with the header, this leads to a 128
+	 * byte payload in total.
+	 */
+
+	buffer_len = ARRAY_SIZE(buffer) - sizeof(struct surface_hid_buffer_slice);
+
+	rqst.target_category = shid->uid.category;
+	rqst.target_id = shid->uid.target;
+	rqst.command_id = SURFACE_HID_CID_GET_DESCRIPTOR;
+	rqst.instance_id = shid->uid.instance;
+	rqst.flags = SSAM_REQUEST_HAS_RESPONSE;
+	rqst.length = sizeof(struct surface_hid_buffer_slice);
+	rqst.payload = buffer;
+
+	rsp.capacity = ARRAY_SIZE(buffer);
+	rsp.pointer = buffer;
+
+	slice = (struct surface_hid_buffer_slice *)buffer;
+	slice->entry = entry;
+	slice->end = 0;
+
+	offset = 0;
+	length = buffer_len;
+
+	while (!slice->end && offset < len) {
+		put_unaligned_le32(offset, &slice->offset);
+		put_unaligned_le32(length, &slice->length);
+
+		rsp.length = 0;
+
+		status = ssam_retry(ssam_request_sync_onstack, shid->ctrl,
+				    &rqst, &rsp, sizeof(*slice));
+		if (status)
+			return status;
+
+		offset = get_unaligned_le32(&slice->offset);
+		length = get_unaligned_le32(&slice->length);
+
+		// don't mess stuff up in case we receive garbage
+		if (length > buffer_len || offset > len)
+			return -EPROTO;
+
+		if (offset + length > len)
+			length = len - offset;
+
+		memcpy(buf + offset, &slice->data[0], length);
+
+		offset += length;
+		length = buffer_len;
+	}
+
+	if (offset != len) {
+		dev_err(shid->dev, "unexpected descriptor length: got %u, "
+			"expected %zu\n", offset, len);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+static int ssam_hid_set_raw_report(struct surface_hid_device *shid,
+				   u8 report_id, bool feature, u8 *buf,
+				   size_t len)
+{
+	struct ssam_request rqst;
+	u8 cid;
+
+	if (feature)
+		cid = SURFACE_HID_CID_SET_FEATURE_REPORT;
+	else
+		cid = SURFACE_HID_CID_OUTPUT_REPORT;
+
+	rqst.target_category = shid->uid.category;
+	rqst.target_id = shid->uid.target;
+	rqst.instance_id = shid->uid.instance;
+	rqst.command_id = cid;
+	rqst.flags = 0;
+	rqst.length = len;
+	rqst.payload = buf;
+
+	buf[0] = report_id;
+
+	return ssam_retry(ssam_request_sync, shid->ctrl, &rqst, NULL);
+}
+
+static int ssam_hid_get_raw_report(struct surface_hid_device *shid,
+				   u8 report_id, u8 *buf, size_t len)
+{
+	struct ssam_request rqst;
+	struct ssam_response rsp;
+
+	rqst.target_category = shid->uid.category;
+	rqst.target_id = shid->uid.target;
+	rqst.instance_id = shid->uid.instance;
+	rqst.command_id = SURFACE_HID_CID_GET_FEATURE_REPORT;
+	rqst.flags = 0;
+	rqst.length = sizeof(report_id);
+	rqst.payload = &report_id;
+
+	rsp.capacity = len;
+	rsp.length = 0;
+	rsp.pointer = buf;
+
+	return ssam_retry(ssam_request_sync_onstack, shid->ctrl, &rqst, &rsp,
+			  sizeof(report_id));
+}
+
+static u32 ssam_hid_event_fn(struct ssam_event_notifier *nf,
+			     const struct ssam_event *event)
+{
+	struct surface_hid_device *shid;
+	int status;
+
+	shid = container_of(nf, struct surface_hid_device, notif);
+
+	if (event->command_id != 0x00)
+		return 0;
+
+	status = hid_input_report(shid->hid, HID_INPUT_REPORT,
+				  (u8 *)&event->data[0], event->length, 0);
+
+	return ssam_notifier_from_errno(status) | SSAM_NOTIF_HANDLED;
+}
+
+
+/* -- Transport driver (HID). ----------------------------------------------- */
+
+static int shid_output_report(struct surface_hid_device *shid, u8 report_id,
+			      u8 *data, size_t len)
+{
+	int status;
+
+	status =  ssam_hid_set_raw_report(shid, report_id, false, data, len);
+	return status >= 0 ? len : status;
+}
+
+static int shid_get_feature_report(struct surface_hid_device *shid,
+				   u8 report_id, u8 *data, size_t len)
+{
+	int status;
+
+	status = ssam_hid_get_raw_report(shid, report_id, data, len);
+	return status >= 0 ? len : status;
+}
+
+static int shid_set_feature_report(struct surface_hid_device *shid,
+				   u8 report_id, u8 *data, size_t len)
+{
+	int status;
+
+	status =  ssam_hid_set_raw_report(shid, report_id, true, data, len);
+	return status >= 0 ? len : status;
+}
+
+#endif /* CONFIG_SURFACE_AGGREGATOR_BUS */
+
+
+/* -- SAM interface (KBD). -------------------------------------------------- */
+
+#define KBD_FEATURE_REPORT_SIZE		7  // 6 + report ID
+
+enum surface_kbd_cid {
+	SURFACE_KBD_CID_GET_DESCRIPTOR     = 0x00,
+	SURFACE_KBD_CID_SET_CAPSLOCK_LED   = 0x01,
+	SURFACE_KBD_CID_EVT_INPUT_GENERIC  = 0x03,
+	SURFACE_KBD_CID_EVT_INPUT_HOTKEYS  = 0x04,
+	SURFACE_KBD_CID_GET_FEATURE_REPORT = 0x0b,
+};
+
+static int ssam_kbd_get_descriptor(struct surface_hid_device *shid, u8 entry,
+				   u8 *buf, size_t len)
+{
+	struct ssam_request rqst;
+	struct ssam_response rsp;
+	int status;
+
+	rqst.target_category = shid->uid.category;
+	rqst.target_id = shid->uid.target;
+	rqst.command_id = SURFACE_KBD_CID_GET_DESCRIPTOR;
+	rqst.instance_id = shid->uid.instance;
+	rqst.flags = SSAM_REQUEST_HAS_RESPONSE;
+	rqst.length = sizeof(entry);
+	rqst.payload = &entry;
+
+	rsp.capacity = len;
+	rsp.length = 0;
+	rsp.pointer = buf;
+
+	status = ssam_retry(ssam_request_sync_onstack, shid->ctrl, &rqst, &rsp,
+			    sizeof(entry));
+	if (status)
+		return status;
+
+	if (rsp.length != len) {
+		dev_err(shid->dev, "invalid descriptor length: got %zu, "
+			"expected, %zu\n", rsp.length, len);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+static int ssam_kbd_set_caps_led(struct surface_hid_device *shid, bool value)
+{
+	struct ssam_request rqst;
+	u8 value_u8 = value;
+
+	rqst.target_category = shid->uid.category;
+	rqst.target_id = shid->uid.target;
+	rqst.command_id = SURFACE_KBD_CID_SET_CAPSLOCK_LED;
+	rqst.instance_id = shid->uid.instance;
+	rqst.flags = 0;
+	rqst.length = sizeof(value_u8);
+	rqst.payload = &value_u8;
+
+	return ssam_retry(ssam_request_sync_onstack, shid->ctrl, &rqst, NULL,
+			  sizeof(value_u8));
+}
+
+static int ssam_kbd_get_feature_report(struct surface_hid_device *shid, u8 *buf,
+				       size_t len)
+{
+	struct ssam_request rqst;
+	struct ssam_response rsp;
+	u8 payload = 0;
+	int status;
+
+	rqst.target_category = shid->uid.category;
+	rqst.target_id = shid->uid.target;
+	rqst.command_id = SURFACE_KBD_CID_GET_FEATURE_REPORT;
+	rqst.instance_id = shid->uid.instance;
+	rqst.flags = SSAM_REQUEST_HAS_RESPONSE;
+	rqst.length = sizeof(payload);
+	rqst.payload = &payload;
+
+	rsp.capacity = len;
+	rsp.length = 0;
+	rsp.pointer = buf;
+
+	status = ssam_retry(ssam_request_sync_onstack, shid->ctrl, &rqst, &rsp,
+			    sizeof(payload));
+	if (status)
+		return status;
+
+	if (rsp.length != len) {
+		dev_err(shid->dev, "invalid feature report length: got %zu, "
+			"expected, %zu\n", rsp.length, len);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+static bool ssam_kbd_is_input_event(const struct ssam_event *event)
+{
+	if (event->command_id == SURFACE_KBD_CID_EVT_INPUT_GENERIC)
+		return true;
+
+	if (event->command_id == SURFACE_KBD_CID_EVT_INPUT_HOTKEYS)
+		return true;
+
+	return false;
+}
+
+static u32 ssam_kbd_event_fn(struct ssam_event_notifier *nf,
+				const struct ssam_event *event)
+{
+	struct surface_hid_device *shid;
+	int status;
+
+	shid = container_of(nf, struct surface_hid_device, notif);
+
+	/*
+	 * Check against device UID manually, as registry and device target
+	 * category doesn't line up.
+	 */
+
+	if (shid->uid.category != event->target_category)
+		return 0;
+
+	if (shid->uid.target != event->target_id)
+		return 0;
+
+	if (shid->uid.instance != event->instance_id)
+		return 0;
+
+	if (!ssam_kbd_is_input_event(event))
+		return 0;
+
+	status = hid_input_report(shid->hid, HID_INPUT_REPORT,
+				  (u8 *)&event->data[0], event->length, 0);
+
+	return ssam_notifier_from_errno(status) | SSAM_NOTIF_HANDLED;
+}
+
+
+/* -- Transport driver (KBD). ----------------------------------------------- */
+
+static int skbd_get_caps_led_value(struct hid_device *hid, u8 report_id,
+				   u8 *data, size_t len)
+{
+	struct hid_field *field;
+	unsigned int offset, size;
+	int i;
+
+	// get led field
+	field = hidinput_get_led_field(hid);
+	if (!field)
+		return -ENOENT;
+
+	// check if we got the correct report
+	if (len != hid_report_len(field->report))
+		return -ENOENT;
+
+	if (report_id != field->report->id)
+		return -ENOENT;
+
+	// get caps lock led index
+	for (i = 0; i < field->report_count; i++)
+		if ((field->usage[i].hid & 0xffff) == 0x02)
+			break;
+
+	if (i == field->report_count)
+		return -ENOENT;
+
+	// extract value
+	size = field->report_size;
+	offset = field->report_offset + i * size;
+	return !!hid_field_extract(hid, data + 1, size, offset);
+}
+
+static int skbd_output_report(struct surface_hid_device *shid, u8 report_id,
+			      u8 *data, size_t len)
+{
+	int caps_led;
+	int status;
+
+	caps_led = skbd_get_caps_led_value(shid->hid, report_id, data, len);
+	if (caps_led < 0)
+		return -EIO;	// only caps output reports are supported
+
+	status = ssam_kbd_set_caps_led(shid, caps_led);
+	if (status < 0)
+		return status;
+
+	return len;
+}
+
+static int skbd_get_feature_report(struct surface_hid_device *shid,
+				   u8 report_id, u8 *data, size_t len)
+{
+	u8 report[KBD_FEATURE_REPORT_SIZE];
+	int status;
+
+	/*
+	 * The keyboard only has a single hard-coded read-only feature report
+	 * of size KBD_FEATURE_REPORT_SIZE. Try to load it and compare its
+	 * report ID against the requested one.
+	 */
+
+	if (len < ARRAY_SIZE(report))
+		return -ENOSPC;
+
+	status = ssam_kbd_get_feature_report(shid, report, ARRAY_SIZE(report));
+	if (status < 0)
+		return status;
+
+	if (report_id != report[0])
+		return -ENOENT;
+
+	memcpy(data, report, ARRAY_SIZE(report));
+	return len;
+}
+
+static int skbd_set_feature_report(struct surface_hid_device *shid,
+				   u8 report_id, u8 *data, size_t len)
+{
+	return -EIO;
+}
+
+
+/* -- Device descriptor access. --------------------------------------------- */
+
+static int surface_hid_load_hid_descriptor(struct surface_hid_device *shid)
+{
+	int status;
+
+	status = shid->ops.get_descriptor(shid, SURFACE_HID_DESC_HID,
+			(u8 *)&shid->hid_desc, sizeof(shid->hid_desc));
+	if (status)
+		return status;
+
+	if (shid->hid_desc.desc_len != sizeof(shid->hid_desc)) {
+		dev_err(shid->dev, "unexpected HID descriptor length: got %u, "
+			"expected %zu\n", shid->hid_desc.desc_len,
+			sizeof(shid->hid_desc));
+		return -EPROTO;
+	}
+
+	if (shid->hid_desc.desc_type != HID_DT_HID) {
+		dev_err(shid->dev, "unexpected HID descriptor type: got 0x%x, "
+			"expected 0x%x\n", shid->hid_desc.desc_type,
+			HID_DT_HID);
+		return -EPROTO;
+	}
+
+	if (shid->hid_desc.num_descriptors != 1) {
+		dev_err(shid->dev, "unexpected number of descriptors: got %u, "
+			"expected 1\n", shid->hid_desc.num_descriptors);
+		return -EPROTO;
+	}
+
+	if (shid->hid_desc.report_desc_type != HID_DT_REPORT) {
+		dev_err(shid->dev, "unexpected report descriptor type: got 0x%x, "
+			"expected 0x%x\n", shid->hid_desc.report_desc_type,
+			HID_DT_REPORT);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+static int surface_hid_load_device_attributes(struct surface_hid_device *shid)
+{
+	int status;
+
+	status = shid->ops.get_descriptor(shid, SURFACE_HID_DESC_ATTRS,
+			(u8 *)&shid->attrs, sizeof(shid->attrs));
+	if (status)
+		return status;
+
+	if (get_unaligned_le32(&shid->attrs.length) != sizeof(shid->attrs)) {
+		dev_err(shid->dev, "unexpected attribute length: got %u, "
+			"expected %zu\n", get_unaligned_le32(&shid->attrs.length),
+			sizeof(shid->attrs));
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+
+/* -- Transport driver (common). -------------------------------------------- */
+
+static int surface_hid_start(struct hid_device *hid)
+{
+	struct surface_hid_device *shid = hid->driver_data;
+
+	return ssam_notifier_register(shid->ctrl, &shid->notif);
+}
+
+static void surface_hid_stop(struct hid_device *hid)
+{
+	struct surface_hid_device *shid = hid->driver_data;
+
+	// Note: This call will log errors for us, so ignore them here.
+	ssam_notifier_unregister(shid->ctrl, &shid->notif);
+}
+
+static int surface_hid_open(struct hid_device *hid)
+{
+	return 0;
+}
+
+static void surface_hid_close(struct hid_device *hid)
+{
+}
+
+static int surface_hid_parse(struct hid_device *hid)
+{
+	struct surface_hid_device *shid = hid->driver_data;
+	size_t len = get_unaligned_le16(&shid->hid_desc.report_desc_len);
+	u8 *buf;
+	int status;
+
+	buf = kzalloc(len, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	status = shid->ops.get_descriptor(shid, SURFACE_HID_DESC_REPORT, buf, len);
+	if (!status)
+		status = hid_parse_report(hid, buf, len);
+
+	kfree(buf);
+	return status;
+}
+
+static int surface_hid_raw_request(struct hid_device *hid,
+		unsigned char reportnum, u8 *buf, size_t len,
+		unsigned char rtype, int reqtype)
+{
+	struct surface_hid_device *shid = hid->driver_data;
+
+	if (rtype == HID_OUTPUT_REPORT && reqtype == HID_REQ_SET_REPORT)
+		return shid->ops.output_report(shid, reportnum, buf, len);
+
+	else if (rtype == HID_FEATURE_REPORT && reqtype == HID_REQ_GET_REPORT)
+		return shid->ops.get_feature_report(shid, reportnum, buf, len);
+
+	else if (rtype == HID_FEATURE_REPORT && reqtype == HID_REQ_SET_REPORT)
+		return shid->ops.set_feature_report(shid, reportnum, buf, len);
+
+	return -EIO;
+}
+
+static struct hid_ll_driver surface_hid_ll_driver = {
+	.start       = surface_hid_start,
+	.stop        = surface_hid_stop,
+	.open        = surface_hid_open,
+	.close       = surface_hid_close,
+	.parse       = surface_hid_parse,
+	.raw_request = surface_hid_raw_request,
+};
+
+
+/* -- Common device setup. -------------------------------------------------- */
+
+static int surface_hid_device_add(struct surface_hid_device *shid)
+{
+	int status;
+
+	status = surface_hid_load_hid_descriptor(shid);
+	if (status)
+		return status;
+
+	status = surface_hid_load_device_attributes(shid);
+	if (status)
+		return status;
+
+	shid->hid = hid_allocate_device();
+	if (IS_ERR(shid->hid))
+		return PTR_ERR(shid->hid);
+
+	shid->hid->dev.parent = shid->dev;
+	shid->hid->bus = BUS_HOST;		// TODO: BUS_SURFACE
+	shid->hid->vendor = cpu_to_le16(shid->attrs.vendor);
+	shid->hid->product = cpu_to_le16(shid->attrs.product);
+	shid->hid->version = cpu_to_le16(shid->hid_desc.hid_version);
+	shid->hid->country = shid->hid_desc.country_code;
+
+	snprintf(shid->hid->name, sizeof(shid->hid->name),
+		 "Microsoft Surface %04X:%04X",
+		 shid->hid->vendor, shid->hid->product);
+
+	strlcpy(shid->hid->phys, dev_name(shid->dev), sizeof(shid->hid->phys));
+
+	shid->hid->driver_data = shid;
+	shid->hid->ll_driver = &surface_hid_ll_driver;
+
+	status = hid_add_device(shid->hid);
+	if (status)
+		hid_destroy_device(shid->hid);
+
+	return status;
+}
+
+static void surface_hid_device_destroy(struct surface_hid_device *shid)
+{
+	hid_destroy_device(shid->hid);
+}
+
+
+/* -- PM ops. --------------------------------------------------------------- */
+
+#ifdef CONFIG_PM_SLEEP
+
+static int surface_hid_suspend(struct device *dev)
+{
+	struct surface_hid_device *d = dev_get_drvdata(dev);
+
+	if (d->hid->driver && d->hid->driver->suspend)
+		return d->hid->driver->suspend(d->hid, PMSG_SUSPEND);
+
+	return 0;
+}
+
+static int surface_hid_resume(struct device *dev)
+{
+	struct surface_hid_device *d = dev_get_drvdata(dev);
+
+	if (d->hid->driver && d->hid->driver->resume)
+		return d->hid->driver->resume(d->hid);
+
+	return 0;
+}
+
+static int surface_hid_freeze(struct device *dev)
+{
+	struct surface_hid_device *d = dev_get_drvdata(dev);
+
+	if (d->hid->driver && d->hid->driver->suspend)
+		return d->hid->driver->suspend(d->hid, PMSG_FREEZE);
+
+	return 0;
+}
+
+static int surface_hid_poweroff(struct device *dev)
+{
+	struct surface_hid_device *d = dev_get_drvdata(dev);
+
+	if (d->hid->driver && d->hid->driver->suspend)
+		return d->hid->driver->suspend(d->hid, PMSG_HIBERNATE);
+
+	return 0;
+}
+
+static int surface_hid_restore(struct device *dev)
+{
+	struct surface_hid_device *d = dev_get_drvdata(dev);
+
+	if (d->hid->driver && d->hid->driver->reset_resume)
+		return d->hid->driver->reset_resume(d->hid);
+
+	return 0;
+}
+
+const struct dev_pm_ops surface_hid_pm_ops = {
+	.freeze   = surface_hid_freeze,
+	.thaw     = surface_hid_resume,
+	.suspend  = surface_hid_suspend,
+	.resume   = surface_hid_resume,
+	.poweroff = surface_hid_poweroff,
+	.restore  = surface_hid_restore,
+};
+
+#else /* CONFIG_PM_SLEEP */
+
+const struct dev_pm_ops surface_hid_pm_ops = { };
+
+#endif /* CONFIG_PM_SLEEP */
+
+
+/* -- Driver setup (HID). --------------------------------------------------- */
+
+#ifdef CONFIG_SURFACE_AGGREGATOR_BUS
+
+static int surface_hid_probe(struct ssam_device *sdev)
+{
+	struct surface_hid_device *shid;
+
+	shid = devm_kzalloc(&sdev->dev, sizeof(*shid), GFP_KERNEL);
+	if (!shid)
+		return -ENOMEM;
+
+	shid->dev = &sdev->dev;
+	shid->ctrl = sdev->ctrl;
+	shid->uid = sdev->uid;
+
+	shid->notif.base.priority = 1;
+	shid->notif.base.fn = ssam_hid_event_fn;
+	shid->notif.event.reg = SSAM_EVENT_REGISTRY_REG,
+	shid->notif.event.id.target_category = sdev->uid.category;
+	shid->notif.event.id.instance = sdev->uid.instance;
+	shid->notif.event.mask = SSAM_EVENT_MASK_STRICT;
+	shid->notif.event.flags = 0;
+
+	shid->ops.get_descriptor = ssam_hid_get_descriptor;
+	shid->ops.output_report = shid_output_report;
+	shid->ops.get_feature_report = shid_get_feature_report;
+	shid->ops.set_feature_report = shid_set_feature_report;
+
+	ssam_device_set_drvdata(sdev, shid);
+	return surface_hid_device_add(shid);
+}
+
+static void surface_hid_remove(struct ssam_device *sdev)
+{
+	surface_hid_device_destroy(ssam_device_get_drvdata(sdev));
+}
+
+static const struct ssam_device_id surface_hid_match[] = {
+	{ SSAM_SDEV(HID, 0x02, SSAM_ANY_IID, 0x00) },
+	{ },
+};
+MODULE_DEVICE_TABLE(ssam, surface_hid_match);
+
+static struct ssam_device_driver surface_hid_driver = {
+	.probe = surface_hid_probe,
+	.remove = surface_hid_remove,
+	.match_table = surface_hid_match,
+	.driver = {
+		.name = "surface_hid",
+		.pm = &surface_hid_pm_ops,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+static int surface_hid_driver_register(void)
+{
+	return ssam_device_driver_register(&surface_hid_driver);
+}
+
+static void surface_hid_driver_unregister(void)
+{
+	ssam_device_driver_unregister(&surface_hid_driver);
+}
+
+#else /* CONFIG_SURFACE_AGGREGATOR_BUS */
+
+static int surface_hid_driver_register(void)
+{
+	return 0;
+}
+
+static void surface_hid_driver_unregister(void)
+{
+}
+
+#endif /* CONFIG_SURFACE_AGGREGATOR_BUS */
+
+
+/* -- Driver setup (KBD). --------------------------------------------------- */
+
+static int surface_kbd_probe(struct platform_device *pdev)
+{
+	struct ssam_controller *ctrl;
+	struct surface_hid_device *shid;
+	int status;
+
+	// add device link to EC
+	status = ssam_client_bind(&pdev->dev, &ctrl);
+	if (status)
+		return status == -ENXIO ? -EPROBE_DEFER : status;
+
+	shid = devm_kzalloc(&pdev->dev, sizeof(*shid), GFP_KERNEL);
+	if (!shid)
+		return -ENOMEM;
+
+	shid->dev = &pdev->dev;
+	shid->ctrl = ctrl;
+
+	shid->uid.domain = SSAM_DOMAIN_SERIALHUB;
+	shid->uid.category = SSAM_SSH_TC_KBD;
+	shid->uid.target = 2;
+	shid->uid.instance = 0;
+	shid->uid.function = 0;
+
+	shid->notif.base.priority = 1;
+	shid->notif.base.fn = ssam_kbd_event_fn;
+	shid->notif.event.reg = SSAM_EVENT_REGISTRY_SAM;
+	shid->notif.event.id.target_category = shid->uid.category;
+	shid->notif.event.id.instance = shid->uid.instance;
+	shid->notif.event.mask = SSAM_EVENT_MASK_NONE;
+	shid->notif.event.flags = 0;
+
+	shid->ops.get_descriptor = ssam_kbd_get_descriptor;
+	shid->ops.output_report = skbd_output_report;
+	shid->ops.get_feature_report = skbd_get_feature_report;
+	shid->ops.set_feature_report = skbd_set_feature_report;
+
+	platform_set_drvdata(pdev, shid);
+	return surface_hid_device_add(shid);
+}
+
+static int surface_kbd_remove(struct platform_device *pdev)
+{
+	surface_hid_device_destroy(platform_get_drvdata(pdev));
+	return 0;
+}
+
+static const struct acpi_device_id surface_kbd_match[] = {
+	{ "MSHW0096" },
+	{ },
+};
+MODULE_DEVICE_TABLE(acpi, surface_kbd_match);
+
+static struct platform_driver surface_kbd_driver = {
+	.probe = surface_kbd_probe,
+	.remove = surface_kbd_remove,
+	.driver = {
+		.name = "surface_keyboard",
+		.acpi_match_table = surface_kbd_match,
+		.pm = &surface_hid_pm_ops,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+
+/* -- Module setup. --------------------------------------------------------- */
+
+static int __init surface_hid_init(void)
+{
+	int status;
+
+	status = surface_hid_driver_register();
+	if (status)
+		return status;
+
+	status = platform_driver_register(&surface_kbd_driver);
+	if (status)
+		surface_hid_driver_unregister();
+
+	return status;
+}
+module_init(surface_hid_init);
+
+static void __exit surface_hid_exit(void)
+{
+	platform_driver_unregister(&surface_kbd_driver);
+	surface_hid_driver_unregister();
+}
+module_exit(surface_hid_exit);
+
+MODULE_AUTHOR("Blaž Hrastnik <blaz@mxxn.io>");
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("HID transport-/device-driver for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/clients/surface_hotplug.c b/drivers/misc/surface_aggregator/clients/surface_hotplug.c
new file mode 100644
index 000000000000..9afddfc6a358
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_hotplug.c
@@ -0,0 +1,269 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface Book (gen. 2 and later) hot-plug driver.
+ *
+ * Surface Book devices (can) have a hot-pluggable discrete GPU (dGPU). This
+ * driver is responsible for out-of-band hot-plug event signalling on these
+ * devices. It is specifically required when the hot-plug device is in D3cold
+ * and can thus not generate PCIe hot-plug events itself.
+ *
+ * Event signalling is handled via ACPI, which will generate the appropriate
+ * device-check notifications to be picked up by the PCIe hot-plug driver.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/acpi.h>
+#include <linux/gpio.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/platform_device.h>
+
+static const struct acpi_gpio_params shps_base_presence_int   = { 0, 0, false };
+static const struct acpi_gpio_params shps_base_presence       = { 1, 0, false };
+static const struct acpi_gpio_params shps_device_power_int    = { 2, 0, false };
+static const struct acpi_gpio_params shps_device_power        = { 3, 0, false };
+static const struct acpi_gpio_params shps_device_presence_int = { 4, 0, false };
+static const struct acpi_gpio_params shps_device_presence     = { 5, 0, false };
+
+static const struct acpi_gpio_mapping shps_acpi_gpios[] = {
+	{ "base_presence-int-gpio",   &shps_base_presence_int,   1 },
+	{ "base_presence-gpio",       &shps_base_presence,       1 },
+	{ "device_power-int-gpio",    &shps_device_power_int,    1 },
+	{ "device_power-gpio",        &shps_device_power,        1 },
+	{ "device_presence-int-gpio", &shps_device_presence_int, 1 },
+	{ "device_presence-gpio",     &shps_device_presence,     1 },
+	{ },
+};
+
+static const guid_t shps_dsm_guid =
+	GUID_INIT(0x5515a847, 0xed55, 0x4b27, 0x83, 0x52, 0xcd,
+		  0x32, 0x0e, 0x10, 0x36, 0x0a);
+
+#define SHPS_DSM_REVISION		1
+
+enum shps_dsm_fn {
+	SHPS_DSM_FN_PCI_NUM_ENTRIES	= 0x01,
+	SHPS_DSM_FN_PCI_GET_ENTRIES	= 0x02,
+	SHPS_DSM_FN_IRQ_BASE_PRESENCE	= 0x03,
+	SHPS_DSM_FN_IRQ_DEVICE_POWER	= 0x04,
+	SHPS_DSM_FN_IRQ_DEVICE_PRESENCE	= 0x05,
+};
+
+enum shps_irq_type {
+	/* NOTE: Must be in order of DSM function */
+	SHPS_IRQ_TYPE_BASE_PRESENCE	= 0,
+	SHPS_IRQ_TYPE_DEVICE_POWER	= 1,
+	SHPS_IRQ_TYPE_DEVICE_PRESENCE	= 2,
+
+	SHPS_NUM_IRQS,
+};
+
+static const char *const shps_gpio_names[] = {
+	[SHPS_IRQ_TYPE_BASE_PRESENCE]	= "base_presence",
+	[SHPS_IRQ_TYPE_DEVICE_POWER]	= "device_power",
+	[SHPS_IRQ_TYPE_DEVICE_PRESENCE]	= "device_presence",
+};
+
+struct shps_device {
+	struct mutex lock[SHPS_NUM_IRQS];
+	struct gpio_desc *gpio[SHPS_NUM_IRQS];
+	unsigned int irq[SHPS_NUM_IRQS];
+};
+
+#define SHPS_IRQ_NOT_PRESENT		((unsigned int)-1)
+
+static void shps_dsm_notify_irq(struct platform_device *pdev,
+				enum shps_irq_type type)
+{
+	struct shps_device *sdev = platform_get_drvdata(pdev);
+	acpi_handle handle = ACPI_HANDLE(&pdev->dev);
+	union acpi_object *result;
+	union acpi_object param;
+	int value;
+
+	mutex_lock(&sdev->lock[type]);
+
+	value = gpiod_get_value_cansleep(sdev->gpio[type]);
+	if (value < 0) {
+		mutex_unlock(&sdev->lock[type]);
+		dev_err(&pdev->dev, "failed to get gpio: %d (irq=%d)\n",
+			type, value);
+		return;
+	}
+
+	dev_dbg(&pdev->dev, "IRQ notification via DSM (irq=%d, value=%d)\n",
+		type, value);
+
+	param.type = ACPI_TYPE_INTEGER;
+	param.integer.value = value;
+
+	result = acpi_evaluate_dsm(handle, &shps_dsm_guid, SHPS_DSM_REVISION,
+			SHPS_DSM_FN_IRQ_BASE_PRESENCE + type, &param);
+
+	if (!result) {
+		mutex_unlock(&sdev->lock[type]);
+		dev_err(&pdev->dev, "IRQ notification via DSM failed"
+			" (irq=%d, gpio=%d)\n", type, value);
+		return;
+	}
+
+	if (result->type != ACPI_TYPE_BUFFER) {
+		dev_err(&pdev->dev, "IRQ notification via DSM failed:"
+			" unexpected result type (irq=%d, gpio=%d)\n",
+			type, value);
+	}
+
+	if (result->buffer.length != 1 || result->buffer.pointer[0] != 0) {
+		dev_err(&pdev->dev, "IRQ notification via DSM failed:"
+			"unexpected result value (irq=%d, gpio=%d)\n",
+			type, value);
+	}
+
+	mutex_unlock(&sdev->lock[type]);
+	ACPI_FREE(result);
+}
+
+static irqreturn_t shps_handle_irq(int irq, void *data)
+{
+	struct platform_device *pdev = data;
+	struct shps_device *sdev = platform_get_drvdata(pdev);
+	int type;
+
+	// figure out which IRQ we're handling
+	for (type = 0; type < SHPS_NUM_IRQS; type++)
+		if (irq == sdev->irq[type])
+			break;
+
+	// we should have found our interrupt, if not: this is a bug
+	if (WARN(type >= SHPS_NUM_IRQS, "invalid IRQ number: %d\n", irq))
+		return IRQ_HANDLED;
+
+	// forward interrupt to ACPI via DSM
+	shps_dsm_notify_irq(pdev, type);
+	return IRQ_HANDLED;
+}
+
+static int shps_setup_irq(struct platform_device *pdev, enum shps_irq_type type)
+{
+	struct shps_device *sdev = platform_get_drvdata(pdev);
+	struct gpio_desc *gpiod;
+	acpi_handle handle = ACPI_HANDLE(&pdev->dev);
+	const char *irq_name;
+	const int dsm = SHPS_DSM_FN_IRQ_BASE_PRESENCE + type;
+	int status, irq;
+
+	// initialize as "not present"
+	sdev->gpio[type] = NULL;
+	sdev->irq[type] = SHPS_IRQ_NOT_PRESENT;
+
+	// only set up interrupts that we actually need
+	if (!acpi_check_dsm(handle, &shps_dsm_guid, SHPS_DSM_REVISION, BIT(dsm))) {
+		dev_dbg(&pdev->dev, "IRQ notification via DSM not present (irq=%d)\n",
+			type);
+		return 0;
+	}
+
+	gpiod = devm_gpiod_get(&pdev->dev, shps_gpio_names[type], GPIOD_ASIS);
+	if (IS_ERR(gpiod))
+		return PTR_ERR(gpiod);
+
+	irq = gpiod_to_irq(gpiod);
+	if (irq < 0)
+		return irq;
+
+	irq_name = devm_kasprintf(&pdev->dev, GFP_KERNEL, "shps-irq-%d", type);
+	if (!irq_name)
+		return -ENOMEM;
+
+	status = devm_request_threaded_irq(&pdev->dev, irq, NULL, shps_handle_irq,
+			IRQF_ONESHOT | IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
+			irq_name, pdev);
+	if (status)
+		return status;
+
+	dev_dbg(&pdev->dev, "set up irq %d as type %d\n", irq, type);
+
+	sdev->gpio[type] = gpiod;
+	sdev->irq[type] = irq;
+
+	return 0;
+}
+
+static int surface_hotplug_probe(struct platform_device *pdev)
+{
+	struct shps_device *sdev;
+	int status, i;
+
+	if (gpiod_count(&pdev->dev, NULL) < 0) {
+		dev_err(&pdev->dev, "gpiod_count returned < 0\n");
+		return -ENODEV;
+	}
+
+	status = devm_acpi_dev_add_driver_gpios(&pdev->dev, shps_acpi_gpios);
+	if (status) {
+		dev_err(&pdev->dev, "failed to add gpios: %d\n", status);
+		return status;
+	}
+
+	sdev = devm_kzalloc(&pdev->dev, sizeof(*sdev), GFP_KERNEL);
+	if (!sdev)
+		return -ENOMEM;
+
+	platform_set_drvdata(pdev, sdev);
+
+	// set up IRQs
+	for (i = 0; i < SHPS_NUM_IRQS; i++) {
+		mutex_init(&sdev->lock[i]);
+
+		status = shps_setup_irq(pdev, i);
+		if (status) {
+			dev_err(&pdev->dev, "failed to set up IRQ %d: %d\n",
+				i, status);
+			return status;
+		}
+	}
+
+	// ensure everything is up-to-date
+	for (i = 0; i < SHPS_NUM_IRQS; i++)
+		if (sdev->irq[i] != SHPS_IRQ_NOT_PRESENT)
+			shps_dsm_notify_irq(pdev, i);
+
+	return 0;
+}
+
+static int surface_hotplug_remove(struct platform_device *pdev)
+{
+	struct shps_device *sdev = platform_get_drvdata(pdev);
+	int i;
+
+	// ensure that IRQs have been fully handled and won't trigger any more
+	for (i = 0; i < SHPS_NUM_IRQS; i++)
+		if (sdev->irq[i] != SHPS_IRQ_NOT_PRESENT)
+			disable_irq(sdev->irq[i]);
+
+	return 0;
+}
+
+static const struct acpi_device_id surface_hotplug_acpi_match[] = {
+	{ "MSHW0153", 0 },
+	{ },
+};
+MODULE_DEVICE_TABLE(acpi, surface_hotplug_acpi_match);
+
+static struct platform_driver surface_hotplug_driver = {
+	.probe = surface_hotplug_probe,
+	.remove = surface_hotplug_remove,
+	.driver = {
+		.name = "surface_hotplug",
+		.acpi_match_table = surface_hotplug_acpi_match,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+module_platform_driver(surface_hotplug_driver);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Surface Hot-Plug Signalling Driver for Surface Book Devices");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/clients/surface_perfmode.c b/drivers/misc/surface_aggregator/clients/surface_perfmode.c
new file mode 100644
index 000000000000..e13f4995b28b
--- /dev/null
+++ b/drivers/misc/surface_aggregator/clients/surface_perfmode.c
@@ -0,0 +1,122 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface performance-mode driver.
+ *
+ * Proides a user-space interface for the performance mode control provided by
+ * the Surface System Aggregator Module (SSAM), influencing cooling behavior
+ * of the device and potentially managing power limits.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <asm/unaligned.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/sysfs.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/device.h>
+
+enum sam_perf_mode {
+	SAM_PERF_MODE_NORMAL  = 1,
+	SAM_PERF_MODE_BATTERY = 2,
+	SAM_PERF_MODE_PERF1   = 3,
+	SAM_PERF_MODE_PERF2   = 4,
+
+	__SAM_PERF_MODE__MIN  = 1,
+	__SAM_PERF_MODE__MAX  = 4,
+};
+
+struct ssam_perf_info {
+	__le32 mode;
+	__le16 unknown1;
+	__le16 unknown2;
+} __packed;
+
+static SSAM_DEFINE_SYNC_REQUEST_CL_R(ssam_tmp_perf_mode_get, struct ssam_perf_info, {
+	.target_category = SSAM_SSH_TC_TMP,
+	.command_id      = 0x02,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_CL_W(__ssam_tmp_perf_mode_set, __le32, {
+	.target_category = SSAM_SSH_TC_TMP,
+	.command_id      = 0x03,
+});
+
+static int ssam_tmp_perf_mode_set(struct ssam_device *sdev, u32 mode)
+{
+	__le32 mode_le = cpu_to_le32(mode);
+
+	if (mode < __SAM_PERF_MODE__MIN || mode > __SAM_PERF_MODE__MAX)
+		return -EINVAL;
+
+	return ssam_retry(__ssam_tmp_perf_mode_set, sdev, &mode_le);
+}
+
+static ssize_t perf_mode_show(struct device *dev, struct device_attribute *attr,
+			      char *data)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+	struct ssam_perf_info info;
+	int status;
+
+	status = ssam_retry(ssam_tmp_perf_mode_get, sdev, &info);
+	if (status) {
+		dev_err(dev, "failed to get current performance mode: %d\n",
+			status);
+		return -EIO;
+	}
+
+	return sprintf(data, "%d\n", le32_to_cpu(info.mode));
+}
+
+static ssize_t perf_mode_store(struct device *dev, struct device_attribute *attr,
+			       const char *data, size_t count)
+{
+	struct ssam_device *sdev = to_ssam_device(dev);
+	int perf_mode;
+	int status;
+
+	status = kstrtoint(data, 0, &perf_mode);
+	if (status < 0)
+		return status;
+
+	status = ssam_tmp_perf_mode_set(sdev, perf_mode);
+	if (status < 0)
+		return status;
+
+	return count;
+}
+
+static const DEVICE_ATTR_RW(perf_mode);
+
+static int surface_sam_sid_perfmode_probe(struct ssam_device *sdev)
+{
+	return sysfs_create_file(&sdev->dev.kobj, &dev_attr_perf_mode.attr);
+}
+
+static void surface_sam_sid_perfmode_remove(struct ssam_device *sdev)
+{
+	sysfs_remove_file(&sdev->dev.kobj, &dev_attr_perf_mode.attr);
+}
+
+static const struct ssam_device_id ssam_perfmode_match[] = {
+	{ SSAM_SDEV(TMP, 0x01, 0x00, 0x01) },
+	{ },
+};
+MODULE_DEVICE_TABLE(ssam, ssam_perfmode_match);
+
+static struct ssam_device_driver surface_sam_sid_perfmode = {
+	.probe = surface_sam_sid_perfmode_probe,
+	.remove = surface_sam_sid_perfmode_remove,
+	.match_table = ssam_perfmode_match,
+	.driver = {
+		.name = "surface_performance_mode",
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+module_ssam_device_driver(surface_sam_sid_perfmode);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Performance mode interface for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/controller.c b/drivers/misc/surface_aggregator/controller.c
new file mode 100644
index 000000000000..feee7c7c5945
--- /dev/null
+++ b/drivers/misc/surface_aggregator/controller.c
@@ -0,0 +1,2557 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Main SSAM/SSH controller structure and functionality.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/acpi.h>
+#include <linux/atomic.h>
+#include <linux/completion.h>
+#include <linux/gpio/consumer.h>
+#include <linux/interrupt.h>
+#include <linux/kref.h>
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/rbtree.h>
+#include <linux/rwsem.h>
+#include <linux/serdev.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/srcu.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/serial_hub.h>
+
+#include "controller.h"
+#include "ssh_msgb.h"
+#include "ssh_request_layer.h"
+
+#include "trace.h"
+
+
+/* -- Safe counters. -------------------------------------------------------- */
+
+/**
+ * ssh_seq_reset() - Reset/initialize sequence ID counter.
+ * @c: The counter to reset.
+ */
+static void ssh_seq_reset(struct ssh_seq_counter *c)
+{
+	WRITE_ONCE(c->value, 0);
+}
+
+/**
+ * ssh_seq_next() - Get next sequence ID.
+ * @c: The counter providing the sequence IDs.
+ *
+ * Return: Retunrs the next sequence ID of the counter.
+ */
+static u8 ssh_seq_next(struct ssh_seq_counter *c)
+{
+	u8 old = READ_ONCE(c->value);
+	u8 new = old + 1;
+	u8 ret;
+
+	while (unlikely((ret = cmpxchg(&c->value, old, new)) != old)) {
+		old = ret;
+		new = old + 1;
+	}
+
+	return old;
+}
+
+/**
+ * ssh_rqid_reset() - Reset/initialize request ID counter.
+ * @c: The counter to reset.
+ */
+static void ssh_rqid_reset(struct ssh_rqid_counter *c)
+{
+	WRITE_ONCE(c->value, 0);
+}
+
+/**
+ * ssh_rqid_next() - Get next request ID.
+ * @c: The counter providing the request IDs.
+ *
+ * Return: Returns the next request ID of the counter, skipping any reserved
+ * request IDs.
+ */
+static u16 ssh_rqid_next(struct ssh_rqid_counter *c)
+{
+	u16 old = READ_ONCE(c->value);
+	u16 new = ssh_rqid_next_valid(old);
+	u16 ret;
+
+	while (unlikely((ret = cmpxchg(&c->value, old, new)) != old)) {
+		old = ret;
+		new = ssh_rqid_next_valid(old);
+	}
+
+	return old;
+}
+
+
+/* -- Event notifier/callbacks. --------------------------------------------- */
+/*
+ * The notifier system is based on linux/notifier.h, specifically the SRCU
+ * implementation. The difference to that is, that some bits of the notifier
+ * call return value can be tracked across multiple calls. This is done so that
+ * handling of events can be tracked and a warning can be issued in case an
+ * event goes unhandled. The idea of that waring is that it should help discover
+ * and identify new/currently unimplemented features.
+ */
+
+
+/**
+ * ssam_event_matches_notifier() - Test if an event matches a notifier;
+ * @notif: The event notifier to test against.
+ * @event: The event to test.
+ *
+ * Return: Returns %true iff the given event matches the given notifier
+ * according to the rules set in the notifier's event mask, %false otherwise.
+ */
+static bool ssam_event_matches_notifier(
+		const struct ssam_event_notifier *notif,
+		const struct ssam_event *event)
+{
+	bool match = notif->event.id.target_category == event->target_category;
+
+	if (notif->event.mask & SSAM_EVENT_MASK_TARGET)
+		match &= notif->event.reg.target_id == event->target_id;
+
+	if (notif->event.mask & SSAM_EVENT_MASK_INSTANCE)
+		match &= notif->event.id.instance == event->instance_id;
+
+	return match;
+}
+
+/**
+ * ssam_nfblk_call_chain() - Call event notifier callbacks of the given chain.
+ * @nh:    The notifier head for which the notifier callbacks should be called.
+ * @event: The event data provided to the callbacks.
+ *
+ * Call all registered notifier callbacks in order of their priority until
+ * either no notifier is left or a notifier returns a value with the
+ * %SSAM_NOTIF_STOP bit set. Note that this bit is automatically set via
+ * ssam_notifier_from_errno() on any non-zero error value.
+ *
+ * Return: Returns the notifier status value, which contains the notifier
+ * status bits (%SSAM_NOTIF_HANDLED and %SSAM_NOTIF_STOP) as well as a
+ * potential error value returned from the last executed notifier callback.
+ * Use ssam_notifier_to_errno() to convert this value to the original error
+ * value.
+ */
+static int ssam_nfblk_call_chain(struct ssam_nf_head *nh, struct ssam_event *event)
+{
+	struct ssam_notifier_block *nb, *next_nb;
+	struct ssam_event_notifier *nf;
+	int ret = 0, idx;
+
+	idx = srcu_read_lock(&nh->srcu);
+
+	nb = rcu_dereference_raw(nh->head);
+	while (nb) {
+		nf = container_of(nb, struct ssam_event_notifier, base);
+		next_nb = rcu_dereference_raw(nb->next);
+
+		if (ssam_event_matches_notifier(nf, event)) {
+			ret = (ret & SSAM_NOTIF_STATE_MASK) | nb->fn(nf, event);
+			if (ret & SSAM_NOTIF_STOP)
+				break;
+		}
+
+		nb = next_nb;
+	}
+
+	srcu_read_unlock(&nh->srcu, idx);
+	return ret;
+}
+
+/**
+ * __ssam_nfblk_insert() - Insert a new notifier block into the given notifier
+ * list.
+ * @nh: The notifier head into which the block should be inserted.
+ * @nb: The notifier block to add.
+ *
+ * Note: This function must be synchronized by the caller with respect to other
+ * insert and/or remove calls.
+ *
+ * Return: Returns zero on success, %-EINVAL if the notifier block has already
+ * been registered.
+ */
+static int __ssam_nfblk_insert(struct ssam_nf_head *nh, struct ssam_notifier_block *nb)
+{
+	struct ssam_notifier_block **link = &nh->head;
+
+	while ((*link) != NULL) {
+		if (unlikely((*link) == nb)) {
+			WARN(1, "double register detected");
+			return -EINVAL;
+		}
+
+		if (nb->priority > (*link)->priority)
+			break;
+
+		link = &((*link)->next);
+	}
+
+	nb->next = *link;
+	rcu_assign_pointer(*link, nb);
+
+	return 0;
+}
+
+/**
+ * __ssam_nfblk_find_link() - Find a notifier block link on the given list.
+ * @nh: The notifier head on wich the search should be conducted.
+ * @nb: The notifier block to search for.
+ *
+ * Note: This function must be synchronized by the caller with respect to
+ * insert and/or remove calls.
+ *
+ * Return: Returns a pointer to the link (i.e. pointer pointing) to the given
+ * notifier block, from the previous node in the list, or %NULL if the given
+ * notifier block is not contained in the notifier list.
+ */
+static struct ssam_notifier_block **__ssam_nfblk_find_link(
+		struct ssam_nf_head *nh, struct ssam_notifier_block *nb)
+{
+	struct ssam_notifier_block **link = &nh->head;
+
+	while ((*link) != NULL) {
+		if ((*link) == nb)
+			return link;
+
+		link = &((*link)->next);
+	}
+
+	return NULL;
+}
+
+/**
+ * __ssam_nfblk_erase() - Erase a notifier block link in the given notifier
+ * list.
+ * @link: The link to be erased.
+ *
+ * Note: This function must be synchronized by the caller with respect to
+ * other insert and/or remove/erase/find calls. The caller _must_ ensure SRCU
+ * synchronization by calling synchronize_srcu() with ``nh->srcu`` after
+ * leaving the critical section, to ensure that the removed notifier block is
+ * not in use any more.
+ */
+static void __ssam_nfblk_erase(struct ssam_notifier_block **link)
+{
+	rcu_assign_pointer(*link, (*link)->next);
+}
+
+
+/**
+ * __ssam_nfblk_remove() - Remove a notifier block from the given notifier list.
+ * @nh: The notifier head from which the block should be removed.
+ * @nb: The notifier block to remove.
+ *
+ * Note: This function must be synchronized by the caller with respect to
+ * other insert and/or remove calls. On success, the caller *must* ensure SRCU
+ * synchronization by calling synchronize_srcu() with ``nh->srcu`` after
+ * leaving the critical section, to ensure that the removed notifier block is
+ * not in use any more.
+ *
+ * Return: Returns zero on success, %-ENOENT if the specified notifier block
+ * could not be found on the notifier list.
+ */
+static int __ssam_nfblk_remove(struct ssam_nf_head *nh,
+			       struct ssam_notifier_block *nb)
+{
+	struct ssam_notifier_block **link;
+
+	link = __ssam_nfblk_find_link(nh, nb);
+	if (!link)
+		return -ENOENT;
+
+	__ssam_nfblk_erase(link);
+	return 0;
+}
+
+/**
+ * ssam_nf_head_init() - Initialize the given notifier head.
+ * @nh: The notifier head to initialize.
+ */
+static int ssam_nf_head_init(struct ssam_nf_head *nh)
+{
+	int status;
+
+	status = init_srcu_struct(&nh->srcu);
+	if (status)
+		return status;
+
+	nh->head = NULL;
+	return 0;
+}
+
+/**
+ * ssam_nf_head_destroy() - Deinitialize the given notifier head.
+ * @nh: The notifier head to deinitialize.
+ */
+static void ssam_nf_head_destroy(struct ssam_nf_head *nh)
+{
+	cleanup_srcu_struct(&nh->srcu);
+}
+
+
+/* -- Event/notification registry. ------------------------------------------ */
+
+/**
+ * struct ssam_nf_refcount_key - Key used for event activation reference
+ * counting.
+ * @reg: The registry via which the event is enabled/disabled.
+ * @id:  The ID uniquely describing the event.
+ */
+struct ssam_nf_refcount_key {
+	struct ssam_event_registry reg;
+	struct ssam_event_id id;
+};
+
+/**
+ * struct ssam_nf_refcount_entry - RB-tree entry for referecnce counting event
+ * activations.
+ * @node:     The node of this entry in the rb-tree.
+ * @key:      The key of the event.
+ * @refcount: The reference-count of the event.
+ * @flags:    The flags used when enabling the event.
+ */
+struct ssam_nf_refcount_entry {
+	struct rb_node node;
+	struct ssam_nf_refcount_key key;
+	int refcount;
+	u8 flags;
+};
+
+
+/**
+ * ssam_nf_refcount_inc() - Increment reference-/activation-count of the given
+ * event.
+ * @nf:  The notifier system reference.
+ * @reg: The registry used to enable/disable the event.
+ * @id:  The event ID.
+ *
+ * Increments the reference-/activation-count associated with the specified
+ * event type/ID, allocating a new entry for this event ID if necessary. A
+ * newly allocated entry will have a refcount of one.
+ *
+ * Return: Returns the refcount entry on success. Returns ``ERR_PTR(-ENOSPC)``
+ * if there have already been %INT_MAX events of the specified ID and type
+ * registered, or ``ERR_PTR(-ENOMEM)`` if the entry could not be allocated.
+ */
+static struct ssam_nf_refcount_entry *ssam_nf_refcount_inc(
+		struct ssam_nf *nf, struct ssam_event_registry reg,
+		struct ssam_event_id id)
+{
+	struct ssam_nf_refcount_entry *entry;
+	struct ssam_nf_refcount_key key;
+	struct rb_node **link = &nf->refcount.rb_node;
+	struct rb_node *parent = NULL;
+	int cmp;
+
+	key.reg = reg;
+	key.id = id;
+
+	while (*link) {
+		entry = rb_entry(*link, struct ssam_nf_refcount_entry, node);
+		parent = *link;
+
+		cmp = memcmp(&key, &entry->key, sizeof(key));
+		if (cmp < 0) {
+			link = &(*link)->rb_left;
+		} else if (cmp > 0) {
+			link = &(*link)->rb_right;
+		} else if (entry->refcount < INT_MAX) {
+			entry->refcount++;
+			return entry;
+		} else {
+			return ERR_PTR(-ENOSPC);
+		}
+	}
+
+	entry = kzalloc(sizeof(*entry), GFP_KERNEL);
+	if (!entry)
+		return ERR_PTR(-ENOMEM);
+
+	entry->key = key;
+	entry->refcount = 1;
+
+	rb_link_node(&entry->node, parent, link);
+	rb_insert_color(&entry->node, &nf->refcount);
+
+	return entry;
+}
+
+/**
+ * ssam_nf_refcount_dec() - Decrement reference-/activation-count of the given
+ * event.
+ * @nf:  The notifier system reference.
+ * @reg: The registry used to enable/disable the event.
+ * @id:  The event ID.
+ *
+ * Decrements the reference-/activation-count of the specified event,
+ * returning its entry. If the returned entry has a refcount of zero, the
+ * caller is responsible for freeing it using kfree().
+ *
+ * Return: Returns the refcount entry on success or %NULL if the entry has not
+ * been found.
+ */
+static struct ssam_nf_refcount_entry *ssam_nf_refcount_dec(
+		struct ssam_nf *nf, struct ssam_event_registry reg,
+		struct ssam_event_id id)
+{
+	struct ssam_nf_refcount_entry *entry;
+	struct ssam_nf_refcount_key key;
+	struct rb_node *node = nf->refcount.rb_node;
+	int cmp;
+
+	key.reg = reg;
+	key.id = id;
+
+	while (node) {
+		entry = rb_entry(node, struct ssam_nf_refcount_entry, node);
+
+		cmp = memcmp(&key, &entry->key, sizeof(key));
+		if (cmp < 0) {
+			node = node->rb_left;
+		} else if (cmp > 0) {
+			node = node->rb_right;
+		} else {
+			entry->refcount--;
+			if (entry->refcount == 0)
+				rb_erase(&entry->node, &nf->refcount);
+
+			return entry;
+		}
+	}
+
+	return NULL;
+}
+
+/**
+ * ssam_nf_refcount_empty() - Test if the notification system has any
+ * enabled/active events.
+ * @nf: The notification system.
+ */
+static bool ssam_nf_refcount_empty(struct ssam_nf *nf)
+{
+	return RB_EMPTY_ROOT(&nf->refcount);
+}
+
+/**
+ * ssam_nf_call() - Call notification callbacks for the provided event.
+ * @nf:    The notifier system
+ * @dev:   The associated device, only used for logging.
+ * @rqid:  The request ID of the event.
+ * @event: The event provided to the callbacks.
+ *
+ * Executa registered callbacks in order of their priority until either no
+ * callback is left or a callback returned a value with the %SSAM_NOTIF_STOP
+ * bit set. Note that this bit is set automatically when converting non.zero
+ * error values via ssam_notifier_from_errno() to notifier values.
+ *
+ * Also note that any callback that could handle an event should return a value
+ * with bit %SSAM_NOTIF_HANDLED set, indicating that the event does not go
+ * unhandled/ignored. In case no registered callback could handle an event,
+ * this function will emit a warning.
+ *
+ * In case a callback failed, this function will emit an error message.
+ */
+static void ssam_nf_call(struct ssam_nf *nf, struct device *dev, u16 rqid,
+			 struct ssam_event *event)
+{
+	struct ssam_nf_head *nf_head;
+	int status, nf_ret;
+
+	if (!ssh_rqid_is_event(rqid)) {
+		dev_warn(dev, "event: unsupported rqid: 0x%04x\n", rqid);
+		return;
+	}
+
+	nf_head = &nf->head[ssh_rqid_to_event(rqid)];
+	nf_ret = ssam_nfblk_call_chain(nf_head, event);
+	status = ssam_notifier_to_errno(nf_ret);
+
+	if (status < 0) {
+		dev_err(dev, "event: error handling event: %d "
+			"(tc: 0x%02x, tid: 0x%02x, cid: 0x%02x, iid: 0x%02x)\n",
+			status, event->target_category, event->target_id,
+			event->command_id, event->instance_id);
+	}
+
+	if (!(nf_ret & SSAM_NOTIF_HANDLED)) {
+		dev_warn(dev, "event: unhandled event (rqid: 0x%02x, "
+			 "tc: 0x%02x, tid: 0x%02x, cid: 0x%02x, iid: 0x%02x)\n",
+			 rqid, event->target_category, event->target_id,
+			 event->command_id, event->instance_id);
+	}
+}
+
+/**
+ * ssam_nf_init() - Initialize the notifier system.
+ * @nf: The notifier system to initialize.
+ */
+static int ssam_nf_init(struct ssam_nf *nf)
+{
+	int i, status;
+
+	for (i = 0; i < SSH_NUM_EVENTS; i++) {
+		status = ssam_nf_head_init(&nf->head[i]);
+		if (status)
+			break;
+	}
+
+	if (status) {
+		for (i = i - 1; i >= 0; i--)
+			ssam_nf_head_destroy(&nf->head[i]);
+
+		return status;
+	}
+
+	mutex_init(&nf->lock);
+	return 0;
+}
+
+/**
+ * ssam_nf_destroy() - Deinitialize the notifier system.
+ * @nf: The notifier system to deinitialize.
+ */
+static void ssam_nf_destroy(struct ssam_nf *nf)
+{
+	int i;
+
+	for (i = 0; i < SSH_NUM_EVENTS; i++)
+		ssam_nf_head_destroy(&nf->head[i]);
+
+	mutex_destroy(&nf->lock);
+}
+
+
+/* -- Event/async request completion system. -------------------------------- */
+
+#define SSAM_CPLT_WQ_NAME	"ssam_cpltq"
+
+/*
+ * SSAM_EVENT_ITEM_CACHE_PAYLOAD_LEN - Maximum payload length for a cached
+ * &struct ssam_event_item.
+ *
+ * This length has been chosen to be accommodate standard touchpad and
+ * keyboard input events. Events with larger payloads will be allocated
+ * separately.
+ */
+#define SSAM_EVENT_ITEM_CACHE_PAYLOAD_LEN	32
+
+static struct kmem_cache *ssam_event_item_cache;
+
+/**
+ * ssam_event_item_cache_init() - Initialize the event item cache.
+ */
+int ssam_event_item_cache_init(void)
+{
+	const unsigned int size = sizeof(struct ssam_event_item)
+				  + SSAM_EVENT_ITEM_CACHE_PAYLOAD_LEN;
+	const unsigned int align = __alignof__(struct ssam_event_item);
+	struct kmem_cache *cache;
+
+	cache = kmem_cache_create("ssam_event_item", size, align, 0, NULL);
+	if (!cache)
+		return -ENOMEM;
+
+	ssam_event_item_cache = cache;
+	return 0;
+}
+
+/**
+ * ssam_event_item_cache_destroy() - Deinitialize the event item cache.
+ */
+void ssam_event_item_cache_destroy(void)
+{
+	kmem_cache_destroy(ssam_event_item_cache);
+	ssam_event_item_cache = NULL;
+}
+
+static void __ssam_event_item_free_cached(struct ssam_event_item *item)
+{
+	kmem_cache_free(ssam_event_item_cache, item);
+}
+
+static void __ssam_event_item_free_generic(struct ssam_event_item *item)
+{
+	kfree(item);
+}
+
+/**
+ * ssam_event_item_free() - Free the provided event item.
+ * @item: The event item to free.
+ */
+static void ssam_event_item_free(struct ssam_event_item *item)
+{
+	trace_ssam_event_item_free(item);
+	item->ops.free(item);
+}
+
+/**
+ * ssam_event_item_alloc() - Allocate an event item with the given payload size.
+ * @len:   The event payload length.
+ * @flags: The flags used for allocation.
+ *
+ * Allocate an event item with the given payload size, preferring allocation
+ * from the event item cache if the payload is small enough (i.e. smaller than
+ * %SSAM_EVENT_ITEM_CACHE_PAYLOAD_LEN). Sets the item operations and payload
+ * length values. The item free callback (``ops.free``) should not be
+ * overwritten after this call.
+ *
+ * Return: Returns the newly allocated event item.
+ */
+static struct ssam_event_item *ssam_event_item_alloc(size_t len, gfp_t flags)
+{
+	struct ssam_event_item *item;
+
+	if (len <= SSAM_EVENT_ITEM_CACHE_PAYLOAD_LEN) {
+		item = kmem_cache_alloc(ssam_event_item_cache, GFP_KERNEL);
+		if (!item)
+			return NULL;
+
+		item->ops.free = __ssam_event_item_free_cached;
+	} else {
+		item = kzalloc(sizeof(*item) + len, GFP_KERNEL);
+		if (!item)
+			return NULL;
+
+		item->ops.free = __ssam_event_item_free_generic;
+	}
+
+	item->event.length = len;
+
+	trace_ssam_event_item_alloc(item, len);
+	return item;
+}
+
+
+/**
+ * ssam_event_queue_push() - Push an event item to the event queue.
+ * @q:    The event queue.
+ * @item: The item to add.
+ */
+static void ssam_event_queue_push(struct ssam_event_queue *q,
+				  struct ssam_event_item *item)
+{
+	spin_lock(&q->lock);
+	list_add_tail(&item->node, &q->head);
+	spin_unlock(&q->lock);
+}
+
+/**
+ * ssam_event_queue_pop() - Pop the next event item from the event queue.
+ * @q: The event queue.
+ *
+ * Returns and removes the next event item from the queue. Returns %NULL If
+ * there is no event item left.
+ */
+static struct ssam_event_item *ssam_event_queue_pop(struct ssam_event_queue *q)
+{
+	struct ssam_event_item *item;
+
+	spin_lock(&q->lock);
+	item = list_first_entry_or_null(&q->head, struct ssam_event_item, node);
+	if (item)
+		list_del(&item->node);
+	spin_unlock(&q->lock);
+
+	return item;
+}
+
+/**
+ * ssam_event_queue_is_empty() - Check if the event queue is empty.
+ * @q: The event queue.
+ */
+static bool ssam_event_queue_is_empty(struct ssam_event_queue *q)
+{
+	bool empty;
+
+	spin_lock(&q->lock);
+	empty = list_empty(&q->head);
+	spin_unlock(&q->lock);
+
+	return empty;
+}
+
+/**
+ * ssam_cplt_get_event_queue() - Get the event queue for the given parameters.
+ * @cplt: The completion system on which to look for the queue.
+ * @tid:  The target ID of the queue.
+ * @rqid: The request ID representing the event ID for which to get the queue.
+ *
+ * Return: Returns the event queue corresponding to the event type described
+ * by the given parameters. If the request ID does not represent an event,
+ * this function returns %NULL. If the target ID is not supported, this
+ * function will fall back to the default target ID (``tid = 1``).
+ */
+static struct ssam_event_queue *ssam_cplt_get_event_queue(
+		struct ssam_cplt *cplt, u8 tid, u16 rqid)
+{
+	u16 event = ssh_rqid_to_event(rqid);
+	u16 tidx = ssh_tid_to_index(tid);
+
+	if (!ssh_rqid_is_event(rqid)) {
+		dev_err(cplt->dev, "event: unsupported request ID: 0x%04x\n", rqid);
+		return NULL;
+	}
+
+	if (!ssh_tid_is_valid(tid)) {
+		dev_warn(cplt->dev, "event: unsupported target ID: %u\n", tid);
+		tidx = 0;
+	}
+
+	return &cplt->event.target[tidx].queue[event];
+}
+
+/**
+ * ssam_cplt_submit() - Submit a work item to the compeltion system workqueue.
+ * @cplt: The completion system.
+ * @work: The work item to submit.
+ */
+static bool ssam_cplt_submit(struct ssam_cplt *cplt, struct work_struct *work)
+{
+	return queue_work(cplt->wq, work);
+}
+
+/**
+ * ssam_cplt_submit_event() - Submit an event to the completion system.
+ * @cplt: The completion system.
+ * @item: The event item to submit.
+ *
+ * Submits the event to the completion system by queuing it on the event item
+ * queue and queuing the respective event queue work item on the completion
+ * workqueue, which will eventually complete the event.
+ *
+ * Return: Returns zero on success, %-EINVAL if there is no event queue that
+ * can handle the given event item.
+ */
+static int ssam_cplt_submit_event(struct ssam_cplt *cplt,
+				  struct ssam_event_item *item)
+{
+	struct ssam_event_queue *evq;
+
+	evq = ssam_cplt_get_event_queue(cplt, item->event.target_id, item->rqid);
+	if (!evq)
+		return -EINVAL;
+
+	ssam_event_queue_push(evq, item);
+	ssam_cplt_submit(cplt, &evq->work);
+	return 0;
+}
+
+/**
+ * ssam_cplt_flush() - Flush the completion system.
+ * @cplt: The completion system.
+ *
+ * Flush the completion system by waiting until all currently submitted work
+ * items have been completed.
+ *
+ * Note: This function does not guarantee that all events will have been
+ * handled once this call terminates. In case of a larger number of
+ * to-be-completed events, the event queue work function may re-schedule its
+ * work item, which this flush operation will ignore.
+ *
+ * This operation is only intended to, during normal operation prior to
+ * shutdown, try to complete most events and requests to get them out of the
+ * system while the system is still fully operational. It does not aim to
+ * provide any guraantee that all of them have been handled.
+ */
+static void ssam_cplt_flush(struct ssam_cplt *cplt)
+{
+	flush_workqueue(cplt->wq);
+}
+
+static void ssam_event_queue_work_fn(struct work_struct *work)
+{
+	struct ssam_event_queue *queue;
+	struct ssam_event_item *item;
+	struct ssam_nf *nf;
+	struct device *dev;
+	int i;
+
+	queue = container_of(work, struct ssam_event_queue, work);
+	nf = &queue->cplt->event.notif;
+	dev = queue->cplt->dev;
+
+	// limit number of processed events to avoid livelocking
+	for (i = 0; i < 10; i++) {
+		item = ssam_event_queue_pop(queue);
+		if (item == NULL)
+			return;
+
+		ssam_nf_call(nf, dev, item->rqid, &item->event);
+		ssam_event_item_free(item);
+	}
+
+	if (!ssam_event_queue_is_empty(queue))
+		ssam_cplt_submit(queue->cplt, &queue->work);
+}
+
+/**
+ * ssam_event_queue_init() - Initialize an event queue.
+ * @cplt: The completion system on which the queue resides.
+ * @evq:  The event queue to initialize.
+ */
+static void ssam_event_queue_init(struct ssam_cplt *cplt,
+				  struct ssam_event_queue *evq)
+{
+	evq->cplt = cplt;
+	spin_lock_init(&evq->lock);
+	INIT_LIST_HEAD(&evq->head);
+	INIT_WORK(&evq->work, ssam_event_queue_work_fn);
+}
+
+/**
+ * ssam_cplt_init() - Initialize completion system.
+ * @cplt: The completion system to initialize.
+ * @dev:  The device used for logging.
+ */
+static int ssam_cplt_init(struct ssam_cplt *cplt, struct device *dev)
+{
+	struct ssam_event_target *target;
+	int status, c, i;
+
+	cplt->dev = dev;
+
+	cplt->wq = create_workqueue(SSAM_CPLT_WQ_NAME);
+	if (!cplt->wq)
+		return -ENOMEM;
+
+	for (c = 0; c < ARRAY_SIZE(cplt->event.target); c++) {
+		target = &cplt->event.target[c];
+
+		for (i = 0; i < ARRAY_SIZE(target->queue); i++)
+			ssam_event_queue_init(cplt, &target->queue[i]);
+	}
+
+	status = ssam_nf_init(&cplt->event.notif);
+	if (status)
+		destroy_workqueue(cplt->wq);
+
+	return status;
+}
+
+/**
+ * ssam_cplt_destroy() - Deinitialize the completion system.
+ * @cplt: The completion system to deinitialize.
+ *
+ * Deinitialize the given completion system and ensure that all pending, i.e.
+ * yet-to-be-completed, event items and requests have been handled.
+ */
+static void ssam_cplt_destroy(struct ssam_cplt *cplt)
+{
+	/*
+	 * Note: destroy_workqueue ensures that all currently queued work will
+	 * be fully completed and the workqueue drained. This means that this
+	 * call will inherently also free any queued ssam_event_items, thus we
+	 * don't have to take care of that here explicitly.
+	 */
+	destroy_workqueue(cplt->wq);
+	ssam_nf_destroy(&cplt->event.notif);
+}
+
+
+/* -- Main SSAM device structures. ------------------------------------------ */
+
+/**
+ * ssam_controller_device() - Get the &struct device associated with this
+ * controller.
+ * @c: The controller for which to get the device.
+ *
+ * Return: Returns the &struct device associated with this controller,
+ * providing its lower-level transport.
+ */
+struct device *ssam_controller_device(struct ssam_controller *c)
+{
+	return ssh_rtl_get_device(&c->rtl);
+}
+EXPORT_SYMBOL_GPL(ssam_controller_device);
+
+static void __ssam_controller_release(struct kref *kref)
+{
+	struct ssam_controller *ctrl = to_ssam_controller(kref, kref);
+
+	ssam_controller_destroy(ctrl);
+	kfree(ctrl);
+}
+
+/**
+ * ssam_controller_get() - Increment reference count of controller.
+ * @c: The controller.
+ *
+ * Return: Returns the controller provided as input.
+ */
+struct ssam_controller *ssam_controller_get(struct ssam_controller *c)
+{
+	if (c)
+		kref_get(&c->kref);
+	return c;
+}
+EXPORT_SYMBOL_GPL(ssam_controller_get);
+
+/**
+ * ssam_controller_put() - Decrement reference count of controller.
+ * @c: The controller.
+ */
+void ssam_controller_put(struct ssam_controller *c)
+{
+	if (c)
+		kref_put(&c->kref, __ssam_controller_release);
+}
+EXPORT_SYMBOL_GPL(ssam_controller_put);
+
+
+/**
+ * ssam_controller_statelock() - Lock the controller against state transitions.
+ * @c: The controller to lock.
+ *
+ * Lock the controller against state transitions. Holding this lock guarantees
+ * that the controller will not transition between states, i.e. if the
+ * controller is in state "started", when this lock has been acquired, it will
+ * remain in this state at least until the lock has been released.
+ *
+ * Multiple clients may concurrently hold this lock. In other words: The
+ * ``statelock`` functions represent the read-lock part of a r/w-semaphore.
+ * Actions causing state transitions of the controller must be executed while
+ * holding the write-part of this r/w-semaphore (see ssam_controller_lock()
+ * and ssam_controller_unlock() for that).
+ *
+ * See ssam_controller_stateunlock() for the corresponding unlock function.
+ */
+void ssam_controller_statelock(struct ssam_controller *c)
+{
+	down_read(&c->lock);
+}
+EXPORT_SYMBOL_GPL(ssam_controller_statelock);
+
+/**
+ * ssam_controller_stateunlock() - Unlock controller state transitions.
+ * @c: The controller to unlock.
+ *
+ * See ssam_controller_statelock() for the corresponding lock function.
+ */
+void ssam_controller_stateunlock(struct ssam_controller *c)
+{
+	up_read(&c->lock);
+}
+EXPORT_SYMBOL_GPL(ssam_controller_stateunlock);
+
+/**
+ * ssam_controller_lock() - Acquire the main controller lock.
+ * @c: The controller to lock.
+ *
+ * This lock must be held for any state transitions, including transition to
+ * suspend/resumed states and during shutdown. See ssam_controller_statelock()
+ * for more details on controller locking.
+ *
+ * See ssam_controller_unlock() for the corresponding unlock function.
+ */
+void ssam_controller_lock(struct ssam_controller *c)
+{
+	down_write(&c->lock);
+}
+
+/*
+ * ssam_controller_unlock() - Release the main controller lock.
+ * @c: The controller to unlock.
+ *
+ * See ssam_controller_lock() for the corresponding lock function.
+ */
+void ssam_controller_unlock(struct ssam_controller *c)
+{
+	up_write(&c->lock);
+}
+
+
+static void ssam_handle_event(struct ssh_rtl *rtl,
+			      const struct ssh_command *cmd,
+			      const struct ssam_span *data)
+{
+	struct ssam_controller *ctrl = to_ssam_controller(rtl, rtl);
+	struct ssam_event_item *item;
+
+	item = ssam_event_item_alloc(data->len, GFP_KERNEL);
+	if (!item)
+		return;
+
+	item->rqid = get_unaligned_le16(&cmd->rqid);
+	item->event.target_category = cmd->tc;
+	item->event.target_id = cmd->tid_in;
+	item->event.command_id = cmd->cid;
+	item->event.instance_id = cmd->iid;
+	memcpy(&item->event.data[0], data->ptr, data->len);
+
+	WARN_ON(ssam_cplt_submit_event(&ctrl->cplt, item));
+}
+
+static const struct ssh_rtl_ops ssam_rtl_ops = {
+	.handle_event = ssam_handle_event,
+};
+
+
+static bool ssam_notifier_empty(struct ssam_controller *ctrl);
+static void ssam_notifier_unregister_all(struct ssam_controller *ctrl);
+
+
+#define SSAM_SSH_DSM_REVISION	0
+static const guid_t SSAM_SSH_DSM_GUID = GUID_INIT(0xd5e383e1, 0xd892, 0x4a76,
+		0x89, 0xfc, 0xf6, 0xaa, 0xae, 0x7e, 0xd5, 0xb5);
+
+enum ssh_dsm_fn {
+	SSH_DSM_FN_SSH_POWER_PROFILE             = 0x05,
+	SSH_DSM_FN_SCREEN_ON_SLEEP_IDLE_TIMEOUT  = 0x06,
+	SSH_DSM_FN_SCREEN_OFF_SLEEP_IDLE_TIMEOUT = 0x07,
+	SSH_DSM_FN_D3_CLOSES_HANDLE              = 0x08,
+	SSH_DSM_FN_SSH_BUFFER_SIZE               = 0x09,
+};
+
+static int ssam_dsm_get_functions(acpi_handle handle, u64 *funcs)
+{
+	union acpi_object *obj;
+	u64 mask = 0;
+	int i;
+
+	*funcs = 0;
+
+	if (!acpi_has_method(handle, "_DSM"))
+		return 0;
+
+	obj = acpi_evaluate_dsm_typed(handle, &SSAM_SSH_DSM_GUID,
+				      SSAM_SSH_DSM_REVISION, 0, NULL,
+				      ACPI_TYPE_BUFFER);
+	if (!obj)
+		return -EFAULT;
+
+	for (i = 0; i < obj->buffer.length && i < 8; i++)
+		mask |= (((u64)obj->buffer.pointer[i]) << (i * 8));
+
+	if (mask & 0x01)
+		*funcs = mask;
+
+	ACPI_FREE(obj);
+	return 0;
+}
+
+static int ssam_dsm_load_u32(acpi_handle handle, u64 funcs, u64 func, u32 *ret)
+{
+	union acpi_object *obj;
+	u64 val;
+
+	if (!(funcs & BIT(func)))
+		return 0;
+
+	obj = acpi_evaluate_dsm_typed(handle, &SSAM_SSH_DSM_GUID,
+				      SSAM_SSH_DSM_REVISION, func, NULL,
+				      ACPI_TYPE_INTEGER);
+	if (!obj)
+		return -EFAULT;
+
+	val = obj->integer.value;
+	ACPI_FREE(obj);
+
+	if (val > U32_MAX)
+		return -ERANGE;
+
+	*ret = val;
+	return 0;
+}
+
+/**
+ * ssam_controller_caps_load_from_acpi() - Load controller capabilities from
+ * ACPI _DSM.
+ * @handle: The handle of the ACPI controller/SSH device.
+ * @caps:   Where to store the capabilities in.
+ *
+ * Initializes the given controller capabilities with default values, then
+ * checks and, if the respective _DSM functions are available, loads the
+ * actual capabilities from the _DSM.
+ *
+ * Return: Returns zero on success, a negative error code on failure.
+ */
+static int ssam_controller_caps_load_from_acpi(
+		acpi_handle handle, struct ssam_controller_caps *caps)
+{
+	u32 d3_closes_handle = false;
+	u64 funcs;
+	int status;
+
+	// set defaults
+	caps->ssh_power_profile = (u32)-1;
+	caps->screen_on_sleep_idle_timeout = (u32)-1;
+	caps->screen_off_sleep_idle_timeout = (u32)-1;
+	caps->d3_closes_handle = false;
+	caps->ssh_buffer_size = (u32)-1;
+
+	status = ssam_dsm_get_functions(handle, &funcs);
+	if (status)
+		return status;
+
+	status = ssam_dsm_load_u32(handle, funcs, SSH_DSM_FN_SSH_POWER_PROFILE,
+				   &caps->ssh_power_profile);
+	if (status)
+		return status;
+
+	status = ssam_dsm_load_u32(handle, funcs,
+				   SSH_DSM_FN_SCREEN_ON_SLEEP_IDLE_TIMEOUT,
+				   &caps->screen_on_sleep_idle_timeout);
+	if (status)
+		return status;
+
+	status = ssam_dsm_load_u32(handle, funcs,
+				   SSH_DSM_FN_SCREEN_OFF_SLEEP_IDLE_TIMEOUT,
+				   &caps->screen_off_sleep_idle_timeout);
+	if (status)
+		return status;
+
+	status = ssam_dsm_load_u32(handle, funcs, SSH_DSM_FN_D3_CLOSES_HANDLE,
+				   &d3_closes_handle);
+	if (status)
+		return status;
+
+	caps->d3_closes_handle = !!d3_closes_handle;
+
+	status = ssam_dsm_load_u32(handle, funcs, SSH_DSM_FN_SSH_BUFFER_SIZE,
+				   &caps->ssh_buffer_size);
+	if (status)
+		return status;
+
+	return 0;
+}
+
+/**
+ * ssam_controller_init() - Initialize SSAM controller.
+ * @ctrl:   The controller to initialize.
+ * @serdev: The serial device representing the underlying data transport.
+ *
+ * Initializes the given controller. Does neither start receiver nor
+ * transmitter threads. After this call, the controller has to be hooked up to
+ * the serdev core separately via &struct serdev_device_ops, relaying calls to
+ * ssam_controller_receive_buf() and ssam_controller_write_wakeup(). Once the
+ * controller has been hooked up, transmitter and receiver threads may be
+ * started via ssam_controller_start(). These setup steps need to be completed
+ * before controller can be used for requests.
+ */
+int ssam_controller_init(struct ssam_controller *ctrl,
+			 struct serdev_device *serdev)
+{
+	acpi_handle handle = ACPI_HANDLE(&serdev->dev);
+	int status;
+
+	init_rwsem(&ctrl->lock);
+	kref_init(&ctrl->kref);
+
+	status = ssam_controller_caps_load_from_acpi(handle, &ctrl->caps);
+	if (status)
+		return status;
+
+	dev_dbg(&serdev->dev,
+		"device capabilities:\n"
+		"  ssh_power_profile:             %u\n"
+		"  ssh_buffer_size:               %u\n"
+		"  screen_on_sleep_idle_timeout:  %u\n"
+		"  screen_off_sleep_idle_timeout: %u\n"
+		"  d3_closes_handle:              %u\n",
+		ctrl->caps.ssh_power_profile,
+		ctrl->caps.ssh_buffer_size,
+		ctrl->caps.screen_on_sleep_idle_timeout,
+		ctrl->caps.screen_off_sleep_idle_timeout,
+		ctrl->caps.d3_closes_handle);
+
+	ssh_seq_reset(&ctrl->counter.seq);
+	ssh_rqid_reset(&ctrl->counter.rqid);
+
+	// initialize event/request completion system
+	status = ssam_cplt_init(&ctrl->cplt, &serdev->dev);
+	if (status)
+		return status;
+
+	// initialize request and packet transport layers
+	status = ssh_rtl_init(&ctrl->rtl, serdev, &ssam_rtl_ops);
+	if (status) {
+		ssam_cplt_destroy(&ctrl->cplt);
+		return status;
+	}
+
+	/*
+	 * Set state via write_once even though we expect to be in an
+	 * exclusive context, due to smoke-testing in
+	 * ssam_request_sync_submit().
+	 */
+	WRITE_ONCE(ctrl->state, SSAM_CONTROLLER_INITIALIZED);
+	return 0;
+}
+
+/**
+ * ssam_controller_start() - Start the receiver and transmitter threads of the
+ * controller.
+ * @ctrl: The controller.
+ *
+ * Note: When this function is called, the controller shouldbe properly hooked
+ * up to the serdev core via &struct serdev_device_ops. Please refert to
+ * ssam_controller_init() for more details on controller initialization.
+ *
+ * This function must be called from an exclusive context with regards to the
+ * state, if necessary, by locking the controller via ssam_controller_lock().
+ */
+int ssam_controller_start(struct ssam_controller *ctrl)
+{
+	int status;
+
+	if (ctrl->state != SSAM_CONTROLLER_INITIALIZED)
+		return -EINVAL;
+
+	status = ssh_rtl_start(&ctrl->rtl);
+	if (status)
+		return status;
+
+	/*
+	 * Set state via write_once even though we expect to be locked/in an
+	 * exclusive context, due to smoke-testing in
+	 * ssam_request_sync_submit().
+	 */
+	WRITE_ONCE(ctrl->state, SSAM_CONTROLLER_STARTED);
+	return 0;
+}
+
+/**
+ * ssam_controller_shutdown() - Shut down the controller.
+ * @ctrl: The controller.
+ *
+ * Shuts down the controller by flushing all pending requests and stopping the
+ * transmitter and receiver threads. All requests submitted after this call
+ * will fail with %-ESHUTDOWN. While it is discouraged to do so, this function
+ * is safe to use in parallel with ongoing request submission.
+ *
+ * In the course of this shutdown procedure, all currently registered
+ * notifiers will be unregistered. It is, however, strongly recommended to not
+ * rely on this behavior, and instead the party registring the notifier should
+ * unregister it before the controller gets shut down, e.g. via the SSAM bus
+ * which guarantees client devices to be removed before a shutdown.
+ *
+ * Note that events may still be pending after this call, but due to the
+ * notifiers being unregistered, the will be dropped when the controller is
+ * subsequently being destroyed via ssam_controller_destroy().
+ *
+ * This function must be called from an exclusive context with regards to the
+ * state, if necessary, by locking the controller via ssam_controller_lock().
+ */
+void ssam_controller_shutdown(struct ssam_controller *ctrl)
+{
+	enum ssam_controller_state s = ctrl->state;
+	int status;
+
+	if (s == SSAM_CONTROLLER_UNINITIALIZED || s == SSAM_CONTROLLER_STOPPED)
+		return;
+
+	// try to flush pending events and requests while everything still works
+	status = ssh_rtl_flush(&ctrl->rtl, msecs_to_jiffies(5000));
+	if (status) {
+		ssam_err(ctrl, "failed to flush request transport layer: %d\n",
+			 status);
+	}
+
+	// try to flush out all currently completing requests and events
+	ssam_cplt_flush(&ctrl->cplt);
+
+	/*
+	 * We expect all notifiers to have been removed by the respective client
+	 * driver that set them up at this point. If this warning occurs, some
+	 * client driver has not done that...
+	 */
+	WARN_ON(!ssam_notifier_empty(ctrl));
+
+	/*
+	 * Nevertheless, we should still take care of drivers that don't behave
+	 * well. Thus disable all enabled events, unregister all notifiers.
+	 */
+	ssam_notifier_unregister_all(ctrl);
+
+	// cancel rem. requests, ensure no new ones can be queued, stop threads
+	ssh_rtl_shutdown(&ctrl->rtl);
+
+	/*
+	 * Set state via write_once even though we expect to be locked/in an
+	 * exclusive context, due to smoke-testing in
+	 * ssam_request_sync_submit().
+	 */
+	WRITE_ONCE(ctrl->state, SSAM_CONTROLLER_STOPPED);
+	ctrl->rtl.ptl.serdev = NULL;
+}
+
+/**
+ * ssam_controller_destroy() - Destroy the controller and free its resources.
+ * @ctrl: The controller.
+ *
+ * Ensures that all resources associated with the controller get freed. This
+ * function should only be called after the controller has been stopped via
+ * ssam_controller_shutdown(). In general, this function should not be called
+ * directly. The only valid place to call this function direclty is during
+ * initialization, before the controller has been fully initialized and passed
+ * to other processes. This function is called automatically when the
+ * reference count of the controller reaches zero.
+ *
+ * Must be called from an exclusive context with regards to the controller
+ * state.
+ */
+void ssam_controller_destroy(struct ssam_controller *ctrl)
+{
+	if (ctrl->state == SSAM_CONTROLLER_UNINITIALIZED)
+		return;
+
+	WARN_ON(ctrl->state != SSAM_CONTROLLER_STOPPED);
+
+	/*
+	 * Note: New events could still have been received after the previous
+	 * flush in ssam_controller_shutdown, before the request transport layer
+	 * has been shut down. At this point, after the shutdown, we can be sure
+	 * that no new events will be queued. The call to ssam_cplt_destroy will
+	 * ensure that those remaining are being completed and freed.
+	 */
+
+	// actually free resources
+	ssam_cplt_destroy(&ctrl->cplt);
+	ssh_rtl_destroy(&ctrl->rtl);
+
+	/*
+	 * Set state via write_once even though we expect to be locked/in an
+	 * exclusive context, due to smoke-testing in
+	 * ssam_request_sync_submit().
+	 */
+	WRITE_ONCE(ctrl->state, SSAM_CONTROLLER_UNINITIALIZED);
+}
+
+/**
+ * ssam_controller_suspend() - Suspend the controller.
+ * @ctrl: The controller to suspend.
+ *
+ * Marks the controller as suspended. Note that display-off and D0-exit
+ * notifications have to be sent manually before transitioning the controller
+ * into the suspended state via this function.
+ *
+ * See ssam_controller_resume() for the corresponding resume function.
+ *
+ * Return: Returns %-EINVAL if the controller is currently not in the
+ * "started" state.
+ */
+int ssam_controller_suspend(struct ssam_controller *ctrl)
+{
+	ssam_controller_lock(ctrl);
+
+	if (ctrl->state != SSAM_CONTROLLER_STARTED) {
+		ssam_controller_unlock(ctrl);
+		return -EINVAL;
+	}
+
+	ssam_dbg(ctrl, "pm: suspending controller\n");
+
+	/*
+	 * Set state via write_once even though we're locked, due to
+	 * smoke-testing in ssam_request_sync_submit().
+	 */
+	WRITE_ONCE(ctrl->state, SSAM_CONTROLLER_SUSPENDED);
+
+	ssam_controller_unlock(ctrl);
+	return 0;
+}
+
+/**
+ * ssam_controller_resume() - Resume the controller from suspend.
+ * @ctrl: The controller to resume.
+ *
+ * Resume the controller from the suspended state it was put into via
+ * ssam_controller_suspend(). This function does not issue display-on and
+ * D0-entry notifications. If required, those have to be sent manually after
+ * this call.
+ *
+ * Return: Returns %-EINVAL if the controller is currently not suspended.
+ */
+int ssam_controller_resume(struct ssam_controller *ctrl)
+{
+	ssam_controller_lock(ctrl);
+
+	if (ctrl->state != SSAM_CONTROLLER_SUSPENDED) {
+		ssam_controller_unlock(ctrl);
+		return -EINVAL;
+	}
+
+	ssam_dbg(ctrl, "pm: resuming controller\n");
+
+	/*
+	 * Set state via write_once even though we're locked, due to
+	 * smoke-testing in ssam_request_sync_submit().
+	 */
+	WRITE_ONCE(ctrl->state, SSAM_CONTROLLER_STARTED);
+
+	ssam_controller_unlock(ctrl);
+	return 0;
+}
+
+
+/* -- Top-level request interface ------------------------------------------- */
+
+/**
+ * ssam_request_write_data() - Construct and write SAM request message to
+ * buffer.
+ * @buf:  The buffer to write the data to.
+ * @ctrl: The controller via which the request will be sent.
+ * @spec: The request data and specification.
+ *
+ * Constructs a SAM/SSH request message and writes it to the provided buffer.
+ * The request and transport counters, specifically RQID and SEQ, will be set
+ * in this call. These counters are obtained from the controller. It is thus
+ * only valid to send the resulting message via the controller specified here.
+ *
+ * For calculation of the required buffer size, refer to the
+ * SSH_COMMAND_MESSAGE_LENGTH() macro.
+ *
+ * Return: Returns the number of bytes used in the buffer on success. Returns
+ * %-EINVAL if the payload length provided in the request specification is too
+ * large (larger than %SSH_COMMAND_MAX_PAYLOAD_SIZE) or if the provided buffer
+ * is too small.
+ */
+ssize_t ssam_request_write_data(struct ssam_span *buf,
+				struct ssam_controller *ctrl,
+				const struct ssam_request *spec)
+{
+	struct msgbuf msgb;
+	u16 rqid;
+	u8 seq;
+
+	if (spec->length > SSH_COMMAND_MAX_PAYLOAD_SIZE)
+		return -EINVAL;
+
+	if (SSH_COMMAND_MESSAGE_LENGTH(spec->length) > buf->len)
+		return -EINVAL;
+
+	msgb_init(&msgb, buf->ptr, buf->len);
+	seq = ssh_seq_next(&ctrl->counter.seq);
+	rqid = ssh_rqid_next(&ctrl->counter.rqid);
+	msgb_push_cmd(&msgb, seq, rqid, spec);
+
+	return msgb_bytes_used(&msgb);
+}
+EXPORT_SYMBOL_GPL(ssam_request_write_data);
+
+
+static void ssam_request_sync_complete(struct ssh_request *rqst,
+				       const struct ssh_command *cmd,
+				       const struct ssam_span *data, int status)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+	struct ssam_request_sync *r;
+
+	r = container_of(rqst, struct ssam_request_sync, base);
+	r->status = status;
+
+	if (r->resp)
+		r->resp->length = 0;
+
+	if (status) {
+		rtl_dbg_cond(rtl, "rsp: request failed: %d\n", status);
+		return;
+	}
+
+	if (!data)	// handle requests without a response
+		return;
+
+	if (!r->resp || !r->resp->pointer) {
+		if (data->len)
+			rtl_warn(rtl, "rsp: no response buffer provided, dropping data\n");
+		return;
+	}
+
+	if (data->len > r->resp->capacity) {
+		rtl_err(rtl, "rsp: response buffer too small, capacity: %zu bytes,"
+			" got: %zu bytes\n", r->resp->capacity, data->len);
+		r->status = -ENOSPC;
+		return;
+	}
+
+	r->resp->length = data->len;
+	memcpy(r->resp->pointer, data->ptr, data->len);
+}
+
+static void ssam_request_sync_release(struct ssh_request *rqst)
+{
+	complete_all(&container_of(rqst, struct ssam_request_sync, base)->comp);
+}
+
+static const struct ssh_request_ops ssam_request_sync_ops = {
+	.release = ssam_request_sync_release,
+	.complete = ssam_request_sync_complete,
+};
+
+
+/**
+ * ssam_request_sync_alloc() - Allocate a synchronous request.
+ * @payload_len: The length of the request payload.
+ * @flags:       Flags used for allocation.
+ * @rqst:        Where to store the pointer to the allocated request.
+ * @buffer:      Where to store the buffer descriptor for the message buffer of
+ *               the request.
+ *
+ * Allocates a synchronous request with corresponding message buffer. The
+ * request still needs to be initialized ssam_request_sync_init() before
+ * it can be submitted, and the message buffer data must still be set to the
+ * returned buffer via ssam_request_sync_set_data() after it has been filled,
+ * if need be with adjusted message length.
+ *
+ * After use, the request and its corresponding message buffer should be freed
+ * via ssam_request_sync_free(). The buffer must not be freed separately.
+ *
+ * Return: Returns zero on success, %-ENOMEM if the request could not be
+ * allocated.
+ */
+int ssam_request_sync_alloc(size_t payload_len, gfp_t flags,
+			    struct ssam_request_sync **rqst,
+			    struct ssam_span *buffer)
+{
+	size_t msglen = SSH_COMMAND_MESSAGE_LENGTH(payload_len);
+
+	*rqst = kzalloc(sizeof(**rqst) + msglen, flags);
+	if (!*rqst)
+		return -ENOMEM;
+
+	buffer->ptr = (u8 *)(*rqst + 1);
+	buffer->len = msglen;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(ssam_request_sync_alloc);
+
+/**
+ * ssam_request_sync_free() - Free a synchronous request.
+ * @rqst: The request to free.
+ *
+ * Free a synchronous request and its corresponding buffer allocated with
+ * ssam_request_sync_alloc(). Do not use for requests allocated on the stack
+ * or via any other function.
+ *
+ * Warning: The caller must ensure that the request is not in use any more.
+ * I.e. the caller must ensure that it has the only reference to the request
+ * and the request is not currently pending. This means that the caller has
+ * either never submitted the request, request submission has failed, or the
+ * caller has waited until the submitted request has been completed via
+ * ssam_request_sync_wait().
+ */
+void ssam_request_sync_free(struct ssam_request_sync *rqst)
+{
+	kfree(rqst);
+}
+EXPORT_SYMBOL_GPL(ssam_request_sync_free);
+
+/**
+ * ssam_request_sync_init() - Initialize a synchronous request struct.
+ * @rqst:  The request to initialize.
+ * @flags: The request flags.
+ *
+ * Initializes the given request struct. Does not initialize the request
+ * message data. This has to be done explicitly after this call via
+ * ssam_request_sync_set_data() and the actual message data has to be written
+ * via ssam_request_write_data().
+ */
+void ssam_request_sync_init(struct ssam_request_sync *rqst,
+			    enum ssam_request_flags flags)
+{
+	ssh_request_init(&rqst->base, flags, &ssam_request_sync_ops);
+	init_completion(&rqst->comp);
+	rqst->resp = NULL;
+	rqst->status = 0;
+}
+EXPORT_SYMBOL_GPL(ssam_request_sync_init);
+
+/**
+ * ssam_request_sync_submit() - Submit a synchronous request.
+ * @ctrl: The controller with which to submit the request.
+ * @rqst: The request to submit.
+ *
+ * Submit a synchronous request. The request has to be initialized and
+ * properly set up, including response buffer (may be %NULL if no response is
+ * expected) and command message data. This function does not wait for the
+ * request to be completed.
+ *
+ * If this function succeeds, ssam_request_sync_wait() must be used to ensure
+ * that the request has been completed before the response data can be
+ * accessed and/or the request can be freed. On failure, the request may
+ * immediately be freed.
+ *
+ * This function may only be used if the controller is active, i.e. has been
+ * initialized and not suspended.
+ */
+int ssam_request_sync_submit(struct ssam_controller *ctrl,
+			     struct ssam_request_sync *rqst)
+{
+	int status;
+
+	/*
+	 * This is only a superficial check. In general, the caller needs to
+	 * ensure that the controller is initialized and is not (and does not
+	 * get) suspended during use, i.e. until the request has been completed
+	 * (if _absolutely_ necessary, by use of ssam_controller_statelock/
+	 * ssam_controller_stateunlock, but something like ssam_client_link
+	 * should be preferred as this needs to last until the request has been
+	 * completed).
+	 *
+	 * Note that it is actually safe to use this function while the
+	 * controller is in the process of being shut down (as ssh_rtl_submit
+	 * is safe with regards to this), but it is generally discouraged to do
+	 * so.
+	 */
+	if (WARN_ON(READ_ONCE(ctrl->state) != SSAM_CONTROLLER_STARTED)) {
+		ssh_request_put(&rqst->base);
+		return -ENXIO;
+	}
+
+	status = ssh_rtl_submit(&ctrl->rtl, &rqst->base);
+	ssh_request_put(&rqst->base);
+
+	return status;
+}
+EXPORT_SYMBOL_GPL(ssam_request_sync_submit);
+
+/**
+ * ssam_request_sync() - Execute a synchronous request.
+ * @ctrl: The controller via which the request will be submitted.
+ * @spec: The request specification and payload.
+ * @rsp:  The response buffer.
+ *
+ * Allocates a synchronous request with its message data buffer on the heap
+ * via ssam_request_sync_alloc(), fully intializes it via the provided request
+ * specification, submits it, and finally waits for its completion before
+ * freeing it and returning its status.
+ *
+ * Return: Returns the status of the request or any failure during setup.
+ */
+int ssam_request_sync(struct ssam_controller *ctrl,
+		      const struct ssam_request *spec,
+		      struct ssam_response *rsp)
+{
+	struct ssam_request_sync *rqst;
+	struct ssam_span buf;
+	ssize_t len;
+	int status;
+
+	status = ssam_request_sync_alloc(spec->length, GFP_KERNEL, &rqst, &buf);
+	if (status)
+		return status;
+
+	ssam_request_sync_init(rqst, spec->flags);
+	ssam_request_sync_set_resp(rqst, rsp);
+
+	len = ssam_request_write_data(&buf, ctrl, spec);
+	if (len < 0) {
+		ssam_request_sync_free(rqst);
+		return len;
+	}
+
+	ssam_request_sync_set_data(rqst, buf.ptr, len);
+
+	status = ssam_request_sync_submit(ctrl, rqst);
+	if (!status)
+		status = ssam_request_sync_wait(rqst);
+
+	ssam_request_sync_free(rqst);
+	return status;
+}
+EXPORT_SYMBOL_GPL(ssam_request_sync);
+
+/**
+ * ssam_request_sync_with_buffer() - Execute a synchronous request with the
+ * provided buffer as backend for the message buffer.
+ * @ctrl: The controller via which the request will be submitted.
+ * @spec: The request specification and payload.
+ * @rsp:  The response buffer.
+ * @buf:  The buffer for the request message data.
+ *
+ * Allocates a synchronous request struct on the stack, fully initializes it
+ * using the provided buffer as message data buffer, submits it, and then
+ * waits for its completion before returning its staus. The
+ * SSH_COMMAND_MESSAGE_LENGTH() macro can be used to compute the required
+ * message buffer size.
+ *
+ * This function does essentially the same as ssam_request_sync(), but instead
+ * of dynamically allocating the request and message data buffer, it uses the
+ * provided message data buffer and stores the (small) request struct on the
+ * heap.
+ *
+ * Return: Returns the status of the request or any failure during setup.
+ */
+int ssam_request_sync_with_buffer(struct ssam_controller *ctrl,
+				  const struct ssam_request *spec,
+				  struct ssam_response *rsp,
+				  struct ssam_span *buf)
+{
+	struct ssam_request_sync rqst;
+	ssize_t len;
+	int status;
+
+	ssam_request_sync_init(&rqst, spec->flags);
+	ssam_request_sync_set_resp(&rqst, rsp);
+
+	len = ssam_request_write_data(buf, ctrl, spec);
+	if (len < 0)
+		return len;
+
+	ssam_request_sync_set_data(&rqst, buf->ptr, len);
+
+	status = ssam_request_sync_submit(ctrl, &rqst);
+	if (!status)
+		status = ssam_request_sync_wait(&rqst);
+
+	return status;
+}
+EXPORT_SYMBOL_GPL(ssam_request_sync_with_buffer);
+
+
+/* -- Internal SAM requests. ------------------------------------------------ */
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_ssh_get_firmware_version, __le32, {
+	.target_category = SSAM_SSH_TC_SAM,
+	.target_id       = 0x01,
+	.command_id      = 0x13,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_ssh_notif_display_off, u8, {
+	.target_category = SSAM_SSH_TC_SAM,
+	.target_id       = 0x01,
+	.command_id      = 0x15,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_ssh_notif_display_on, u8, {
+	.target_category = SSAM_SSH_TC_SAM,
+	.target_id       = 0x01,
+	.command_id      = 0x16,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_ssh_notif_d0_exit, u8, {
+	.target_category = SSAM_SSH_TC_SAM,
+	.target_id       = 0x01,
+	.command_id      = 0x33,
+	.instance_id     = 0x00,
+});
+
+static SSAM_DEFINE_SYNC_REQUEST_R(ssam_ssh_notif_d0_entry, u8, {
+	.target_category = SSAM_SSH_TC_SAM,
+	.target_id       = 0x01,
+	.command_id      = 0x34,
+	.instance_id     = 0x00,
+});
+
+/**
+ * struct ssh_notification_params - Command payload to enable/disable SSH
+ * notifications.
+ * @target_category: The target category for which notifications should be
+ *                   enabled/disabled.
+ * @flags:           Flags determining how notifications are being sent.
+ * @request_id:      The request ID that is used to send these notifications.
+ * @instance_id:     The specific instance in the given target category for
+ *                   which notifications should be enabled.
+ */
+struct ssh_notification_params {
+	u8 target_category;
+	u8 flags;
+	__le16 request_id;
+	u8 instance_id;
+} __packed;
+
+static_assert(sizeof(struct ssh_notification_params) == 5);
+
+/**
+ * ssam_ssh_event_enable() - Enable SSH event.
+ * @ctrl:  The controller for which to enable the event.
+ * @reg:   The event registry describing what request to use for enabling and
+ *         disabling the event.
+ * @id:    The event identifier.
+ * @flags: The event flags.
+ *
+ * This is a wrapper for the raw SAM request to enable an event, thus it does
+ * not handle referecnce counting for enable/disable of events. If an event
+ * has already been enabled, the EC will ignore this request.
+ *
+ * Return: Returns the status of the executed SAM request (zero on success and
+ * negative on direct failure) or %-EPROTO if the request response indicates a
+ * failure.
+ */
+static int ssam_ssh_event_enable(struct ssam_controller *ctrl,
+				 struct ssam_event_registry reg,
+				 struct ssam_event_id id, u8 flags)
+{
+	struct ssh_notification_params params;
+	struct ssam_request rqst;
+	struct ssam_response result;
+	int status;
+
+	u16 rqid = ssh_tc_to_rqid(id.target_category);
+	u8 buf[1] = { 0x00 };
+
+	// only allow RQIDs that lie within event spectrum
+	if (!ssh_rqid_is_event(rqid))
+		return -EINVAL;
+
+	params.target_category = id.target_category;
+	params.instance_id = id.instance;
+	params.flags = flags;
+	put_unaligned_le16(rqid, &params.request_id);
+
+	rqst.target_category = reg.target_category;
+	rqst.target_id = reg.target_id;
+	rqst.command_id = reg.cid_enable;
+	rqst.instance_id = 0x00;
+	rqst.flags = SSAM_REQUEST_HAS_RESPONSE;
+	rqst.length = sizeof(params);
+	rqst.payload = (u8 *)&params;
+
+	result.capacity = ARRAY_SIZE(buf);
+	result.length = 0;
+	result.pointer = buf;
+
+	status = ssam_retry(ssam_request_sync_onstack, ctrl, &rqst, &result,
+			    sizeof(params));
+	if (status) {
+		ssam_err(ctrl, "failed to enable event source (tc: 0x%02x, "
+			 "iid: 0x%02x, reg: 0x%02x)\n", id.target_category,
+			 id.instance, reg.target_category);
+	}
+
+	if (buf[0] != 0x00) {
+		ssam_err(ctrl, "unexpected result while enabling event source: "
+			 "0x%02x (tc: 0x%02x, iid: 0x%02x, reg: 0x%02x)\n",
+			 buf[0], id.target_category, id.instance,
+			 reg.target_category);
+		return -EPROTO;
+	}
+
+	return status;
+
+}
+
+/**
+ * ssam_ssh_event_disable() - Disable SSH event.
+ * @ctrl:  The controller for which to disable the event.
+ * @reg:   The event registry describing what request to use for enabling and
+ *         disabling the event (must be same as used when enabling the event).
+ * @id:    The event identifier.
+ * @flags: The event flags (likely ignored for disabling of events).
+ *
+ * This is a wrapper for the raw SAM request to disable an event, thus it does
+ * not handle reference counting for enable/disable of events. If an event has
+ * already been disabled, the EC will ignore this request.
+ *
+ * Return: Returns the status of the executed SAM request (zero on success and
+ * negative on direct failure) or %-EPROTO if the request response indicates a
+ * failure.
+ */
+static int ssam_ssh_event_disable(struct ssam_controller *ctrl,
+				  struct ssam_event_registry reg,
+				  struct ssam_event_id id, u8 flags)
+{
+	struct ssh_notification_params params;
+	struct ssam_request rqst;
+	struct ssam_response result;
+	int status;
+
+	u16 rqid = ssh_tc_to_rqid(id.target_category);
+	u8 buf[1] = { 0x00 };
+
+	// only allow RQIDs that lie within event spectrum
+	if (!ssh_rqid_is_event(rqid))
+		return -EINVAL;
+
+	params.target_category = id.target_category;
+	params.instance_id = id.instance;
+	params.flags = flags;
+	put_unaligned_le16(rqid, &params.request_id);
+
+	rqst.target_category = reg.target_category;
+	rqst.target_id = reg.target_id;
+	rqst.command_id = reg.cid_disable;
+	rqst.instance_id = 0x00;
+	rqst.flags = SSAM_REQUEST_HAS_RESPONSE;
+	rqst.length = sizeof(params);
+	rqst.payload = (u8 *)&params;
+
+	result.capacity = ARRAY_SIZE(buf);
+	result.length = 0;
+	result.pointer = buf;
+
+	status = ssam_retry(ssam_request_sync_onstack, ctrl, &rqst, &result,
+			    sizeof(params));
+	if (status) {
+		ssam_err(ctrl, "failed to disable event source (tc: 0x%02x, "
+			 "iid: 0x%02x, reg: 0x%02x)\n", id.target_category,
+			 id.instance, reg.target_category);
+	}
+
+	if (buf[0] != 0x00) {
+		ssam_err(ctrl, "unexpected result while disabling event source: "
+			 "0x%02x (tc: 0x%02x, iid: 0x%02x, reg: 0x%02x)\n",
+			 buf[0], id.target_category, id.instance,
+			 reg.target_category);
+		return -EPROTO;
+	}
+
+	return status;
+}
+
+
+/* -- Wrappers for internal SAM requests. ----------------------------------- */
+
+/**
+ * ssam_get_firmware_version() - Get the SAM/EC firmware version.
+ * @ctrl:    The controller.
+ * @version: Where to store the version number.
+ *
+ * Return: Returns zero on success or the status of the executed SAM request
+ * if that request failed.
+ */
+int ssam_get_firmware_version(struct ssam_controller *ctrl, u32 *version)
+{
+	__le32 __version;
+	int status;
+
+	status = ssam_retry(ssam_ssh_get_firmware_version, ctrl, &__version);
+	if (status)
+		return status;
+
+	*version = le32_to_cpu(__version);
+	return 0;
+}
+
+/**
+ * ssam_ctrl_notif_display_off() - Notify EC that the display has been turned
+ * off.
+ * @ctrl: The controller.
+ *
+ * Notify the EC that the display has been turned off and the driver may enter
+ * a lower-power state. This will prevent events from being sent directly.
+ * Rather, the EC signals an event by pulling the wakeup GPIO high for as long
+ * as there are pending events. The events then need to be manually released,
+ * one by one, via the GPIO callback request. All pending events accumulated
+ * during this state can also be released by issuing the display-on
+ * notification, e.g. via ssam_ctrl_notif_display_on(), which will also reset
+ * the GPIO.
+ *
+ * On some devices, specifically ones with an integrated keyboard, the keyboard
+ * backlight will be turned off by this call.
+ *
+ * This function will only send the display-off notification command if
+ * display noticications are supported by the EC. Currently all known devices
+ * support these notification.
+ *
+ * Use ssam_ctrl_notif_display_on() to reverse the effects of this function.
+ *
+ * Return: Returns zero on success or if no request has been executed, the
+ * status of the executed SAM request if that request failed, or %-EPROTO if
+ * an unexpected response has been received.
+ */
+int ssam_ctrl_notif_display_off(struct ssam_controller *ctrl)
+{
+	int status;
+	u8 response;
+
+	ssam_dbg(ctrl, "pm: notifying display off\n");
+
+	status = ssam_retry(ssam_ssh_notif_display_off, ctrl, &response);
+	if (status)
+		return status;
+
+	if (response != 0) {
+		ssam_err(ctrl, "unexpected response from display-off notification: 0x%02x\n",
+			 response);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+/**
+ * ssam_ctrl_notif_display_on() - Notify EC that the display has been turned on.
+ * @ctrl: The controller.
+ *
+ * Notify the EC that the display has been turned back on and the driver has
+ * exited its lower-power state. This notification is the counterpart to the
+ * display-off notification sent via ssam_ctrl_notif_display_off() and will
+ * reverse its effects, including resetting events to their default behavior.
+ *
+ * This function will only send the display-on notification command if display
+ * noticications are supported by the EC. Currently all known devices support
+ * these notification.
+ *
+ * See ssam_ctrl_notif_display_off() for more details.
+ *
+ * Return: Returns zero on success or if no request has been executed, the
+ * status of the executed SAM request if that request failed, or %-EPROTO if
+ * an unexpected response has been received.
+ */
+int ssam_ctrl_notif_display_on(struct ssam_controller *ctrl)
+{
+	int status;
+	u8 response;
+
+	ssam_dbg(ctrl, "pm: notifying display on\n");
+
+	status = ssam_retry(ssam_ssh_notif_display_on, ctrl, &response);
+	if (status)
+		return status;
+
+	if (response != 0) {
+		ssam_err(ctrl, "unexpected response from display-on notification: 0x%02x\n",
+			 response);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+/**
+ * ssam_ctrl_notif_d0_exit() - Notify EC that the driver/device exits the D0
+ * power state.
+ * @ctrl: The controller
+ *
+ * Notifies the EC that the driver prepares to exit the D0 power state in
+ * favor of a lower-power state. Exact effects of this function related to the
+ * EC are currently unknown.
+ *
+ * This function will only send the D0-exit notification command if D0-state
+ * noticications are supported by the EC. Only newer Surface generations
+ * support these notifications.
+ *
+ * Use ssam_ctrl_notif_d0_entry() to reverse the effects of this function.
+ *
+ * Return: Returns zero on success or if no request has been executed, the
+ * status of the executed SAM request if that request failed, or %-EPROTO if
+ * an unexpected response has been received.
+ */
+int ssam_ctrl_notif_d0_exit(struct ssam_controller *ctrl)
+{
+	int status;
+	u8 response;
+
+	if (!ctrl->caps.d3_closes_handle)
+		return 0;
+
+	ssam_dbg(ctrl, "pm: notifying D0 exit\n");
+
+	status = ssam_retry(ssam_ssh_notif_d0_exit, ctrl, &response);
+	if (status)
+		return status;
+
+	if (response != 0) {
+		ssam_err(ctrl, "unexpected response from D0-exit notification:"
+			 " 0x%02x\n", response);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+/**
+ * ssam_ctrl_notif_d0_entry() - Notify EC that the driver/device enters the D0
+ * power state.
+ * @ctrl: The controller
+ *
+ * Notifies the EC that the driver has exited a lower-power state and entered
+ * the D0 power state. Exact effects of this function related to the EC are
+ * currently unknown.
+ *
+ * This function will only send the D0-entry notification command if D0-state
+ * noticications are supported by the EC. Only newer Surface generations
+ * support these notifications.
+ *
+ * See ssam_ctrl_notif_d0_exit() for more details.
+ *
+ * Return: Returns zero on success or if no request has been executed, the
+ * status of the executed SAM request if that request failed, or %-EPROTO if
+ * an unexpected response has been received.
+ */
+int ssam_ctrl_notif_d0_entry(struct ssam_controller *ctrl)
+{
+	int status;
+	u8 response;
+
+	if (!ctrl->caps.d3_closes_handle)
+		return 0;
+
+	ssam_dbg(ctrl, "pm: notifying D0 entry\n");
+
+	status = ssam_retry(ssam_ssh_notif_d0_entry, ctrl, &response);
+	if (status)
+		return status;
+
+	if (response != 0) {
+		ssam_err(ctrl, "unexpected response from D0-entry notification:"
+			 " 0x%02x\n", response);
+		return -EPROTO;
+	}
+
+	return 0;
+}
+
+
+/* -- Top-level event registry interface. ----------------------------------- */
+
+/**
+ * ssam_notifier_register() - Register an event notifier.
+ * @ctrl: The controller to register the notifier on.
+ * @n:    The event notifier to register.
+ *
+ * Register an event notifier and increment the usage counter of the
+ * associated SAM event. If the event was previously not enabled, it will be
+ * enabled during this call.
+ *
+ * Return: Returns zero on success, %-ENOSPC if there have already been
+ * %INT_MAX notifiers for the event ID/type associated with the notifier block
+ * registered, %-ENOMEM if the corresponding event entry could not be
+ * allocated. If this is the first time that a notifier block is registered
+ * for the specific associated event, returns the status of the event-enable
+ * EC-command.
+ */
+int ssam_notifier_register(struct ssam_controller *ctrl,
+			   struct ssam_event_notifier *n)
+{
+	u16 rqid = ssh_tc_to_rqid(n->event.id.target_category);
+	struct ssam_nf_refcount_entry *entry;
+	struct ssam_nf_head *nf_head;
+	struct ssam_nf *nf;
+	int status;
+
+	if (!ssh_rqid_is_event(rqid))
+		return -EINVAL;
+
+	nf = &ctrl->cplt.event.notif;
+	nf_head = &nf->head[ssh_rqid_to_event(rqid)];
+
+	mutex_lock(&nf->lock);
+
+	entry = ssam_nf_refcount_inc(nf, n->event.reg, n->event.id);
+	if (IS_ERR(entry)) {
+		mutex_unlock(&nf->lock);
+		return PTR_ERR(entry);
+	}
+
+	ssam_dbg(ctrl, "enabling event (reg: 0x%02x, tc: 0x%02x, iid: 0x%02x,"
+		 " rc: %d)\n", n->event.reg.target_category,
+		 n->event.id.target_category, n->event.id.instance,
+		 entry->refcount);
+
+	status = __ssam_nfblk_insert(nf_head, &n->base);
+	if (status) {
+		entry = ssam_nf_refcount_dec(nf, n->event.reg, n->event.id);
+		if (entry->refcount == 0)
+			kfree(entry);
+
+		mutex_unlock(&nf->lock);
+		return status;
+	}
+
+	if (entry->refcount == 1) {
+		status = ssam_ssh_event_enable(ctrl, n->event.reg, n->event.id,
+					       n->event.flags);
+		if (status) {
+			__ssam_nfblk_remove(nf_head, &n->base);
+			kfree(ssam_nf_refcount_dec(nf, n->event.reg, n->event.id));
+			mutex_unlock(&nf->lock);
+			synchronize_srcu(&nf_head->srcu);
+			return status;
+		}
+
+		entry->flags = n->event.flags;
+
+	} else if (entry->flags != n->event.flags) {
+		ssam_warn(ctrl, "inconsistent flags when enabling event: got 0x%02x,"
+			  " expected 0x%02x (reg: 0x%02x, tc: 0x%02x, iid: 0x%02x)",
+			  n->event.flags, entry->flags, n->event.reg.target_category,
+			  n->event.id.target_category, n->event.id.instance);
+	}
+
+	mutex_unlock(&nf->lock);
+	return 0;
+
+}
+EXPORT_SYMBOL_GPL(ssam_notifier_register);
+
+/**
+ * ssam_notifier_unregister() - Unregister an event notifier.
+ * @ctrl: The controller the notifier has been registered on.
+ * @n:    The event notifier to unregister.
+ *
+ * Unregister an event notifier and decrement the usage counter of the
+ * associated SAM event. If the usage counter reaches zero, the event will be
+ * disabled.
+ *
+ * Return: Returns zero on success, %-ENOENT if the given notifier block has
+ * not been registered on the controller. If the given notifier block was the
+ * last one associated with its specific event, returns the status of the
+ * event-disable EC-command.
+ */
+int ssam_notifier_unregister(struct ssam_controller *ctrl,
+			     struct ssam_event_notifier *n)
+{
+	u16 rqid = ssh_tc_to_rqid(n->event.id.target_category);
+	struct ssam_notifier_block **link;
+	struct ssam_nf_refcount_entry *entry;
+	struct ssam_nf_head *nf_head;
+	struct ssam_nf *nf;
+	int status = 0;
+
+	if (!ssh_rqid_is_event(rqid))
+		return -EINVAL;
+
+	nf = &ctrl->cplt.event.notif;
+	nf_head = &nf->head[ssh_rqid_to_event(rqid)];
+
+	mutex_lock(&nf->lock);
+
+	link = __ssam_nfblk_find_link(nf_head, &n->base);
+	if (!link) {
+		mutex_unlock(&nf->lock);
+		return -ENOENT;
+	}
+
+	entry = ssam_nf_refcount_dec(nf, n->event.reg, n->event.id);
+	if (WARN_ON(!entry)) {
+		mutex_unlock(&nf->lock);
+		return -ENOENT;
+	}
+
+	ssam_dbg(ctrl, "disabling event (reg: 0x%02x, tc: 0x%02x, iid: 0x%02x,"
+		 " rc: %d)\n", n->event.reg.target_category,
+		 n->event.id.target_category, n->event.id.instance,
+		 entry->refcount);
+
+	if (entry->flags != n->event.flags) {
+		ssam_warn(ctrl, "inconsistent flags when enabling event: got 0x%02x,"
+			  " expected 0x%02x (reg: 0x%02x, tc: 0x%02x, iid: 0x%02x)",
+			  n->event.flags, entry->flags, n->event.reg.target_category,
+			  n->event.id.target_category, n->event.id.instance);
+	}
+
+	if (entry->refcount == 0) {
+		status = ssam_ssh_event_disable(ctrl, n->event.reg, n->event.id,
+						n->event.flags);
+		kfree(entry);
+	}
+
+	__ssam_nfblk_erase(link);
+	mutex_unlock(&nf->lock);
+	synchronize_srcu(&nf_head->srcu);
+
+	return status;
+}
+EXPORT_SYMBOL_GPL(ssam_notifier_unregister);
+
+/**
+ * ssam_notifier_disable_registered() - Disable events for all registered
+ * notifiers.
+ * @ctrl: The controller for which to disable the notifiers/events.
+ *
+ * Disables events for all currently registered notifiers. In case of an error
+ * (EC command failing), all previously disabled events will be restored and
+ * the error code returned.
+ *
+ * This function is intended to disable all events prior to hibenration entry.
+ * See ssam_notifier_restore_registered() to restore/re-enable all events
+ * disabled with this function.
+ *
+ * Note that this function will not disable events for notifiers registered
+ * after calling this function. It should thus be made sure that no new
+ * notifiers are going to be added after this call and before the corresponding
+ * call to ssam_notifier_restore_registered().
+ *
+ * Return: Returns zero on success. In case of failure returns the error code
+ * returned by the failed EC command to disable an event.
+ */
+int ssam_notifier_disable_registered(struct ssam_controller *ctrl)
+{
+	struct ssam_nf *nf = &ctrl->cplt.event.notif;
+	struct rb_node *n;
+	int status;
+
+	mutex_lock(&nf->lock);
+	for (n = rb_first(&nf->refcount); n != NULL; n = rb_next(n)) {
+		struct ssam_nf_refcount_entry *e;
+
+		e = rb_entry(n, struct ssam_nf_refcount_entry, node);
+		status = ssam_ssh_event_disable(ctrl, e->key.reg,
+						e->key.id, e->flags);
+		if (status)
+			goto err;
+	}
+	mutex_unlock(&nf->lock);
+
+	return 0;
+
+err:
+	for (n = rb_prev(n); n != NULL; n = rb_prev(n)) {
+		struct ssam_nf_refcount_entry *e;
+
+		e = rb_entry(n, struct ssam_nf_refcount_entry, node);
+		ssam_ssh_event_enable(ctrl, e->key.reg, e->key.id, e->flags);
+	}
+	mutex_unlock(&nf->lock);
+
+	return status;
+}
+
+/**
+ * ssam_notifier_restore_registered() - Restore/re-enable events for all
+ * registered notifiers.
+ * @ctrl: The controller for which to restore the notifiers/events.
+ *
+ * Restores/re-enables all events for which notifiers have been registered on
+ * the given controller. In case of a failure, the error is logged and the
+ * function continues to try and enable the remaining events.
+ *
+ * This function is intended to restore/re-enable all registered events after
+ * hibernation. See ssam_notifier_disable_registered() for the counter part
+ * disabling the events and more details.
+ */
+void ssam_notifier_restore_registered(struct ssam_controller *ctrl)
+{
+	struct ssam_nf *nf = &ctrl->cplt.event.notif;
+	struct rb_node *n;
+
+	mutex_lock(&nf->lock);
+	for (n = rb_first(&nf->refcount); n != NULL; n = rb_next(n)) {
+		struct ssam_nf_refcount_entry *e;
+
+		e = rb_entry(n, struct ssam_nf_refcount_entry, node);
+
+		// ignore errors, will get logged in call
+		ssam_ssh_event_enable(ctrl, e->key.reg, e->key.id, e->flags);
+	}
+	mutex_unlock(&nf->lock);
+}
+
+/**
+ * ssam_notifier_empty() - Check if there are any registered notifiers.
+ * @ctrl: The controller to check on.
+ *
+ * Return: Returns %true if there are currently no notifiers registered on the
+ * controller, %false otherwise.
+ */
+static bool ssam_notifier_empty(struct ssam_controller *ctrl)
+{
+	struct ssam_nf *nf = &ctrl->cplt.event.notif;
+	bool result;
+
+	mutex_lock(&nf->lock);
+	result = ssam_nf_refcount_empty(nf);
+	mutex_unlock(&nf->lock);
+
+	return result;
+}
+
+/**
+ * ssam_notifier_unregister_all() - Unregister all currently registered
+ * notifiers.
+ * @ctrl: The controller to unregister the notifiers on.
+ *
+ * Unregisters all currently registered notifiers. This function is used to
+ * ensure that all notifiers will be unregistered and assocaited
+ * entries/resources freed when the controller is being shut down.
+ */
+static void ssam_notifier_unregister_all(struct ssam_controller *ctrl)
+{
+	struct ssam_nf *nf = &ctrl->cplt.event.notif;
+	struct ssam_nf_refcount_entry *e, *n;
+
+	mutex_lock(&nf->lock);
+	rbtree_postorder_for_each_entry_safe(e, n, &nf->refcount, node) {
+		// ignore errors, will get logged in call
+		ssam_ssh_event_disable(ctrl, e->key.reg, e->key.id, e->flags);
+		kfree(e);
+	}
+	nf->refcount = RB_ROOT;
+	mutex_unlock(&nf->lock);
+}
+
+
+/* -- Wakeup IRQ. ----------------------------------------------------------- */
+
+static irqreturn_t ssam_irq_handle(int irq, void *dev_id)
+{
+	struct ssam_controller *ctrl = dev_id;
+
+	ssam_dbg(ctrl, "pm: wake irq triggered\n");
+
+	/*
+	 * Note: Proper wakeup detection is currently unimplemented.
+	 *       When the EC is in display-off or any other non-D0 state, it
+	 *       does not send events/notifications to the host. Instead it
+	 *       signals that there are events available via the wakeup IRQ.
+	 *       This driver is responsible for calling back to the EC to
+	 *       release these events one-by-one.
+	 *
+	 *       This IRQ should not cause a full system resume by its own.
+	 *       Instead, events should be handled by their respective subsystem
+	 *       drivers, which in turn should signal whether a full system
+	 *       resume should be performed.
+	 *
+	 * TODO: Send GPIO callback command repeatedly to EC until callback
+	 *       returns 0x00. Return flag of callback is "has more events".
+	 *       Each time the command is sent, one event is "released". Once
+	 *       all events have been released (return = 0x00), the GPIO is
+	 *       re-armed. Detect wakeup events during this process, go back to
+	 *       sleep if no wakeup event has been received.
+	 */
+
+	return IRQ_HANDLED;
+}
+
+/**
+ * ssam_irq_setup() - Set up SAM EC wakeup-GPIO interrupt.
+ * @ctrl: The controller for which the IRQ should be set up.
+ *
+ * Set up an IRQ for the wakeup-GPIO pin of the SAM EC. This IRQ can be used
+ * to wake the device from a low power state.
+ *
+ * Note that this IRQ can only be triggered while the EC is in the display-off
+ * state. In this state, events are not sent to the host in the usual way.
+ * Instead the wakeup-GPIO gets pulled to "high" as long as there are pending
+ * events and these events need to be released one-by-one via the GPIO
+ * callback request, either until there are no events left and the GPIO is
+ * reset, or all at once by transitioning the EC out of the display-off state,
+ * which will also clear the GPIO.
+ *
+ * Not all events, however, should trigger a full system wakeup. Instead the
+ * driver should, if necessary, inspect and forward each event to the
+ * corresponding subsystem, which in turn should decide if the system needs to
+ * be woken up. This logic has not been implemented yet, thus wakeup by this
+ * IRQ should be disabled by default to avoid spurious wake-ups, caused, for
+ * example, by the remaining battery percentage changing. Refer to comments in
+ * this function and comments in the corresponding IRQ handler for more
+ * details on how this should be implemented.
+ *
+ * See also ssam_ctrl_notif_display_off() and ssam_ctrl_notif_display_off()
+ * for functions to transition the EC into and out of the display-off state as
+ * well as more details on it.
+ *
+ * The IRQ is disabled by default and has to be enabled before it can wake up
+ * the device from suspend via ssam_irq_arm_for_wakeup(). On teardown, the IRQ
+ * should be freed via ssam_irq_free().
+ */
+int ssam_irq_setup(struct ssam_controller *ctrl)
+{
+	struct device *dev = ssam_controller_device(ctrl);
+	struct gpio_desc *gpiod;
+	int irq;
+	int status;
+
+	/*
+	 * The actual GPIO interrupt is declared in ACPI as TRIGGER_HIGH.
+	 * However, the GPIO line only gets reset by sending the GPIO callback
+	 * command to SAM (or alternatively the display-on notification). As
+	 * proper handling for this interrupt is not implemented yet, leaving
+	 * the IRQ at TRIGGER_HIGH would cause an IRQ storm (as the callback
+	 * never gets sent and thus the line never gets reset). To avoid this,
+	 * mark the IRQ as TRIGGER_RISING for now, only creating a single
+	 * interrupt, and let the SAM resume callback during the controller
+	 * resume process clear it.
+	 */
+	const int irqf = IRQF_SHARED | IRQF_ONESHOT | IRQF_TRIGGER_RISING;
+
+	gpiod = gpiod_get(dev, "ssam_wakeup-int", GPIOD_ASIS);
+	if (IS_ERR(gpiod))
+		return PTR_ERR(gpiod);
+
+	irq = gpiod_to_irq(gpiod);
+	gpiod_put(gpiod);
+
+	if (irq < 0)
+		return irq;
+
+	status = request_threaded_irq(irq, NULL, ssam_irq_handle, irqf,
+				      "ssam_wakeup", ctrl);
+	if (status)
+		return status;
+
+	ctrl->irq.num = irq;
+	disable_irq(ctrl->irq.num);
+	return 0;
+}
+
+/**
+ * ssam_irq_free() - Free SAM EC wakeup-GPIO interrupt.
+ * @ctrl: The controller for which the IRQ should be freed.
+ *
+ * Free the wakeup-GPIO IRQ previously set-up via ssam_irq_setup().
+ */
+void ssam_irq_free(struct ssam_controller *ctrl)
+{
+	free_irq(ctrl->irq.num, ctrl);
+	ctrl->irq.num = -1;
+}
+
+/**
+ * ssam_irq_arm_for_wakeup() - Arm the EC IRQ for wakeup, if enabled.
+ * @ctrl: The controller for which the IRQ should be armed.
+ *
+ * Sets up the IRQ so that it can be used to wake the device. Specifically,
+ * this function enables the irq and then, if the device is allowed to wake up
+ * the system, calls enable_irq_wake(). See ssam_irq_disarm_wakeup() for the
+ * corresponding function to disable the IRQ.
+ *
+ * This function is intended to arm the IRQ before entering S2idle suspend.
+ *
+ * Note: calls to ssam_irq_arm_for_wakeup() and ssam_irq_disarm_wakeup() must
+ * be balanced.
+ */
+int ssam_irq_arm_for_wakeup(struct ssam_controller *ctrl)
+{
+	struct device *dev = ssam_controller_device(ctrl);
+	int status;
+
+	enable_irq(ctrl->irq.num);
+	if (device_may_wakeup(dev)) {
+		status = enable_irq_wake(ctrl->irq.num);
+		if (status) {
+			ssam_err(ctrl, "failed to enable wake IRQ: %d\n", status);
+			disable_irq(ctrl->irq.num);
+			return status;
+		}
+
+		ctrl->irq.wakeup_enabled = true;
+	} else {
+		ctrl->irq.wakeup_enabled = false;
+	}
+
+	return 0;
+}
+
+/**
+ * ssam_irq_disarm_wakeup() - Disarm the wakeup IRQ.
+ * @ctrl: The controller for which the IRQ should be disarmed.
+ *
+ * Disarm the IRQ previously set up for wake via ssam_irq_arm_for_wakeup().
+ *
+ * This function is intended to disarm the IRQ after exiting S2idle suspend.
+ *
+ * Note: calls to ssam_irq_arm_for_wakeup() and ssam_irq_disarm_wakeup() must
+ * be balanced.
+ */
+void ssam_irq_disarm_wakeup(struct ssam_controller *ctrl)
+{
+	int status;
+
+	if (ctrl->irq.wakeup_enabled) {
+		status = disable_irq_wake(ctrl->irq.num);
+		if (status)
+			ssam_err(ctrl, "failed to disable wake IRQ: %d\n", status);
+
+		ctrl->irq.wakeup_enabled = false;
+	}
+	disable_irq(ctrl->irq.num);
+}
diff --git a/drivers/misc/surface_aggregator/controller.h b/drivers/misc/surface_aggregator/controller.h
new file mode 100644
index 000000000000..96e2b87a25d9
--- /dev/null
+++ b/drivers/misc/surface_aggregator/controller.h
@@ -0,0 +1,288 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Main SSAM/SSH controller structure and functionality.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _SURFACE_AGGREGATOR_CONTROLLER_H
+#define _SURFACE_AGGREGATOR_CONTROLLER_H
+
+#include <linux/kref.h>
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/rbtree.h>
+#include <linux/rwsem.h>
+#include <linux/serdev.h>
+#include <linux/spinlock.h>
+#include <linux/srcu.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/serial_hub.h>
+
+#include "ssh_request_layer.h"
+
+
+/* -- Safe counters. -------------------------------------------------------- */
+
+/**
+ * struct ssh_seq_counter - Safe counter for SSH sequence IDs.
+ * @value: The current counter value.
+ */
+struct ssh_seq_counter {
+	u8 value;
+};
+
+/**
+ * struct ssh_rqid_counter - Safe counter for SSH request IDs.
+ * @value: The current counter value.
+ */
+struct ssh_rqid_counter {
+	u16 value;
+};
+
+
+/* -- Event/notification system. -------------------------------------------- */
+
+/**
+ * struct ssam_nf_head - Notifier head for SSAM events.
+ * @srcu: The SRCU struct for synchronization.
+ * @head: Head-pointer for the single-linked list of notifier blocks registered
+ *        under this head.
+ */
+struct ssam_nf_head {
+	struct srcu_struct srcu;
+	struct ssam_notifier_block __rcu *head;
+};
+
+/**
+ * struct ssam_nf - Notifier callback- and activation-registry for SSAM events.
+ * @lock:     Lock guarding (de-)registration of notifier blocks. Note: This
+ *            lock does not need to be held for notifier calls, only
+ *            registration and deregistration.
+ * @refcount: The root of the RB-tree used for reference-counting enabled
+ *            events/notifications.
+ * @head:     The list of notifier heads for event/notifiaction callbacks.
+ */
+struct ssam_nf {
+	struct mutex lock;
+	struct rb_root refcount;
+	struct ssam_nf_head head[SSH_NUM_EVENTS];
+};
+
+
+/* -- Event/async request completion system. -------------------------------- */
+
+struct ssam_cplt;
+
+/**
+ * struct ssam_event_item - Struct for event queuing and completion.
+ * @node:     The node in the queue.
+ * @rqid:     The request ID of the event.
+ * @ops:      Instance specific functions.
+ * @ops.free: Callback for freeing this event item.
+ * @event:    Actual event data.
+ */
+struct ssam_event_item {
+	struct list_head node;
+	u16 rqid;
+
+	struct {
+		void (*free)(struct ssam_event_item *event);
+	} ops;
+
+	struct ssam_event event;	// must be last
+};
+
+/**
+ * struct ssam_event_queue - Queue for completing received events.
+ * @cplt: Reference to the completion system on which this queue is active.
+ * @lock: The lock for any operation on the queue.
+ * @head: The list-head of the queue.
+ * @work: The &struct work_struct performing completion work for this queue.
+ */
+struct ssam_event_queue {
+	struct ssam_cplt *cplt;
+
+	spinlock_t lock;
+	struct list_head head;
+	struct work_struct work;
+};
+
+/**
+ * struct ssam_event_target - Set of queues for a single SSH target ID.
+ * @queue: The array of queues, one queue per event ID.
+ */
+struct ssam_event_target {
+	struct ssam_event_queue queue[SSH_NUM_EVENTS];
+};
+
+/**
+ * struct ssam_cplt - SSAM event/async request completion system.
+ * @dev:          The device with which this system is associated. Only used
+ *                for logging.
+ * @wq:           The &struct workqueue_struct on which all completion work
+ *                items are queued.
+ * @event:        Event completion management.
+ * @event.target: Array of &struct ssam_event_target, one for each target.
+ * @event.notif:  Notifier callbacks and event activation reference counting.
+ */
+struct ssam_cplt {
+	struct device *dev;
+	struct workqueue_struct *wq;
+
+	struct {
+		struct ssam_event_target target[SSH_NUM_TARGETS];
+		struct ssam_nf notif;
+	} event;
+};
+
+
+/* -- Main SSAM device structures. ------------------------------------------ */
+
+/**
+ * enum ssam_controller_state - State values for &struct ssam_controller.
+ * @SSAM_CONTROLLER_UNINITIALIZED:
+ *	The controller has not been initialized yet or has been deinitialized.
+ * @SSAM_CONTROLLER_INITIALIZED:
+ *	The controller is initialized, but has not been started yet.
+ * @SSAM_CONTROLLER_STARTED:
+ *	The controller has been started and is ready to use.
+ * @SSAM_CONTROLLER_STOPPED:
+ *	The controller has been stopped.
+ * @SSAM_CONTROLLER_SUSPENDED:
+ *	The controller has been suspended.
+ */
+enum ssam_controller_state {
+	SSAM_CONTROLLER_UNINITIALIZED,
+	SSAM_CONTROLLER_INITIALIZED,
+	SSAM_CONTROLLER_STARTED,
+	SSAM_CONTROLLER_STOPPED,
+	SSAM_CONTROLLER_SUSPENDED,
+};
+
+/**
+ * struct ssam_controller_caps - Controller device capabilities.
+ * @ssh_power_profile:             SSH power profile.
+ * @ssh_buffer_size:               SSH driver UART buffer size.
+ * @screen_on_sleep_idle_timeout:  SAM UART screen-on sleep idle timeout.
+ * @screen_off_sleep_idle_timeout: SAM UART screen-off sleep idle timeout.
+ * @d3_closes_handle:              SAM closes UART handle in D3.
+ *
+ * Controller and SSH device capabilities found in ACPI.
+ */
+struct ssam_controller_caps {
+	u32 ssh_power_profile;
+	u32 ssh_buffer_size;
+	u32 screen_on_sleep_idle_timeout;
+	u32 screen_off_sleep_idle_timeout;
+	u32 d3_closes_handle:1;
+};
+
+/**
+ * struct ssam_controller - SSAM controller device.
+ * @kref:  Reference count of the controller.
+ * @lock:  Main lock for the controller, used to guard state changes.
+ * @state: Controller state.
+ * @rtl:   Request transport layer for SSH I/O.
+ * @cplt:  Completion system for SSH/SSAM events and asynchronous requests.
+ * @counter:      Safe SSH message ID counters.
+ * @counter.seq:  Sequence ID counter.
+ * @counter.rqid: Request ID counter.
+ * @irq:          Wakeup IRQ resources.
+ * @irq.num:      The wakeup IRQ number.
+ * @irq.wakeup_enabled: Whether wakeup by IRQ is enabled during suspend.
+ * @caps: The controller device capabilities.
+ */
+struct ssam_controller {
+	struct kref kref;
+
+	struct rw_semaphore lock;
+	enum ssam_controller_state state;
+
+	struct ssh_rtl rtl;
+	struct ssam_cplt cplt;
+
+	struct {
+		struct ssh_seq_counter seq;
+		struct ssh_rqid_counter rqid;
+	} counter;
+
+	struct {
+		int num;
+		bool wakeup_enabled;
+	} irq;
+
+	struct ssam_controller_caps caps;
+};
+
+#define to_ssam_controller(ptr, member) \
+	container_of(ptr, struct ssam_controller, member)
+
+#define ssam_dbg(ctrl, fmt, ...)  rtl_dbg(&(ctrl)->rtl, fmt, ##__VA_ARGS__)
+#define ssam_info(ctrl, fmt, ...) rtl_info(&(ctrl)->rtl, fmt, ##__VA_ARGS__)
+#define ssam_warn(ctrl, fmt, ...) rtl_warn(&(ctrl)->rtl, fmt, ##__VA_ARGS__)
+#define ssam_err(ctrl, fmt, ...)  rtl_err(&(ctrl)->rtl, fmt, ##__VA_ARGS__)
+
+
+/**
+ * ssam_controller_receive_buf() - Provide input-data to the controller.
+ * @ctrl: The controller.
+ * @buf:  The input buffer.
+ * @n:    The number of bytes in the input buffer.
+ *
+ * Provide input data to be evaluated by the controller, which has been
+ * received via the lower-level transport.
+ *
+ * Return: Returns the number of bytes consumed, or, if the packet transport
+ * layer of the controller has been shut down, %-ESHUTDOWN.
+ */
+static inline
+int ssam_controller_receive_buf(struct ssam_controller *ctrl,
+				const unsigned char *buf, size_t n)
+{
+	return ssh_ptl_rx_rcvbuf(&ctrl->rtl.ptl, buf, n);
+}
+
+/**
+ * ssam_controller_write_wakeup() - Notify the controller that the underlying
+ * device has space available for data to be written.
+ * @ctrl: The controller.
+ */
+static inline void ssam_controller_write_wakeup(struct ssam_controller *ctrl)
+{
+	ssh_ptl_tx_wakeup(&ctrl->rtl.ptl);
+}
+
+
+int ssam_controller_init(struct ssam_controller *ctrl, struct serdev_device *s);
+int ssam_controller_start(struct ssam_controller *ctrl);
+void ssam_controller_shutdown(struct ssam_controller *ctrl);
+void ssam_controller_destroy(struct ssam_controller *ctrl);
+
+int ssam_notifier_disable_registered(struct ssam_controller *ctrl);
+void ssam_notifier_restore_registered(struct ssam_controller *ctrl);
+
+int ssam_irq_setup(struct ssam_controller *ctrl);
+void ssam_irq_free(struct ssam_controller *ctrl);
+int ssam_irq_arm_for_wakeup(struct ssam_controller *ctrl);
+void ssam_irq_disarm_wakeup(struct ssam_controller *ctrl);
+
+void ssam_controller_lock(struct ssam_controller *c);
+void ssam_controller_unlock(struct ssam_controller *c);
+
+int ssam_get_firmware_version(struct ssam_controller *ctrl, u32 *version);
+int ssam_ctrl_notif_display_off(struct ssam_controller *ctrl);
+int ssam_ctrl_notif_display_on(struct ssam_controller *ctrl);
+int ssam_ctrl_notif_d0_exit(struct ssam_controller *ctrl);
+int ssam_ctrl_notif_d0_entry(struct ssam_controller *ctrl);
+
+int ssam_controller_suspend(struct ssam_controller *ctrl);
+int ssam_controller_resume(struct ssam_controller *ctrl);
+
+int ssam_event_item_cache_init(void);
+void ssam_event_item_cache_destroy(void);
+
+#endif /* _SURFACE_AGGREGATOR_CONTROLLER_H */
diff --git a/drivers/misc/surface_aggregator/core.c b/drivers/misc/surface_aggregator/core.c
new file mode 100644
index 000000000000..6aaac82b9091
--- /dev/null
+++ b/drivers/misc/surface_aggregator/core.c
@@ -0,0 +1,831 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Surface Serial Hub (SSH) driver for communication with the Surface/System
+ * Aggregator Module (SSAM/SAM).
+ *
+ * Provides access to a SAM-over-SSH connected EC via a controller device.
+ * Handles communication via requests as well as enabling, disabling, and
+ * relaying of events.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <linux/acpi.h>
+#include <linux/atomic.h>
+#include <linux/completion.h>
+#include <linux/gpio/consumer.h>
+#include <linux/kernel.h>
+#include <linux/kref.h>
+#include <linux/module.h>
+#include <linux/pm.h>
+#include <linux/serdev.h>
+#include <linux/sysfs.h>
+
+#include <linux/surface_aggregator/controller.h>
+
+#include "bus.h"
+#include "controller.h"
+
+#define CREATE_TRACE_POINTS
+#include "trace.h"
+
+
+/* -- Static controller reference. ------------------------------------------ */
+
+/*
+ * Main controller reference. The corresponding lock must be held while
+ * accessing (reading/writing) the reference.
+ */
+static struct ssam_controller *__ssam_controller;
+static DEFINE_SPINLOCK(__ssam_controller_lock);
+
+/**
+ * ssam_get_controller() - Get reference to SSAM controller.
+ *
+ * Returns a reference to the SSAM controller of the system or %NULL if there
+ * is none, it hasn't been set up yet, or it has already been unregistered.
+ * This function automatically increments the reference count of the
+ * controller, thus the calling party must ensure that ssam_controller_put()
+ * is called when it doesn't need the controller any more.
+ */
+struct ssam_controller *ssam_get_controller(void)
+{
+	struct ssam_controller *ctrl;
+
+	spin_lock(&__ssam_controller_lock);
+
+	ctrl = __ssam_controller;
+	if (!ctrl)
+		goto out;
+
+	if (WARN_ON(!kref_get_unless_zero(&ctrl->kref)))
+		ctrl = NULL;
+
+out:
+	spin_unlock(&__ssam_controller_lock);
+	return ctrl;
+}
+EXPORT_SYMBOL_GPL(ssam_get_controller);
+
+/**
+ * ssam_try_set_controller() - Try to set the main controller reference.
+ * @ctrl: The controller to which the reference should point.
+ *
+ * Set the main controller reference to the given pointer if the reference
+ * hasn't been set already.
+ *
+ * Return: Returns zero on success or %-EBUSY if the reference has already
+ * been set.
+ */
+static int ssam_try_set_controller(struct ssam_controller *ctrl)
+{
+	int status = 0;
+
+	spin_lock(&__ssam_controller_lock);
+	if (!__ssam_controller)
+		__ssam_controller = ctrl;
+	else
+		status = -EBUSY;
+	spin_unlock(&__ssam_controller_lock);
+
+	return status;
+}
+
+/**
+ * ssam_clear_controller() - Remove/clear the main controller reference.
+ *
+ * Clears the main controller reference, i.e. sets it to %NULL. This function
+ * should be called before the controller is shut down.
+ */
+static void ssam_clear_controller(void)
+{
+	spin_lock(&__ssam_controller_lock);
+	__ssam_controller = NULL;
+	spin_unlock(&__ssam_controller_lock);
+}
+
+
+/**
+ * ssam_client_link() - Link an arbitrary client device to the controller.
+ * @c: The controller to link to.
+ * @client: The client device.
+ *
+ * Link an arbitrary client device to the controller by creating a device link
+ * between it as consumer and the controller device as provider. This function
+ * can be used for non-SSAM devices (or SSAM devices not registered as child
+ * under the controller) to guarantee that the controller is valid for as long
+ * as the driver of the client device is bound, and that proper suspend and
+ * resume ordering is guaranteed.
+ *
+ * The device link does not have to be destructed manually. It is removed
+ * automatically once the driver of the client device unbinds.
+ *
+ * Return: Returns zero on success, %-ENXIO if the controller is not ready or
+ * going to be removed soon, or %-ENOMEM if the device link could not be
+ * created for other reasons.
+ */
+int ssam_client_link(struct ssam_controller *c, struct device *client)
+{
+	const u32 flags = DL_FLAG_PM_RUNTIME | DL_FLAG_AUTOREMOVE_CONSUMER;
+	struct device_link *link;
+	struct device *ctrldev;
+
+	ssam_controller_statelock(c);
+
+	if (c->state != SSAM_CONTROLLER_STARTED) {
+		ssam_controller_stateunlock(c);
+		return -ENXIO;
+	}
+
+	ctrldev = ssam_controller_device(c);
+	if (!ctrldev) {
+		ssam_controller_stateunlock(c);
+		return -ENXIO;
+	}
+
+	link = device_link_add(client, ctrldev, flags);
+	if (!link) {
+		ssam_controller_stateunlock(c);
+		return -ENOMEM;
+	}
+
+	/*
+	 * Return -ENXIO if supplier driver is on its way to be removed. In this
+	 * case, the controller won't be around for much longer and the device
+	 * link is not going to save us any more, as unbinding is already in
+	 * progress.
+	 */
+	if (READ_ONCE(link->status) == DL_STATE_SUPPLIER_UNBIND) {
+		ssam_controller_stateunlock(c);
+		return -ENXIO;
+	}
+
+	ssam_controller_stateunlock(c);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(ssam_client_link);
+
+/**
+ * ssam_client_bind() - Bind an arbitrary client device to the controller.
+ * @client: The client device.
+ * @ctrl: A pointer to where the controller reference should be returned.
+ *
+ * Link an arbitrary client device to the controller by creating a device link
+ * between it as consumer and the main controller device as provider. This
+ * function can be used for non-SSAM devices to guarantee that the controller
+ * returned by this function is valid for as long as the driver of the client
+ * device is bound, and that proper suspend and resume ordering is guaranteed.
+ *
+ * This function does essentially the same as ssam_client_link(), except that
+ * it first fetches the main controller reference, then creates the link, and
+ * finally returns this reference in the @ctrl parameter. Note that this
+ * function does not increment the reference counter of the controller, as,
+ * due to the link, the controller lifetime is assured as long as the driver
+ * of the client device is bound.
+ *
+ * It is not valid to use the controller reference obtained by this method
+ * outside of the driver bound to the client device at the time of calling
+ * this function, without first incrementing the reference count of the
+ * controller via ssam_controller_get(). Even after doing this, care must be
+ * taken that requests are only submitted and notifiers are only
+ * (un-)registered when the controller is active and not suspended. In other
+ * words: The device link only lives as long as the client driver is bound and
+ * any guarantees enforced by this link (e.g. active controller state) can
+ * only be relied upon as long as this link exists and may need to be enforced
+ * in other ways afterwards.
+ *
+ * The created device link does not have to be destructed manually. It is
+ * removed automatically once the driver of the client device unbinds.
+ *
+ * Return: Returns zero on success, %-ENXIO if the controller is not present,
+ * not ready or going to be removed soon, or %-ENOMEM if the device link could
+ * not be created for other reasons.
+ */
+int ssam_client_bind(struct device *client, struct ssam_controller **ctrl)
+{
+	struct ssam_controller *c;
+	int status;
+
+	c = ssam_get_controller();
+	if (!c)
+		return -ENXIO;
+
+	status = ssam_client_link(c, client);
+
+	/*
+	 * Note that we can drop our controller reference in both success and
+	 * failure cases: On success, we have bound the controller lifetime
+	 * inherently to the client driver lifetime, i.e. it the controller is
+	 * now guaranteed to outlive the client driver. On failure, we're not
+	 * going to use the controller any more.
+	 */
+	ssam_controller_put(c);
+
+	*ctrl = status == 0 ? c : NULL;
+	return status;
+}
+EXPORT_SYMBOL_GPL(ssam_client_bind);
+
+
+/* -- Glue layer (serdev_device -> ssam_controller). ------------------------ */
+
+static int ssam_receive_buf(struct serdev_device *dev, const unsigned char *buf,
+			    size_t n)
+{
+	struct ssam_controller *ctrl;
+
+	ctrl = serdev_device_get_drvdata(dev);
+	return ssam_controller_receive_buf(ctrl, buf, n);
+}
+
+static void ssam_write_wakeup(struct serdev_device *dev)
+{
+	ssam_controller_write_wakeup(serdev_device_get_drvdata(dev));
+}
+
+static const struct serdev_device_ops ssam_serdev_ops = {
+	.receive_buf = ssam_receive_buf,
+	.write_wakeup = ssam_write_wakeup,
+};
+
+
+/* -- SysFS and misc. ------------------------------------------------------- */
+
+static int ssam_log_firmware_version(struct ssam_controller *ctrl)
+{
+	u32 version, a, b, c;
+	int status;
+
+	status = ssam_get_firmware_version(ctrl, &version);
+	if (status)
+		return status;
+
+	a = (version >> 24) & 0xff;
+	b = ((version >> 8) & 0xffff);
+	c = version & 0xff;
+
+	ssam_info(ctrl, "SAM firmware version: %u.%u.%u\n", a, b, c);
+	return 0;
+}
+
+static ssize_t firmware_version_show(struct device *dev,
+				     struct device_attribute *attr, char *buf)
+{
+	struct ssam_controller *ctrl = dev_get_drvdata(dev);
+	u32 version, a, b, c;
+	int status;
+
+	status = ssam_get_firmware_version(ctrl, &version);
+	if (status < 0)
+		return status;
+
+	a = (version >> 24) & 0xff;
+	b = ((version >> 8) & 0xffff);
+	c = version & 0xff;
+
+	return snprintf(buf, PAGE_SIZE - 1, "%u.%u.%u\n", a, b, c);
+}
+static DEVICE_ATTR_RO(firmware_version);
+
+static struct attribute *ssam_sam_attrs[] = {
+	&dev_attr_firmware_version.attr,
+	NULL,
+};
+
+static const struct attribute_group ssam_sam_group = {
+	.name = "sam",
+	.attrs = ssam_sam_attrs,
+};
+
+
+/* -- ACPI based device setup. ---------------------------------------------- */
+
+static acpi_status ssam_serdev_setup_via_acpi_crs(struct acpi_resource *rsc,
+						  void *ctx)
+{
+	struct serdev_device *serdev = ctx;
+	struct acpi_resource_common_serialbus *serial;
+	struct acpi_resource_uart_serialbus *uart;
+	bool flow_control;
+	int status = 0;
+
+	if (rsc->type != ACPI_RESOURCE_TYPE_SERIAL_BUS)
+		return AE_OK;
+
+	serial = &rsc->data.common_serial_bus;
+	if (serial->type != ACPI_RESOURCE_SERIAL_TYPE_UART)
+		return AE_OK;
+
+	uart = &rsc->data.uart_serial_bus;
+
+	// set up serdev device
+	serdev_device_set_baudrate(serdev, uart->default_baud_rate);
+
+	// serdev currently only supports RTSCTS flow control
+	if (uart->flow_control & (~((u8) ACPI_UART_FLOW_CONTROL_HW))) {
+		dev_warn(&serdev->dev, "setup: unsupported flow control (value: 0x%02x)\n",
+			 uart->flow_control);
+	}
+
+	// set RTSCTS flow control
+	flow_control = uart->flow_control & ACPI_UART_FLOW_CONTROL_HW;
+	serdev_device_set_flow_control(serdev, flow_control);
+
+	// serdev currently only supports EVEN/ODD parity
+	switch (uart->parity) {
+	case ACPI_UART_PARITY_NONE:
+		status = serdev_device_set_parity(serdev, SERDEV_PARITY_NONE);
+		break;
+	case ACPI_UART_PARITY_EVEN:
+		status = serdev_device_set_parity(serdev, SERDEV_PARITY_EVEN);
+		break;
+	case ACPI_UART_PARITY_ODD:
+		status = serdev_device_set_parity(serdev, SERDEV_PARITY_ODD);
+		break;
+	default:
+		dev_warn(&serdev->dev, "setup: unsupported parity (value: 0x%02x)\n",
+			 uart->parity);
+		break;
+	}
+
+	if (status) {
+		dev_err(&serdev->dev, "setup: failed to set parity (value: 0x%02x,"
+			" error: %d)\n", uart->parity, status);
+		return AE_ERROR;
+	}
+
+	return AE_CTRL_TERMINATE;       // we've found the resource and are done
+}
+
+static acpi_status ssam_serdev_setup_via_acpi(acpi_handle handle,
+					      struct serdev_device *serdev)
+{
+	return acpi_walk_resources(handle, METHOD_NAME__CRS,
+				   ssam_serdev_setup_via_acpi_crs, serdev);
+}
+
+
+/* -- Power management. ----------------------------------------------------- */
+
+static void ssam_serial_hub_shutdown(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	/*
+	 * Try to disable notifiers, signal display-off and D0-exit, ignore any
+	 * errors.
+	 *
+	 * Note: It has not been established yet if this is actually
+	 * necessary/useful for shutdown.
+	 */
+
+	status = ssam_notifier_disable_registered(c);
+	if (status) {
+		ssam_err(c, "pm: failed to disable notifiers for shutdown: %d\n",
+			 status);
+	}
+
+	status = ssam_ctrl_notif_display_off(c);
+	if (status)
+		ssam_err(c, "pm: display-off notification failed: %d\n", status);
+
+	status = ssam_ctrl_notif_d0_exit(c);
+	if (status)
+		ssam_err(c, "pm: D0-exit notification failed: %d\n", status);
+}
+
+#ifdef CONFIG_PM_SLEEP
+
+static int ssam_serial_hub_pm_prepare(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	/*
+	 * Try to signal display-off, This will quiesce events.
+	 *
+	 * Note: Signalling display-off/display-on should normally be done from
+	 * some sort of display state notifier. As that is not available, signal
+	 * it here.
+	 */
+
+	status = ssam_ctrl_notif_display_off(c);
+	if (status)
+		ssam_err(c, "pm: display-off notification failed: %d\n", status);
+
+	return status;
+}
+
+static void ssam_serial_hub_pm_complete(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	/*
+	 * Try to signal display-on. This will restore events.
+	 *
+	 * Note: Signalling display-off/display-on should normally be done from
+	 * some sort of display state notifier. As that is not available, signal
+	 * it here.
+	 */
+
+	status = ssam_ctrl_notif_display_on(c);
+	if (status)
+		ssam_err(c, "pm: display-on notification failed: %d\n", status);
+}
+
+static int ssam_serial_hub_pm_suspend(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	/*
+	 * Try to signal D0-exit, enable IRQ wakeup if specified. Abort on
+	 * error.
+	 */
+
+	status = ssam_ctrl_notif_d0_exit(c);
+	if (status) {
+		ssam_err(c, "pm: D0-exit notification failed: %d\n", status);
+		goto err_notif;
+	}
+
+	status = ssam_irq_arm_for_wakeup(c);
+	if (status)
+		goto err_irq;
+
+	WARN_ON(ssam_controller_suspend(c));
+	return 0;
+
+err_irq:
+	ssam_ctrl_notif_d0_entry(c);
+err_notif:
+	ssam_ctrl_notif_display_on(c);
+	return status;
+}
+
+static int ssam_serial_hub_pm_resume(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	WARN_ON(ssam_controller_resume(c));
+
+	/*
+	 * Try to disable IRQ wakeup (if specified) and signal D0-entry. In
+	 * case of errors, log them and try to restore normal operation state
+	 * as far as possible.
+	 *
+	 * Note: Signalling display-off/display-on should normally be done from
+	 * some sort of display state notifier. As that is not available, signal
+	 * it here.
+	 */
+
+	ssam_irq_disarm_wakeup(c);
+
+	status = ssam_ctrl_notif_d0_entry(c);
+	if (status)
+		ssam_err(c, "pm: D0-entry notification failed: %d\n", status);
+
+	return 0;
+}
+
+static int ssam_serial_hub_pm_freeze(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	/*
+	 * During hibernation image creation, we only have to ensure that the
+	 * EC doesn't send us any events. This is done via the display-off
+	 * and D0-exit notifications. Note that this sets up the wakeup IRQ
+	 * on the EC side, however, we have disabled it by default on our side
+	 * and won't enable it here.
+	 *
+	 * See ssam_serial_hub_poweroff() for more details on the hibernation
+	 * process.
+	 */
+
+	status = ssam_ctrl_notif_d0_exit(c);
+	if (status) {
+		ssam_err(c, "pm: D0-exit notification failed: %d\n", status);
+		ssam_ctrl_notif_display_on(c);
+		return status;
+	}
+
+	WARN_ON(ssam_controller_suspend(c));
+	return 0;
+}
+
+static int ssam_serial_hub_pm_thaw(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	WARN_ON(ssam_controller_resume(c));
+
+	status = ssam_ctrl_notif_d0_entry(c);
+	if (status)
+		ssam_err(c, "pm: D0-exit notification failed: %d\n", status);
+
+	return status;
+}
+
+static int ssam_serial_hub_pm_poweroff(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	/*
+	 * When entering hibernation and powering off the system, the EC, at
+	 * least on some models, may disable events. Without us taking care of
+	 * that, this leads to events not being enabled/restored when the
+	 * system resumes from hibernation, resulting SAM-HID subsystem devices
+	 * (i.e. keyboard, touchpad) not working, AC-plug/AC-unplug events being
+	 * gone, etc.
+	 *
+	 * To avoid these issues, we disable all registered events here (this is
+	 * likely not actually required) and restore them during the drivers PM
+	 * restore callback.
+	 *
+	 * Wakeup from the EC interrupt is not supported during hibernation,
+	 * so don't arm the IRQ here.
+	 */
+
+	status = ssam_notifier_disable_registered(c);
+	if (status) {
+		ssam_err(c, "pm: failed to disable notifiers for hibernation: %d\n",
+			 status);
+		return status;
+	}
+
+	status = ssam_ctrl_notif_d0_exit(c);
+	if (status) {
+		ssam_err(c, "pm: D0-exit notification failed: %d\n", status);
+		ssam_notifier_restore_registered(c);
+		return status;
+	}
+
+	WARN_ON(ssam_controller_suspend(c));
+	return 0;
+}
+
+static int ssam_serial_hub_pm_restore(struct device *dev)
+{
+	struct ssam_controller *c = dev_get_drvdata(dev);
+	int status;
+
+	/*
+	 * Ignore but log errors, try to restore state as much as possible in
+	 * case of failures. See ssam_serial_hub_poweroff() for more details on
+	 * the hibernation process.
+	 */
+
+	WARN_ON(ssam_controller_resume(c));
+
+	status = ssam_ctrl_notif_d0_entry(c);
+	if (status)
+		ssam_err(c, "pm: D0-entry notification failed: %d\n", status);
+
+	ssam_notifier_restore_registered(c);
+	return 0;
+}
+
+static const struct dev_pm_ops ssam_serial_hub_pm_ops = {
+	.prepare  = ssam_serial_hub_pm_prepare,
+	.complete = ssam_serial_hub_pm_complete,
+	.suspend  = ssam_serial_hub_pm_suspend,
+	.resume   = ssam_serial_hub_pm_resume,
+	.freeze   = ssam_serial_hub_pm_freeze,
+	.thaw     = ssam_serial_hub_pm_thaw,
+	.poweroff = ssam_serial_hub_pm_poweroff,
+	.restore  = ssam_serial_hub_pm_restore,
+};
+
+#else /* CONFIG_PM_SLEEP */
+
+static const struct dev_pm_ops ssam_serial_hub_pm_ops = { };
+
+#endif /* CONFIG_PM_SLEEP */
+
+
+/* -- Device/driver setup. -------------------------------------------------- */
+
+static const struct acpi_gpio_params gpio_ssam_wakeup_int = { 0, 0, false };
+static const struct acpi_gpio_params gpio_ssam_wakeup     = { 1, 0, false };
+
+static const struct acpi_gpio_mapping ssam_acpi_gpios[] = {
+	{ "ssam_wakeup-int-gpio", &gpio_ssam_wakeup_int, 1 },
+	{ "ssam_wakeup-gpio",     &gpio_ssam_wakeup,     1 },
+	{ },
+};
+
+static int ssam_serial_hub_probe(struct serdev_device *serdev)
+{
+	struct ssam_controller *ctrl;
+	acpi_handle *ssh = ACPI_HANDLE(&serdev->dev);
+	acpi_status astatus;
+	int status;
+
+	if (gpiod_count(&serdev->dev, NULL) < 0)
+		return -ENODEV;
+
+	status = devm_acpi_dev_add_driver_gpios(&serdev->dev, ssam_acpi_gpios);
+	if (status)
+		return status;
+
+	// allocate controller
+	ctrl = kzalloc(sizeof(*ctrl), GFP_KERNEL);
+	if (!ctrl)
+		return -ENOMEM;
+
+	// initialize controller
+	status = ssam_controller_init(ctrl, serdev);
+	if (status)
+		goto err_ctrl_init;
+
+	// set up serdev device
+	serdev_device_set_drvdata(serdev, ctrl);
+	serdev_device_set_client_ops(serdev, &ssam_serdev_ops);
+	status = serdev_device_open(serdev);
+	if (status)
+		goto err_devopen;
+
+	astatus = ssam_serdev_setup_via_acpi(ssh, serdev);
+	if (ACPI_FAILURE(astatus)) {
+		status = -ENXIO;
+		goto err_devinit;
+	}
+
+	// start controller
+	status = ssam_controller_start(ctrl);
+	if (status)
+		goto err_devinit;
+
+	// initial SAM requests: log version, notify default/init power states
+	status = ssam_log_firmware_version(ctrl);
+	if (status)
+		goto err_initrq;
+
+	status = ssam_ctrl_notif_d0_entry(ctrl);
+	if (status)
+		goto err_initrq;
+
+	status = ssam_ctrl_notif_display_on(ctrl);
+	if (status)
+		goto err_initrq;
+
+	status = sysfs_create_group(&serdev->dev.kobj, &ssam_sam_group);
+	if (status)
+		goto err_initrq;
+
+	// setup IRQ
+	status = ssam_irq_setup(ctrl);
+	if (status)
+		goto err_irq;
+
+	// finally, set main controller reference
+	status = ssam_try_set_controller(ctrl);
+	if (WARN_ON(status))	// currently, we're the only provider
+		goto err_mainref;
+
+	/*
+	 * TODO: The EC can wake up the system via the associated GPIO interrupt
+	 *       in multiple situations. One of which is the remaining battery
+	 *       capacity falling below a certain threshold. Normally, we should
+	 *       use the device_init_wakeup function, however, the EC also seems
+	 *       to have other reasons for waking up the system and it seems
+	 *       that Windows has additional checks whether the system should be
+	 *       resumed. In short, this causes some spurious unwanted wake-ups.
+	 *       For now let's thus default power/wakeup to false.
+	 */
+	device_set_wakeup_capable(&serdev->dev, true);
+	acpi_walk_dep_device_list(ssh);
+
+	return 0;
+
+err_mainref:
+	ssam_irq_free(ctrl);
+err_irq:
+	sysfs_remove_group(&serdev->dev.kobj, &ssam_sam_group);
+err_initrq:
+	ssam_controller_shutdown(ctrl);
+err_devinit:
+	serdev_device_close(serdev);
+err_devopen:
+	ssam_controller_destroy(ctrl);
+err_ctrl_init:
+	kfree(ctrl);
+	return status;
+}
+
+static void ssam_serial_hub_remove(struct serdev_device *serdev)
+{
+	struct ssam_controller *ctrl = serdev_device_get_drvdata(serdev);
+	int status;
+
+	// clear static reference, so that no one else can get a new one
+	ssam_clear_controller();
+
+	ssam_irq_free(ctrl);
+	sysfs_remove_group(&serdev->dev.kobj, &ssam_sam_group);
+	ssam_controller_lock(ctrl);
+
+	// remove all client devices
+	ssam_controller_remove_clients(ctrl);
+
+	// act as if suspending to disable events
+	status = ssam_ctrl_notif_display_off(ctrl);
+	if (status) {
+		dev_err(&serdev->dev, "display-off notification failed: %d\n",
+			status);
+	}
+
+	status = ssam_ctrl_notif_d0_exit(ctrl);
+	if (status) {
+		dev_err(&serdev->dev, "D0-exit notification failed: %d\n",
+			status);
+	}
+
+	// shut down controller and remove serdev device reference from it
+	ssam_controller_shutdown(ctrl);
+
+	// shut down actual transport
+	serdev_device_wait_until_sent(serdev, 0);
+	serdev_device_close(serdev);
+
+	// drop our controller reference
+	ssam_controller_unlock(ctrl);
+	ssam_controller_put(ctrl);
+
+	device_set_wakeup_capable(&serdev->dev, false);
+}
+
+
+static const struct acpi_device_id ssam_serial_hub_match[] = {
+	{ "MSHW0084", 0 },
+	{ },
+};
+MODULE_DEVICE_TABLE(acpi, ssam_serial_hub_match);
+
+static struct serdev_device_driver ssam_serial_hub = {
+	.probe = ssam_serial_hub_probe,
+	.remove = ssam_serial_hub_remove,
+	.driver = {
+		.name = "surface_serial_hub",
+		.acpi_match_table = ssam_serial_hub_match,
+		.pm = &ssam_serial_hub_pm_ops,
+		.shutdown = ssam_serial_hub_shutdown,
+		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
+	},
+};
+
+
+/* -- Module setup. --------------------------------------------------------- */
+
+static int __init ssam_core_init(void)
+{
+	int status;
+
+	status = ssam_bus_register();
+	if (status)
+		goto err_bus;
+
+	status = ssh_ctrl_packet_cache_init();
+	if (status)
+		goto err_cpkg;
+
+	status = ssam_event_item_cache_init();
+	if (status)
+		goto err_evitem;
+
+	status = serdev_device_driver_register(&ssam_serial_hub);
+	if (status)
+		goto err_register;
+
+	return 0;
+
+err_register:
+	ssam_event_item_cache_destroy();
+err_evitem:
+	ssh_ctrl_packet_cache_destroy();
+err_cpkg:
+	ssam_bus_unregister();
+err_bus:
+	return status;
+}
+module_init(ssam_core_init);
+
+static void __exit ssam_core_exit(void)
+{
+	serdev_device_driver_unregister(&ssam_serial_hub);
+	ssam_event_item_cache_destroy();
+	ssh_ctrl_packet_cache_destroy();
+	ssam_bus_unregister();
+}
+module_exit(ssam_core_exit);
+
+MODULE_AUTHOR("Maximilian Luz <luzmaximilian@gmail.com>");
+MODULE_DESCRIPTION("Subsystem and Surface Serial Hub driver for Surface System Aggregator Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/misc/surface_aggregator/ssh_msgb.h b/drivers/misc/surface_aggregator/ssh_msgb.h
new file mode 100644
index 000000000000..7c29e7d7028a
--- /dev/null
+++ b/drivers/misc/surface_aggregator/ssh_msgb.h
@@ -0,0 +1,201 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * SSH message builder functions.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _SURFACE_AGGREGATOR_SSH_MSGB_H
+#define _SURFACE_AGGREGATOR_SSH_MSGB_H
+
+#include <asm/unaligned.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/controller.h>
+#include <linux/surface_aggregator/serial_hub.h>
+
+
+/**
+ * struct msgbuf - Buffer struct to construct SSH messages.
+ * @begin: Pointer to the beginning of the allocated buffer space.
+ * @end:   Pointer to the end (one past last element) of the allocated buffer
+ *         space.
+ * @ptr:   Pointer to the first free element in the buffer.
+ */
+struct msgbuf {
+	u8 *begin;
+	u8 *end;
+	u8 *ptr;
+};
+
+/**
+ * msgb_init() - Initialize the given message buffer struct.
+ * @msgb: The buffer struct to initialize
+ * @ptr:  Pointer to the underlying memory by which the buffer will be backed.
+ * @cap:  Size of the underlying memory.
+ *
+ * Initialize the given message buffer struct using the provided memory as
+ * backing.
+ */
+static inline void msgb_init(struct msgbuf *msgb, u8 *ptr, size_t cap)
+{
+	msgb->begin = ptr;
+	msgb->end = ptr + cap;
+	msgb->ptr = ptr;
+}
+
+/**
+ * msgb_bytes_used() - Return the current number of bytes used in the buffer.
+ * @msgb: The message buffer.
+ */
+static inline size_t msgb_bytes_used(const struct msgbuf *msgb)
+{
+	return msgb->ptr - msgb->begin;
+}
+
+/**
+ * msgb_push_u16() - Push a u16 value to the buffer.
+ * @msgb:  The message buffer.
+ * @value: The value to push to the buffer.
+ */
+static inline void msgb_push_u16(struct msgbuf *msgb, u16 value)
+{
+	if (WARN_ON(msgb->ptr + sizeof(u16) > msgb->end))
+		return;
+
+	put_unaligned_le16(value, msgb->ptr);
+	msgb->ptr += sizeof(u16);
+}
+
+/**
+ * msgb_push_syn() - Push SSH SYN bytes to the buffer.
+ * @msgb: The message buffer.
+ */
+static inline void msgb_push_syn(struct msgbuf *msgb)
+{
+	msgb_push_u16(msgb, SSH_MSG_SYN);
+}
+
+/**
+ * msgb_push_buf() - Push raw data to the buffer.
+ * @msgb: The message buffer.
+ * @buf:  The data to push to the buffer.
+ * @len:  The length of the data to push to the buffer.
+ */
+static inline void msgb_push_buf(struct msgbuf *msgb, const u8 *buf, size_t len)
+{
+	msgb->ptr = memcpy(msgb->ptr, buf, len) + len;
+}
+
+/**
+ * msgb_push_crc() - Compute CRC and push it to the buffer.
+ * @msgb: The message buffer.
+ * @buf:  The data for which the CRC should be computed.
+ * @len:  The length of the data for which the CRC should be computed.
+ */
+static inline void msgb_push_crc(struct msgbuf *msgb, const u8 *buf, size_t len)
+{
+	msgb_push_u16(msgb, ssh_crc(buf, len));
+}
+
+/**
+ * msgb_push_frame() - Push a SSH message frame header to the buffer.
+ * @msgb: The message buffer
+ * @ty:   The type of the frame.
+ * @len:  The length of the payload of the frame.
+ * @seq:  The sequence ID of the frame/packet.
+ */
+static inline void msgb_push_frame(struct msgbuf *msgb, u8 ty, u16 len, u8 seq)
+{
+	struct ssh_frame *frame = (struct ssh_frame *)msgb->ptr;
+	const u8 *const begin = msgb->ptr;
+
+	if (WARN_ON(msgb->ptr + sizeof(*frame) > msgb->end))
+		return;
+
+	frame->type = ty;
+	put_unaligned_le16(len, &frame->len);
+	frame->seq  = seq;
+
+	msgb->ptr += sizeof(*frame);
+	msgb_push_crc(msgb, begin, msgb->ptr - begin);
+}
+
+/**
+ * msgb_push_ack() - Push a SSH ACK frame to the buffer.
+ * @msgb: The message buffer
+ * @seq:  The sequence ID of the frame/packet to be ACKed.
+ */
+static inline void msgb_push_ack(struct msgbuf *msgb, u8 seq)
+{
+	// SYN
+	msgb_push_syn(msgb);
+
+	// ACK-type frame + CRC
+	msgb_push_frame(msgb, SSH_FRAME_TYPE_ACK, 0x00, seq);
+
+	// payload CRC (ACK-type frames do not have a payload)
+	msgb_push_crc(msgb, msgb->ptr, 0);
+}
+
+/**
+ * msgb_push_nak() - Push a SSH NAK frame to the buffer.
+ * @msgb: The message buffer
+ */
+static inline void msgb_push_nak(struct msgbuf *msgb)
+{
+	// SYN
+	msgb_push_syn(msgb);
+
+	// NAK-type frame + CRC
+	msgb_push_frame(msgb, SSH_FRAME_TYPE_NAK, 0x00, 0x00);
+
+	// payload CRC (ACK-type frames do not have a payload)
+	msgb_push_crc(msgb, msgb->ptr, 0);
+}
+
+/**
+ * msgb_push_cmd() - Push a SSH command frame with payload to the buffer.
+ * @msgb: The message buffer.
+ * @seq:  The sequence ID (SEQ) of the frame/packet.
+ * @rqid: The request ID (RQID) of the request contained in the frame.
+ * @rqst: The request to wrap in the frame.
+ */
+static inline void msgb_push_cmd(struct msgbuf *msgb, u8 seq, u16 rqid,
+				 const struct ssam_request *rqst)
+{
+	struct ssh_command *cmd;
+	const u8 *cmd_begin;
+	const u8 type = SSH_FRAME_TYPE_DATA_SEQ;
+
+	// SYN
+	msgb_push_syn(msgb);
+
+	// command frame + crc
+	msgb_push_frame(msgb, type, sizeof(*cmd) + rqst->length, seq);
+
+	// frame payload: command struct + payload
+	if (WARN_ON(msgb->ptr + sizeof(*cmd) > msgb->end))
+		return;
+
+	cmd_begin = msgb->ptr;
+	cmd = (struct ssh_command *)msgb->ptr;
+
+	cmd->type    = SSH_PLD_TYPE_CMD;
+	cmd->tc      = rqst->target_category;
+	cmd->tid_out = rqst->target_id;
+	cmd->tid_in  = 0x00;
+	cmd->iid     = rqst->instance_id;
+	put_unaligned_le16(rqid, &cmd->rqid);
+	cmd->cid     = rqst->command_id;
+
+	msgb->ptr += sizeof(*cmd);
+
+	// command payload
+	msgb_push_buf(msgb, rqst->payload, rqst->length);
+
+	// crc for command struct + payload
+	msgb_push_crc(msgb, cmd_begin, msgb->ptr - cmd_begin);
+}
+
+#endif /* _SURFACE_AGGREGATOR_SSH_MSGB_H */
diff --git a/drivers/misc/surface_aggregator/ssh_packet_layer.c b/drivers/misc/surface_aggregator/ssh_packet_layer.c
new file mode 100644
index 000000000000..2d72e8c02842
--- /dev/null
+++ b/drivers/misc/surface_aggregator/ssh_packet_layer.c
@@ -0,0 +1,2009 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * SSH packet transport layer.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <asm/unaligned.h>
+#include <linux/atomic.h>
+#include <linux/error-injection.h>
+#include <linux/jiffies.h>
+#include <linux/kfifo.h>
+#include <linux/kref.h>
+#include <linux/kthread.h>
+#include <linux/ktime.h>
+#include <linux/list.h>
+#include <linux/serdev.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/workqueue.h>
+
+#include <linux/surface_aggregator/serial_hub.h>
+
+#include "ssh_msgb.h"
+#include "ssh_packet_layer.h"
+#include "ssh_parser.h"
+
+#include "trace.h"
+
+
+/*
+ * To simplify reasoning about the code below, we define a few concepts. The
+ * system below is similar to a state-machine for packets, however, there are
+ * too many states to explicitly write them down. To (somewhat) manage the
+ * states and packets we rely on flags, reference counting, and some simple
+ * concepts. State transitions are triggered by actions.
+ *
+ * >> Actions <<
+ *
+ * - submit
+ * - transmission start (process next item in queue)
+ * - transmission finished (guaranteed to never be parallel to transmission
+ *   start)
+ * - ACK received
+ * - NAK received (this is equivalent to issuing re-submit for all pending
+ *   packets)
+ * - timeout (this is equivalent to re-issuing a submit or canceling)
+ * - cancel (non-pending and pending)
+ *
+ * >> Data Structures, Packet Ownership, General Overview <<
+ *
+ * The code below employs two main data structures: The packet queue,
+ * containing all packets scheduled for transmission, and the set of pending
+ * packets, containing all packets awaiting an ACK.
+ *
+ * Shared ownership of a packet is controlled via reference counting. Inside
+ * the transport system are a total of five packet owners:
+ *
+ * - the packet queue,
+ * - the pending set,
+ * - the transmitter thread,
+ * - the receiver thread (via ACKing), and
+ * - the timeout work item.
+ *
+ * Normal operation is as follows: The initial reference of the packet is
+ * obtained by submitting the packet and queueing it. The receiver thread
+ * takes packets from the queue. By doing this, it does not increment the
+ * refcount but takes over the reference (removing it from the queue). If the
+ * packet is sequenced (i.e. needs to be ACKed by the client), the transmitter
+ * thread sets-up the timeout and adds the packet to the pending set before
+ * starting to transmit it. As the timeout is handled by a reaper task, no
+ * additional reference for it is needed. After the transmit is done, the
+ * reference held by the transmitter thread is dropped. If the packet is
+ * unsequenced (i.e. does not need an ACK), the packet is completed by the
+ * transmitter thread before dropping that reference.
+ *
+ * On receival of an ACK, the receiver thread removes and obtains the
+ * reference to the packet from the pending set. The receiver thread will then
+ * complete the packet and drop its reference.
+ *
+ * On receival of a NAK, the receiver thread re-submits all currently pending
+ * packets.
+ *
+ * Packet timeouts are detected by the timeout reaper. This is a task,
+ * scheduled depending on the earliest packet timeout expiration date,
+ * checking all currently pending packets if their timeout has expired. If the
+ * timeout of a packet has expired, it is re-submitted and the number of tries
+ * of this packet is incremented. If this number reaches its limit, the packet
+ * will be completed with a failure.
+ *
+ * On transmission failure (such as repeated packet timeouts), the completion
+ * callback is immediately run by on thread on which the error was detected.
+ *
+ * To ensure that a packet eventually leaves the system it is marked as
+ * "locked" directly before it is going to be completed or when it is
+ * canceled. Marking a packet as "locked" has the effect that passing and
+ * creating new references of the packet is disallowed. This means that the
+ * packet cannot be added to the queue, the pending set, and the timeout, or
+ * be picked up by the transmitter thread or receiver thread. To remove a
+ * packet from the system it has to be marked as locked and subsequently all
+ * references from the data structures (queue, pending) have to be removed.
+ * References held by threads will eventually be dropped automatically as
+ * their execution progresses.
+ *
+ * Note that the packet completion callback is, in case of success and for a
+ * sequenced packet, guaranteed to run on the receiver thread, thus providing
+ * a way to reliably identify responses to the packet. The packet completion
+ * callback is only run once and it does not indicate that the packet has
+ * fully left the system (for this, one should rely on the release method,
+ * triggered when the reference count of the packet reaches zero). In case of
+ * re-submission (and with somewhat unlikely timing), it may be possible that
+ * the packet is being re-transmitted while the completion callback runs.
+ * Completion will occur both on success and internal error, as well as when
+ * the packet is canceled.
+ *
+ * >> Flags <<
+ *
+ * Flags are used to indicate the state and progression of a packet. Some flags
+ * have stricter guarantees than other:
+ *
+ * - locked
+ *   Indicates if the packet is locked. If the packet is locked, passing and/or
+ *   creating additional references to the packet is forbidden. The packet thus
+ *   may not be queued, dequeued, or removed or added to the pending set. Note
+ *   that the packet state flags may still change (e.g. it may be marked as
+ *   ACKed, transmitted, ...).
+ *
+ * - completed
+ *   Indicates if the packet completion callback has been executed or is about
+ *   to be executed. This flag is used to ensure that the packet completion
+ *   callback is only run once.
+ *
+ * - queued
+ *   Indicates if a packet is present in the submission queue or not. This flag
+ *   must only be modified with the queue lock held, and must be coherent to the
+ *   presence of the packet in the queue.
+ *
+ * - pending
+ *   Indicates if a packet is present in the set of pending packets or not.
+ *   This flag must only be modified with the pending lock held, and must be
+ *   coherent to the presence of the packet in the pending set.
+ *
+ * - transmitting
+ *   Indicates if the packet is currently transmitting. In case of
+ *   re-transmissions, it is only safe to wait on the "transmitted" completion
+ *   after this flag has been set. The completion will be set both in success
+ *   and error case.
+ *
+ * - transmitted
+ *   Indicates if the packet has been transmitted. This flag is not cleared by
+ *   the system, thus it indicates the first transmission only.
+ *
+ * - acked
+ *   Indicates if the packet has been acknowledged by the client. There are no
+ *   other guarantees given. For example, the packet may still be canceled
+ *   and/or the completion may be triggered an error even though this bit is
+ *   set. Rely on the status provided to the completion callback instead.
+ *
+ * - canceled
+ *   Indicates if the packet has been canceled from the outside. There are no
+ *   other guarantees given. Specifically, the packet may be completed by
+ *   another part of the system before the cancellation attempts to complete it.
+ *
+ * >> General Notes <<
+ *
+ * To avoid deadlocks, if both queue and pending locks are required, the
+ * pending lock must be acquired before the queue lock.
+ */
+
+/*
+ * SSH_PTL_MAX_PACKET_TRIES - Maximum transmission attempts for packet.
+ *
+ * Maximum number of transmission attempts per sequenced packet in case of
+ * time-outs. Must be smaller than 16. If the packet times out after this
+ * amount of tries, the packet will be completed with %-ETIMEDOUT as status
+ * code.
+ */
+#define SSH_PTL_MAX_PACKET_TRIES		3
+
+/*
+ * SSH_PTL_PACKET_TIMEOUT - Packet timeout.
+ *
+ * Timeout as ktime_t delta for ACKs. If we have not received an ACK in this
+ * time-frame after starting transmission, the packet will be re-submitted.
+ */
+#define SSH_PTL_PACKET_TIMEOUT			ms_to_ktime(1000)
+
+/*
+ * SSH_PTL_PACKET_TIMEOUT_RESOLUTION - Packet timeout granularity.
+ *
+ * Time-resolution for timeouts. Should be larger than one jiffy to avoid
+ * direct re-scheduling of reaper work_struct.
+ */
+#define SSH_PTL_PACKET_TIMEOUT_RESOLUTION	ms_to_ktime(max(2000 / HZ, 50))
+
+/*
+ * SSH_PTL_MAX_PENDING - Maximum number of pending packets.
+ *
+ * Maximum number of sequenced packets concurrently waiting for an ACK.
+ * Packets marked as blocking will not be transmitted while this limit is
+ * reached.
+ */
+#define SSH_PTL_MAX_PENDING			1
+
+/*
+ * SSH_PTL_RX_BUF_LEN - Evaluation-buffer size in bytes.
+ */
+#define SSH_PTL_RX_BUF_LEN			4096
+
+/*
+ * SSH_PTL_RX_FIFO_LEN - Fifo input-buffer size in bytes.
+ */
+#define SSH_PTL_RX_FIFO_LEN			4096
+
+
+#ifdef CONFIG_SURFACE_AGGREGATOR_ERROR_INJECTION
+
+/**
+ * ssh_ptl_should_drop_ack_packet() - Error injection hook to drop ACK packets.
+ *
+ * Useful to test detection and handling of automated re-transmits by the EC.
+ * Specifically of packets that the EC consideres not-ACKed but the driver
+ * already consideres ACKed (due to dropped ACK). In this case, the EC
+ * re-transmits the packet-to-be-ACKed and the driver should detect it as
+ * duplicate/already handled. Note that the driver should still send an ACK
+ * for the re-transmitted packet.
+ */
+static noinline bool ssh_ptl_should_drop_ack_packet(void)
+{
+	return false;
+}
+ALLOW_ERROR_INJECTION(ssh_ptl_should_drop_ack_packet, TRUE);
+
+/**
+ * ssh_ptl_should_drop_nak_packet() - Error injection hook to drop NAK packets.
+ *
+ * Useful to test/force automated (timeout-based) re-transmit by the EC.
+ * Specifically, packets that have not reached the driver completely/with valid
+ * checksums. Only useful in combination with receival of (injected) bad data.
+ */
+static noinline bool ssh_ptl_should_drop_nak_packet(void)
+{
+	return false;
+}
+ALLOW_ERROR_INJECTION(ssh_ptl_should_drop_nak_packet, TRUE);
+
+/**
+ * ssh_ptl_should_drop_dsq_packet() - Error injection hook to drop sequenced
+ * data packet.
+ *
+ * Useful to test re-transmit timeout of the driver. If the data packet has not
+ * been ACKed after a certain time, the driver should re-transmit the packet up
+ * to limited number of times defined in SSH_PTL_MAX_PACKET_TRIES.
+ */
+static noinline bool ssh_ptl_should_drop_dsq_packet(void)
+{
+	return false;
+}
+ALLOW_ERROR_INJECTION(ssh_ptl_should_drop_dsq_packet, TRUE);
+
+/**
+ * ssh_ptl_should_fail_write() - Error injection hook to make
+ * serdev_device_write() fail.
+ *
+ * Hook to simulate errors in serdev_device_write when transmitting packets.
+ */
+static noinline int ssh_ptl_should_fail_write(void)
+{
+	return 0;
+}
+ALLOW_ERROR_INJECTION(ssh_ptl_should_fail_write, ERRNO);
+
+/**
+ * ssh_ptl_should_corrupt_tx_data() - Error injection hook to simualte invalid
+ * data being sent to the EC.
+ *
+ * Hook to simulate corrupt/invalid data being sent from host (driver) to EC.
+ * Causes the packet data to be actively corrupted by overwriting it with
+ * pre-defined values, such that it becomes invalid, causing the EC to respond
+ * with a NAK packet. Useful to test handling of NAK packets received by the
+ * driver.
+ */
+static noinline bool ssh_ptl_should_corrupt_tx_data(void)
+{
+	return false;
+}
+ALLOW_ERROR_INJECTION(ssh_ptl_should_corrupt_tx_data, TRUE);
+
+/**
+ * ssh_ptl_should_corrupt_rx_syn() - Error injection hook to simulate invalid
+ * data being sent by the EC.
+ *
+ * Hook to simulate invalid SYN bytes, i.e. an invalid start of messages and
+ * test handling thereof in the driver.
+ */
+static noinline bool ssh_ptl_should_corrupt_rx_syn(void)
+{
+	return false;
+}
+ALLOW_ERROR_INJECTION(ssh_ptl_should_corrupt_rx_syn, TRUE);
+
+/**
+ * ssh_ptl_should_corrupt_rx_data() - Error injection hook to simulate invalid
+ * data being sent by the EC.
+ *
+ * Hook to simulate invalid data/checksum of the message frame and test handling
+ * thereof in the driver.
+ */
+static noinline bool ssh_ptl_should_corrupt_rx_data(void)
+{
+	return false;
+}
+ALLOW_ERROR_INJECTION(ssh_ptl_should_corrupt_rx_data, TRUE);
+
+
+static bool __ssh_ptl_should_drop_ack_packet(struct ssh_packet *packet)
+{
+	if (likely(!ssh_ptl_should_drop_ack_packet()))
+		return false;
+
+	trace_ssam_ei_tx_drop_ack_packet(packet);
+	ptl_info(packet->ptl, "packet error injection: dropping ACK packet %p\n",
+		 packet);
+
+	return true;
+}
+
+static bool __ssh_ptl_should_drop_nak_packet(struct ssh_packet *packet)
+{
+	if (likely(!ssh_ptl_should_drop_nak_packet()))
+		return false;
+
+	trace_ssam_ei_tx_drop_nak_packet(packet);
+	ptl_info(packet->ptl, "packet error injection: dropping NAK packet %p\n",
+		 packet);
+
+	return true;
+}
+
+static bool __ssh_ptl_should_drop_dsq_packet(struct ssh_packet *packet)
+{
+	if (likely(!ssh_ptl_should_drop_dsq_packet()))
+		return false;
+
+	trace_ssam_ei_tx_drop_dsq_packet(packet);
+	ptl_info(packet->ptl,
+		"packet error injection: dropping sequenced data packet %p\n",
+		 packet);
+
+	return true;
+}
+
+static bool ssh_ptl_should_drop_packet(struct ssh_packet *packet)
+{
+	// ignore packets that don't carry any data (i.e. flush)
+	if (!packet->data.ptr || !packet->data.len)
+		return false;
+
+	switch (packet->data.ptr[SSH_MSGOFFSET_FRAME(type)]) {
+	case SSH_FRAME_TYPE_ACK:
+		return __ssh_ptl_should_drop_ack_packet(packet);
+
+	case SSH_FRAME_TYPE_NAK:
+		return __ssh_ptl_should_drop_nak_packet(packet);
+
+	case SSH_FRAME_TYPE_DATA_SEQ:
+		return __ssh_ptl_should_drop_dsq_packet(packet);
+
+	default:
+		return false;
+	}
+}
+
+static int ssh_ptl_write_buf(struct ssh_ptl *ptl, struct ssh_packet *packet,
+			     const unsigned char *buf, size_t count)
+{
+	int status;
+
+	status = ssh_ptl_should_fail_write();
+	if (unlikely(status)) {
+		trace_ssam_ei_tx_fail_write(packet, status);
+		ptl_info(packet->ptl,
+			 "packet error injection: simulating transmit error %d,"
+			 " packet %p\n", status, packet);
+
+		return status;
+	}
+
+	return serdev_device_write_buf(ptl->serdev, buf, count);
+}
+
+static void ssh_ptl_tx_inject_invalid_data(struct ssh_packet *packet)
+{
+	// ignore packets that don't carry any data (i.e. flush)
+	if (!packet->data.ptr || !packet->data.len)
+		return;
+
+	// only allow sequenced data packets to be modified
+	if (packet->data.ptr[SSH_MSGOFFSET_FRAME(type)] != SSH_FRAME_TYPE_DATA_SEQ)
+		return;
+
+	if (likely(!ssh_ptl_should_corrupt_tx_data()))
+		return;
+
+	trace_ssam_ei_tx_corrupt_data(packet);
+	ptl_info(packet->ptl,
+		 "packet error injection: simulating invalid transmit data on packet %p\n",
+		 packet);
+
+	/*
+	 * NB: The value 0xb3 has been chosen more or less randomly so that it
+	 * doesn't have any (major) overlap with the SYN bytes (aa 55) and is
+	 * non-trivial (i.e. non-zero, non-0xff).
+	 */
+	memset(packet->data.ptr, 0xb3, packet->data.len);
+}
+
+static void ssh_ptl_rx_inject_invalid_syn(struct ssh_ptl *ptl,
+					  struct ssam_span *data)
+{
+	struct ssam_span frame;
+
+	// check if there actually is something to corrupt
+	if (!sshp_find_syn(data, &frame))
+		return;
+
+	if (likely(!ssh_ptl_should_corrupt_rx_syn()))
+		return;
+
+	trace_ssam_ei_rx_corrupt_syn("data_length", data->len);
+
+	data->ptr[1] = 0xb3;	// set second byte of SYN to "random" value
+}
+
+static void ssh_ptl_rx_inject_invalid_data(struct ssh_ptl *ptl,
+					   struct ssam_span *frame)
+{
+	size_t payload_len, message_len;
+	struct ssh_frame *sshf;
+
+	// ignore incomplete messages, will get handled once it's complete
+	if (frame->len < SSH_MESSAGE_LENGTH(0))
+		return;
+
+	// ignore incomplete messages, part 2
+	payload_len = get_unaligned_le16(&frame->ptr[SSH_MSGOFFSET_FRAME(len)]);
+	message_len = SSH_MESSAGE_LENGTH(payload_len);
+	if (frame->len < message_len)
+		return;
+
+	if (likely(!ssh_ptl_should_corrupt_rx_data()))
+		return;
+
+	sshf = (struct ssh_frame *)&frame->ptr[SSH_MSGOFFSET_FRAME(type)];
+	trace_ssam_ei_rx_corrupt_data(sshf);
+
+	/*
+	 * Flip bits in first byte of payload checksum. This is basically
+	 * equivalent to a payload/frame data error without us having to worry
+	 * about (the, arguably pretty small, probability of) accidental
+	 * checksum collisions.
+	 */
+	frame->ptr[frame->len - 2] = ~frame->ptr[frame->len - 2];
+}
+
+#else /* CONFIG_SURFACE_AGGREGATOR_ERROR_INJECTION */
+
+static inline bool ssh_ptl_should_drop_packet(struct ssh_packet *packet)
+{
+	return false;
+}
+
+static inline int ssh_ptl_write_buf(struct ssh_ptl *ptl,
+				    struct ssh_packet *packet,
+				    const unsigned char *buf,
+				    size_t count)
+{
+	return serdev_device_write_buf(ptl->serdev, buf, count);
+}
+
+static inline void ssh_ptl_tx_inject_invalid_data(struct ssh_packet *packet)
+{
+}
+
+static inline void ssh_ptl_rx_inject_invalid_syn(struct ssh_ptl *ptl,
+						 struct ssam_span *data)
+{
+}
+
+static inline void ssh_ptl_rx_inject_invalid_data(struct ssh_ptl *ptl,
+						  struct ssam_span *frame)
+{
+}
+
+#endif /* CONFIG_SURFACE_AGGREGATOR_ERROR_INJECTION */
+
+
+static void __ssh_ptl_packet_release(struct kref *kref)
+{
+	struct ssh_packet *p = container_of(kref, struct ssh_packet, refcnt);
+
+	trace_ssam_packet_release(p);
+
+	ptl_dbg_cond(p->ptl, "ptl: releasing packet %p\n", p);
+	p->ops->release(p);
+}
+
+/**
+ * ssh_packet_get() - Increment reference count of packet.
+ * @packet: The packet to increment the reference count of.
+ *
+ * Increments the reference count of the given packet. See ssh_packet_put()
+ * for the counter-part of this function.
+ *
+ * Return: Returns the packet provided as input.
+ */
+struct ssh_packet *ssh_packet_get(struct ssh_packet *packet)
+{
+	if (packet)
+		kref_get(&packet->refcnt);
+	return packet;
+}
+EXPORT_SYMBOL_GPL(ssh_packet_get);
+
+/**
+ * ssh_packet_put() - Decrement reference count of packet.
+ * @packet: The packet to decrement the reference count of.
+ *
+ * If the reference count reaches zero, the ``release`` callback specified in
+ * the packet's &struct ssh_packet_ops, i.e. ``packet->ops->release``, will be
+ * called.
+ *
+ * See ssh_packet_get() for the counter-part of this function.
+ */
+void ssh_packet_put(struct ssh_packet *packet)
+{
+	if (packet)
+		kref_put(&packet->refcnt, __ssh_ptl_packet_release);
+}
+EXPORT_SYMBOL_GPL(ssh_packet_put);
+
+static u8 ssh_packet_get_seq(struct ssh_packet *packet)
+{
+	return packet->data.ptr[SSH_MSGOFFSET_FRAME(seq)];
+}
+
+
+/**
+ * ssh_packet_init() - Initialize SSH packet.
+ * @packet:   The packet to initialize.
+ * @type:     Type-flags of the packet.
+ * @priority: Priority of the packet. See SSH_PACKET_PRIORITY() for details.
+ * @ops:      Packet operations.
+ *
+ * Initializes the given SSH packet. Sets the transmission buffer pointer to
+ * %NULL and the transmission buffer length to zero. For data-type packets,
+ * this buffer has to be set separately via ssh_packet_set_data() before
+ * submission, and must contain a valid SSH message, i.e. frame with optional
+ * payload of any type.
+ */
+void ssh_packet_init(struct ssh_packet *packet, unsigned long type,
+		     u8 priority, const struct ssh_packet_ops *ops)
+{
+	kref_init(&packet->refcnt);
+
+	packet->ptl = NULL;
+	INIT_LIST_HEAD(&packet->queue_node);
+	INIT_LIST_HEAD(&packet->pending_node);
+
+	packet->state = type & SSH_PACKET_FLAGS_TY_MASK;
+	packet->priority = priority;
+	packet->timestamp = KTIME_MAX;
+
+	packet->data.ptr = NULL;
+	packet->data.len = 0;
+
+	packet->ops = ops;
+}
+
+
+static struct kmem_cache *ssh_ctrl_packet_cache;
+
+/**
+ * ssh_ctrl_packet_cache_init() - Initialize the control packet cache.
+ */
+int ssh_ctrl_packet_cache_init(void)
+{
+	const unsigned int size = sizeof(struct ssh_packet) + SSH_MSG_LEN_CTRL;
+	const unsigned int align = __alignof__(struct ssh_packet);
+	struct kmem_cache *cache;
+
+	cache = kmem_cache_create("ssam_ctrl_packet", size, align, 0, NULL);
+	if (!cache)
+		return -ENOMEM;
+
+	ssh_ctrl_packet_cache = cache;
+	return 0;
+}
+
+/**
+ * ssh_ctrl_packet_cache_destroy() - Deinitialize the control packet cache.
+ */
+void ssh_ctrl_packet_cache_destroy(void)
+{
+	kmem_cache_destroy(ssh_ctrl_packet_cache);
+	ssh_ctrl_packet_cache = NULL;
+}
+
+/**
+ * ssh_ctrl_packet_alloc() - Allocate packet from control packet cache.
+ * @packet: Where the pointer to the newly allocated packet should be stored.
+ * @buffer: The buffer corresponding to this packet.
+ * @flags:  Flags used for allocation.
+ *
+ * Allocates a packet and corresponding transport buffer from the control
+ * packet cache. Sets the packet's buffer reference to the allocated buffer.
+ * The packet must be freed via ssh_ctrl_packet_free(), which will also free
+ * the corresponding buffer. The corresponding buffer must not be freed
+ * separately. Intended to be used with %ssh_ptl_ctrl_packet_ops as packet
+ * operations.
+ *
+ * Return: Returns zero on success, %-ENOMEM if the allocation failed.
+ */
+static int ssh_ctrl_packet_alloc(struct ssh_packet **packet,
+				 struct ssam_span *buffer, gfp_t flags)
+{
+	*packet = kmem_cache_alloc(ssh_ctrl_packet_cache, flags);
+	if (!*packet)
+		return -ENOMEM;
+
+	buffer->ptr = (u8 *)(*packet + 1);
+	buffer->len = SSH_MSG_LEN_CTRL;
+
+	trace_ssam_ctrl_packet_alloc(*packet, buffer->len);
+	return 0;
+}
+
+/**
+ * ssh_ctrl_packet_free() - Free packet allocated from control packet cache.
+ * @p: The packet to free.
+ */
+static void ssh_ctrl_packet_free(struct ssh_packet *p)
+{
+	trace_ssam_ctrl_packet_free(p);
+	kmem_cache_free(ssh_ctrl_packet_cache, p);
+}
+
+static const struct ssh_packet_ops ssh_ptl_ctrl_packet_ops = {
+	.complete = NULL,
+	.release = ssh_ctrl_packet_free,
+};
+
+
+static void ssh_ptl_timeout_reaper_mod(struct ssh_ptl *ptl, ktime_t now,
+				       ktime_t expires)
+{
+	unsigned long delta = msecs_to_jiffies(ktime_ms_delta(expires, now));
+	ktime_t aexp = ktime_add(expires, SSH_PTL_PACKET_TIMEOUT_RESOLUTION);
+	ktime_t old_exp, old_act;
+
+	// re-adjust / schedule reaper if it is above resolution delta
+	old_act = READ_ONCE(ptl->rtx_timeout.expires);
+	if (ktime_after(aexp, old_act))
+		return;
+
+	do {
+		old_exp = old_act;
+		old_act = cmpxchg64(&ptl->rtx_timeout.expires, old_exp, expires);
+	} while (old_exp != old_act && ktime_before(aexp, old_act));
+
+	// if we updated the reaper expiration, modify work timeout
+	if (old_exp == old_act && old_act != expires)
+		mod_delayed_work(system_wq, &ptl->rtx_timeout.reaper, delta);
+}
+
+static void ssh_ptl_timeout_start(struct ssh_packet *packet)
+{
+	struct ssh_ptl *ptl = packet->ptl;
+	ktime_t timestamp = ktime_get_coarse_boottime();
+	ktime_t timeout = ptl->rtx_timeout.timeout;
+
+	if (test_bit(SSH_PACKET_SF_LOCKED_BIT, &packet->state))
+		return;
+
+	WRITE_ONCE(packet->timestamp, timestamp);
+	/*
+	 * Ensure timestamp is set before starting the reaper. Paired with
+	 * implicit barrier following check on ssh_packet_get_expiration() in
+	 * ssh_ptl_timeout_reap().
+	 */
+	smp_mb__after_atomic();
+
+	ssh_ptl_timeout_reaper_mod(packet->ptl, timestamp, timestamp + timeout);
+}
+
+
+static void ssh_packet_next_try(struct ssh_packet *p)
+{
+	u8 priority = READ_ONCE(p->priority);
+	u8 base = ssh_packet_priority_get_base(priority);
+	u8 try = ssh_packet_priority_get_try(priority);
+
+	WRITE_ONCE(p->priority, __SSH_PACKET_PRIORITY(base, try + 1));
+}
+
+/* must be called with queue lock held */
+static struct list_head *__ssh_ptl_queue_find_entrypoint(struct ssh_packet *p)
+{
+	struct list_head *head;
+	u8 priority = READ_ONCE(p->priority);
+
+	/*
+	 * We generally assume that there are less control (ACK/NAK) packets and
+	 * re-submitted data packets as there are normal data packets (at least
+	 * in situations in which many packets are queued; if there aren't many
+	 * packets queued the decision on how to iterate should be basically
+	 * irrellevant; the number of control/data packets is more or less
+	 * limited via the maximum number of pending packets). Thus, when
+	 * inserting a control or re-submitted data packet, (determined by their
+	 * priority), we search from front to back. Normal data packets are,
+	 * usually queued directly at the tail of the queue, so for those search
+	 * from back to front.
+	 */
+
+	if (priority > SSH_PACKET_PRIORITY(DATA, 0)) {
+		list_for_each(head, &p->ptl->queue.head) {
+			p = list_entry(head, struct ssh_packet, queue_node);
+
+			if (READ_ONCE(p->priority) < priority)
+				break;
+		}
+	} else {
+		list_for_each_prev(head, &p->ptl->queue.head) {
+			p = list_entry(head, struct ssh_packet, queue_node);
+
+			if (READ_ONCE(p->priority) >= priority) {
+				head = head->next;
+				break;
+			}
+		}
+	}
+
+
+	return head;
+}
+
+/* must be called with queue lock held */
+static int __ssh_ptl_queue_push(struct ssh_packet *packet)
+{
+	struct ssh_ptl *ptl = packet->ptl;
+	struct list_head *head;
+
+
+	if (test_bit(SSH_PTL_SF_SHUTDOWN_BIT, &ptl->state))
+		return -ESHUTDOWN;
+
+	// avoid further transitions when cancelling/completing
+	if (test_bit(SSH_PACKET_SF_LOCKED_BIT, &packet->state))
+		return -EINVAL;
+
+	// if this packet has already been queued, do not add it
+	if (test_and_set_bit(SSH_PACKET_SF_QUEUED_BIT, &packet->state))
+		return -EALREADY;
+
+	head = __ssh_ptl_queue_find_entrypoint(packet);
+
+	list_add_tail(&ssh_packet_get(packet)->queue_node, head);
+	return 0;
+}
+
+static int ssh_ptl_queue_push(struct ssh_packet *packet)
+{
+	int status;
+
+	spin_lock(&packet->ptl->queue.lock);
+	status = __ssh_ptl_queue_push(packet);
+	spin_unlock(&packet->ptl->queue.lock);
+
+	return status;
+}
+
+static void ssh_ptl_queue_remove(struct ssh_packet *packet)
+{
+	struct ssh_ptl *ptl = packet->ptl;
+
+	spin_lock(&ptl->queue.lock);
+
+	if (!test_and_clear_bit(SSH_PACKET_SF_QUEUED_BIT, &packet->state)) {
+		spin_unlock(&ptl->queue.lock);
+		return;
+	}
+
+	list_del(&packet->queue_node);
+
+	spin_unlock(&ptl->queue.lock);
+	ssh_packet_put(packet);
+}
+
+
+static void ssh_ptl_pending_push(struct ssh_packet *packet)
+{
+	struct ssh_ptl *ptl = packet->ptl;
+
+	spin_lock(&ptl->pending.lock);
+
+	// if we are cancelling/completing this packet, do not add it
+	if (test_bit(SSH_PACKET_SF_LOCKED_BIT, &packet->state)) {
+		spin_unlock(&ptl->pending.lock);
+		return;
+	}
+
+	// in case it is already pending (e.g. re-submission), do not add it
+	if (test_and_set_bit(SSH_PACKET_SF_PENDING_BIT, &packet->state)) {
+		spin_unlock(&ptl->pending.lock);
+		return;
+	}
+
+	atomic_inc(&ptl->pending.count);
+	list_add_tail(&ssh_packet_get(packet)->pending_node, &ptl->pending.head);
+
+	spin_unlock(&ptl->pending.lock);
+}
+
+static void ssh_ptl_pending_remove(struct ssh_packet *packet)
+{
+	struct ssh_ptl *ptl = packet->ptl;
+
+	spin_lock(&ptl->pending.lock);
+
+	if (!test_and_clear_bit(SSH_PACKET_SF_PENDING_BIT, &packet->state)) {
+		spin_unlock(&ptl->pending.lock);
+		return;
+	}
+
+	list_del(&packet->pending_node);
+	atomic_dec(&ptl->pending.count);
+
+	spin_unlock(&ptl->pending.lock);
+
+	ssh_packet_put(packet);
+}
+
+
+/* warning: does not check/set "completed" bit */
+static void __ssh_ptl_complete(struct ssh_packet *p, int status)
+{
+	struct ssh_ptl *ptl = READ_ONCE(p->ptl);
+
+	trace_ssam_packet_complete(p, status);
+	ptl_dbg_cond(ptl, "ptl: completing packet %p (status: %d)\n", p, status);
+
+	if (p->ops->complete)
+		p->ops->complete(p, status);
+}
+
+static void ssh_ptl_remove_and_complete(struct ssh_packet *p, int status)
+{
+	/*
+	 * A call to this function should in general be preceded by
+	 * set_bit(SSH_PACKET_SF_LOCKED_BIT, &p->flags) to avoid re-adding the
+	 * packet to the structures it's going to be removed from.
+	 *
+	 * The set_bit call does not need explicit memory barriers as the
+	 * implicit barrier of the test_and_set_bit() call below ensure that the
+	 * flag is visible before we actually attempt to remove the packet.
+	 */
+
+	if (test_and_set_bit(SSH_PACKET_SF_COMPLETED_BIT, &p->state))
+		return;
+
+	ssh_ptl_queue_remove(p);
+	ssh_ptl_pending_remove(p);
+
+	__ssh_ptl_complete(p, status);
+}
+
+
+static bool ssh_ptl_tx_can_process(struct ssh_packet *packet)
+{
+	struct ssh_ptl *ptl = packet->ptl;
+
+	if (test_bit(SSH_PACKET_TY_FLUSH_BIT, &packet->state))
+		return !atomic_read(&ptl->pending.count);
+
+	// we can alwas process non-blocking packets
+	if (!test_bit(SSH_PACKET_TY_BLOCKING_BIT, &packet->state))
+		return true;
+
+	// if we are already waiting for this packet, send it again
+	if (test_bit(SSH_PACKET_SF_PENDING_BIT, &packet->state))
+		return true;
+
+	// otherwise: check if we have the capacity to send
+	return atomic_read(&ptl->pending.count) < SSH_PTL_MAX_PENDING;
+}
+
+static struct ssh_packet *ssh_ptl_tx_pop(struct ssh_ptl *ptl)
+{
+	struct ssh_packet *packet = ERR_PTR(-ENOENT);
+	struct ssh_packet *p, *n;
+
+	spin_lock(&ptl->queue.lock);
+	list_for_each_entry_safe(p, n, &ptl->queue.head, queue_node) {
+		/*
+		 * If we are cancelling or completing this packet, ignore it.
+		 * It's going to be removed from this queue shortly.
+		 */
+		if (test_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state))
+			continue;
+
+		/*
+		 * Packets should be ordered non-blocking/to-be-resent first.
+		 * If we cannot process this packet, assume that we can't
+		 * process any following packet either and abort.
+		 */
+		if (!ssh_ptl_tx_can_process(p)) {
+			packet = ERR_PTR(-EBUSY);
+			break;
+		}
+
+		/*
+		 * We are allowed to change the state now. Remove it from the
+		 * queue and mark it as being transmitted. Note that we cannot
+		 * add it to the set of pending packets yet, as queue locks must
+		 * always be acquired before packet locks (otherwise we might
+		 * run into a deadlock).
+		 */
+
+		list_del(&p->queue_node);
+
+		set_bit(SSH_PACKET_SF_TRANSMITTING_BIT, &p->state);
+		// ensure that state never gets zero
+		smp_mb__before_atomic();
+		clear_bit(SSH_PACKET_SF_QUEUED_BIT, &p->state);
+
+		packet = p;
+		break;
+	}
+	spin_unlock(&ptl->queue.lock);
+
+	return packet;
+}
+
+static struct ssh_packet *ssh_ptl_tx_next(struct ssh_ptl *ptl)
+{
+	struct ssh_packet *p;
+
+	p = ssh_ptl_tx_pop(ptl);
+	if (IS_ERR(p))
+		return p;
+
+	if (test_bit(SSH_PACKET_TY_SEQUENCED_BIT, &p->state)) {
+		ptl_dbg(ptl, "ptl: transmitting sequenced packet %p\n", p);
+		ssh_ptl_pending_push(p);
+		ssh_ptl_timeout_start(p);
+	} else {
+		ptl_dbg(ptl, "ptl: transmitting non-sequenced packet %p\n", p);
+	}
+
+	/*
+	 * Update number of tries. This directly influences the priority in case
+	 * the packet is re-submitted (e.g. via timeout/NAK). Note that this is
+	 * the only place where we update the priority in-flight. As this runs
+	 * only on the tx-thread, this read-modify-write procedure is safe.
+	 */
+	ssh_packet_next_try(p);
+
+	return p;
+}
+
+static void ssh_ptl_tx_compl_success(struct ssh_packet *packet)
+{
+	struct ssh_ptl *ptl = packet->ptl;
+
+	ptl_dbg(ptl, "ptl: successfully transmitted packet %p\n", packet);
+
+	// transition state to "transmitted"
+	set_bit(SSH_PACKET_SF_TRANSMITTED_BIT, &packet->state);
+	// ensure that state never gets zero
+	smp_mb__before_atomic();
+	clear_bit(SSH_PACKET_SF_TRANSMITTING_BIT, &packet->state);
+
+	// if the packet is unsequenced, we're done: lock and complete
+	if (!test_bit(SSH_PACKET_TY_SEQUENCED_BIT, &packet->state)) {
+		set_bit(SSH_PACKET_SF_LOCKED_BIT, &packet->state);
+		ssh_ptl_remove_and_complete(packet, 0);
+	}
+
+	/*
+	 * Notify that a packet transmission has finished. In general we're only
+	 * waiting for one packet (if any), so wake_up_all should be fine.
+	 */
+	wake_up_all(&ptl->tx.packet_wq);
+}
+
+static void ssh_ptl_tx_compl_error(struct ssh_packet *packet, int status)
+{
+	// transmission failure: lock the packet and try to complete it
+	set_bit(SSH_PACKET_SF_LOCKED_BIT, &packet->state);
+	// ensure that state never gets zero
+	smp_mb__before_atomic();
+	clear_bit(SSH_PACKET_SF_TRANSMITTING_BIT, &packet->state);
+
+	ptl_err(packet->ptl, "ptl: transmission error: %d\n", status);
+	ptl_dbg(packet->ptl, "ptl: failed to transmit packet: %p\n", packet);
+
+	ssh_ptl_remove_and_complete(packet, status);
+
+	/*
+	 * Notify that a packet transmission has finished. In general we're only
+	 * waiting for one packet (if any), so wake_up_all should be fine.
+	 */
+	wake_up_all(&packet->ptl->tx.packet_wq);
+}
+
+static void ssh_ptl_tx_threadfn_wait(struct ssh_ptl *ptl)
+{
+	wait_event_interruptible(ptl->tx.thread_wq,
+		READ_ONCE(ptl->tx.thread_signal) || kthread_should_stop());
+	WRITE_ONCE(ptl->tx.thread_signal, false);
+}
+
+static int ssh_ptl_tx_threadfn(void *data)
+{
+	struct ssh_ptl *ptl = data;
+
+	while (!kthread_should_stop()) {
+		unsigned char *buf;
+		bool drop = false;
+		size_t len = 0;
+		int status = 0;
+
+		// if we don't have a packet, get the next and add it to pending
+		if (IS_ERR_OR_NULL(ptl->tx.packet)) {
+			ptl->tx.packet = ssh_ptl_tx_next(ptl);
+			ptl->tx.offset = 0;
+
+			// if no packet can be processed, we are done
+			if (IS_ERR(ptl->tx.packet)) {
+				ssh_ptl_tx_threadfn_wait(ptl);
+				continue;
+			}
+		}
+
+		// error injection: drop packet to simulate transmission problem
+		if (ptl->tx.offset == 0)
+			drop = ssh_ptl_should_drop_packet(ptl->tx.packet);
+
+		// error injection: simulate invalid packet data
+		if (ptl->tx.offset == 0 && !drop)
+			ssh_ptl_tx_inject_invalid_data(ptl->tx.packet);
+
+		// note: flush-packets don't have any data
+		if (likely(ptl->tx.packet->data.ptr && !drop)) {
+			buf = ptl->tx.packet->data.ptr + ptl->tx.offset;
+			len = ptl->tx.packet->data.len - ptl->tx.offset;
+
+			ptl_dbg(ptl, "tx: sending data (length: %zu)\n", len);
+			print_hex_dump_debug("tx: ", DUMP_PREFIX_OFFSET, 16, 1,
+					     buf, len, false);
+
+			status = ssh_ptl_write_buf(ptl, ptl->tx.packet, buf, len);
+		}
+
+		if (status < 0) {
+			// complete packet with error
+			ssh_ptl_tx_compl_error(ptl->tx.packet, status);
+			ssh_packet_put(ptl->tx.packet);
+			ptl->tx.packet = NULL;
+
+		} else if (status == len) {
+			// complete packet and/or mark as transmitted
+			ssh_ptl_tx_compl_success(ptl->tx.packet);
+			ssh_packet_put(ptl->tx.packet);
+			ptl->tx.packet = NULL;
+
+		} else {	// need more buffer space
+			ptl->tx.offset += status;
+			ssh_ptl_tx_threadfn_wait(ptl);
+		}
+	}
+
+	// cancel active packet before we actually stop
+	if (!IS_ERR_OR_NULL(ptl->tx.packet)) {
+		ssh_ptl_tx_compl_error(ptl->tx.packet, -ESHUTDOWN);
+		ssh_packet_put(ptl->tx.packet);
+		ptl->tx.packet = NULL;
+	}
+
+	return 0;
+}
+
+/**
+ * ssh_ptl_tx_wakeup() - Wake up packet transmitter thread.
+ * @ptl: The packet transport layer.
+ *
+ * Wakes up the packet transmitter thread. If the packet transport layer has
+ * been shut down, calls to this function will be ignored.
+ */
+void ssh_ptl_tx_wakeup(struct ssh_ptl *ptl)
+{
+	if (test_bit(SSH_PTL_SF_SHUTDOWN_BIT, &ptl->state))
+		return;
+
+	WRITE_ONCE(ptl->tx.thread_signal, true);
+	/*
+	 * Ensure that the signal is set before we wake the transmitter
+	 * thread to prevent lost updates: If the signal is not set,
+	 * when the thread checks it in ssh_ptl_tx_threadfn_wait(), it
+	 * may go back to sleep.
+	 */
+	smp_mb__after_atomic();
+	wake_up(&ptl->tx.thread_wq);
+}
+
+/**
+ * ssh_ptl_tx_start() - Start packet transmitter thread.
+ * @ptl: The packet transport layer.
+ *
+ * Return: Returns zero on success, a negative error code on failure.
+ */
+int ssh_ptl_tx_start(struct ssh_ptl *ptl)
+{
+	ptl->tx.thread = kthread_run(ssh_ptl_tx_threadfn, ptl,
+				     "ssam_serial_hub-tx");
+	if (IS_ERR(ptl->tx.thread))
+		return PTR_ERR(ptl->tx.thread);
+
+	return 0;
+}
+
+/**
+ * ssh_ptl_tx_stop() - Stop packet transmitter thread.
+ * @ptl: The packet transport layer.
+ *
+ * Return: Returns zero on success, a negative error code on failure.
+ */
+int ssh_ptl_tx_stop(struct ssh_ptl *ptl)
+{
+	int status = 0;
+
+	if (ptl->tx.thread) {
+		status = kthread_stop(ptl->tx.thread);
+		ptl->tx.thread = NULL;
+	}
+
+	return status;
+}
+
+
+static struct ssh_packet *ssh_ptl_ack_pop(struct ssh_ptl *ptl, u8 seq_id)
+{
+	struct ssh_packet *packet = ERR_PTR(-ENOENT);
+	struct ssh_packet *p, *n;
+
+	spin_lock(&ptl->pending.lock);
+	list_for_each_entry_safe(p, n, &ptl->pending.head, pending_node) {
+		/*
+		 * We generally expect packets to be in order, so first packet
+		 * to be added to pending is first to be sent, is first to be
+		 * ACKed.
+		 */
+		if (unlikely(ssh_packet_get_seq(p) != seq_id))
+			continue;
+
+		/*
+		 * In case we receive an ACK while handling a transmission
+		 * error completion. The packet will be removed shortly.
+		 */
+		if (unlikely(test_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state))) {
+			packet = ERR_PTR(-EPERM);
+			break;
+		}
+
+		/*
+		 * Mark the packet as ACKed and remove it from pending by
+		 * removing its node and decrementing the pending counter.
+		 */
+		set_bit(SSH_PACKET_SF_ACKED_BIT, &p->state);
+		// ensure that state never gets zero
+		smp_mb__before_atomic();
+		clear_bit(SSH_PACKET_SF_PENDING_BIT, &p->state);
+
+		atomic_dec(&ptl->pending.count);
+		list_del(&p->pending_node);
+		packet = p;
+
+		break;
+	}
+	spin_unlock(&ptl->pending.lock);
+
+	return packet;
+}
+
+static void ssh_ptl_wait_until_transmitted(struct ssh_packet *packet)
+{
+	wait_event(packet->ptl->tx.packet_wq,
+		   test_bit(SSH_PACKET_SF_TRANSMITTED_BIT, &packet->state)
+		   || test_bit(SSH_PACKET_SF_LOCKED_BIT, &packet->state));
+}
+
+static void ssh_ptl_acknowledge(struct ssh_ptl *ptl, u8 seq)
+{
+	struct ssh_packet *p;
+	int status = 0;
+
+	p = ssh_ptl_ack_pop(ptl, seq);
+	if (IS_ERR(p)) {
+		if (PTR_ERR(p) == -ENOENT) {
+			/*
+			 * The packet has not been found in the set of pending
+			 * packets.
+			 */
+			ptl_warn(ptl, "ptl: received ACK for non-pending packet\n");
+		} else {
+			/*
+			 * The packet is pending, but we are not allowed to take
+			 * it because it has been locked.
+			 */
+			WARN_ON(PTR_ERR(p) != -EPERM);
+		}
+		return;
+	}
+
+	ptl_dbg(ptl, "ptl: received ACK for packet %p\n", p);
+
+	/*
+	 * It is possible that the packet has been transmitted, but the state
+	 * has not been updated from "transmitting" to "transmitted" yet.
+	 * In that case, we need to wait for this transition to occur in order
+	 * to determine between success or failure.
+	 */
+	if (test_bit(SSH_PACKET_SF_TRANSMITTING_BIT, &p->state))
+		ssh_ptl_wait_until_transmitted(p);
+
+	/*
+	 * The packet will already be locked in case of a transmission error or
+	 * cancellation. Let the transmitter or cancellation issuer complete the
+	 * packet.
+	 */
+	if (unlikely(test_and_set_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state))) {
+		ssh_packet_put(p);
+		return;
+	}
+
+	if (unlikely(!test_bit(SSH_PACKET_SF_TRANSMITTED_BIT, &p->state))) {
+		ptl_err(ptl, "ptl: received ACK before packet had been fully transmitted\n");
+		status = -EREMOTEIO;
+	}
+
+	ssh_ptl_remove_and_complete(p, status);
+	ssh_packet_put(p);
+
+	if (atomic_read(&ptl->pending.count) < SSH_PTL_MAX_PENDING)
+		ssh_ptl_tx_wakeup(ptl);
+}
+
+
+/**
+ * ssh_ptl_submit() - Submit a packet to the transport layer.
+ * @ptl: The packet transport layer to submit the packet to.
+ * @p:   The packet to submit.
+ *
+ * Submits a new packet to the transport layer, queuing it to be sent. This
+ * function should not be used for re-submission.
+ *
+ * Return: Returns zero on success, %-EINVAL if a packet field is invalid or
+ * the packet has been canceled prior to submission, %-EALREADY if the packet
+ * has already been submitted, or %-ESHUTDOWN if the packet transport layer
+ * has been shut down.
+ */
+int ssh_ptl_submit(struct ssh_ptl *ptl, struct ssh_packet *p)
+{
+	struct ssh_ptl *ptl_old;
+	int status;
+
+	trace_ssam_packet_submit(p);
+
+	// validate packet fields
+	if (test_bit(SSH_PACKET_TY_FLUSH_BIT, &p->state)) {
+		if (p->data.ptr || test_bit(SSH_PACKET_TY_SEQUENCED_BIT, &p->state))
+			return -EINVAL;
+	} else if (!p->data.ptr) {
+		return -EINVAL;
+	}
+
+	/*
+	 * The ptl reference only gets set on or before the first submission.
+	 * After the first submission, it has to be read-only.
+	 */
+	ptl_old = READ_ONCE(p->ptl);
+	if (ptl_old == NULL)
+		WRITE_ONCE(p->ptl, ptl);
+	else if (WARN_ON(ptl_old != ptl))
+		return -EALREADY;	// submitted on different PTL
+
+	status = ssh_ptl_queue_push(p);
+	if (status)
+		return status;
+
+	if (!test_bit(SSH_PACKET_TY_BLOCKING_BIT, &p->state)
+	    || (atomic_read(&ptl->pending.count) < SSH_PTL_MAX_PENDING))
+		ssh_ptl_tx_wakeup(ptl);
+
+	return 0;
+}
+
+/* must be called with pending lock held */
+static int __ssh_ptl_resubmit(struct ssh_packet *packet)
+{
+	int status;
+
+	trace_ssam_packet_resubmit(packet);
+
+	spin_lock(&packet->ptl->queue.lock);
+
+	status = __ssh_ptl_queue_push(packet);
+	if (status) {
+		/*
+		 * An error here indicates that the packet has either already
+		 * been queued, been locked, or the transport layer is being
+		 * shut down. In all cases: Ignore the error.
+		 */
+		spin_unlock(&packet->ptl->queue.lock);
+		return status;
+	}
+
+	/*
+	 * Reset the timestamp. This must be called and executed before the
+	 * pending lock is released. The lock release should be a sufficient
+	 * barrier for this operation, thus there is no need to manually add
+	 * one here.
+	 */
+	WRITE_ONCE(packet->timestamp, KTIME_MAX);
+
+	spin_unlock(&packet->ptl->queue.lock);
+	return 0;
+}
+
+static void ssh_ptl_resubmit_pending(struct ssh_ptl *ptl)
+{
+	struct ssh_packet *p;
+	bool resub = false;
+	u8 try;
+
+	/*
+	 * Note: We deliberately do not remove/attempt to cancel and complete
+	 * packets that are out of tires in this function. The packet will be
+	 * eventually canceled and completed by the timeout. Removing the packet
+	 * here could lead to overly eager cancellation if the packet has not
+	 * been re-transmitted yet but the tries-counter already updated (i.e
+	 * ssh_ptl_tx_next() removed the packet from the queue and updated the
+	 * counter, but re-transmission for the last try has not actually
+	 * started yet).
+	 */
+
+	spin_lock(&ptl->pending.lock);
+
+	// re-queue all pending packets
+	list_for_each_entry(p, &ptl->pending.head, pending_node) {
+		// avoid further transitions if locked
+		if (test_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state))
+			continue;
+
+		// do not re-schedule if packet is out of tries
+		try = ssh_packet_priority_get_try(READ_ONCE(p->priority));
+		if (try >= SSH_PTL_MAX_PACKET_TRIES)
+			continue;
+
+		/*
+		 * Submission fails if the packet has been locked, is already
+		 * queued, or the layer is being shut down. No need to
+		 * re-schedule tx-thread in those cases.
+		 */
+		if (!__ssh_ptl_resubmit(p))
+			resub = true;
+	}
+
+	spin_unlock(&ptl->pending.lock);
+
+	if (resub)
+		ssh_ptl_tx_wakeup(ptl);
+}
+
+/**
+ * ssh_ptl_cancel() - Cancel a packet.
+ * @p: The packet to cancel.
+ *
+ * Cancels a packet. There are no guarantees on when completion and release
+ * callbacks will be called. This may occur during execution of this function
+ * or may occur at any point later.
+ *
+ * Note that it is not guaranteed that the packet will actually be cancelled
+ * if the packet is concurrently completed by another process. The only
+ * guarantee of this function is that the packet will be completed (with
+ * success, failure, or cancellation) and released from the transport layer in
+ * a reasonable time-frame.
+ *
+ * May be called before the packet has been submitted, in which case any later
+ * packet submission fails.
+ */
+void ssh_ptl_cancel(struct ssh_packet *p)
+{
+	if (test_and_set_bit(SSH_PACKET_SF_CANCELED_BIT, &p->state))
+		return;
+
+	trace_ssam_packet_cancel(p);
+
+	/*
+	 * Lock packet and commit with memory barrier. If this packet has
+	 * already been locked, it's going to be removed and completed by
+	 * another party, which should have precedence.
+	 */
+	if (test_and_set_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state))
+		return;
+
+	/*
+	 * By marking the packet as locked and employing the implicit memory
+	 * barrier of test_and_set_bit, we have guaranteed that, at this point,
+	 * the packet cannot be added to the queue any more.
+	 *
+	 * In case the packet has never been submitted, packet->ptl is NULL. If
+	 * the packet is currently being submitted, packet->ptl may be NULL or
+	 * non-NULL. Due marking the packet as locked above and committing with
+	 * the memory barrier, we have guaranteed that, if packet->ptl is NULL,
+	 * the packet will never be added to the queue. If packet->ptl is
+	 * non-NULL, we don't have any guarantees.
+	 */
+
+	if (READ_ONCE(p->ptl)) {
+		ssh_ptl_remove_and_complete(p, -ECANCELED);
+
+		if (atomic_read(&p->ptl->pending.count) < SSH_PTL_MAX_PENDING)
+			ssh_ptl_tx_wakeup(p->ptl);
+
+	} else if (!test_and_set_bit(SSH_PACKET_SF_COMPLETED_BIT, &p->state)) {
+		__ssh_ptl_complete(p, -ECANCELED);
+	}
+}
+
+
+static ktime_t ssh_packet_get_expiration(struct ssh_packet *p, ktime_t timeout)
+{
+	ktime_t timestamp = READ_ONCE(p->timestamp);
+
+	if (timestamp != KTIME_MAX)
+		return ktime_add(timestamp, timeout);
+	else
+		return KTIME_MAX;
+}
+
+static void ssh_ptl_timeout_reap(struct work_struct *work)
+{
+	struct ssh_ptl *ptl = to_ssh_ptl(work, rtx_timeout.reaper.work);
+	struct ssh_packet *p, *n;
+	LIST_HEAD(claimed);
+	ktime_t now = ktime_get_coarse_boottime();
+	ktime_t timeout = ptl->rtx_timeout.timeout;
+	ktime_t next = KTIME_MAX;
+	bool resub = false;
+
+	trace_ssam_ptl_timeout_reap("pending", atomic_read(&ptl->pending.count));
+
+	/*
+	 * Mark reaper as "not pending". This is done before checking any
+	 * packets to avoid lost-update type problems.
+	 */
+	WRITE_ONCE(ptl->rtx_timeout.expires, KTIME_MAX);
+	/*
+	 * Ensure that the reaper is marked as deactivated before we continue
+	 * checking packets to prevent lost-update problems when a packet is
+	 * added to the pending set and ssh_ptl_timeout_reaper_mod is called
+	 * during execution of the part below.
+	 */
+	smp_mb__after_atomic();
+
+	spin_lock(&ptl->pending.lock);
+
+	list_for_each_entry_safe(p, n, &ptl->pending.head, pending_node) {
+		ktime_t expires = ssh_packet_get_expiration(p, timeout);
+		u8 try;
+
+		/*
+		 * Check if the timeout hasn't expired yet. Find out next
+		 * expiration date to be handled after this run.
+		 */
+		if (ktime_after(expires, now)) {
+			next = ktime_before(expires, next) ? expires : next;
+			continue;
+		}
+
+		// check if we still have some tries left
+		try = ssh_packet_priority_get_try(READ_ONCE(p->priority));
+		if (likely(try < SSH_PTL_MAX_PACKET_TRIES)) {
+			trace_ssam_packet_timeout(p);
+
+			/*
+			 * Submission fails if the packet has been locked, is
+			 * already queued, or the layer is being shut down.
+			 * No need to re-schedule tx-thread in those cases.
+			 */
+			if (!__ssh_ptl_resubmit(p))
+				resub = true;
+
+			continue;
+		}
+
+		// no more tries left: cancel the packet
+
+		// if someone else has locked the packet already, don't use it
+		if (test_and_set_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state))
+			continue;
+
+		trace_ssam_packet_timeout(p);
+
+		/*
+		 * We have now marked the packet as locked. Thus it cannot be
+		 * added to the pending list again after we've removed it here.
+		 * We can therefore re-use the pending_node of this packet
+		 * temporarily.
+		 */
+
+		clear_bit(SSH_PACKET_SF_PENDING_BIT, &p->state);
+
+		atomic_dec(&ptl->pending.count);
+		list_del(&p->pending_node);
+
+		list_add_tail(&p->pending_node, &claimed);
+	}
+
+	spin_unlock(&ptl->pending.lock);
+
+	// cancel and complete the packet
+	list_for_each_entry_safe(p, n, &claimed, pending_node) {
+		if (!test_and_set_bit(SSH_PACKET_SF_COMPLETED_BIT, &p->state)) {
+			ssh_ptl_queue_remove(p);
+			__ssh_ptl_complete(p, -ETIMEDOUT);
+		}
+
+		// drop the reference we've obtained by removing it from pending
+		list_del(&p->pending_node);
+		ssh_packet_put(p);
+	}
+
+	// ensure that reaper doesn't run again immediately
+	next = max(next, ktime_add(now, SSH_PTL_PACKET_TIMEOUT_RESOLUTION));
+	if (next != KTIME_MAX)
+		ssh_ptl_timeout_reaper_mod(ptl, now, next);
+
+	if (resub)
+		ssh_ptl_tx_wakeup(ptl);
+}
+
+
+static bool ssh_ptl_rx_retransmit_check(struct ssh_ptl *ptl, u8 seq)
+{
+	int i;
+
+	// check if SEQ has been seen recently (i.e. packet was re-transmitted)
+	for (i = 0; i < ARRAY_SIZE(ptl->rx.blocked.seqs); i++) {
+		if (likely(ptl->rx.blocked.seqs[i] != seq))
+			continue;
+
+		ptl_dbg(ptl, "ptl: ignoring repeated data packet\n");
+		return true;
+	}
+
+	// update list of blocked seuence IDs
+	ptl->rx.blocked.seqs[ptl->rx.blocked.offset] = seq;
+	ptl->rx.blocked.offset = (ptl->rx.blocked.offset + 1)
+				  % ARRAY_SIZE(ptl->rx.blocked.seqs);
+
+	return false;
+}
+
+static void ssh_ptl_rx_dataframe(struct ssh_ptl *ptl,
+				 const struct ssh_frame *frame,
+				 const struct ssam_span *payload)
+{
+	if (ssh_ptl_rx_retransmit_check(ptl, frame->seq))
+		return;
+
+	ptl->ops.data_received(ptl, payload);
+}
+
+static void ssh_ptl_send_ack(struct ssh_ptl *ptl, u8 seq)
+{
+	struct ssh_packet *packet;
+	struct ssam_span buf;
+	struct msgbuf msgb;
+	int status;
+
+	status = ssh_ctrl_packet_alloc(&packet, &buf, GFP_KERNEL);
+	if (status) {
+		ptl_err(ptl, "ptl: failed to allocate ACK packet\n");
+		return;
+	}
+
+	ssh_packet_init(packet, 0, SSH_PACKET_PRIORITY(ACK, 0),
+			&ssh_ptl_ctrl_packet_ops);
+
+	msgb_init(&msgb, buf.ptr, buf.len);
+	msgb_push_ack(&msgb, seq);
+	ssh_packet_set_data(packet, msgb.begin, msgb_bytes_used(&msgb));
+
+	ssh_ptl_submit(ptl, packet);
+	ssh_packet_put(packet);
+}
+
+static void ssh_ptl_send_nak(struct ssh_ptl *ptl)
+{
+	struct ssh_packet *packet;
+	struct ssam_span buf;
+	struct msgbuf msgb;
+	int status;
+
+	status = ssh_ctrl_packet_alloc(&packet, &buf, GFP_KERNEL);
+	if (status) {
+		ptl_err(ptl, "ptl: failed to allocate NAK packet\n");
+		return;
+	}
+
+	ssh_packet_init(packet, 0, SSH_PACKET_PRIORITY(NAK, 0),
+			&ssh_ptl_ctrl_packet_ops);
+
+	msgb_init(&msgb, buf.ptr, buf.len);
+	msgb_push_nak(&msgb);
+	ssh_packet_set_data(packet, msgb.begin, msgb_bytes_used(&msgb));
+
+	ssh_ptl_submit(ptl, packet);
+	ssh_packet_put(packet);
+}
+
+static size_t ssh_ptl_rx_eval(struct ssh_ptl *ptl, struct ssam_span *source)
+{
+	struct ssh_frame *frame;
+	struct ssam_span payload;
+	struct ssam_span aligned;
+	bool syn_found;
+	int status;
+
+	// error injection: modify data to simulate corrupt SYN bytes
+	ssh_ptl_rx_inject_invalid_syn(ptl, source);
+
+	// find SYN
+	syn_found = sshp_find_syn(source, &aligned);
+
+	if (unlikely(aligned.ptr - source->ptr) > 0) {
+		ptl_warn(ptl, "rx: parser: invalid start of frame, skipping\n");
+
+		/*
+		 * Notes:
+		 * - This might send multiple NAKs in case the communication
+		 *   starts with an invalid SYN and is broken down into multiple
+		 *   pieces. This should generally be handled fine, we just
+		 *   might receive duplicate data in this case, which is
+		 *   detected when handling data frames.
+		 * - This path will also be executed on invalid CRCs: When an
+		 *   invalid CRC is encountered, the code below will skip data
+		 *   until direclty after the SYN. This causes the search for
+		 *   the next SYN, which is generally not placed directly after
+		 *   the last one.
+		 *
+		 *   Open question: Should we send this in case of invalid
+		 *   payload CRCs if the frame-type is nonsequential (current
+		 *   implementation) or should we drop that frame without
+		 *   telling the EC?
+		 */
+		ssh_ptl_send_nak(ptl);
+	}
+
+	if (unlikely(!syn_found))
+		return aligned.ptr - source->ptr;
+
+	// error injection: modify data to simulate corruption
+	ssh_ptl_rx_inject_invalid_data(ptl, &aligned);
+
+	// parse and validate frame
+	status = sshp_parse_frame(&ptl->serdev->dev, &aligned, &frame, &payload,
+				  SSH_PTL_RX_BUF_LEN);
+	if (status)	// invalid frame: skip to next syn
+		return aligned.ptr - source->ptr + sizeof(u16);
+	if (!frame)	// not enough data
+		return aligned.ptr - source->ptr;
+
+	trace_ssam_rx_frame_received(frame);
+
+	switch (frame->type) {
+	case SSH_FRAME_TYPE_ACK:
+		ssh_ptl_acknowledge(ptl, frame->seq);
+		break;
+
+	case SSH_FRAME_TYPE_NAK:
+		ssh_ptl_resubmit_pending(ptl);
+		break;
+
+	case SSH_FRAME_TYPE_DATA_SEQ:
+		ssh_ptl_send_ack(ptl, frame->seq);
+		fallthrough;
+
+	case SSH_FRAME_TYPE_DATA_NSQ:
+		ssh_ptl_rx_dataframe(ptl, frame, &payload);
+		break;
+
+	default:
+		ptl_warn(ptl, "ptl: received frame with unknown type 0x%02x\n",
+			 frame->type);
+		break;
+	}
+
+	return aligned.ptr - source->ptr + SSH_MESSAGE_LENGTH(frame->len);
+}
+
+static int ssh_ptl_rx_threadfn(void *data)
+{
+	struct ssh_ptl *ptl = data;
+
+	while (true) {
+		struct ssam_span span;
+		size_t offs = 0;
+		size_t n;
+
+		wait_event_interruptible(ptl->rx.wq,
+					 !kfifo_is_empty(&ptl->rx.fifo)
+					 || kthread_should_stop());
+		if (kthread_should_stop())
+			break;
+
+		// copy from fifo to evaluation buffer
+		n = sshp_buf_read_from_fifo(&ptl->rx.buf, &ptl->rx.fifo);
+
+		ptl_dbg(ptl, "rx: received data (size: %zu)\n", n);
+		print_hex_dump_debug("rx: ", DUMP_PREFIX_OFFSET, 16, 1,
+				     ptl->rx.buf.ptr + ptl->rx.buf.len - n,
+				     n, false);
+
+		// parse until we need more bytes or buffer is empty
+		while (offs < ptl->rx.buf.len) {
+			sshp_buf_span_from(&ptl->rx.buf, offs, &span);
+			n = ssh_ptl_rx_eval(ptl, &span);
+			if (n == 0)
+				break;	// need more bytes
+
+			offs += n;
+		}
+
+		// throw away the evaluated parts
+		sshp_buf_drop(&ptl->rx.buf, offs);
+	}
+
+	return 0;
+}
+
+static void ssh_ptl_rx_wakeup(struct ssh_ptl *ptl)
+{
+	wake_up(&ptl->rx.wq);
+}
+
+/**
+ * ssh_ptl_rx_start() - Start packet transport layer receiver thread.
+ * @ptl: The packet transport layer.
+ *
+ * Return: Returns zero on success, a negative error code on failure.
+ */
+int ssh_ptl_rx_start(struct ssh_ptl *ptl)
+{
+	if (ptl->rx.thread)
+		return 0;
+
+	ptl->rx.thread = kthread_run(ssh_ptl_rx_threadfn, ptl,
+				     "ssam_serial_hub-rx");
+	if (IS_ERR(ptl->rx.thread))
+		return PTR_ERR(ptl->rx.thread);
+
+	return 0;
+}
+
+/**
+ * ssh_ptl_rx_stop() - Stop packet transport layer receiver thread.
+ * @ptl: The packet transport layer.
+ *
+ * Return: Returns zero on success, a negative error code on failure.
+ */
+int ssh_ptl_rx_stop(struct ssh_ptl *ptl)
+{
+	int status = 0;
+
+	if (ptl->rx.thread) {
+		status = kthread_stop(ptl->rx.thread);
+		ptl->rx.thread = NULL;
+	}
+
+	return status;
+}
+
+/**
+ * ssh_ptl_rx_rcvbuf() - Push data from lower-layer transport to the packet
+ * layer.
+ * @ptl: The packet transport layer.
+ * @buf: Pointer to the data to push to the layer.
+ * @n:   Size of the data to push to the layer, in bytes.
+ *
+ * Pushes data from a lower-layer transport to the receiver fifo buffer of the
+ * packet layer and notifies the reveiver thread. Calls to this function are
+ * ignored once the packet layer has been shut down.
+ *
+ * Return: Returns the number of bytes transferred (positive or zero) on
+ * success. Returns %-ESHUTDOWN if the packet layer has been shut down.
+ */
+int ssh_ptl_rx_rcvbuf(struct ssh_ptl *ptl, const u8 *buf, size_t n)
+{
+	int used;
+
+	if (test_bit(SSH_PTL_SF_SHUTDOWN_BIT, &ptl->state))
+		return -ESHUTDOWN;
+
+	used = kfifo_in(&ptl->rx.fifo, buf, n);
+	if (used)
+		ssh_ptl_rx_wakeup(ptl);
+
+	return used;
+}
+
+
+/**
+ * ssh_ptl_shutdown() - Shut down the packet transport layer.
+ * @ptl: The packet transport layer.
+ *
+ * Shuts down the packet transport layer, removing and canceling all queued
+ * and pending packets. Packets canceled by this operation will be completed
+ * with %-ESHUTDOWN as status. Receiver and transmitter threads will be
+ * stopped.
+ *
+ * As a result of this function, the transport layer will be marked as shut
+ * down. Submission of packets after the transport layer has been shut down
+ * will fail with %-ESHUTDOWN.
+ */
+void ssh_ptl_shutdown(struct ssh_ptl *ptl)
+{
+	LIST_HEAD(complete_q);
+	LIST_HEAD(complete_p);
+	struct ssh_packet *p, *n;
+	int status;
+
+	// ensure that no new packets (including ACK/NAK) can be submitted
+	set_bit(SSH_PTL_SF_SHUTDOWN_BIT, &ptl->state);
+	/*
+	 * Ensure that the layer gets marked as shut-down before actually
+	 * stopping it. In combination with the check in ssh_ptl_queue_push(),
+	 * this guarantees that no new packets can be added and all already
+	 * queued packets are properly cancelled. In combination with the check
+	 * in ssh_ptl_rx_rcvbuf(), this guarantees that received data is
+	 * properly cut off.
+	 */
+	smp_mb__after_atomic();
+
+	status = ssh_ptl_rx_stop(ptl);
+	if (status)
+		ptl_err(ptl, "ptl: failed to stop receiver thread\n");
+
+	status = ssh_ptl_tx_stop(ptl);
+	if (status)
+		ptl_err(ptl, "ptl: failed to stop transmitter thread\n");
+
+	cancel_delayed_work_sync(&ptl->rtx_timeout.reaper);
+
+	/*
+	 * At this point, all threads have been stopped. This means that the
+	 * only references to packets from inside the system are in the queue
+	 * and pending set.
+	 *
+	 * Note: We still need locks here because someone could still be
+	 * cancelling packets.
+	 *
+	 * Note 2: We can re-use queue_node (or pending_node) if we mark the
+	 * packet as locked an then remove it from the queue (or pending set
+	 * respecitvely). Marking the packet as locked avoids re-queueing
+	 * (which should already be prevented by having stopped the treads...)
+	 * and not setting QUEUED_BIT (or PENDING_BIT) prevents removal from a
+	 * new list via other threads (e.g. canellation).
+	 *
+	 * Note 3: There may be overlap between complete_p and complete_q.
+	 * This is handled via test_and_set_bit() on the "completed" flag
+	 * (also handles cancellation).
+	 */
+
+	// mark queued packets as locked and move them to complete_q
+	spin_lock(&ptl->queue.lock);
+	list_for_each_entry_safe(p, n, &ptl->queue.head, queue_node) {
+		set_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state);
+		// ensure that state does not get zero
+		smp_mb__before_atomic();
+		clear_bit(SSH_PACKET_SF_QUEUED_BIT, &p->state);
+
+		list_del(&p->queue_node);
+		list_add_tail(&p->queue_node, &complete_q);
+	}
+	spin_unlock(&ptl->queue.lock);
+
+	// mark pending packets as locked and move them to complete_p
+	spin_lock(&ptl->pending.lock);
+	list_for_each_entry_safe(p, n, &ptl->pending.head, pending_node) {
+		set_bit(SSH_PACKET_SF_LOCKED_BIT, &p->state);
+		// ensure that state does not get zero
+		smp_mb__before_atomic();
+		clear_bit(SSH_PACKET_SF_PENDING_BIT, &p->state);
+
+		list_del(&p->pending_node);
+		list_add_tail(&p->pending_node, &complete_q);
+	}
+	atomic_set(&ptl->pending.count, 0);
+	spin_unlock(&ptl->pending.lock);
+
+	// complete and drop packets on complete_q
+	list_for_each_entry(p, &complete_q, queue_node) {
+		if (!test_and_set_bit(SSH_PACKET_SF_COMPLETED_BIT, &p->state))
+			__ssh_ptl_complete(p, -ESHUTDOWN);
+
+		ssh_packet_put(p);
+	}
+
+	// complete and drop packets on complete_p
+	list_for_each_entry(p, &complete_p, pending_node) {
+		if (!test_and_set_bit(SSH_PACKET_SF_COMPLETED_BIT, &p->state))
+			__ssh_ptl_complete(p, -ESHUTDOWN);
+
+		ssh_packet_put(p);
+	}
+
+	/*
+	 * At this point we have guaranteed that the system doesn't reference
+	 * any packets any more.
+	 */
+}
+
+/**
+ * ssh_ptl_init() - Initialize packet transport layer.
+ * @ptl:    The packet transport layer to initialize.
+ * @serdev: The underlying serial device, i.e. the lower-level transport.
+ * @ops:    Packet layer operations.
+ *
+ * Initializes the given packet transport layer. Transmitter and receiver
+ * threads must be started separately via ssh_ptl_tx_start() and
+ * ssh_ptl_rx_start(), after the packet-layer has been initialized and the
+ * lower-level transport layer has been set up.
+ *
+ * Return: Returns zero on success and a nonzero error code on failure.
+ */
+int ssh_ptl_init(struct ssh_ptl *ptl, struct serdev_device *serdev,
+		 struct ssh_ptl_ops *ops)
+{
+	int i, status;
+
+	ptl->serdev = serdev;
+	ptl->state = 0;
+
+	spin_lock_init(&ptl->queue.lock);
+	INIT_LIST_HEAD(&ptl->queue.head);
+
+	spin_lock_init(&ptl->pending.lock);
+	INIT_LIST_HEAD(&ptl->pending.head);
+	atomic_set_release(&ptl->pending.count, 0);
+
+	ptl->tx.thread = NULL;
+	ptl->tx.thread_signal = false;
+	ptl->tx.packet = NULL;
+	ptl->tx.offset = 0;
+	init_waitqueue_head(&ptl->tx.thread_wq);
+	init_waitqueue_head(&ptl->tx.packet_wq);
+
+	ptl->rx.thread = NULL;
+	init_waitqueue_head(&ptl->rx.wq);
+
+	ptl->rtx_timeout.timeout = SSH_PTL_PACKET_TIMEOUT;
+	ptl->rtx_timeout.expires = KTIME_MAX;
+	INIT_DELAYED_WORK(&ptl->rtx_timeout.reaper, ssh_ptl_timeout_reap);
+
+	ptl->ops = *ops;
+
+	// initialize list of recent/blocked SEQs with invalid sequence IDs
+	for (i = 0; i < ARRAY_SIZE(ptl->rx.blocked.seqs); i++)
+		ptl->rx.blocked.seqs[i] = 0xFFFF;
+	ptl->rx.blocked.offset = 0;
+
+	status = kfifo_alloc(&ptl->rx.fifo, SSH_PTL_RX_FIFO_LEN, GFP_KERNEL);
+	if (status)
+		return status;
+
+	status = sshp_buf_alloc(&ptl->rx.buf, SSH_PTL_RX_BUF_LEN, GFP_KERNEL);
+	if (status)
+		kfifo_free(&ptl->rx.fifo);
+
+	return status;
+}
+
+/**
+ * ssh_ptl_destroy() - Deinitialize packet transport layer.
+ * @ptl: The packet transport layer to deinitialize.
+ *
+ * Deinitializes the given packet transport layer and frees resources
+ * associated with it. If receiver and/or transmitter threads have been
+ * started, the layer must first be shut down via ssh_ptl_shutdown() before
+ * this function can be called.
+ */
+void ssh_ptl_destroy(struct ssh_ptl *ptl)
+{
+	kfifo_free(&ptl->rx.fifo);
+	sshp_buf_free(&ptl->rx.buf);
+}
diff --git a/drivers/misc/surface_aggregator/ssh_packet_layer.h b/drivers/misc/surface_aggregator/ssh_packet_layer.h
new file mode 100644
index 000000000000..f3d8a85389d5
--- /dev/null
+++ b/drivers/misc/surface_aggregator/ssh_packet_layer.h
@@ -0,0 +1,175 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * SSH packet transport layer.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _SURFACE_AGGREGATOR_SSH_PACKET_LAYER_H
+#define _SURFACE_AGGREGATOR_SSH_PACKET_LAYER_H
+
+#include <linux/atomic.h>
+#include <linux/kfifo.h>
+#include <linux/ktime.h>
+#include <linux/list.h>
+#include <linux/serdev.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+#include <linux/wait.h>
+#include <linux/workqueue.h>
+
+#include <linux/surface_aggregator/serial_hub.h>
+#include "ssh_parser.h"
+
+
+/**
+ * enum ssh_ptl_state_flags - State-flags for &struct ssh_ptl.
+ *
+ * @SSH_PTL_SF_SHUTDOWN_BIT:
+ *	Indicates that the packet transport layer has been shut down or is
+ *	being shut down and should not accept any new packets/data.
+ */
+enum ssh_ptl_state_flags {
+	SSH_PTL_SF_SHUTDOWN_BIT,
+};
+
+/**
+ * struct ssh_ptl_ops - Callback operations for packet transport layer.
+ * @data_received: Function called when a data-packet has been received. Both,
+ *                 the packet layer on which the packet has been received and
+ *                 the packet's payload data are provided to this function.
+ */
+struct ssh_ptl_ops {
+	void (*data_received)(struct ssh_ptl *p, const struct ssam_span *data);
+};
+
+/**
+ * struct ssh_ptl - SSH packet transport layer.
+ * @serdev:        Serial device providing the underlying data transport.
+ * @state:         State(-flags) of the transport layer.
+ * @queue:         Packet submission queue.
+ * @queue.lock:    Lock for modifying the packet submission queue.
+ * @queue.head:    List-head of the packet submission queue.
+ * @pending:       Set/list of pending packets.
+ * @pending.lock:  Lock for modifying the pending set.
+ * @pending.head:  List-head of the pending set/list.
+ * @pending.count: Number of currently pending packets.
+ * @tx:            Transmitter subsystem.
+ * @tx.thread_signal: Signal notifying transmitter thread of data to be sent.
+ * @tx.thread:     Transmitter thread.
+ * @tx.thread_wq:  Waitqueue-head for transmitter thread.
+ * @tx.packet_wq:  Waitqueue-head for packet transmit completion.
+ * @tx.packet:     Currently sent packet.
+ * @tx.offset:     Data-offset into the packet currently being transmitted.
+ * @rx:            Receiver subsystem.
+ * @rx.thread:     Receiver thread.
+ * @rx.wq:         Waitqueue-head for receiver thread.
+ * @rx.fifo:       Buffer for receiving data/pushing data to receiver thread.
+ * @rx.buf:        Buffer for evaluating data on receiver thread.
+ * @rx.blocked:    List of recent/blocked sequence IDs to detect retransmission.
+ * @rx.blocked.seqs:   Array of blocked sequence IDs.
+ * @rx.blocked.offset: Offset indicating where a new ID should be inserted.
+ * @rtx_timeout:   Retransmission timeout subsystem.
+ * @rtx_timeout.timeout: Timeout inverval for retransmission.
+ * @rtx_timeout.expires: Time specifying when the reaper work is next scheduled.
+ * @rtx_timeout.reaper:  Work performing timeout checks and subsequent actions.
+ * @ops:           Packet layer operations.
+ */
+struct ssh_ptl {
+	struct serdev_device *serdev;
+	unsigned long state;
+
+	struct {
+		spinlock_t lock;
+		struct list_head head;
+	} queue;
+
+	struct {
+		spinlock_t lock;
+		struct list_head head;
+		atomic_t count;
+	} pending;
+
+	struct {
+		bool thread_signal;
+		struct task_struct *thread;
+		struct wait_queue_head thread_wq;
+		struct wait_queue_head packet_wq;
+		struct ssh_packet *packet;
+		size_t offset;
+	} tx;
+
+	struct {
+		struct task_struct *thread;
+		struct wait_queue_head wq;
+		struct kfifo fifo;
+		struct sshp_buf buf;
+
+		struct {
+			u16 seqs[8];
+			u16 offset;
+		} blocked;
+	} rx;
+
+	struct {
+		ktime_t timeout;
+		ktime_t expires;
+		struct delayed_work reaper;
+	} rtx_timeout;
+
+	struct ssh_ptl_ops ops;
+};
+
+
+#define __ssam_prcond(func, p, fmt, ...)		\
+	do {						\
+		if ((p))				\
+			func((p), fmt, ##__VA_ARGS__);	\
+	} while (0)
+
+#define ptl_dbg(p, fmt, ...)  dev_dbg(&(p)->serdev->dev, fmt, ##__VA_ARGS__)
+#define ptl_info(p, fmt, ...) dev_info(&(p)->serdev->dev, fmt, ##__VA_ARGS__)
+#define ptl_warn(p, fmt, ...) dev_warn(&(p)->serdev->dev, fmt, ##__VA_ARGS__)
+#define ptl_err(p, fmt, ...)  dev_err(&(p)->serdev->dev, fmt, ##__VA_ARGS__)
+#define ptl_dbg_cond(p, fmt, ...) __ssam_prcond(ptl_dbg, p, fmt, ##__VA_ARGS__)
+
+#define to_ssh_ptl(ptr, member) \
+	container_of(ptr, struct ssh_ptl, member)
+
+
+int ssh_ptl_init(struct ssh_ptl *ptl, struct serdev_device *serdev,
+		 struct ssh_ptl_ops *ops);
+
+void ssh_ptl_destroy(struct ssh_ptl *ptl);
+
+/**
+ * ssh_ptl_get_device() - Get device associated with packet transport layer.
+ * @ptl: The packet transport layer.
+ *
+ * Return: Returns the device on which the given packet transport layer builds
+ * upon.
+ */
+static inline struct device *ssh_ptl_get_device(struct ssh_ptl *ptl)
+{
+	return ptl->serdev ? &ptl->serdev->dev : NULL;
+}
+
+int ssh_ptl_tx_start(struct ssh_ptl *ptl);
+int ssh_ptl_tx_stop(struct ssh_ptl *ptl);
+int ssh_ptl_rx_start(struct ssh_ptl *ptl);
+int ssh_ptl_rx_stop(struct ssh_ptl *ptl);
+void ssh_ptl_shutdown(struct ssh_ptl *ptl);
+
+int ssh_ptl_submit(struct ssh_ptl *ptl, struct ssh_packet *p);
+void ssh_ptl_cancel(struct ssh_packet *p);
+
+int ssh_ptl_rx_rcvbuf(struct ssh_ptl *ptl, const u8 *buf, size_t n);
+void ssh_ptl_tx_wakeup(struct ssh_ptl *ptl);
+
+void ssh_packet_init(struct ssh_packet *packet, unsigned long type,
+		     u8 priority, const struct ssh_packet_ops *ops);
+
+int ssh_ctrl_packet_cache_init(void);
+void ssh_ctrl_packet_cache_destroy(void);
+
+#endif /* _SURFACE_AGGREGATOR_SSH_PACKET_LAYER_H */
diff --git a/drivers/misc/surface_aggregator/ssh_parser.c b/drivers/misc/surface_aggregator/ssh_parser.c
new file mode 100644
index 000000000000..575cbc039ad7
--- /dev/null
+++ b/drivers/misc/surface_aggregator/ssh_parser.c
@@ -0,0 +1,229 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * SSH message parser.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <asm/unaligned.h>
+#include <linux/compiler.h>
+#include <linux/device.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/serial_hub.h>
+#include "ssh_parser.h"
+
+
+/**
+ * sshp_validate_crc() - Validate a CRC in raw message data.
+ * @src: The span of data over which the CRC should be computed.
+ * @crc: The pointer to the expected u16 CRC value.
+ *
+ * Computes the CRC of the provided data span (@src), compares it to the CRC
+ * stored at the given address (@crc), and returns the result of this
+ * comparison, i.e. %true iff equal. This function is intended to run on raw
+ * input/message data.
+ *
+ * Return: Returns %true iff the computed CRC matches the stored CRC, %false
+ * otherwise.
+ */
+static bool sshp_validate_crc(const struct ssam_span *src, const u8 *crc)
+{
+	u16 actual = ssh_crc(src->ptr, src->len);
+	u16 expected = get_unaligned_le16(crc);
+
+	return actual == expected;
+}
+
+/**
+ * sshp_starts_with_syn() - Check if the given data starts with SSH SYN bytes.
+ * @src: The data span to check the start of.
+ */
+static bool sshp_starts_with_syn(const struct ssam_span *src)
+{
+	return src->len >= 2 && get_unaligned_le16(src->ptr) == SSH_MSG_SYN;
+}
+
+/**
+ * sshp_find_syn() - Find SSH SYN bytes in the given data span.
+ * @src: The data span to search in.
+ * @rem: The span (output) indicating the remaining data, starting with SSH
+ *       SYN bytes, if found.
+ *
+ * Search for SSH SYN bytes in the given source span. If found, set the @rem
+ * span to the remaining data, starting with the first SYN bytes and capped by
+ * the source span length, and return %true. This function does not copy
+ * any data, but rather only sets pointers to the respecitve start addresses
+ * and length values.
+ *
+ * If no SSH SYN bytes could be found, set the @rem span to the zero-length
+ * span at the end of the source span and return %false.
+ *
+ * If partial SSH SYN bytes could be found at the end of the source span, set
+ * the @rem span to cover these partial SYN bytes, capped by the end of the
+ * source span, and return %false. This function should then be re-run once
+ * more data is available.
+ *
+ * Return: Returns %true iff a complete SSG SYN sequence could be found,
+ * %false otherwise.
+ */
+bool sshp_find_syn(const struct ssam_span *src, struct ssam_span *rem)
+{
+	size_t i;
+
+	for (i = 0; i < src->len - 1; i++) {
+		if (likely(get_unaligned_le16(src->ptr + i) == SSH_MSG_SYN)) {
+			rem->ptr = src->ptr + i;
+			rem->len = src->len - i;
+			return true;
+		}
+	}
+
+	if (unlikely(src->ptr[src->len - 1] == (SSH_MSG_SYN & 0xff))) {
+		rem->ptr = src->ptr + src->len - 1;
+		rem->len = 1;
+		return false;
+	}
+
+	rem->ptr = src->ptr + src->len;
+	rem->len = 0;
+	return false;
+}
+
+/**
+ * sshp_parse_frame() - Parse SSH frame.
+ * @dev: The device used for logging.
+ * @source: The source to parse from.
+ * @frame: The parsed frame (output).
+ * @payload: The parsed payload (output).
+ * @maxlen: The maximum supported message length.
+ *
+ * Parses and validates a SSH frame, including its payload, from the given
+ * source. Sets the provided @frame pointer to the start of the frame and
+ * writes the limits of the frame payload to the provided @payload span
+ * pointer.
+ *
+ * This function does not copy any data, but rather only validates the message
+ * data and sets pointers (and length values) to indicate the respective parts.
+ *
+ * If no complete SSH frame could be found, the frame pointer will be set to
+ * the %NULL pointer and the payload span will be set to the null span (start
+ * pointer %NULL, size zero).
+ *
+ * Return: Returns zero on success or if the frame is incomplete, %-ENOMSG if
+ * the start of the message is invalid, %-EBADMSG if any (frame-header or
+ * payload) CRC is ivnalid, or %-EMSGSIZE if the SSH message is bigger than
+ * the maximum message length specified in the @maxlen parameter.
+ */
+int sshp_parse_frame(const struct device *dev, const struct ssam_span *source,
+		     struct ssh_frame **frame, struct ssam_span *payload,
+		     size_t maxlen)
+{
+	struct ssam_span sf;
+	struct ssam_span sp;
+
+	// initialize output
+	*frame = NULL;
+	payload->ptr = NULL;
+	payload->len = 0;
+
+	if (!sshp_starts_with_syn(source)) {
+		dev_warn(dev, "rx: parser: invalid start of frame\n");
+		return -ENOMSG;
+	}
+
+	// check for minimum packet length
+	if (unlikely(source->len < SSH_MESSAGE_LENGTH(0))) {
+		dev_dbg(dev, "rx: parser: not enough data for frame\n");
+		return 0;
+	}
+
+	// pin down frame
+	sf.ptr = source->ptr + sizeof(u16);
+	sf.len = sizeof(struct ssh_frame);
+
+	// validate frame CRC
+	if (unlikely(!sshp_validate_crc(&sf, sf.ptr + sf.len))) {
+		dev_warn(dev, "rx: parser: invalid frame CRC\n");
+		return -EBADMSG;
+	}
+
+	// ensure packet does not exceed maximum length
+	sp.len = get_unaligned_le16(&((struct ssh_frame *)sf.ptr)->len);
+	if (unlikely(sp.len + SSH_MESSAGE_LENGTH(0) > maxlen)) {
+		dev_warn(dev, "rx: parser: frame too large: %u bytes\n",
+			 ((struct ssh_frame *)sf.ptr)->len);
+		return -EMSGSIZE;
+	}
+
+	// pin down payload
+	sp.ptr = sf.ptr + sf.len + sizeof(u16);
+
+	// check for frame + payload length
+	if (source->len < SSH_MESSAGE_LENGTH(sp.len)) {
+		dev_dbg(dev, "rx: parser: not enough data for payload\n");
+		return 0;
+	}
+
+	// validate payload crc
+	if (unlikely(!sshp_validate_crc(&sp, sp.ptr + sp.len))) {
+		dev_warn(dev, "rx: parser: invalid payload CRC\n");
+		return -EBADMSG;
+	}
+
+	*frame = (struct ssh_frame *)sf.ptr;
+	*payload = sp;
+
+	dev_dbg(dev, "rx: parser: valid frame found (type: 0x%02x, len: %u)\n",
+		(*frame)->type, (*frame)->len);
+
+	return 0;
+}
+
+/**
+ * sshp_parse_command() - Parse SSH command frame payload.
+ * @dev: The device used for logging.
+ * @source: The source to parse from.
+ * @command: The parsed command (output).
+ * @command_data: The parsed command data/payload (output).
+ *
+ * Parses and validates a SSH command frame payload. Sets the @command pointer
+ * to the command header and the @command_data span to the command data (i.e.
+ * payload of the command). This will result in a zero-length span if the
+ * command does not have any associated data/payload. This function does not
+ * check the frame-payload-type field, which should be checked by the caller
+ * before calling this function.
+ *
+ * The @source parameter should be the complete frame payload, e.g. returned
+ * by the sshp_parse_frame() command.
+ *
+ * This function does not copy any data, but rather only validates the frame
+ * payload data and sets pointers (and length values) to indicate the
+ * respective parts.
+ *
+ * Return: Returns zero on success or %-ENOMSG if @source does not represent a
+ * valid command-type frame payload, i.e. is too short.
+ */
+int sshp_parse_command(const struct device *dev, const struct ssam_span *source,
+		       struct ssh_command **command,
+		       struct ssam_span *command_data)
+{
+	// check for minimum length
+	if (unlikely(source->len < sizeof(struct ssh_command))) {
+		*command = NULL;
+		command_data->ptr = NULL;
+		command_data->len = 0;
+
+		dev_err(dev, "rx: parser: command payload is too short\n");
+		return -ENOMSG;
+	}
+
+	*command = (struct ssh_command *)source->ptr;
+	command_data->ptr = source->ptr + sizeof(struct ssh_command);
+	command_data->len = source->len - sizeof(struct ssh_command);
+
+	dev_dbg(dev, "rx: parser: valid command found (tc: 0x%02x, cid: 0x%02x)\n",
+		(*command)->tc, (*command)->cid);
+
+	return 0;
+}
diff --git a/drivers/misc/surface_aggregator/ssh_parser.h b/drivers/misc/surface_aggregator/ssh_parser.h
new file mode 100644
index 000000000000..71c43ab07bf6
--- /dev/null
+++ b/drivers/misc/surface_aggregator/ssh_parser.h
@@ -0,0 +1,157 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * SSH message parser.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _SURFACE_AGGREGATOR_SSH_PARSER_H
+#define _SURFACE_AGGREGATOR_SSH_PARSER_H
+
+#include <linux/device.h>
+#include <linux/kfifo.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/serial_hub.h>
+
+
+/**
+ * struct sshp_buf - Parser buffer for SSH messages.
+ * @ptr: Pointer to the beginning of the buffer.
+ * @len: Number of bytes used in the buffer.
+ * @cap: Maximum capacity of the buffer.
+ */
+struct sshp_buf {
+	u8    *ptr;
+	size_t len;
+	size_t cap;
+};
+
+/**
+ * sshp_buf_init() - Initialize a SSH parser buffer.
+ * @buf: The buffer to initialize.
+ * @ptr: The memory backing the buffer.
+ * @cap: The length of the memory backing the buffer, i.e. its capacity.
+ *
+ * Initializes the buffer with the given memory as backing and set its used
+ * length to zero.
+ */
+static inline void sshp_buf_init(struct sshp_buf *buf, u8 *ptr, size_t cap)
+{
+	buf->ptr = ptr;
+	buf->len = 0;
+	buf->cap = cap;
+}
+
+/**
+ * sshp_buf_alloc() - Allocate and initialize a SSH parser buffer.
+ * @buf:   The buffer to initialize/allocate to.
+ * @cap:   The desired capacity of the buffer.
+ * @flags: The flags used for allocating the memory.
+ *
+ * Allocates @cap bytes and initializes the provided buffer struct with the
+ * allocated memory.
+ *
+ * Return: Returns zero on success and %-ENOMEM if allocation failed.
+ */
+static inline int sshp_buf_alloc(struct sshp_buf *buf, size_t cap, gfp_t flags)
+{
+	u8 *ptr;
+
+	ptr = kzalloc(cap, flags);
+	if (!ptr)
+		return -ENOMEM;
+
+	sshp_buf_init(buf, ptr, cap);
+	return 0;
+
+}
+
+/**
+ * sshp_buf_free() - Free a SSH parser buffer.
+ * @buf: The buffer to free.
+ *
+ * Frees a SSH parser buffer by freeing the memory backing it and then
+ * resetting its pointer to %NULL and length and capacity to zero. Intended to
+ * free a buffer previously allocated with sshp_buf_alloc().
+ */
+static inline void sshp_buf_free(struct sshp_buf *buf)
+{
+	kfree(buf->ptr);
+	buf->ptr = NULL;
+	buf->len = 0;
+	buf->cap = 0;
+}
+
+/**
+ * sshp_buf_drop() - Drop data from the beginning of the buffer.
+ * @buf: The buffer to drop data from.
+ * @n:   The number of bytes to drop.
+ *
+ * Drops the first @n bytes from the buffer. Re-aligns any remaining data to
+ * the beginning of the buffer.
+ */
+static inline void sshp_buf_drop(struct sshp_buf *buf, size_t n)
+{
+	memmove(buf->ptr, buf->ptr + n, buf->len - n);
+	buf->len -= n;
+}
+
+/**
+ * sshp_buf_read_from_fifo() - Transfer data from a fifo to the buffer.
+ * @buf:  The buffer to write the data into.
+ * @fifo: The fifo to read the data from.
+ *
+ * Transfers the data contained in the fifo to the buffer, removing it from
+ * the fifo. This function will try to transfer as much data as possible,
+ * limited either by the remaining space in the buffer or by the number of
+ * bytes available in the fifo.
+ *
+ * Return: Returns the number of bytes transferred.
+ */
+static inline size_t sshp_buf_read_from_fifo(struct sshp_buf *buf,
+					     struct kfifo *fifo)
+{
+	size_t n;
+
+	n =  kfifo_out(fifo, buf->ptr + buf->len, buf->cap - buf->len);
+	buf->len += n;
+
+	return n;
+}
+
+/**
+ * sshp_buf_span_from() - Initialize a span from the given buffer and offset.
+ * @buf:    The buffer to create the span from.
+ * @offset: The offset in the buffer at which the span should start.
+ * @span:   The span to initialize (output).
+ *
+ * Initializes the provided span to point to the memory at the given offset in
+ * the buffer, with the length of the span being capped by the number of bytes
+ * used in the buffer after the offset (i.e. bytes remaining after the
+ * offset).
+ *
+ * Warning: This function does not validate that @offset is less than or equal
+ * to the number of bytes used in the buffer or the buffer capacity. This must
+ * be guaranteed by the caller.
+ */
+static inline void sshp_buf_span_from(struct sshp_buf *buf, size_t offset,
+				      struct ssam_span *span)
+{
+	span->ptr = buf->ptr + offset;
+	span->len = buf->len - offset;
+}
+
+
+bool sshp_find_syn(const struct ssam_span *src, struct ssam_span *rem);
+
+int sshp_parse_frame(const struct device *dev, const struct ssam_span *source,
+		     struct ssh_frame **frame, struct ssam_span *payload,
+		     size_t maxlen);
+
+int sshp_parse_command(const struct device *dev, const struct ssam_span *source,
+		       struct ssh_command **command,
+		       struct ssam_span *command_data);
+
+#endif /* _SURFACE_AGGREGATOR_SSH_PARSER_h */
diff --git a/drivers/misc/surface_aggregator/ssh_request_layer.c b/drivers/misc/surface_aggregator/ssh_request_layer.c
new file mode 100644
index 000000000000..f47bd949b6c3
--- /dev/null
+++ b/drivers/misc/surface_aggregator/ssh_request_layer.c
@@ -0,0 +1,1254 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * SSH request transport layer.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#include <asm/unaligned.h>
+#include <linux/atomic.h>
+#include <linux/completion.h>
+#include <linux/error-injection.h>
+#include <linux/ktime.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+
+#include <linux/surface_aggregator/serial_hub.h>
+#include <linux/surface_aggregator/controller.h>
+
+#include "ssh_packet_layer.h"
+#include "ssh_request_layer.h"
+
+#include "trace.h"
+
+
+/*
+ * SSH_RTL_REQUEST_TIMEOUT - Request timeout.
+ *
+ * Timeout as ktime_t delta for request responses. If we have not received a
+ * response in this time-frame after finishing the underlying packet
+ * transmission, the request will be completed with %-ETIMEDOUT as status
+ * code.
+ */
+#define SSH_RTL_REQUEST_TIMEOUT			ms_to_ktime(3000)
+
+/*
+ * SSH_RTL_REQUEST_TIMEOUT_RESOLUTION - Request timeout granularity.
+ *
+ * Time-resolution for timeouts. Should be larger than one jiffy to avoid
+ * direct re-scheduling of reaper work_struct.
+ */
+#define SSH_RTL_REQUEST_TIMEOUT_RESOLUTION	ms_to_ktime(max(2000 / HZ, 50))
+
+/*
+ * SSH_RTL_MAX_PENDING - Maximum number of pending requests.
+ *
+ * Maximum number of requests concurrently waiting to be completed (i.e.
+ * waiting for the corresponding packet transmission to finish if they don't
+ * have a response or waiting for a response if they have one).
+ */
+#define SSH_RTL_MAX_PENDING		3
+
+
+#ifdef CONFIG_SURFACE_AGGREGATOR_ERROR_INJECTION
+
+/**
+ * ssh_rtl_should_drop_response() - Error injection hook to drop request
+ * responses.
+ *
+ * Useful to cause request transmission timeouts in the driver by dropping the
+ * response to a request.
+ */
+static noinline bool ssh_rtl_should_drop_response(void)
+{
+	return false;
+}
+ALLOW_ERROR_INJECTION(ssh_rtl_should_drop_response, TRUE);
+
+#else
+
+static inline bool ssh_rtl_should_drop_response(void)
+{
+	return false;
+}
+
+#endif
+
+
+static u16 ssh_request_get_rqid(struct ssh_request *rqst)
+{
+	return get_unaligned_le16(rqst->packet.data.ptr
+				  + SSH_MSGOFFSET_COMMAND(rqid));
+}
+
+static u32 ssh_request_get_rqid_safe(struct ssh_request *rqst)
+{
+	if (!rqst->packet.data.ptr)
+		return (u32)-1;
+
+	return ssh_request_get_rqid(rqst);
+}
+
+
+static void ssh_rtl_queue_remove(struct ssh_request *rqst)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+
+	spin_lock(&rtl->queue.lock);
+
+	if (!test_and_clear_bit(SSH_REQUEST_SF_QUEUED_BIT, &rqst->state)) {
+		spin_unlock(&rtl->queue.lock);
+		return;
+	}
+
+	list_del(&rqst->node);
+
+	spin_unlock(&rtl->queue.lock);
+	ssh_request_put(rqst);
+}
+
+static bool ssh_rtl_queue_empty(struct ssh_rtl *rtl)
+{
+	bool empty;
+
+	spin_lock(&rtl->queue.lock);
+	empty = list_empty(&rtl->queue.head);
+	spin_unlock(&rtl->queue.lock);
+
+	return empty;
+}
+
+
+static void ssh_rtl_pending_remove(struct ssh_request *rqst)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+
+	spin_lock(&rtl->pending.lock);
+
+	if (!test_and_clear_bit(SSH_REQUEST_SF_PENDING_BIT, &rqst->state)) {
+		spin_unlock(&rtl->pending.lock);
+		return;
+	}
+
+	atomic_dec(&rtl->pending.count);
+	list_del(&rqst->node);
+
+	spin_unlock(&rtl->pending.lock);
+
+	ssh_request_put(rqst);
+}
+
+static int ssh_rtl_tx_pending_push(struct ssh_request *rqst)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+
+	spin_lock(&rtl->pending.lock);
+
+	if (test_bit(SSH_REQUEST_SF_LOCKED_BIT, &rqst->state)) {
+		spin_unlock(&rtl->pending.lock);
+		return -EINVAL;
+	}
+
+	if (test_and_set_bit(SSH_REQUEST_SF_PENDING_BIT, &rqst->state)) {
+		spin_unlock(&rtl->pending.lock);
+		return -EALREADY;
+	}
+
+	atomic_inc(&rtl->pending.count);
+	list_add_tail(&ssh_request_get(rqst)->node, &rtl->pending.head);
+
+	spin_unlock(&rtl->pending.lock);
+	return 0;
+}
+
+
+static void ssh_rtl_complete_with_status(struct ssh_request *rqst, int status)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+
+	trace_ssam_request_complete(rqst, status);
+
+	// rtl/ptl may not be set if we're cancelling before submitting
+	rtl_dbg_cond(rtl, "rtl: completing request (rqid: 0x%04x, status: %d)\n",
+		     ssh_request_get_rqid_safe(rqst), status);
+
+	rqst->ops->complete(rqst, NULL, NULL, status);
+}
+
+static void ssh_rtl_complete_with_rsp(struct ssh_request *rqst,
+				      const struct ssh_command *cmd,
+				      const struct ssam_span *data)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+
+	trace_ssam_request_complete(rqst, 0);
+
+	rtl_dbg(rtl, "rtl: completing request with response (rqid: 0x%04x)\n",
+		ssh_request_get_rqid(rqst));
+
+	rqst->ops->complete(rqst, cmd, data, 0);
+}
+
+
+static bool ssh_rtl_tx_can_process(struct ssh_request *rqst)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+
+	if (test_bit(SSH_REQUEST_TY_FLUSH_BIT, &rqst->state))
+		return !atomic_read(&rtl->pending.count);
+
+	return atomic_read(&rtl->pending.count) < SSH_RTL_MAX_PENDING;
+}
+
+static struct ssh_request *ssh_rtl_tx_next(struct ssh_rtl *rtl)
+{
+	struct ssh_request *rqst = ERR_PTR(-ENOENT);
+	struct ssh_request *p, *n;
+
+	spin_lock(&rtl->queue.lock);
+
+	// find first non-locked request and remove it
+	list_for_each_entry_safe(p, n, &rtl->queue.head, node) {
+		if (unlikely(test_bit(SSH_REQUEST_SF_LOCKED_BIT, &p->state)))
+			continue;
+
+		if (!ssh_rtl_tx_can_process(p)) {
+			rqst = ERR_PTR(-EBUSY);
+			break;
+		}
+
+		// remove from queue and mark as transmitting
+		set_bit(SSH_REQUEST_SF_TRANSMITTING_BIT, &p->state);
+		// ensure state never gets zero
+		smp_mb__before_atomic();
+		clear_bit(SSH_REQUEST_SF_QUEUED_BIT, &p->state);
+
+		list_del(&p->node);
+
+		rqst = p;
+		break;
+	}
+
+	spin_unlock(&rtl->queue.lock);
+	return rqst;
+}
+
+static int ssh_rtl_tx_try_process_one(struct ssh_rtl *rtl)
+{
+	struct ssh_request *rqst;
+	int status;
+
+	// get and prepare next request for transmit
+	rqst = ssh_rtl_tx_next(rtl);
+	if (IS_ERR(rqst))
+		return PTR_ERR(rqst);
+
+	// add to/mark as pending
+	status = ssh_rtl_tx_pending_push(rqst);
+	if (status) {
+		ssh_request_put(rqst);
+		return -EAGAIN;
+	}
+
+	// submit packet
+	status = ssh_ptl_submit(&rtl->ptl, &rqst->packet);
+	if (status == -ESHUTDOWN) {
+		/*
+		 * Packet has been refused due to the packet layer shutting
+		 * down. Complete it here.
+		 */
+		set_bit(SSH_REQUEST_SF_LOCKED_BIT, &rqst->state);
+		/*
+		 * Note: A barrier is not required here, as there are only two
+		 * references in the system at this point: The one that we have,
+		 * and the other one that belongs to the pending set. Due to the
+		 * request being marked as "transmitting", our process is the
+		 * only one allowed to remove the pending node and change the
+		 * state. Normally, the task would fall to the packet callback,
+		 * but as this is a path where submission failed, this callback
+		 * will never be executed.
+		 */
+
+		ssh_rtl_pending_remove(rqst);
+		ssh_rtl_complete_with_status(rqst, -ESHUTDOWN);
+
+		ssh_request_put(rqst);
+		return -ESHUTDOWN;
+
+	} else if (status) {
+		/*
+		 * If submitting the packet failed and the packet layer isn't
+		 * shutting down, the packet has either been submmitted/queued
+		 * before (-EALREADY, which cannot happen as we have guaranteed
+		 * that requests cannot be re-submitted), or the packet was
+		 * marked as locked (-EINVAL). To mark the packet locked at this
+		 * stage, the request, and thus the packets itself, had to have
+		 * been canceled. Simply drop the reference. Cancellation itself
+		 * will remove it from the set of pending requests.
+		 */
+
+		WARN_ON(status != -EINVAL);
+
+		ssh_request_put(rqst);
+		return -EAGAIN;
+	}
+
+	ssh_request_put(rqst);
+	return 0;
+}
+
+static bool ssh_rtl_tx_schedule(struct ssh_rtl *rtl)
+{
+	if (atomic_read(&rtl->pending.count) >= SSH_RTL_MAX_PENDING)
+		return false;
+
+	if (ssh_rtl_queue_empty(rtl))
+		return false;
+
+	return schedule_work(&rtl->tx.work);
+}
+
+static void ssh_rtl_tx_work_fn(struct work_struct *work)
+{
+	struct ssh_rtl *rtl = to_ssh_rtl(work, tx.work);
+	int i, status;
+
+	/*
+	 * Try to be nice and not block/live-lock the workqueue: Run a maximum
+	 * of 10 tries, then re-submit if necessary. This should not be
+	 * necessary for normal execution, but guarantee it anyway.
+	 */
+	for (i = 0; i < 10; i++) {
+		status = ssh_rtl_tx_try_process_one(rtl);
+		if (status == -ENOENT || status == -EBUSY)
+			return;		// no more requests to process
+
+		if (status == -ESHUTDOWN) {
+			/*
+			 * Packet system shutting down. No new packets can be
+			 * transmitted. Return silently, the party initiating
+			 * the shutdown should handle the rest.
+			 */
+			return;
+		}
+
+		WARN_ON(status != 0 && status != -EAGAIN);
+	}
+
+	// out of tries, reschedule
+	ssh_rtl_tx_schedule(rtl);
+}
+
+
+/**
+ * ssh_rtl_submit() - Submit a request to the transport layer.
+ * @rtl:  The request transport layer.
+ * @rqst: The request to submit.
+ *
+ * Submits a request to the transport layer. A single request may not be
+ * submitted multiple times without reinitializing it.
+ *
+ * Return: Returns zero on success, %-EINVAL if the request type is invalid or
+ * the request has been canceled prior to submission, %-EALREADY if the
+ * request has already been submitted, or %-ESHUTDOWN in case the request
+ * transport layer has been shut down.
+ */
+int ssh_rtl_submit(struct ssh_rtl *rtl, struct ssh_request *rqst)
+{
+	trace_ssam_request_submit(rqst);
+
+	/*
+	 * Ensure that requests expecting a response are sequenced. If this
+	 * invariant ever changes, see the comment in ssh_rtl_complete() on what
+	 * is required to be changed in the code.
+	 */
+	if (test_bit(SSH_REQUEST_TY_HAS_RESPONSE_BIT, &rqst->state))
+		if (!test_bit(SSH_PACKET_TY_SEQUENCED_BIT, &rqst->packet.state))
+			return -EINVAL;
+
+	spin_lock(&rtl->queue.lock);
+
+	/*
+	 * Try to set ptl and check if this request has already been submitted.
+	 *
+	 * Must be inside lock as we might run into a lost update problem
+	 * otherwise: If this were outside of the lock, cancellation in
+	 * ssh_rtl_cancel_nonpending() may run after we've set the ptl
+	 * reference but before we enter the lock. In that case, we'd detect
+	 * that the request is being added to the queue and would try to remove
+	 * it from that, but removal might fail because it hasn't actually been
+	 * added yet. By putting this cmpxchg in the critical section, we
+	 * ensure that the queuing detection only triggers when we are already
+	 * in the critical section and the remove process will wait until the
+	 * push operation has been completed (via lock) due to that. Only then,
+	 * we can safely try to remove it.
+	 */
+	if (cmpxchg(&rqst->packet.ptl, NULL, &rtl->ptl) != NULL) {
+		spin_unlock(&rtl->queue.lock);
+		return -EALREADY;
+	}
+
+	/*
+	 * Ensure that we set ptl reference before we continue modifying state.
+	 * This is required for non-pending cancellation. This barrier is paired
+	 * with the one in ssh_rtl_cancel_nonpending().
+	 *
+	 * By setting the ptl reference before we test for "locked", we can
+	 * check if the "locked" test may have already run. See comments in
+	 * ssh_rtl_cancel_nonpending() for more detail.
+	 */
+	smp_mb__after_atomic();
+
+	if (test_bit(SSH_RTL_SF_SHUTDOWN_BIT, &rtl->state)) {
+		spin_unlock(&rtl->queue.lock);
+		return -ESHUTDOWN;
+	}
+
+	if (test_bit(SSH_REQUEST_SF_LOCKED_BIT, &rqst->state)) {
+		spin_unlock(&rtl->queue.lock);
+		return -EINVAL;
+	}
+
+	set_bit(SSH_REQUEST_SF_QUEUED_BIT, &rqst->state);
+	list_add_tail(&ssh_request_get(rqst)->node, &rtl->queue.head);
+
+	spin_unlock(&rtl->queue.lock);
+
+	ssh_rtl_tx_schedule(rtl);
+	return 0;
+}
+
+static void ssh_rtl_timeout_reaper_mod(struct ssh_rtl *rtl, ktime_t now,
+				       ktime_t expires)
+{
+	unsigned long delta = msecs_to_jiffies(ktime_ms_delta(expires, now));
+	ktime_t aexp = ktime_add(expires, SSH_RTL_REQUEST_TIMEOUT_RESOLUTION);
+	ktime_t old_exp, old_act;
+
+	// re-adjust / schedule reaper if it is above resolution delta
+	old_act = READ_ONCE(rtl->rtx_timeout.expires);
+	if (ktime_after(aexp, old_act))
+		return;
+
+	do {
+		old_exp = old_act;
+		old_act = cmpxchg64(&rtl->rtx_timeout.expires, old_exp, expires);
+	} while (old_exp != old_act && ktime_before(aexp, old_act));
+
+	// if we updated the reaper expiration, modify work timeout
+	if (old_exp == old_act && old_act != expires)
+		mod_delayed_work(system_wq, &rtl->rtx_timeout.reaper, delta);
+}
+
+static void ssh_rtl_timeout_start(struct ssh_request *rqst)
+{
+	struct ssh_rtl *rtl = ssh_request_rtl(rqst);
+	ktime_t timestamp = ktime_get_coarse_boottime();
+	ktime_t timeout = rtl->rtx_timeout.timeout;
+
+	if (test_bit(SSH_REQUEST_SF_LOCKED_BIT, &rqst->state))
+		return;
+
+	WRITE_ONCE(rqst->timestamp, timestamp);
+	/*
+	 * Ensure timestamp is set before starting the reaper. Paired with
+	 * implicit barrier following check on ssh_request_get_expiration() in
+	 * ssh_rtl_timeout_reap.
+	 */
+	smp_mb__after_atomic();
+
+	ssh_rtl_timeout_reaper_mod(rtl, timestamp, timestamp + timeout);
+}
+
+
+static void ssh_rtl_complete(struct ssh_rtl *rtl,
+			     const struct ssh_command *command,
+			     const struct ssam_span *command_data)
+{
+	struct ssh_request *r = NULL;
+	struct ssh_request *p, *n;
+	u16 rqid = get_unaligned_le16(&command->rqid);
+
+	trace_ssam_rx_response_received(command, command_data->len);
+
+	/*
+	 * Get request from pending based on request ID and mark it as response
+	 * received and locked.
+	 */
+	spin_lock(&rtl->pending.lock);
+	list_for_each_entry_safe(p, n, &rtl->pending.head, node) {
+		// we generally expect requests to be processed in order
+		if (unlikely(ssh_request_get_rqid(p) != rqid))
+			continue;
+
+		// simulate response timeout
+		if (ssh_rtl_should_drop_response()) {
+			spin_unlock(&rtl->pending.lock);
+
+			trace_ssam_ei_rx_drop_response(p);
+			rtl_info(rtl, "request error injection: dropping response for request %p\n",
+				 &p->packet);
+			return;
+		}
+
+		/*
+		 * Mark as "response received" and "locked" as we're going to
+		 * complete it.
+		 */
+		set_bit(SSH_REQUEST_SF_LOCKED_BIT, &p->state);
+		set_bit(SSH_REQUEST_SF_RSPRCVD_BIT, &p->state);
+		// ensure state never gets zero
+		smp_mb__before_atomic();
+		clear_bit(SSH_REQUEST_SF_PENDING_BIT, &p->state);
+
+		atomic_dec(&rtl->pending.count);
+		list_del(&p->node);
+
+		r = p;
+		break;
+	}
+	spin_unlock(&rtl->pending.lock);
+
+	if (!r) {
+		rtl_warn(rtl, "rtl: dropping unexpected command message (rqid = 0x%04x)\n",
+			 rqid);
+		return;
+	}
+
+	// if the request hasn't been completed yet, we will do this now
+	if (test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state)) {
+		ssh_request_put(r);
+		ssh_rtl_tx_schedule(rtl);
+		return;
+	}
+
+	/*
+	 * Make sure the request has been transmitted. In case of a sequenced
+	 * request, we are guaranteed that the completion callback will run on
+	 * the receiver thread directly when the ACK for the packet has been
+	 * received. Similarly, this function is guaranteed to run on the
+	 * receiver thread. Thus we are guaranteed that if the packet has been
+	 * successfully transmitted and received an ACK, the transmitted flag
+	 * has been set and is visible here.
+	 *
+	 * We are currently not handling unsequenced packets here, as those
+	 * should never expect a response as ensured in ssh_rtl_submit. If this
+	 * ever changes, one would have to test for
+	 *
+	 *	(r->state & (transmitting | transmitted))
+	 *
+	 * on unsequenced packets to determine if they could have been
+	 * transmitted. There are no synchronization guarantees as in the
+	 * sequenced case, since, in this case, the callback function will not
+	 * run on the same thread. Thus an exact determination is impossible.
+	 */
+	if (!test_bit(SSH_REQUEST_SF_TRANSMITTED_BIT, &r->state)) {
+		rtl_err(rtl, "rtl: received response before ACK for request (rqid = 0x%04x)\n",
+			rqid);
+
+		/*
+		 * NB: Timeout has already been canceled, request already been
+		 * removed from pending and marked as locked and completed. As
+		 * we receive a "false" response, the packet might still be
+		 * queued though.
+		 */
+		ssh_rtl_queue_remove(r);
+
+		ssh_rtl_complete_with_status(r, -EREMOTEIO);
+		ssh_request_put(r);
+
+		ssh_rtl_tx_schedule(rtl);
+		return;
+	}
+
+	/*
+	 * NB: Timeout has already been canceled, request already been
+	 * removed from pending and marked as locked and completed. The request
+	 * can also not be queued any more, as it has been marked as
+	 * transmitting and later transmitted. Thus no need to remove it from
+	 * anywhere.
+	 */
+
+	ssh_rtl_complete_with_rsp(r, command, command_data);
+	ssh_request_put(r);
+
+	ssh_rtl_tx_schedule(rtl);
+}
+
+
+static bool ssh_rtl_cancel_nonpending(struct ssh_request *r)
+{
+	struct ssh_rtl *rtl;
+	unsigned long flags, fixed;
+	bool remove;
+
+	/*
+	 * Handle unsubmitted request: Try to mark the packet as locked,
+	 * expecting the state to be zero (i.e. unsubmitted). Note that, if
+	 * setting the state worked, we might still be adding the packet to the
+	 * queue in a currently executing submit call. In that case, however,
+	 * ptl reference must have been set previously, as locked is checked
+	 * after setting ptl. Furthermore, when the ptl reference is set, the
+	 * submission process is guaranteed to have entered the critical
+	 * section. Thus only if we successfully locked this request and ptl is
+	 * NULL, we have successfully removed the request, i.e. we are
+	 * guaranteed that, due to the "locked" check in ssh_rtl_submit(), the
+	 * packet will never be added. Otherwise, we need to try and grab it
+	 * from the queue, where we are now guaranteed that the packet is or has
+	 * been due to the critical section.
+	 *
+	 * Note that if the CMPXCHG fails, we are guaranteed that ptl has
+	 * been set and is non-NULL, as states can only be nonzero after this
+	 * has been set. Also note that we need to fetch the static (type) flags
+	 * to ensure that they don't cause the cmpxchg to fail.
+	 */
+	fixed = READ_ONCE(r->state) & SSH_REQUEST_FLAGS_TY_MASK;
+	flags = cmpxchg(&r->state, fixed, SSH_REQUEST_SF_LOCKED_BIT);
+
+	/*
+	 * Force correct ordering with regards to state and ptl reference access
+	 * to safe-guard cancellation to concurrent submission against a
+	 * lost-update problem. First try to exchange state, then also check
+	 * ptl if that worked. This barrier is paired with the
+	 * one in ssh_rtl_submit().
+	 */
+	smp_mb__after_atomic();
+
+	if (flags == fixed && !READ_ONCE(r->packet.ptl)) {
+		if (test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+			return true;
+
+		ssh_rtl_complete_with_status(r, -ECANCELED);
+		return true;
+	}
+
+	rtl = ssh_request_rtl(r);
+	spin_lock(&rtl->queue.lock);
+
+	/*
+	 * Note: 1) Requests cannot be re-submitted. 2) If a request is queued,
+	 * it cannot be "transmitting"/"pending" yet. Thus, if we successfully
+	 * remove the request here, we have removed all its occurences in the
+	 * system.
+	 */
+
+	remove = test_and_clear_bit(SSH_REQUEST_SF_QUEUED_BIT, &r->state);
+	if (!remove) {
+		spin_unlock(&rtl->queue.lock);
+		return false;
+	}
+
+	set_bit(SSH_REQUEST_SF_LOCKED_BIT, &r->state);
+	list_del(&r->node);
+
+	spin_unlock(&rtl->queue.lock);
+
+	ssh_request_put(r);	// drop reference obtained from queue
+
+	if (test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+		return true;
+
+	ssh_rtl_complete_with_status(r, -ECANCELED);
+	return true;
+}
+
+static bool ssh_rtl_cancel_pending(struct ssh_request *r)
+{
+	// if the packet is already locked, it's going to be removed shortly
+	if (test_and_set_bit(SSH_REQUEST_SF_LOCKED_BIT, &r->state))
+		return true;
+
+	/*
+	 * Now that we have locked the packet, we have guaranteed that it can't
+	 * be added to the system any more. If ptl is zero, the locked
+	 * check in ssh_rtl_submit() has not been run and any submission,
+	 * currently in progress or called later, won't add the packet. Thus we
+	 * can directly complete it.
+	 *
+	 * The implicit memory barrier of test_and_set_bit() should be enough
+	 * to ensure that the correct order (first lock, then check ptl) is
+	 * ensured. This is paired with the barrier in ssh_rtl_submit().
+	 */
+	if (!READ_ONCE(r->packet.ptl)) {
+		if (test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+			return true;
+
+		ssh_rtl_complete_with_status(r, -ECANCELED);
+		return true;
+	}
+
+	/*
+	 * Try to cancel the packet. If the packet has not been completed yet,
+	 * this will subsequently (and synchronously) call the completion
+	 * callback of the packet, which will complete the request.
+	 */
+	ssh_ptl_cancel(&r->packet);
+
+	/*
+	 * If the packet has been completed with success, i.e. has not been
+	 * canceled by the above call, the request may not have been completed
+	 * yet (may be waiting for a response). Check if we need to do this
+	 * here.
+	 */
+	if (test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+		return true;
+
+	ssh_rtl_queue_remove(r);
+	ssh_rtl_pending_remove(r);
+	ssh_rtl_complete_with_status(r, -ECANCELED);
+
+	return true;
+}
+
+/**
+ * ssh_rtl_cancel() - Cancel request.
+ * @rqst:    The request to cancel.
+ * @pending: Whether to also cancel pending requests.
+ *
+ * Cancels the given request. If @pending is %false, this will not cancel
+ * pending requests, i.e. requests that have already been submitted to the
+ * packet layer but not been completed yet. If @pending is %true, this will
+ * cancel the given request regardless of the state it is in.
+ *
+ * If the request has been canceled by calling this function, both completion
+ * and release callbacks of the request will be executed in a reasonable
+ * time-frame. This may happen during execution of this function, however,
+ * there is no guarantee for this. For example, a request currently
+ * transmitting will be canceled/completed only after transmission has
+ * completed, and the respective callbacks will be executed on the transmitter
+ * thread, which may happen during, but also some time after execution of the
+ * cancel function.
+ *
+ * Return: Returns %true iff the given request has been canceled or completed,
+ * either by this function or prior to calling this function, %false
+ * otherwise. If @pending is %true, this function will always return %true.
+ */
+bool ssh_rtl_cancel(struct ssh_request *rqst, bool pending)
+{
+	struct ssh_rtl *rtl;
+	bool canceled;
+
+	if (test_and_set_bit(SSH_REQUEST_SF_CANCELED_BIT, &rqst->state))
+		return true;
+
+	trace_ssam_request_cancel(rqst);
+
+	if (pending)
+		canceled = ssh_rtl_cancel_pending(rqst);
+	else
+		canceled = ssh_rtl_cancel_nonpending(rqst);
+
+	// note: rtl may be NULL if request has not been submitted yet
+	rtl = ssh_request_rtl(rqst);
+	if (canceled && rtl)
+		ssh_rtl_tx_schedule(rtl);
+
+	return canceled;
+}
+
+
+static void ssh_rtl_packet_callback(struct ssh_packet *p, int status)
+{
+	struct ssh_request *r = to_ssh_request(p);
+
+	if (unlikely(status)) {
+		set_bit(SSH_REQUEST_SF_LOCKED_BIT, &r->state);
+
+		if (test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+			return;
+
+		/*
+		 * The packet may get cancelled even though it has not been
+		 * submitted yet. The request may still be queued. Check the
+		 * queue and remove it if necessary. As the timeout would have
+		 * been started in this function on success, there's no need to
+		 * cancel it here.
+		 */
+		ssh_rtl_queue_remove(r);
+		ssh_rtl_pending_remove(r);
+		ssh_rtl_complete_with_status(r, status);
+
+		ssh_rtl_tx_schedule(ssh_request_rtl(r));
+		return;
+	}
+
+	// update state: mark as transmitted and clear transmitting
+	set_bit(SSH_REQUEST_SF_TRANSMITTED_BIT, &r->state);
+	// ensure state never gets zero
+	smp_mb__before_atomic();
+	clear_bit(SSH_REQUEST_SF_TRANSMITTING_BIT, &r->state);
+
+	// if we expect a response, we just need to start the timeout
+	if (test_bit(SSH_REQUEST_TY_HAS_RESPONSE_BIT, &r->state)) {
+		ssh_rtl_timeout_start(r);
+		return;
+	}
+
+	/*
+	 * If we don't expect a response, lock, remove, and complete the
+	 * request. Note that, at this point, the request is guaranteed to have
+	 * left the queue and no timeout has been started. Thus we only need to
+	 * remove it from pending. If the request has already been completed (it
+	 * may have been canceled) return.
+	 */
+
+	set_bit(SSH_REQUEST_SF_LOCKED_BIT, &r->state);
+	if (test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+		return;
+
+	ssh_rtl_pending_remove(r);
+	ssh_rtl_complete_with_status(r, 0);
+
+	ssh_rtl_tx_schedule(ssh_request_rtl(r));
+}
+
+
+static ktime_t ssh_request_get_expiration(struct ssh_request *r, ktime_t timeout)
+{
+	ktime_t timestamp = READ_ONCE(r->timestamp);
+
+	if (timestamp != KTIME_MAX)
+		return ktime_add(timestamp, timeout);
+	else
+		return KTIME_MAX;
+}
+
+static void ssh_rtl_timeout_reap(struct work_struct *work)
+{
+	struct ssh_rtl *rtl = to_ssh_rtl(work, rtx_timeout.reaper.work);
+	struct ssh_request *r, *n;
+	LIST_HEAD(claimed);
+	ktime_t now = ktime_get_coarse_boottime();
+	ktime_t timeout = rtl->rtx_timeout.timeout;
+	ktime_t next = KTIME_MAX;
+
+	trace_ssam_rtl_timeout_reap("pending", atomic_read(&rtl->pending.count));
+
+	/*
+	 * Mark reaper as "not pending". This is done before checking any
+	 * requests to avoid lost-update type problems.
+	 */
+	WRITE_ONCE(rtl->rtx_timeout.expires, KTIME_MAX);
+	/*
+	 * Ensure that the reaper is marked as deactivated before we continue
+	 * checking requests to prevent lost-update problems when a request is
+	 * added to the pending set and ssh_rtl_timeout_reaper_mod is called
+	 * during execution of the part below.
+	 */
+	smp_mb__after_atomic();
+
+	spin_lock(&rtl->pending.lock);
+	list_for_each_entry_safe(r, n, &rtl->pending.head, node) {
+		ktime_t expires = ssh_request_get_expiration(r, timeout);
+
+		/*
+		 * Check if the timeout hasn't expired yet. Find out next
+		 * expiration date to be handled after this run.
+		 */
+		if (ktime_after(expires, now)) {
+			next = ktime_before(expires, next) ? expires : next;
+			continue;
+		}
+
+		// avoid further transitions if locked
+		if (test_and_set_bit(SSH_REQUEST_SF_LOCKED_BIT, &r->state))
+			continue;
+
+		/*
+		 * We have now marked the packet as locked. Thus it cannot be
+		 * added to the pending or queued lists again after we've
+		 * removed it here. We can therefore re-use the node of this
+		 * packet temporarily.
+		 */
+
+		clear_bit(SSH_REQUEST_SF_PENDING_BIT, &r->state);
+
+		atomic_dec(&rtl->pending.count);
+		list_del(&r->node);
+
+		list_add_tail(&r->node, &claimed);
+	}
+	spin_unlock(&rtl->pending.lock);
+
+	// cancel and complete the request
+	list_for_each_entry_safe(r, n, &claimed, node) {
+		trace_ssam_request_timeout(r);
+
+		/*
+		 * At this point we've removed the packet from pending. This
+		 * means that we've obtained the last (only) reference of the
+		 * system to it. Thus we can just complete it.
+		 */
+		if (!test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+			ssh_rtl_complete_with_status(r, -ETIMEDOUT);
+
+		// drop the reference we've obtained by removing it from pending
+		list_del(&r->node);
+		ssh_request_put(r);
+	}
+
+	// ensure that reaper doesn't run again immediately
+	next = max(next, ktime_add(now, SSH_RTL_REQUEST_TIMEOUT_RESOLUTION));
+	if (next != KTIME_MAX)
+		ssh_rtl_timeout_reaper_mod(rtl, now, next);
+
+	ssh_rtl_tx_schedule(rtl);
+}
+
+
+static void ssh_rtl_rx_event(struct ssh_rtl *rtl, const struct ssh_command *cmd,
+			     const struct ssam_span *data)
+{
+	trace_ssam_rx_event_received(cmd, data->len);
+
+	rtl_dbg(rtl, "rtl: handling event (rqid: 0x%04x)\n",
+		get_unaligned_le16(&cmd->rqid));
+
+	rtl->ops.handle_event(rtl, cmd, data);
+}
+
+static void ssh_rtl_rx_command(struct ssh_ptl *p, const struct ssam_span *data)
+{
+	struct ssh_rtl *rtl = to_ssh_rtl(p, ptl);
+	struct device *dev = &p->serdev->dev;
+	struct ssh_command *command;
+	struct ssam_span command_data;
+
+	if (sshp_parse_command(dev, data, &command, &command_data))
+		return;
+
+	if (ssh_rqid_is_event(get_unaligned_le16(&command->rqid)))
+		ssh_rtl_rx_event(rtl, command, &command_data);
+	else
+		ssh_rtl_complete(rtl, command, &command_data);
+}
+
+static void ssh_rtl_rx_data(struct ssh_ptl *p, const struct ssam_span *data)
+{
+	if (!data->len) {
+		ptl_err(p, "rtl: rx: no data frame payload\n");
+		return;
+	}
+
+	switch (data->ptr[0]) {
+	case SSH_PLD_TYPE_CMD:
+		ssh_rtl_rx_command(p, data);
+		break;
+
+	default:
+		ptl_err(p, "rtl: rx: unknown frame payload type (type: 0x%02x)\n",
+			data->ptr[0]);
+		break;
+	}
+}
+
+
+static void ssh_rtl_packet_release(struct ssh_packet *p)
+{
+	struct ssh_request *rqst;
+
+	rqst = to_ssh_request(p);
+	rqst->ops->release(rqst);
+}
+
+static const struct ssh_packet_ops ssh_rtl_packet_ops = {
+	.complete = ssh_rtl_packet_callback,
+	.release = ssh_rtl_packet_release,
+};
+
+/**
+ * ssh_request_init() - Initialize SSH request.
+ * @rqst:  The request to initialize.
+ * @flags: Request flags, determining the type of the request.
+ * @ops:   Request operations.
+ *
+ * Initializes the given SSH request and underlying packet. Sets the message
+ * buffer pointer to %NULL and the message buffer length to zero. This buffer
+ * has to be set separately via ssh_request_set_data() before submission and
+ * must contain a valid SSH request message.
+ */
+void ssh_request_init(struct ssh_request *rqst, enum ssam_request_flags flags,
+		      const struct ssh_request_ops *ops)
+{
+	unsigned long type = BIT(SSH_PACKET_TY_BLOCKING_BIT);
+
+	if (!(flags & SSAM_REQUEST_UNSEQUENCED))
+		type |= BIT(SSH_PACKET_TY_SEQUENCED_BIT);
+
+	ssh_packet_init(&rqst->packet, type, SSH_PACKET_PRIORITY(DATA, 0),
+			&ssh_rtl_packet_ops);
+
+	INIT_LIST_HEAD(&rqst->node);
+
+	rqst->state = 0;
+	if (flags & SSAM_REQUEST_HAS_RESPONSE)
+		rqst->state |= BIT(SSH_REQUEST_TY_HAS_RESPONSE_BIT);
+
+	rqst->timestamp = KTIME_MAX;
+	rqst->ops = ops;
+}
+
+
+/**
+ * ssh_rtl_init() - Initialize request transport layer.
+ * @rtl:    The request transport layer to initialize.
+ * @serdev: The underlying serial device, i.e. the lower-level transport.
+ * @ops:    Request transport layer operations.
+ *
+ * Initializes the given request transport layer and associated packet
+ * transport layer. Transmitter and receiver threads must be started
+ * separately via ssh_rtl_tx_start() and ssh_rtl_rx_start(), after the
+ * request-layer has been initialized and the lower-level serial device layer
+ * has been set up.
+ *
+ * Return: Returns zero on success and a nonzero error code on failure.
+ */
+int ssh_rtl_init(struct ssh_rtl *rtl, struct serdev_device *serdev,
+		 const struct ssh_rtl_ops *ops)
+{
+	struct ssh_ptl_ops ptl_ops;
+	int status;
+
+	ptl_ops.data_received = ssh_rtl_rx_data;
+
+	status = ssh_ptl_init(&rtl->ptl, serdev, &ptl_ops);
+	if (status)
+		return status;
+
+	spin_lock_init(&rtl->queue.lock);
+	INIT_LIST_HEAD(&rtl->queue.head);
+
+	spin_lock_init(&rtl->pending.lock);
+	INIT_LIST_HEAD(&rtl->pending.head);
+	atomic_set_release(&rtl->pending.count, 0);
+
+	INIT_WORK(&rtl->tx.work, ssh_rtl_tx_work_fn);
+
+	rtl->rtx_timeout.timeout = SSH_RTL_REQUEST_TIMEOUT;
+	rtl->rtx_timeout.expires = KTIME_MAX;
+	INIT_DELAYED_WORK(&rtl->rtx_timeout.reaper, ssh_rtl_timeout_reap);
+
+	rtl->ops = *ops;
+
+	return 0;
+}
+
+/**
+ * ssh_rtl_destroy() - Deinitialize request transport layer.
+ * @rtl: The request transport layer to deinitialize.
+ *
+ * Deinitializes the given request transport layer and frees resources
+ * associated with it. If receiver and/or transmitter threads have been
+ * started, the layer must first be shut down via ssh_rtl_shutdown() before
+ * this function can be called.
+ */
+void ssh_rtl_destroy(struct ssh_rtl *rtl)
+{
+	ssh_ptl_destroy(&rtl->ptl);
+}
+
+/**
+ * ssh_rtl_tx_start() - Start request transmitter and receiver.
+ * @rtl: The request transport layer.
+ *
+ * Return: Returns zero on success, a negative error code on failure.
+ */
+int ssh_rtl_start(struct ssh_rtl *rtl)
+{
+	int status;
+
+	status = ssh_ptl_tx_start(&rtl->ptl);
+	if (status)
+		return status;
+
+	ssh_rtl_tx_schedule(rtl);
+
+	status = ssh_ptl_rx_start(&rtl->ptl);
+	if (status) {
+		ssh_rtl_flush(rtl, msecs_to_jiffies(5000));
+		ssh_ptl_tx_stop(&rtl->ptl);
+		return status;
+	}
+
+	return 0;
+}
+
+struct ssh_flush_request {
+	struct ssh_request base;
+	struct completion completion;
+	int status;
+};
+
+static void ssh_rtl_flush_request_complete(struct ssh_request *r,
+					   const struct ssh_command *cmd,
+					   const struct ssam_span *data,
+					   int status)
+{
+	struct ssh_flush_request *rqst;
+
+	rqst = container_of(r, struct ssh_flush_request, base);
+	rqst->status = status;
+}
+
+static void ssh_rtl_flush_request_release(struct ssh_request *r)
+{
+	struct ssh_flush_request *rqst;
+
+	rqst = container_of(r, struct ssh_flush_request, base);
+	complete_all(&rqst->completion);
+}
+
+static const struct ssh_request_ops ssh_rtl_flush_request_ops = {
+	.complete = ssh_rtl_flush_request_complete,
+	.release = ssh_rtl_flush_request_release,
+};
+
+/**
+ * ssh_rtl_flush() - Flush the request transport layer.
+ * @rtl:     request transport layer
+ * @timeout: timeout for the flush operation in jiffies
+ *
+ * Queue a special flush request and wait for its completion. This request
+ * will be completed after all other currently queued and pending requests
+ * have been completed. Instead of a normal data packet, this request submits
+ * a special flush packet, meaning that upon completion, also the underlying
+ * packet transport layer has been flushed.
+ *
+ * Flushing the request layer gurarantees that all previously submitted
+ * requests have been fully completed before this call returns. Additinally,
+ * flushing blocks execution of all later submitted requests until the flush
+ * has been completed.
+ *
+ * If the caller ensures that no new requests are submitted after a call to
+ * this function, the request transport layer is guaranteed to have no
+ * remaining requests when this call returns. The same guarantee does not hold
+ * for the packet layer, on which control packets may still be queued after
+ * this call.
+ *
+ * Return: Returns zero on success, %-ETIMEDOUT if the flush timed out and has
+ * been canceled as a result of the timeout, or %-ESHUTDOWN if the packet
+ * and/or request transport layer has been shut down before this call. May
+ * also return %-EINTR if the underlying packet transmission has been
+ * interrupted.
+ */
+int ssh_rtl_flush(struct ssh_rtl *rtl, unsigned long timeout)
+{
+	const unsigned int init_flags = SSAM_REQUEST_UNSEQUENCED;
+	struct ssh_flush_request rqst;
+	int status;
+
+	ssh_request_init(&rqst.base, init_flags, &ssh_rtl_flush_request_ops);
+	rqst.base.packet.state |= BIT(SSH_PACKET_TY_FLUSH_BIT);
+	rqst.base.packet.priority = SSH_PACKET_PRIORITY(FLUSH, 0);
+	rqst.base.state |= BIT(SSH_REQUEST_TY_FLUSH_BIT);
+
+	init_completion(&rqst.completion);
+
+	status = ssh_rtl_submit(rtl, &rqst.base);
+	if (status)
+		return status;
+
+	ssh_request_put(&rqst.base);
+
+	if (!wait_for_completion_timeout(&rqst.completion, timeout)) {
+		ssh_rtl_cancel(&rqst.base, true);
+		wait_for_completion(&rqst.completion);
+	}
+
+	WARN_ON(rqst.status != 0 && rqst.status != -ECANCELED
+		&& rqst.status != -ESHUTDOWN && rqst.status != -EINTR);
+
+	return rqst.status == -ECANCELED ? -ETIMEDOUT : rqst.status;
+}
+
+
+/**
+ * ssh_rtl_shutdown() - Shut down request transport layer.
+ * @rtl: The request transport layer.
+ *
+ * Shuts down the request transport layer, removing and canceling all queued
+ * and pending requests. Requests canceled by this operation will be completed
+ * with %-ESHUTDOWN as status. Receiver and transmitter threads will be
+ * stopped, the lower-level packet layer will be shutdown.
+ *
+ * As a result of this function, the transport layer will be marked as shut
+ * down. Submission of requests after the transport layer has been shut down
+ * will fail with %-ESHUTDOWN.
+ */
+void ssh_rtl_shutdown(struct ssh_rtl *rtl)
+{
+	struct ssh_request *r, *n;
+	LIST_HEAD(claimed);
+	int pending;
+
+	set_bit(SSH_RTL_SF_SHUTDOWN_BIT, &rtl->state);
+	/*
+	 * Ensure that the layer gets marked as shut-down before actually
+	 * stopping it. In combination with the check in ssh_rtl_sunmit(), this
+	 * guarantees that no new requests can be added and all already queued
+	 * requests are properly cancelled.
+	 */
+	smp_mb__after_atomic();
+
+	// remove requests from queue
+	spin_lock(&rtl->queue.lock);
+	list_for_each_entry_safe(r, n, &rtl->queue.head, node) {
+		set_bit(SSH_REQUEST_SF_LOCKED_BIT, &r->state);
+		// ensure state never gets zero
+		smp_mb__before_atomic();
+		clear_bit(SSH_REQUEST_SF_QUEUED_BIT, &r->state);
+
+		list_del(&r->node);
+		list_add_tail(&r->node, &claimed);
+	}
+	spin_unlock(&rtl->queue.lock);
+
+	/*
+	 * We have now guaranteed that the queue is empty and no more new
+	 * requests can be submitted (i.e. it will stay empty). This means that
+	 * calling ssh_rtl_tx_schedule() will not schedule tx.work any more. So
+	 * we can simply call cancel_work_sync() on tx.work here and when that
+	 * returns, we've locked it down. This also means that after this call,
+	 * we don't submit any more packets to the underlying packet layer, so
+	 * we can also shut that down.
+	 */
+
+	cancel_work_sync(&rtl->tx.work);
+	ssh_ptl_shutdown(&rtl->ptl);
+	cancel_delayed_work_sync(&rtl->rtx_timeout.reaper);
+
+	/*
+	 * Shutting down the packet layer should also have caneled all requests.
+	 * Thus the pending set should be empty. Attempt to handle this
+	 * gracefully anyways, even though this should be dead code.
+	 */
+
+	pending = atomic_read(&rtl->pending.count);
+	if (WARN_ON(pending)) {
+		spin_lock(&rtl->pending.lock);
+		list_for_each_entry_safe(r, n, &rtl->pending.head, node) {
+			set_bit(SSH_REQUEST_SF_LOCKED_BIT, &r->state);
+			// ensure state never gets zero
+			smp_mb__before_atomic();
+			clear_bit(SSH_REQUEST_SF_PENDING_BIT, &r->state);
+
+			list_del(&r->node);
+			list_add_tail(&r->node, &claimed);
+		}
+		spin_unlock(&rtl->pending.lock);
+	}
+
+	// finally, cancel and complete the requests we claimed before
+	list_for_each_entry_safe(r, n, &claimed, node) {
+		// test_and_set because we still might compete with cancellation
+		if (!test_and_set_bit(SSH_REQUEST_SF_COMPLETED_BIT, &r->state))
+			ssh_rtl_complete_with_status(r, -ESHUTDOWN);
+
+		// drop reference we've obtained by removing it from the lists
+		list_del(&r->node);
+		ssh_request_put(r);
+	}
+}
diff --git a/drivers/misc/surface_aggregator/ssh_request_layer.h b/drivers/misc/surface_aggregator/ssh_request_layer.h
new file mode 100644
index 000000000000..e945e0532628
--- /dev/null
+++ b/drivers/misc/surface_aggregator/ssh_request_layer.h
@@ -0,0 +1,142 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * SSH request transport layer.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _SURFACE_AGGREGATOR_SSH_REQUEST_LAYER_H
+#define _SURFACE_AGGREGATOR_SSH_REQUEST_LAYER_H
+
+#include <linux/atomic.h>
+#include <linux/ktime.h>
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <linux/workqueue.h>
+
+#include <linux/surface_aggregator/serial_hub.h>
+#include <linux/surface_aggregator/controller.h>
+
+#include "ssh_packet_layer.h"
+
+
+/**
+ * enum ssh_rtl_state_flags - State-flags for &struct ssh_rtl.
+ *
+ * @SSH_RTL_SF_SHUTDOWN_BIT:
+ *	Indicates that the request transport layer has been shut down or is
+ *	being shut down and should not accept any new requests.
+ */
+enum ssh_rtl_state_flags {
+	SSH_RTL_SF_SHUTDOWN_BIT,
+};
+
+/**
+ * struct ssh_rtl_ops - Callback operations for request transport layer.
+ * @handle_event: Function called when a SSH event has been received. The
+ *                specified function takes the request layer, received command
+ *                struct, and corresponding payload as arguments. If the event
+ *                has no payload, the payload span is empty (not %NULL).
+ */
+struct ssh_rtl_ops {
+	void (*handle_event)(struct ssh_rtl *rtl, const struct ssh_command *cmd,
+			     const struct ssam_span *data);
+};
+
+/**
+ * struct ssh_rtl - SSH request transport layer.
+ * @ptl:           Underlying packet transport layer.
+ * @state:         State(-flags) of the transport layer.
+ * @queue:         Request submission queue.
+ * @queue.lock:    Lock for modifying the request submission queue.
+ * @queue.head:    List-head of the request submission queue.
+ * @pending:       Set/list of pending requests.
+ * @pending.lock:  Lock for modifying the request set.
+ * @pending.head:  List-head of the pending set/list.
+ * @pending.count: Number of currently pending requests.
+ * @tx:            Transmitter subsystem.
+ * @tx.work:       Transmitter work item.
+ * @rtx_timeout:   Retransmission timeout subsystem.
+ * @rtx_timeout.timeout: Timeout inverval for retransmission.
+ * @rtx_timeout.expires: Time specifying when the reaper work is next scheduled.
+ * @rtx_timeout.reaper:  Work performing timeout checks and subsequent actions.
+ * @ops:           Request layer operations.
+ */
+struct ssh_rtl {
+	struct ssh_ptl ptl;
+	unsigned long state;
+
+	struct {
+		spinlock_t lock;
+		struct list_head head;
+	} queue;
+
+	struct {
+		spinlock_t lock;
+		struct list_head head;
+		atomic_t count;
+	} pending;
+
+	struct {
+		struct work_struct work;
+	} tx;
+
+	struct {
+		ktime_t timeout;
+		ktime_t expires;
+		struct delayed_work reaper;
+	} rtx_timeout;
+
+	struct ssh_rtl_ops ops;
+};
+
+#define rtl_dbg(r, fmt, ...)  ptl_dbg(&(r)->ptl, fmt, ##__VA_ARGS__)
+#define rtl_info(p, fmt, ...) ptl_info(&(p)->ptl, fmt, ##__VA_ARGS__)
+#define rtl_warn(r, fmt, ...) ptl_warn(&(r)->ptl, fmt, ##__VA_ARGS__)
+#define rtl_err(r, fmt, ...)  ptl_err(&(r)->ptl, fmt, ##__VA_ARGS__)
+#define rtl_dbg_cond(r, fmt, ...) __ssam_prcond(rtl_dbg, r, fmt, ##__VA_ARGS__)
+
+#define to_ssh_rtl(ptr, member) \
+	container_of(ptr, struct ssh_rtl, member)
+
+/**
+ * ssh_rtl_get_device() - Get device associated with request transport layer.
+ * @rtl: The request transport layer.
+ *
+ * Return: Returns the device on which the given request transport layer
+ * builds upon.
+ */
+static inline struct device *ssh_rtl_get_device(struct ssh_rtl *rtl)
+{
+	return ssh_ptl_get_device(&rtl->ptl);
+}
+
+/**
+ * ssh_request_rtl() - Get request transport layer associated with request.
+ * @rqst: The request to get the request transport layer reference for.
+ *
+ * Return: Returns the &struct ssh_rtl associated with the given SSH request.
+ */
+static inline struct ssh_rtl *ssh_request_rtl(struct ssh_request *rqst)
+{
+	struct ssh_ptl *ptl;
+
+	ptl = READ_ONCE(rqst->packet.ptl);
+	return likely(ptl) ? to_ssh_rtl(ptl, ptl) : NULL;
+}
+
+int ssh_rtl_submit(struct ssh_rtl *rtl, struct ssh_request *rqst);
+bool ssh_rtl_cancel(struct ssh_request *rqst, bool pending);
+
+int ssh_rtl_init(struct ssh_rtl *rtl, struct serdev_device *serdev,
+		 const struct ssh_rtl_ops *ops);
+
+int ssh_rtl_start(struct ssh_rtl *rtl);
+int ssh_rtl_flush(struct ssh_rtl *rtl, unsigned long timeout);
+void ssh_rtl_shutdown(struct ssh_rtl *rtl);
+void ssh_rtl_destroy(struct ssh_rtl *rtl);
+
+void ssh_request_init(struct ssh_request *rqst, enum ssam_request_flags flags,
+		      const struct ssh_request_ops *ops);
+
+#endif /* _SURFACE_AGGREGATOR_SSH_REQUEST_LAYER_H */
diff --git a/drivers/misc/surface_aggregator/trace.h b/drivers/misc/surface_aggregator/trace.h
new file mode 100644
index 000000000000..232bf1142aae
--- /dev/null
+++ b/drivers/misc/surface_aggregator/trace.h
@@ -0,0 +1,625 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Trace points for SSAM/SSH.
+ *
+ * Copyright (C) 2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM surface_aggregator
+
+#if !defined(_SURFACE_AGGREGATOR_TRACE_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _SURFACE_AGGREGATOR_TRACE_H
+
+#include <linux/surface_aggregator/serial_hub.h>
+
+#include <asm/unaligned.h>
+#include <linux/tracepoint.h>
+
+
+TRACE_DEFINE_ENUM(SSH_FRAME_TYPE_DATA_SEQ);
+TRACE_DEFINE_ENUM(SSH_FRAME_TYPE_DATA_NSQ);
+TRACE_DEFINE_ENUM(SSH_FRAME_TYPE_ACK);
+TRACE_DEFINE_ENUM(SSH_FRAME_TYPE_NAK);
+
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_LOCKED_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_QUEUED_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_PENDING_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_TRANSMITTING_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_TRANSMITTED_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_ACKED_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_CANCELED_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_SF_COMPLETED_BIT);
+
+TRACE_DEFINE_ENUM(SSH_PACKET_TY_FLUSH_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_TY_SEQUENCED_BIT);
+TRACE_DEFINE_ENUM(SSH_PACKET_TY_BLOCKING_BIT);
+
+TRACE_DEFINE_ENUM(SSH_PACKET_FLAGS_SF_MASK);
+TRACE_DEFINE_ENUM(SSH_PACKET_FLAGS_TY_MASK);
+
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_LOCKED_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_QUEUED_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_PENDING_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_TRANSMITTING_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_TRANSMITTED_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_RSPRCVD_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_CANCELED_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_SF_COMPLETED_BIT);
+
+TRACE_DEFINE_ENUM(SSH_REQUEST_TY_FLUSH_BIT);
+TRACE_DEFINE_ENUM(SSH_REQUEST_TY_HAS_RESPONSE_BIT);
+
+TRACE_DEFINE_ENUM(SSH_REQUEST_FLAGS_SF_MASK);
+TRACE_DEFINE_ENUM(SSH_REQUEST_FLAGS_TY_MASK);
+
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_SAM);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_BAT);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_TMP);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_PMC);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_FAN);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_PoM);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_DBG);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_KBD);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_FWU);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_UNI);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_LPC);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_TCL);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_SFL);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_KIP);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_EXT);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_BLD);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_BAS);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_SEN);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_SRQ);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_MCU);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_HID);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_TCH);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_BKL);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_TAM);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_ACC);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_UFI);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_USC);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_PEN);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_VID);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_AUD);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_SMC);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_KPD);
+TRACE_DEFINE_ENUM(SSAM_SSH_TC_REG);
+
+
+#define SSAM_PTR_UID_LEN		9
+#define SSAM_U8_FIELD_NOT_APPLICABLE	((u16)-1)
+#define SSAM_SEQ_NOT_APPLICABLE		((u16)-1)
+#define SSAM_RQID_NOT_APPLICABLE	((u32)-1)
+#define SSAM_SSH_TC_NOT_APPLICABLE	0
+
+
+#ifndef _SURFACE_AGGREGATOR_TRACE_HELPERS
+#define _SURFACE_AGGREGATOR_TRACE_HELPERS
+
+/**
+ * ssam_trace_ptr_uid() - Convert the pointer to a non-pointer UID string.
+ * @ptr: The pointer to convert.
+ * @uid_str: A buffer of length SSAM_PTR_UID_LEN where the UID will be stored.
+ *
+ * Converts the given pointer into a UID string that is safe to be shared
+ * with userspace and logs, i.e. doesn't give away the real memory location.
+ */
+static inline void ssam_trace_ptr_uid(const void *ptr, char *uid_str)
+{
+	char buf[2 * sizeof(void *) + 1];
+
+	snprintf(buf, ARRAY_SIZE(buf), "%p", ptr);
+	memcpy(uid_str, &buf[ARRAY_SIZE(buf) - SSAM_PTR_UID_LEN],
+	       SSAM_PTR_UID_LEN);
+}
+
+/**
+ * ssam_trace_get_packet_seq() - Read the packet's sequence ID.
+ * @p: The packet.
+ *
+ * Return: Returns the packet's sequence ID (SEQ) field if present, or
+ * %SSAM_SEQ_NOT_APPLICABLE if not (e.g. flush packet).
+ */
+static inline u16 ssam_trace_get_packet_seq(const struct ssh_packet *p)
+{
+	if (!p->data.ptr || p->data.len < SSH_MESSAGE_LENGTH(0))
+		return SSAM_SEQ_NOT_APPLICABLE;
+
+	return p->data.ptr[SSH_MSGOFFSET_FRAME(seq)];
+}
+
+/**
+ * ssam_trace_get_request_id() - Read the packet's request ID.
+ * @p: The packet.
+ *
+ * Return: Returns the packet's request ID (RQID) field if the packet
+ * represents a request with command data, or %SSAM_RQID_NOT_APPLICABLE if not
+ * (e.g. flush request, control packet).
+ */
+static inline u32 ssam_trace_get_request_id(const struct ssh_packet *p)
+{
+	if (!p->data.ptr || p->data.len < SSH_COMMAND_MESSAGE_LENGTH(0))
+		return SSAM_RQID_NOT_APPLICABLE;
+
+	return get_unaligned_le16(&p->data.ptr[SSH_MSGOFFSET_COMMAND(rqid)]);
+}
+
+/**
+ * ssam_trace_get_request_tc() - Read the packet's request target category.
+ * @p: The packet.
+ *
+ * Return: Returns the packet's request target category (TC) field if the
+ * packet represents a request with command data, or %SSAM_TC_NOT_APPLICABLE
+ * if not (e.g. flush request, control packet).
+ */
+static inline u32 ssam_trace_get_request_tc(const struct ssh_packet *p)
+{
+	if (!p->data.ptr || p->data.len < SSH_COMMAND_MESSAGE_LENGTH(0))
+		return SSAM_SSH_TC_NOT_APPLICABLE;
+
+	return get_unaligned_le16(&p->data.ptr[SSH_MSGOFFSET_COMMAND(tc)]);
+}
+
+#endif /* _SURFACE_AGGREGATOR_TRACE_HELPERS */
+
+#define ssam_trace_get_command_field_u8(packet, field) \
+	((!packet || packet->data.len < SSH_COMMAND_MESSAGE_LENGTH(0)) \
+	 ? 0 : p->data.ptr[SSH_MSGOFFSET_COMMAND(field)])
+
+#define ssam_show_generic_u8_field(value)				\
+	__print_symbolic(value,						\
+		{ SSAM_U8_FIELD_NOT_APPLICABLE,		"N/A" }		\
+	)
+
+
+#define ssam_show_frame_type(ty)					\
+	__print_symbolic(ty,						\
+		{ SSH_FRAME_TYPE_DATA_SEQ,		"DSEQ" },	\
+		{ SSH_FRAME_TYPE_DATA_NSQ,		"DNSQ" },	\
+		{ SSH_FRAME_TYPE_ACK,			"ACK"  },	\
+		{ SSH_FRAME_TYPE_NAK,			"NAK"  }	\
+	)
+
+#define ssam_show_packet_type(type)					\
+	__print_flags(flags & SSH_PACKET_FLAGS_TY_MASK, "",		\
+		{ BIT(SSH_PACKET_TY_FLUSH_BIT),		"F" },		\
+		{ BIT(SSH_PACKET_TY_SEQUENCED_BIT),	"S" },		\
+		{ BIT(SSH_PACKET_TY_BLOCKING_BIT),	"B" }		\
+	)
+
+#define ssam_show_packet_state(state)					\
+	__print_flags(flags & SSH_PACKET_FLAGS_SF_MASK, "",		\
+		{ BIT(SSH_PACKET_SF_LOCKED_BIT),	"L" },		\
+		{ BIT(SSH_PACKET_SF_QUEUED_BIT),	"Q" },		\
+		{ BIT(SSH_PACKET_SF_PENDING_BIT),	"P" },		\
+		{ BIT(SSH_PACKET_SF_TRANSMITTING_BIT),	"S" },		\
+		{ BIT(SSH_PACKET_SF_TRANSMITTED_BIT),	"T" },		\
+		{ BIT(SSH_PACKET_SF_ACKED_BIT),		"A" },		\
+		{ BIT(SSH_PACKET_SF_CANCELED_BIT),	"C" },		\
+		{ BIT(SSH_PACKET_SF_COMPLETED_BIT),	"F" }		\
+	)
+
+#define ssam_show_packet_seq(seq)					\
+	__print_symbolic(seq,						\
+		{ SSAM_SEQ_NOT_APPLICABLE,		"N/A" }		\
+	)
+
+
+#define ssam_show_request_type(flags)					\
+	__print_flags(flags & SSH_REQUEST_FLAGS_TY_MASK, "",		\
+		{ BIT(SSH_REQUEST_TY_FLUSH_BIT),	"F" },		\
+		{ BIT(SSH_REQUEST_TY_HAS_RESPONSE_BIT),	"R" }		\
+	)
+
+#define ssam_show_request_state(flags)					\
+	__print_flags(flags & SSH_REQUEST_FLAGS_SF_MASK, "",		\
+		{ BIT(SSH_REQUEST_SF_LOCKED_BIT),	"L" },		\
+		{ BIT(SSH_REQUEST_SF_QUEUED_BIT),	"Q" },		\
+		{ BIT(SSH_REQUEST_SF_PENDING_BIT),	"P" },		\
+		{ BIT(SSH_REQUEST_SF_TRANSMITTING_BIT),	"S" },		\
+		{ BIT(SSH_REQUEST_SF_TRANSMITTED_BIT),	"T" },		\
+		{ BIT(SSH_REQUEST_SF_RSPRCVD_BIT),	"A" },		\
+		{ BIT(SSH_REQUEST_SF_CANCELED_BIT),	"C" },		\
+		{ BIT(SSH_REQUEST_SF_COMPLETED_BIT),	"F" }		\
+	)
+
+#define ssam_show_request_id(rqid)					\
+	__print_symbolic(rqid,						\
+		{ SSAM_RQID_NOT_APPLICABLE,		"N/A" }		\
+	)
+
+#define ssam_show_ssh_tc(rqid)						\
+	__print_symbolic(rqid,						\
+		{ SSAM_SSH_TC_NOT_APPLICABLE,		"N/A" },	\
+		{ SSAM_SSH_TC_SAM,			"SAM" },	\
+		{ SSAM_SSH_TC_BAT,			"BAT" },	\
+		{ SSAM_SSH_TC_TMP,			"TMP" },	\
+		{ SSAM_SSH_TC_PMC,			"PMC" },	\
+		{ SSAM_SSH_TC_FAN,			"FAN" },	\
+		{ SSAM_SSH_TC_PoM,			"PoM" },	\
+		{ SSAM_SSH_TC_DBG,			"DBG" },	\
+		{ SSAM_SSH_TC_KBD,			"KBD" },	\
+		{ SSAM_SSH_TC_FWU,			"FWU" },	\
+		{ SSAM_SSH_TC_UNI,			"UNI" },	\
+		{ SSAM_SSH_TC_LPC,			"LPC" },	\
+		{ SSAM_SSH_TC_TCL,			"TCL" },	\
+		{ SSAM_SSH_TC_SFL,			"SFL" },	\
+		{ SSAM_SSH_TC_KIP,			"KIP" },	\
+		{ SSAM_SSH_TC_EXT,			"EXT" },	\
+		{ SSAM_SSH_TC_BLD,			"BLD" },	\
+		{ SSAM_SSH_TC_BAS,			"BAS" },	\
+		{ SSAM_SSH_TC_SEN,			"SEN" },	\
+		{ SSAM_SSH_TC_SRQ,			"SRQ" },	\
+		{ SSAM_SSH_TC_MCU,			"MCU" },	\
+		{ SSAM_SSH_TC_HID,			"HID" },	\
+		{ SSAM_SSH_TC_TCH,			"TCH" },	\
+		{ SSAM_SSH_TC_BKL,			"BKL" },	\
+		{ SSAM_SSH_TC_TAM,			"TAM" },	\
+		{ SSAM_SSH_TC_ACC,			"ACC" },	\
+		{ SSAM_SSH_TC_UFI,			"UFI" },	\
+		{ SSAM_SSH_TC_USC,			"USC" },	\
+		{ SSAM_SSH_TC_PEN,			"PEN" },	\
+		{ SSAM_SSH_TC_VID,			"VID" },	\
+		{ SSAM_SSH_TC_AUD,			"AUD" },	\
+		{ SSAM_SSH_TC_SMC,			"SMC" },	\
+		{ SSAM_SSH_TC_KPD,			"KPD" },	\
+		{ SSAM_SSH_TC_REG,			"REG" }		\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_frame_class,
+	TP_PROTO(const struct ssh_frame *frame),
+
+	TP_ARGS(frame),
+
+	TP_STRUCT__entry(
+		__field(u8, type)
+		__field(u8, seq)
+		__field(u16, len)
+	),
+
+	TP_fast_assign(
+		__entry->type = frame->type;
+		__entry->seq = frame->seq;
+		__entry->len = get_unaligned_le16(&frame->len);
+	),
+
+	TP_printk("ty=%s, seq=0x%02x, len=%u",
+		ssam_show_frame_type(__entry->type),
+		__entry->seq,
+		__entry->len
+	)
+);
+
+#define DEFINE_SSAM_FRAME_EVENT(name)				\
+	DEFINE_EVENT(ssam_frame_class, ssam_##name,		\
+		TP_PROTO(const struct ssh_frame *frame),	\
+		TP_ARGS(frame)					\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_command_class,
+	TP_PROTO(const struct ssh_command *cmd, u16 len),
+
+	TP_ARGS(cmd, len),
+
+	TP_STRUCT__entry(
+		__field(u16, rqid)
+		__field(u16, len)
+		__field(u8, tc)
+		__field(u8, cid)
+		__field(u8, iid)
+	),
+
+	TP_fast_assign(
+		__entry->rqid = get_unaligned_le16(&cmd->rqid);
+		__entry->tc = cmd->tc;
+		__entry->cid = cmd->cid;
+		__entry->iid = cmd->iid;
+		__entry->len = len;
+	),
+
+	TP_printk("rqid=0x%04x, tc=%s, cid=0x%02x, iid=0x%02x, len=%u",
+		__entry->rqid,
+		ssam_show_ssh_tc(__entry->tc),
+		__entry->cid,
+		__entry->iid,
+		__entry->len
+	)
+);
+
+#define DEFINE_SSAM_COMMAND_EVENT(name)					\
+	DEFINE_EVENT(ssam_command_class, ssam_##name,			\
+		TP_PROTO(const struct ssh_command *cmd, u16 len),	\
+		TP_ARGS(cmd, len)					\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_packet_class,
+	TP_PROTO(const struct ssh_packet *packet),
+
+	TP_ARGS(packet),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, state)
+		__array(char, uid, SSAM_PTR_UID_LEN)
+		__field(u8, priority)
+		__field(u16, length)
+		__field(u16, seq)
+	),
+
+	TP_fast_assign(
+		__entry->state = READ_ONCE(packet->state);
+		ssam_trace_ptr_uid(packet, __entry->uid);
+		__entry->priority = READ_ONCE(packet->priority);
+		__entry->length = packet->data.len;
+		__entry->seq = ssam_trace_get_packet_seq(packet);
+	),
+
+	TP_printk("uid=%s, seq=%s, ty=%s, pri=0x%02x, len=%u, sta=%s",
+		__entry->uid,
+		ssam_show_packet_seq(__entry->seq),
+		ssam_show_packet_type(__entry->state),
+		__entry->priority,
+		__entry->length,
+		ssam_show_packet_state(__entry->state)
+	)
+);
+
+#define DEFINE_SSAM_PACKET_EVENT(name)				\
+	DEFINE_EVENT(ssam_packet_class, ssam_##name,		\
+		TP_PROTO(const struct ssh_packet *packet),	\
+		TP_ARGS(packet)					\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_packet_status_class,
+	TP_PROTO(const struct ssh_packet *packet, int status),
+
+	TP_ARGS(packet, status),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, state)
+		__field(int, status)
+		__array(char, uid, SSAM_PTR_UID_LEN)
+		__field(u8, priority)
+		__field(u16, length)
+		__field(u16, seq)
+	),
+
+	TP_fast_assign(
+		__entry->state = READ_ONCE(packet->state);
+		__entry->status = status;
+		ssam_trace_ptr_uid(packet, __entry->uid);
+		__entry->priority = READ_ONCE(packet->priority);
+		__entry->length = packet->data.len;
+		__entry->seq = ssam_trace_get_packet_seq(packet);
+	),
+
+	TP_printk("uid=%s, seq=%s, ty=%s, pri=0x%02x, len=%u, sta=%s, status=%d",
+		__entry->uid,
+		ssam_show_packet_seq(__entry->seq),
+		ssam_show_packet_type(__entry->state),
+		__entry->priority,
+		__entry->length,
+		ssam_show_packet_state(__entry->state),
+		__entry->status
+	)
+);
+
+#define DEFINE_SSAM_PACKET_STATUS_EVENT(name)				\
+	DEFINE_EVENT(ssam_packet_status_class, ssam_##name,		\
+		TP_PROTO(const struct ssh_packet *packet, int status),	\
+		TP_ARGS(packet, status)					\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_request_class,
+	TP_PROTO(const struct ssh_request *request),
+
+	TP_ARGS(request),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, state)
+		__field(u32, rqid)
+		__array(char, uid, SSAM_PTR_UID_LEN)
+		__field(u8, tc)
+		__field(u16, cid)
+		__field(u16, iid)
+	),
+
+	TP_fast_assign(
+		const struct ssh_packet *p = &request->packet;
+
+		// use packet for UID so we can match requests to packets
+		__entry->state = READ_ONCE(request->state);
+		__entry->rqid = ssam_trace_get_request_id(p);
+		ssam_trace_ptr_uid(p, __entry->uid);
+		__entry->tc = ssam_trace_get_request_tc(p);
+		__entry->cid = ssam_trace_get_command_field_u8(p, cid);
+		__entry->iid = ssam_trace_get_command_field_u8(p, iid);
+	),
+
+	TP_printk("uid=%s, rqid=%s, ty=%s, sta=%s, tc=%s, cid=%s, iid=%s",
+		__entry->uid,
+		ssam_show_request_id(__entry->rqid),
+		ssam_show_request_type(__entry->state),
+		ssam_show_request_state(__entry->state),
+		ssam_show_ssh_tc(__entry->tc),
+		ssam_show_generic_u8_field(__entry->cid),
+		ssam_show_generic_u8_field(__entry->iid)
+	)
+);
+
+#define DEFINE_SSAM_REQUEST_EVENT(name)				\
+	DEFINE_EVENT(ssam_request_class, ssam_##name,		\
+		TP_PROTO(const struct ssh_request *request),	\
+		TP_ARGS(request)				\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_request_status_class,
+	TP_PROTO(const struct ssh_request *request, int status),
+
+	TP_ARGS(request, status),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, state)
+		__field(u32, rqid)
+		__field(int, status)
+		__array(char, uid, SSAM_PTR_UID_LEN)
+		__field(u8, tc)
+		__field(u16, cid)
+		__field(u16, iid)
+	),
+
+	TP_fast_assign(
+		const struct ssh_packet *p = &request->packet;
+
+		// use packet for UID so we can match requests to packets
+		__entry->state = READ_ONCE(request->state);
+		__entry->rqid = ssam_trace_get_request_id(p);
+		__entry->status = status;
+		ssam_trace_ptr_uid(p, __entry->uid);
+		__entry->tc = ssam_trace_get_request_tc(p);
+		__entry->cid = ssam_trace_get_command_field_u8(p, cid);
+		__entry->iid = ssam_trace_get_command_field_u8(p, iid);
+	),
+
+	TP_printk("uid=%s, rqid=%s, ty=%s, sta=%s, tc=%s, cid=%s, iid=%s, status=%d",
+		__entry->uid,
+		ssam_show_request_id(__entry->rqid),
+		ssam_show_request_type(__entry->state),
+		ssam_show_request_state(__entry->state),
+		ssam_show_ssh_tc(__entry->tc),
+		ssam_show_generic_u8_field(__entry->cid),
+		ssam_show_generic_u8_field(__entry->iid),
+		__entry->status
+	)
+);
+
+#define DEFINE_SSAM_REQUEST_STATUS_EVENT(name)				\
+	DEFINE_EVENT(ssam_request_status_class, ssam_##name,		\
+		TP_PROTO(const struct ssh_request *request, int status),\
+		TP_ARGS(request, status)				\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_alloc_class,
+	TP_PROTO(void *ptr, size_t len),
+
+	TP_ARGS(ptr, len),
+
+	TP_STRUCT__entry(
+		__field(size_t, len)
+		__array(char, uid, SSAM_PTR_UID_LEN)
+	),
+
+	TP_fast_assign(
+		__entry->len = len;
+		ssam_trace_ptr_uid(ptr, __entry->uid);
+	),
+
+	TP_printk("uid=%s, len=%zu", __entry->uid, __entry->len)
+);
+
+#define DEFINE_SSAM_ALLOC_EVENT(name)					\
+	DEFINE_EVENT(ssam_alloc_class, ssam_##name,			\
+		TP_PROTO(void *ptr, size_t len),			\
+		TP_ARGS(ptr, len)					\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_free_class,
+	TP_PROTO(void *ptr),
+
+	TP_ARGS(ptr),
+
+	TP_STRUCT__entry(
+		__array(char, uid, SSAM_PTR_UID_LEN)
+	),
+
+	TP_fast_assign(
+		ssam_trace_ptr_uid(ptr, __entry->uid);
+	),
+
+	TP_printk("uid=%s", __entry->uid)
+);
+
+#define DEFINE_SSAM_FREE_EVENT(name)					\
+	DEFINE_EVENT(ssam_free_class, ssam_##name,			\
+		TP_PROTO(void *ptr),					\
+		TP_ARGS(ptr)						\
+	)
+
+
+DECLARE_EVENT_CLASS(ssam_generic_uint_class,
+	TP_PROTO(const char *property, unsigned int value),
+
+	TP_ARGS(property, value),
+
+	TP_STRUCT__entry(
+		__field(unsigned int, value)
+		__string(property, property)
+	),
+
+	TP_fast_assign(
+		__entry->value = value;
+		__assign_str(property, property);
+	),
+
+	TP_printk("%s=%u", __get_str(property), __entry->value)
+);
+
+#define DEFINE_SSAM_GENERIC_UINT_EVENT(name)				\
+	DEFINE_EVENT(ssam_generic_uint_class, ssam_##name,		\
+		TP_PROTO(const char *property, unsigned int value),	\
+		TP_ARGS(property, value)				\
+	)
+
+
+DEFINE_SSAM_FRAME_EVENT(rx_frame_received);
+DEFINE_SSAM_COMMAND_EVENT(rx_response_received);
+DEFINE_SSAM_COMMAND_EVENT(rx_event_received);
+
+DEFINE_SSAM_PACKET_EVENT(packet_release);
+DEFINE_SSAM_PACKET_EVENT(packet_submit);
+DEFINE_SSAM_PACKET_EVENT(packet_resubmit);
+DEFINE_SSAM_PACKET_EVENT(packet_timeout);
+DEFINE_SSAM_PACKET_EVENT(packet_cancel);
+DEFINE_SSAM_PACKET_STATUS_EVENT(packet_complete);
+DEFINE_SSAM_GENERIC_UINT_EVENT(ptl_timeout_reap);
+
+DEFINE_SSAM_REQUEST_EVENT(request_submit);
+DEFINE_SSAM_REQUEST_EVENT(request_timeout);
+DEFINE_SSAM_REQUEST_EVENT(request_cancel);
+DEFINE_SSAM_REQUEST_STATUS_EVENT(request_complete);
+DEFINE_SSAM_GENERIC_UINT_EVENT(rtl_timeout_reap);
+
+DEFINE_SSAM_PACKET_EVENT(ei_tx_drop_ack_packet);
+DEFINE_SSAM_PACKET_EVENT(ei_tx_drop_nak_packet);
+DEFINE_SSAM_PACKET_EVENT(ei_tx_drop_dsq_packet);
+DEFINE_SSAM_PACKET_STATUS_EVENT(ei_tx_fail_write);
+DEFINE_SSAM_PACKET_EVENT(ei_tx_corrupt_data);
+DEFINE_SSAM_GENERIC_UINT_EVENT(ei_rx_corrupt_syn);
+DEFINE_SSAM_FRAME_EVENT(ei_rx_corrupt_data);
+DEFINE_SSAM_REQUEST_EVENT(ei_rx_drop_response);
+
+DEFINE_SSAM_ALLOC_EVENT(ctrl_packet_alloc);
+DEFINE_SSAM_FREE_EVENT(ctrl_packet_free);
+
+DEFINE_SSAM_ALLOC_EVENT(event_item_alloc);
+DEFINE_SSAM_FREE_EVENT(event_item_free);
+
+#endif /* _SURFACE_AGGREGATOR_TRACE_H */
+
+/* This part must be outside protection */
+#undef TRACE_INCLUDE_PATH
+#undef TRACE_INCLUDE_FILE
+
+#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_FILE trace
+
+#include <trace/define_trace.h>
diff --git a/include/linux/mod_devicetable.h b/include/linux/mod_devicetable.h
index a35f82fb36da..9932a610bda0 100644
--- a/include/linux/mod_devicetable.h
+++ b/include/linux/mod_devicetable.h
@@ -838,15 +838,16 @@ struct mhi_device_id {
 
 /* Surface System Aggregator Module */
 
-#define SSAM_MATCH_CHANNEL	0x1
+#define SSAM_MATCH_TARGET	0x1
 #define SSAM_MATCH_INSTANCE	0x2
 #define SSAM_MATCH_FUNCTION	0x4
 
 struct ssam_device_id {
 	__u8 match_flags;
 
+	__u8 domain;
 	__u8 category;
-	__u8 channel;
+	__u8 target;
 	__u8 instance;
 	__u8 function;
 
diff --git a/include/linux/surface_acpi_notify.h b/include/linux/surface_acpi_notify.h
new file mode 100644
index 000000000000..8e3e86c7d78c
--- /dev/null
+++ b/include/linux/surface_acpi_notify.h
@@ -0,0 +1,39 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Interface for Surface ACPI Notify (SAN) driver.
+ *
+ * Provides access to discrete GPU notifications sent from ACPI via the SAN
+ * driver, which are not handled by this driver directly.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _LINUX_SURFACE_ACPI_NOTIFY_H
+#define _LINUX_SURFACE_ACPI_NOTIFY_H
+
+#include <linux/notifier.h>
+#include <linux/types.h>
+
+/**
+ * struct san_dgpu_event - Discrete GPU ACPI event.
+ * @category: Category of the event.
+ * @target:   Target ID of the event source.
+ * @command:  Command ID of the event.
+ * @instance: Instance ID of the event source.
+ * @length:   Length of the event's payload data (in bytes).
+ * @payload:  Pointer to the event's payload data.
+ */
+struct san_dgpu_event {
+	u8 category;
+	u8 target;
+	u8 command;
+	u8 instance;
+	u16 length;
+	u8 *payload;
+};
+
+int san_client_link(struct device *client);
+int san_dgpu_notifier_register(struct notifier_block *nb);
+int san_dgpu_notifier_unregister(struct notifier_block *nb);
+
+#endif /* _LINUX_SURFACE_ACPI_NOTIFY_H */
diff --git a/include/linux/surface_aggregator/controller.h b/include/linux/surface_aggregator/controller.h
new file mode 100644
index 000000000000..d128c68c04e0
--- /dev/null
+++ b/include/linux/surface_aggregator/controller.h
@@ -0,0 +1,832 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Surface System Aggregator Module (SSAM) controller interface.
+ *
+ * Main communication interface for the SSAM EC. Provides a controller
+ * managing access and communication to and from the SSAM EC, as well as main
+ * communication structures and definitions.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _LINUX_SURFACE_AGGREGATOR_CONTROLLER_H
+#define _LINUX_SURFACE_AGGREGATOR_CONTROLLER_H
+
+#include <linux/completion.h>
+#include <linux/device.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/serial_hub.h>
+
+
+/* -- Main data types and definitions --------------------------------------- */
+
+/**
+ * enum ssam_event_flags - Flags for enabling/disabling SAM events
+ * @SSAM_EVENT_SEQUENCED: The event will be sent via a sequenced data frame.
+ */
+enum ssam_event_flags {
+	SSAM_EVENT_SEQUENCED = BIT(0),
+};
+
+/**
+ * struct ssam_event - SAM event sent from the EC to the host.
+ * @target_category: Target category of the event source. See &enum ssam_ssh_tc.
+ * @target_id:       Target ID of the event source.
+ * @command_id:      Command ID of the event.
+ * @instance_id:     Instance ID of the event source.
+ * @length:          Length of the event payload in bytes.
+ * @data:            Event payload data.
+ */
+struct ssam_event {
+	u8 target_category;
+	u8 target_id;
+	u8 command_id;
+	u8 instance_id;
+	u16 length;
+	u8 data[];
+};
+
+/**
+ * enum ssam_request_flags - Flags for SAM requests.
+ *
+ * @SSAM_REQUEST_HAS_RESPONSE:
+ *	Specifies that the request expects a response. If not set, the request
+ *	will be directly completed after its underlying packet has been
+ *	transmitted. If set, the request transport system waits for a response
+ *	of the request.
+ *
+ * @SSAM_REQUEST_UNSEQUENCED:
+ *	Specifies that the request should be transmitted via an unsequenced
+ *	packet. If set, the request must not have a response, meaning that this
+ *	flag and the %SSAM_REQUEST_HAS_RESPONSE flag are mutually exclusive.
+ */
+enum ssam_request_flags {
+	SSAM_REQUEST_HAS_RESPONSE = BIT(0),
+	SSAM_REQUEST_UNSEQUENCED  = BIT(1),
+};
+
+/**
+ * struct ssam_request - SAM request description.
+ * @target_category: Category of the request's target. See &enum ssam_ssh_tc.
+ * @target_id:       ID of the request's target.
+ * @command_id:      Command ID of the request.
+ * @instance_id:     Instance ID of the request's target.
+ * @flags:           Flags for the request. See &enum ssam_request_flags.
+ * @length:          Length of the request payload in bytes.
+ * @payload:         Request payload data.
+ *
+ * This struct fully describes a SAM request with payload. It is intended to
+ * help set up the actual transport struct, e.g. &struct ssam_request_sync,
+ * and specifically its raw message data via ssam_request_write_data().
+ */
+struct ssam_request {
+	u8 target_category;
+	u8 target_id;
+	u8 command_id;
+	u8 instance_id;
+	u16 flags;
+	u16 length;
+	const u8 *payload;
+};
+
+/**
+ * struct ssam_response - Response buffer for SAM request.
+ * @capacity: Capacity of the buffer, in bytes.
+ * @length:   Length of the actual data stored in the memory pointed to by
+ *            @pointer, in bytes. Set by the transport system.
+ * @pointer:  Pointer to the buffer's memory, storing the response payload data.
+ */
+struct ssam_response {
+	size_t capacity;
+	size_t length;
+	u8 *pointer;
+};
+
+struct ssam_controller;
+
+
+struct ssam_controller *ssam_get_controller(void);
+int ssam_client_link(struct ssam_controller *ctrl, struct device *client);
+int ssam_client_bind(struct device *client, struct ssam_controller **ctrl);
+
+struct device *ssam_controller_device(struct ssam_controller *c);
+
+struct ssam_controller *ssam_controller_get(struct ssam_controller *c);
+void ssam_controller_put(struct ssam_controller *c);
+
+void ssam_controller_statelock(struct ssam_controller *c);
+void ssam_controller_stateunlock(struct ssam_controller *c);
+
+ssize_t ssam_request_write_data(struct ssam_span *buf,
+				struct ssam_controller *ctrl,
+				const struct ssam_request *spec);
+
+
+/* -- Synchronous request interface. ---------------------------------------- */
+
+/**
+ * struct ssam_request_sync - Synchronous SAM request struct.
+ * @base:   Underlying SSH request.
+ * @comp:   Completion used to signal full completion of the request. After the
+ *          request has been submitted, this struct may only be modified or
+ *          deallocated after the completion has been signaled.
+ *          request has been submitted,
+ * @resp:   Buffer to store the response.
+ * @status: Status of the request, set after the base request has been
+ *          completed or has failed.
+ */
+struct ssam_request_sync {
+	struct ssh_request base;
+	struct completion comp;
+	struct ssam_response *resp;
+	int status;
+};
+
+int ssam_request_sync_alloc(size_t payload_len, gfp_t flags,
+			    struct ssam_request_sync **rqst,
+			    struct ssam_span *buffer);
+
+void ssam_request_sync_free(struct ssam_request_sync *rqst);
+
+void ssam_request_sync_init(struct ssam_request_sync *rqst,
+			    enum ssam_request_flags flags);
+
+/**
+ * ssam_request_sync_set_data - Set message data of a synchronous request.
+ * @rqst: The request.
+ * @ptr:  Pointer to the request message data.
+ * @len:  Length of the request message data.
+ *
+ * Set the request message data of a synchronous request. The provided buffer
+ * needs to live until the request has been completed.
+ */
+static inline void ssam_request_sync_set_data(struct ssam_request_sync *rqst,
+					      u8 *ptr, size_t len)
+{
+	ssh_request_set_data(&rqst->base, ptr, len);
+}
+
+/**
+ * ssam_request_sync_set_resp - Set response buffer of a synchronous request.
+ * @rqst: The request.
+ * @resp: The response buffer.
+ *
+ * Sets the response buffer ot a synchronous request. This buffer will store
+ * the response of the request after it has been completed. May be %NULL if
+ * no response is expected.
+ */
+static inline void ssam_request_sync_set_resp(struct ssam_request_sync *rqst,
+					      struct ssam_response *resp)
+{
+	rqst->resp = resp;
+}
+
+int ssam_request_sync_submit(struct ssam_controller *ctrl,
+			     struct ssam_request_sync *rqst);
+
+/**
+ * ssam_request_sync_wait - Wait for completion of a synchronous request.
+ * @rqst: The request to wait for.
+ *
+ * Wait for completion and release of a synchronous request. After this
+ * function terminates, the request is guaranteed to have left the transport
+ * system. After successful submission of a request, this function must be
+ * called before accessing the response of the request, freeing the request,
+ * or freeing any of the buffers associated with the request.
+ *
+ * This function must not be called if the request has not been submitted yet
+ * and may lead to a deadlock/infinite wait if a subsequent request submission
+ * fails in that case, due to the completion never triggering.
+ *
+ * Return: Returns the status of the given request, which is set on completion
+ * of the packet. This value is zero on success and negative on failure.
+ */
+static inline int ssam_request_sync_wait(struct ssam_request_sync *rqst)
+{
+	wait_for_completion(&rqst->comp);
+	return rqst->status;
+}
+
+int ssam_request_sync(struct ssam_controller *ctrl,
+		      const struct ssam_request *spec,
+		      struct ssam_response *rsp);
+
+int ssam_request_sync_with_buffer(struct ssam_controller *ctrl,
+				  const struct ssam_request *spec,
+				  struct ssam_response *rsp,
+				  struct ssam_span *buf);
+
+
+/**
+ * ssam_request_sync_onstack - Execute a synchronous request on the stack.
+ * @ctrl: The controller via which the request is submitted.
+ * @rqst: The request specification.
+ * @rsp:  The response buffer.
+ * @payload_len: The (maximum) request payload length.
+ *
+ * Allocates a synchronous request with specified payload length on the stack,
+ * fully intializes it via the provided request specification, submits it, and
+ * finally waits for its completion before returning its status. This helper
+ * macro essentially allocates the request message buffer on the stack and
+ * then calls ssam_request_sync_with_buffer().
+ *
+ * Note: The @payload_len parameter specifies the maximum payload length, used
+ * for buffer allocation. The actual payload length may be smaller.
+ *
+ * Return: Returns the status of the request or any failure during setup, i.e.
+ * zero on success and a negative value on failure.
+ */
+#define ssam_request_sync_onstack(ctrl, rqst, rsp, payload_len)			\
+	({									\
+		u8 __data[SSH_COMMAND_MESSAGE_LENGTH(payload_len)];		\
+		struct ssam_span __buf = { &__data[0], ARRAY_SIZE(__data) };	\
+										\
+		ssam_request_sync_with_buffer(ctrl, rqst, rsp, &__buf);		\
+	})
+
+/**
+ * __ssam_retry - Retry request in case of I/O errors or timeouts.
+ * @request: The request function to execute. Must return an integer.
+ * @n:       Number of tries.
+ * @args:    Arguments for the request function.
+ *
+ * Executes the given request function, i.e. calls @request. In case the
+ * request returns %-EREMOTEIO (indicates I/O error) or -%ETIMEDOUT (request
+ * or underlying packet timed out), @request will be re-executed again, up to
+ * @n times in total.
+ *
+ * Return: Returns the return value of the last execution of @request.
+ */
+#define __ssam_retry(request, n, args...)				\
+	({								\
+		int __i, __s = 0;					\
+									\
+		for (__i = (n); __i > 0; __i--) {			\
+			__s = request(args);				\
+			if (__s != -ETIMEDOUT && __s != -EREMOTEIO)	\
+				break;					\
+		}							\
+		__s;							\
+	})
+
+/**
+ * ssam_retry - Retry request in case of I/O errors or timeouts up to three
+ * times in total.
+ * @request: The request function to execute. Must return an integer.
+ * @args:    Arguments for the request function.
+ *
+ * Executes the given request function, i.e. calls @request. In case the
+ * request returns %-EREMOTEIO (indicates I/O error) or -%ETIMEDOUT (request
+ * or underlying packet timed out), @request will be re-executed again, up to
+ * three times in total.
+ *
+ * See __ssam_retry() for a more generic macro for this purpose.
+ *
+ * Return: Returns the return value of the last execution of @request.
+ */
+#define ssam_retry(request, args...) \
+	__ssam_retry(request, 3, args)
+
+/**
+ * struct ssam_request_spec - Blue-print specification of SAM request.
+ * @target_category: Category of the request's target. See &enum ssam_ssh_tc.
+ * @target_id:       ID of the request's target.
+ * @command_id:      Command ID of the request.
+ * @instance_id:     Instance ID of the request's target.
+ * @flags:           Flags for the request. See &enum ssam_request_flags.
+ *
+ * Blue-print specification for a SAM request. This struct describes the
+ * unique static parameters of a request (i.e. type) without specifying any of
+ * its instance-specific data (e.g. payload). It is intended to be used as base
+ * for defining simple request functions via the
+ * ``SSAM_DEFINE_SYNC_REQUEST_x()`` family of macros.
+ */
+struct ssam_request_spec {
+	u8 target_category;
+	u8 target_id;
+	u8 command_id;
+	u8 instance_id;
+	u8 flags;
+};
+
+/**
+ * struct ssam_request_spec_md - Blue-print specification for multi-device SAM
+ * request.
+ * @target_category: Category of the request's target. See &enum ssam_ssh_tc.
+ * @command_id:      Command ID of the request.
+ * @flags:           Flags for the request. See &enum ssam_request_flags.
+ *
+ * Blue-print specification for a multi-device SAM request, i.e. a request
+ * that is applicable to multiple device instances, described by their
+ * individual target and instance IDs. This struct describes the unique static
+ * parameters of a request (i.e. type) without specifying any of its
+ * instance-specific data (e.g. payload) and without specifying any of its
+ * device specific IDs (i.e. target and instance ID). It is intended to be
+ * used as base for defining simple multi-device request functions via the
+ * ``SSAM_DEFINE_SYNC_REQUEST_MD_x()`` and ``SSAM_DEFINE_SYNC_REQUEST_CL_x()``
+ * families of macros.
+ */
+struct ssam_request_spec_md {
+	u8 target_category;
+	u8 command_id;
+	u8 flags;
+};
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_N() - Define synchronous SAM request function
+ * with neither argument nor return value.
+ * @name: Name of the generated function.
+ * @spec: Specification (&struct ssam_request_spec) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request having neither argument nor return value. The
+ * generated function takes care of setting up the request struct and buffer
+ * allocation, as well as execution of the request itself, returning once the
+ * request has been fully completed. The required transport buffer will be
+ * allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_controller
+ * *ctrl)``, returning the status of the request, which is zero on success and
+ * negative on failure. The ``ctrl`` parameter is the controller via which the
+ * request is being sent.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_N(name, spec...)				\
+	int name(struct ssam_controller *ctrl)					\
+	{									\
+		struct ssam_request_spec s = (struct ssam_request_spec)spec;	\
+		struct ssam_request rqst;					\
+										\
+		rqst.target_category = s.target_category;			\
+		rqst.target_id = s.target_id;					\
+		rqst.command_id = s.command_id;					\
+		rqst.instance_id = s.instance_id;				\
+		rqst.flags = s.flags;						\
+		rqst.length = 0;						\
+		rqst.payload = NULL;						\
+										\
+		return ssam_request_sync_onstack(ctrl, &rqst, NULL, 0);		\
+	}
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_W() - Define synchronous SAM request function with
+ * argument.
+ * @name:  Name of the generated function.
+ * @atype: Type of the request's argument.
+ * @spec:  Specification (&struct ssam_request_spec) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request taking an argument of type @atype and having no
+ * return value. The generated function takes care of setting up the request
+ * struct, buffer allocation, as well as execution of the request itself,
+ * returning once the request has been fully completed. The required transport
+ * buffer will be allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_controller
+ * *ctrl, const atype *arg)``, returning the status of the request, which is
+ * zero on success and negative on failure. The ``ctrl`` parameter is the
+ * controller via which the request is sent. The request argument is specified
+ * via the ``arg`` pointer.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_W(name, atype, spec...)			\
+	int name(struct ssam_controller *ctrl, const atype *arg)		\
+	{									\
+		struct ssam_request_spec s = (struct ssam_request_spec)spec;	\
+		struct ssam_request rqst;					\
+										\
+		rqst.target_category = s.target_category;			\
+		rqst.target_id = s.target_id;					\
+		rqst.command_id = s.command_id;					\
+		rqst.instance_id = s.instance_id;				\
+		rqst.flags = s.flags;						\
+		rqst.length = sizeof(atype);					\
+		rqst.payload = (u8 *)arg;					\
+										\
+		return ssam_request_sync_onstack(ctrl, &rqst, NULL,		\
+						 sizeof(atype));		\
+	}
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_R() - Define synchronous SAM request function with
+ * return value.
+ * @name:  Name of the generated function.
+ * @rtype: Type of the request's return value.
+ * @spec:  Specification (&struct ssam_request_spec) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request taking no argument but having a return value of
+ * type @rtype. The generated function takes care of setting up the request
+ * and response structs, buffer allocation, as well as execution of the
+ * request itself, returning once the request has been fully completed. The
+ * required transport buffer will be allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_controller
+ * *ctrl, rtype *ret)``, returning the status of the request, which is zero on
+ * success and negative on failure. The ``ctrl`` parameter is the controller
+ * via which the request is sent. The request's return value is written to the
+ * memory pointed to by the ``ret`` parameter.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_R(name, rtype, spec...)			\
+	int name(struct ssam_controller *ctrl, rtype *ret)			\
+	{									\
+		struct ssam_request_spec s = (struct ssam_request_spec)spec;	\
+		struct ssam_request rqst;					\
+		struct ssam_response rsp;					\
+		int status;							\
+										\
+		rqst.target_category = s.target_category;			\
+		rqst.target_id = s.target_id;					\
+		rqst.command_id = s.command_id;					\
+		rqst.instance_id = s.instance_id;				\
+		rqst.flags = s.flags | SSAM_REQUEST_HAS_RESPONSE;		\
+		rqst.length = 0;						\
+		rqst.payload = NULL;						\
+										\
+		rsp.capacity = sizeof(rtype);					\
+		rsp.length = 0;							\
+		rsp.pointer = (u8 *)ret;					\
+										\
+		status = ssam_request_sync_onstack(ctrl, &rqst, &rsp, 0);	\
+		if (status)							\
+			return status;						\
+										\
+		if (rsp.length != sizeof(rtype)) {				\
+			struct device *dev = ssam_controller_device(ctrl);	\
+			dev_err(dev, "rqst: invalid response length, expected "	\
+				"%zu, got %zu (tc: 0x%02x, cid: 0x%02x)",	\
+				sizeof(rtype), rsp.length, rqst.target_category,\
+				rqst.command_id);				\
+			return -EIO;						\
+		}								\
+										\
+		return 0;							\
+	}
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_MD_N() - Define synchronous multi-device SAM
+ * request function with neither argument nor return value.
+ * @name: Name of the generated function.
+ * @spec: Specification (&struct ssam_request_spec_md) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request having neither argument nor return value. Device
+ * specifying parameters are not hard-coded, but instead must be provided to
+ * the function. The generated function takes care of setting up the request
+ * struct, buffer allocation, as well as execution of the request itself,
+ * returning once the request has been fully completed. The required transport
+ * buffer will be allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_controller
+ * *ctrl, u8 tid, u8 iid)``, returning the status of the request, which is
+ * zero on success and negative on failure. The ``ctrl`` parameter is the
+ * controller via which the request is sent, ``tid`` the target ID for the
+ * request, and ``iid`` the instance ID.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_MD_N(name, spec...)				\
+	int name(struct ssam_controller *ctrl, u8 tid, u8 iid)			\
+	{									\
+		struct ssam_request_spec_md s					\
+			= (struct ssam_request_spec_md)spec;			\
+		struct ssam_request rqst;					\
+										\
+		rqst.target_category = s.target_category;			\
+		rqst.target_id = tid;						\
+		rqst.command_id = s.command_id;					\
+		rqst.instance_id = iid;						\
+		rqst.flags = s.flags;						\
+		rqst.length = 0;						\
+		rqst.payload = NULL;						\
+										\
+		return ssam_request_sync_onstack(ctrl, &rqst, NULL, 0);		\
+	}
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_MD_W() - Define synchronous multi-device SAM
+ * request function with argument.
+ * @name:  Name of the generated function.
+ * @atype: Type of the request's argument.
+ * @spec:  Specification (&struct ssam_request_spec_md) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request taking an argument of type @atype and having no
+ * return value. Device specifying parameters are not hard-coded, but instead
+ * must be provided to the function. The generated function takes care of
+ * setting up the request struct, buffer allocation, as well as execution of
+ * the request itself, returning once the request has been fully completed.
+ * The required transport buffer will be allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_controller
+ * *ctrl, u8 tid, u8 iid, const atype *arg)``, returning the status of the
+ * request, which is zero on success and negative on failure. The ``ctrl``
+ * parameter is the controller via which the request is sent, ``tid`` the
+ * target ID for the request, and ``iid`` the instance ID. The request argument
+ * is specified via the ``arg`` pointer.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_MD_W(name, atype, spec...)			\
+	int name(struct ssam_controller *ctrl, u8 tid, u8 iid, const atype *arg)\
+	{									\
+		struct ssam_request_spec_md s					\
+			= (struct ssam_request_spec_md)spec;			\
+		struct ssam_request rqst;					\
+										\
+		rqst.target_category = s.target_category;			\
+		rqst.target_id = tid;						\
+		rqst.command_id = s.command_id;					\
+		rqst.instance_id = iid;						\
+		rqst.flags = s.flags;						\
+		rqst.length = sizeof(atype);					\
+		rqst.payload = (u8 *)arg;					\
+										\
+		return ssam_request_sync_onstack(ctrl, &rqst, NULL,		\
+						 sizeof(atype));		\
+	}
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_MD_R() - Define synchronous multi-device SAM
+ * request function with return value.
+ * @name:  Name of the generated function.
+ * @rtype: Type of the request's return value.
+ * @spec:  Specification (&struct ssam_request_spec_md) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request taking no argument but having a return value of
+ * type @rtype. Device specifying parameters are not hard-coded, but instead
+ * must be provided to the function. The generated function takes care of
+ * setting up the request and response structs, buffer allocation, as well as
+ * execution of the request itself, returning once the request has been fully
+ * completed. The required transport buffer will be allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_controller
+ * *ctrl, u8 tid, u8 iid, rtype *ret)``, returning the status of the request,
+ * which is zero on success and negative on failure. The ``ctrl`` parameter is
+ * the controller via which the request is sent, ``tid`` the target ID for the
+ * request, and ``iid`` the instance ID. The request's return value is written
+ * to the memory pointed to by the ``ret`` parameter.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_MD_R(name, rtype, spec...)			\
+	int name(struct ssam_controller *ctrl, u8 tid, u8 iid, rtype *ret)	\
+	{									\
+		struct ssam_request_spec_md s					\
+			= (struct ssam_request_spec_md)spec;			\
+		struct ssam_request rqst;					\
+		struct ssam_response rsp;					\
+		int status;							\
+										\
+		rqst.target_category = s.target_category;			\
+		rqst.target_id = tid;						\
+		rqst.command_id = s.command_id;					\
+		rqst.instance_id = iid;						\
+		rqst.flags = s.flags | SSAM_REQUEST_HAS_RESPONSE;		\
+		rqst.length = 0;						\
+		rqst.payload = NULL;						\
+										\
+		rsp.capacity = sizeof(rtype);					\
+		rsp.length = 0;							\
+		rsp.pointer = (u8 *)ret;					\
+										\
+		status = ssam_request_sync_onstack(ctrl, &rqst, &rsp, 0);	\
+		if (status)							\
+			return status;						\
+										\
+		if (rsp.length != sizeof(rtype)) {				\
+			struct device *dev = ssam_controller_device(ctrl);	\
+			dev_err(dev, "rqst: invalid response length, expected "	\
+				"%zu, got %zu (tc: 0x%02x, cid: 0x%02x)",	\
+				sizeof(rtype), rsp.length, rqst.target_category,\
+				rqst.command_id);				\
+			return -EIO;						\
+		}								\
+										\
+		return 0;							\
+	}
+
+
+/* -- Event notifier/callbacks. --------------------------------------------- */
+
+#define SSAM_NOTIF_STATE_SHIFT		2
+#define SSAM_NOTIF_STATE_MASK		((1 << SSAM_NOTIF_STATE_SHIFT) - 1)
+
+/**
+ * enum ssam_notif_flags - Flags used in return values from SSAM notifier
+ * callback functions.
+ *
+ * @SSAM_NOTIF_HANDLED:
+ *	Indicates that the notification has been handled. This flag should be
+ *	set by the handler if the handler can act/has acted upon the event
+ *	provided to it. This flag should not be set if the handler is not a
+ *	primary handler intended for the provided event.
+ *
+ *	If this flag has not been set by any handler after the notifier chain
+ *	has been traversed, a warning will be emitted, stating that the event
+ *	has not been handled.
+ *
+ * @SSAM_NOTIF_STOP:
+ *	Indicates that the notifier traversal should stop. If this flag is
+ *	returned from a notifier callback, notifier chain traversal will
+ *	immediately stop and any remaining notifiers will not be called. This
+ *	flag is automatically set when ssam_notifier_from_errno() is called
+ *	with a negative error value.
+ */
+enum ssam_notif_flags {
+	SSAM_NOTIF_HANDLED = BIT(0),
+	SSAM_NOTIF_STOP    = BIT(1),
+};
+
+
+struct ssam_event_notifier;
+
+typedef u32 (*ssam_notifier_fn_t)(struct ssam_event_notifier *nf,
+				  const struct ssam_event *event);
+
+/**
+ * struct ssam_notifier_block - Base notifier block for SSAM event
+ * notifications.
+ * @next:     The next notifier block in order of priority.
+ * @fn:       The callback function of this notifier. This function takes the
+ *            respective notifier block and event as input and should return
+ *            a notifier value, which can either be obtained from the flags
+ *            provided in &enum ssam_notif_flags, converted from a standard
+ *            error value via ssam_notifier_from_errno(), or a combination of
+ *            both (e.g. ``ssam_notifier_from_errno(e) | SSAM_NOTIF_HANDLED``).
+ * @priority: Priority value determining the order in which notifier callbacks
+ *            will be called. A higher value means higher priority, i.e. the
+ *            associated callback will be executed earlier than other (lower
+ *            priority) callbacks.
+ */
+struct ssam_notifier_block {
+	struct ssam_notifier_block __rcu *next;
+	ssam_notifier_fn_t fn;
+	int priority;
+};
+
+/**
+ * ssam_notifier_from_errno() - Convert standard error value to notifier
+ * return code.
+ * @err: The error code to convert, must be negative (in case of failure) or
+ *       zero (in case of success).
+ *
+ * Return: Returns the notifier return value obtained by converting the
+ * specified @err value. In case @err is negative, the %SSAM_NOTIF_STOP flag
+ * will be set, causing notifier call chain traversal to abort.
+ */
+static inline u32 ssam_notifier_from_errno(int err)
+{
+	if (WARN_ON(err > 0) || err == 0)
+		return 0;
+	else
+		return ((-err) << SSAM_NOTIF_STATE_SHIFT) | SSAM_NOTIF_STOP;
+}
+
+/**
+ * ssam_notifier_to_errno() - Convert notifier return code to standard error
+ * value.
+ * @ret: The notifier return value to convert.
+ *
+ * Return: Returns the negative error value encoded in @ret or zero if @ret
+ * indicates success.
+ */
+static inline int ssam_notifier_to_errno(u32 ret)
+{
+	return -(ret >> SSAM_NOTIF_STATE_SHIFT);
+}
+
+
+/* -- Event/notification registry. ------------------------------------------ */
+
+/**
+ * struct ssam_event_registry - Registry specification used for enabling events.
+ * @target_category: Target category for the event registry requests.
+ * @target_id:       Target ID for the event registry requests.
+ * @cid_enable:      Command ID for the event-enable request.
+ * @cid_disable:     Command ID for the event-disable request.
+ *
+ * This struct describes a SAM event registry via the minimal collection of
+ * SAM IDs specifying the requests to use for enabling and disabling an event.
+ * The individual event to be enabled/disabled itself is specified via &struct
+ * ssam_event_id.
+ */
+struct ssam_event_registry {
+	u8 target_category;
+	u8 target_id;
+	u8 cid_enable;
+	u8 cid_disable;
+};
+
+/**
+ * struct ssam_event_id - Unique event ID used for enabling events.
+ * @target_category: Target category of the event source.
+ * @instance:        Instance ID of the event source.
+ *
+ * This struct specifies the event to be enabled/disabled via an externally
+ * provided registry. It does not specify the registry to be used itself, this
+ * is done via &struct ssam_event_registry.
+ */
+struct ssam_event_id {
+	u8 target_category;
+	u8 instance;
+};
+
+/**
+ * enum ssam_event_mask - Flags specifying how events are matched to notifiers.
+ *
+ * @SSAM_EVENT_MASK_NONE:
+ *	Run the callback for any event with matching target category. Do not
+ *	do any additional filtering.
+ *
+ * @SSAM_EVENT_MASK_TARGET:
+ *	In addition to filtering by target category, only execute the notifier
+ *	callback for events with a target ID matching to the one of the
+ *	registry used for enabling/disabling the event.
+ *
+ * @SSAM_EVENT_MASK_INSTANCE:
+ *	In addition to filtering by target category, only execute the notifier
+ *	callback for events with an instance ID matching to the instance ID
+ *	used when enabling the event.
+ *
+ * @SSAM_EVENT_MASK_STRICT:
+ *	Do all the filtering above.
+ */
+enum ssam_event_mask {
+	SSAM_EVENT_MASK_TARGET   = BIT(0),
+	SSAM_EVENT_MASK_INSTANCE = BIT(1),
+
+	SSAM_EVENT_MASK_NONE = 0,
+	SSAM_EVENT_MASK_STRICT
+		= SSAM_EVENT_MASK_TARGET
+		| SSAM_EVENT_MASK_INSTANCE,
+};
+
+
+/**
+ * SSAM_EVENT_REGISTRY() - Define a new event registry.
+ * @tc:      Target category for the event registry requests.
+ * @tid:     Target ID for the event registry requests.
+ * @cid_en:  Command ID for the event-enable request.
+ * @cid_dis: Command ID for the event-disable request.
+ *
+ * Return: Returns the &struct ssam_event_registry specified by the given
+ * parameters.
+ */
+#define SSAM_EVENT_REGISTRY(tc, tid, cid_en, cid_dis)	\
+	((struct ssam_event_registry) {			\
+		.target_category = (tc),		\
+		.target_id = (tid),			\
+		.cid_enable = (cid_en),			\
+		.cid_disable = (cid_dis),		\
+	})
+
+#define SSAM_EVENT_REGISTRY_SAM	\
+	SSAM_EVENT_REGISTRY(SSAM_SSH_TC_SAM, 0x01, 0x0b, 0x0c)
+
+#define SSAM_EVENT_REGISTRY_KIP	\
+	SSAM_EVENT_REGISTRY(SSAM_SSH_TC_KIP, 0x02, 0x27, 0x28)
+
+#define SSAM_EVENT_REGISTRY_REG \
+	SSAM_EVENT_REGISTRY(SSAM_SSH_TC_REG, 0x02, 0x01, 0x02)
+
+
+/**
+ * struct ssam_event_notifier - Notifier block for SSAM events.
+ * @base:        The base notifier block with callback function and priority.
+ * @event:       The event for which this block will receive notifications.
+ * @event.reg:   Registry via which the event will be enabled/disabled.
+ * @event.id:    ID specifying the event.
+ * @event.mask:  Flags determining how events are matched to the notifier.
+ * @event.flags: Flags used for enabling the event.
+ */
+struct ssam_event_notifier {
+	struct ssam_notifier_block base;
+
+	struct {
+		struct ssam_event_registry reg;
+		struct ssam_event_id id;
+		enum ssam_event_mask mask;
+		u8 flags;
+	} event;
+};
+
+int ssam_notifier_register(struct ssam_controller *ctrl,
+			   struct ssam_event_notifier *n);
+
+int ssam_notifier_unregister(struct ssam_controller *ctrl,
+			     struct ssam_event_notifier *n);
+
+#endif /* _LINUX_SURFACE_AGGREGATOR_CONTROLLER_H */
diff --git a/include/linux/surface_aggregator/device.h b/include/linux/surface_aggregator/device.h
new file mode 100644
index 000000000000..64b1299d7bd8
--- /dev/null
+++ b/include/linux/surface_aggregator/device.h
@@ -0,0 +1,430 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Surface System Aggregator Module (SSAM) bus and client-device subsystem.
+ *
+ * Main interface for the surface-aggregator bus, surface-aggregator client
+ * devices, and respective drivers building on top of the SSAM controller.
+ * Provides support for non-platform/non-ACPI SSAM clients via dedicated
+ * subsystem.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _LINUX_SURFACE_AGGREGATOR_DEVICE_H
+#define _LINUX_SURFACE_AGGREGATOR_DEVICE_H
+
+#include <linux/device.h>
+#include <linux/mod_devicetable.h>
+#include <linux/types.h>
+
+#include <linux/surface_aggregator/controller.h>
+
+
+/* -- Surface System Aggregator Module bus. --------------------------------- */
+
+/**
+ * enum ssam_device_domain - SAM device domain.
+ * @SSAM_DOMAIN_VIRTUAL:   Virtual device.
+ * @SSAM_DOMAIN_SERIALHUB: Physical dovice connected via Surface Serial Hub.
+ */
+enum ssam_device_domain {
+	SSAM_DOMAIN_VIRTUAL   = 0x00,
+	SSAM_DOMAIN_SERIALHUB = 0x01,
+};
+
+/**
+ * enum ssam_virtual_tc - Target categories for the virtual SAM domain.
+ * @SSAM_VIRTUAL_TC_HUB: Device hub category.
+ */
+enum ssam_virtual_tc {
+	SSAM_VIRTUAL_TC_HUB = 0x00,
+};
+
+/**
+ * struct ssam_device_uid - Unique identifier for SSAM device.
+ * @domain:   Domain of the device.
+ * @category: Target category of the device.
+ * @target:   Target ID of the device.
+ * @instance: Instance ID of the device.
+ * @function: Sub-function of the device. This field can be used to split a
+ *            single SAM device into multiple virtual subdevices to separate
+ *            different functionality of that device and allow one driver per
+ *            such functionality.
+ */
+struct ssam_device_uid {
+	u8 domain;
+	u8 category;
+	u8 target;
+	u8 instance;
+	u8 function;
+};
+
+/*
+ * Special values for device matching.
+ *
+ * These values are intended to be used with SSAM_DEVICE(), SSAM_VDEV(), and
+ * SSAM_SDEV() exclusively. Specifically, they are used to initialize the
+ * match_flags member of the device ID structure. Do not use them directly
+ * with struct ssam_device_id or struct ssam_device_uid.
+ */
+#define SSAM_ANY_TID		0xffff
+#define SSAM_ANY_IID		0xffff
+#define SSAM_ANY_FUN		0xffff
+
+/**
+ * SSAM_DEVICE() - Initialize a &struct ssam_device_id with the given
+ * parameters.
+ * @d:   Domain of the device.
+ * @cat: Target category of the device.
+ * @tid: Target ID of the device.
+ * @iid: Instance ID of the device.
+ * @fun: Sub-function of the device.
+ *
+ * Initializes a &struct ssam_device_id with the given parameters. See &struct
+ * ssam_device_uid for details regarding the parameters. The special values
+ * %SSAM_ANY_TID, %SSAM_ANY_IID, and %SSAM_ANY_FUN can be used to specify that
+ * matching should ignore target ID, instance ID, and/or sub-function,
+ * respectively. This macro initializes the ``match_flags`` field based on the
+ * given parameters.
+ *
+ * Note: The parameters @d and @cat must be valid &u8 values, the parameters
+ * @tid, @iid, and @fun must be either valid &u8 values or %SSAM_ANY_TID,
+ * %SSAM_ANY_IID, or %SSAM_ANY_FUN, respectively. Other non-&u8 values are not
+ * allowed.
+ */
+#define SSAM_DEVICE(d, cat, tid, iid, fun)					\
+	.match_flags = (((tid) != SSAM_ANY_TID) ? SSAM_MATCH_TARGET : 0)	\
+		     | (((iid) != SSAM_ANY_IID) ? SSAM_MATCH_INSTANCE : 0)	\
+		     | (((fun) != SSAM_ANY_FUN) ? SSAM_MATCH_FUNCTION : 0),	\
+	.domain   = d,								\
+	.category = cat,							\
+	.target   = ((tid) != SSAM_ANY_TID) ? (tid) : 0,			\
+	.instance = ((iid) != SSAM_ANY_IID) ? (iid) : 0,			\
+	.function = ((fun) != SSAM_ANY_FUN) ? (fun) : 0				\
+
+/**
+ * SSAM_VDEV() - Initialize a &struct ssam_device_id as virtual device with
+ * the given parameters.
+ * @cat: Target category of the device.
+ * @tid: Target ID of the device.
+ * @iid: Instance ID of the device.
+ * @fun: Sub-function of the device.
+ *
+ * Initializes a &struct ssam_device_id with the given parameters in the
+ * virtual domain. See &struct ssam_device_uid for details regarding the
+ * parameters. The special values %SSAM_ANY_TID, %SSAM_ANY_IID, and
+ * %SSAM_ANY_FUN can be used to specify that matching should ignore target ID,
+ * instance ID, and/or sub-function, respectively. This macro initializes the
+ * ``match_flags`` field based on the given parameters.
+ *
+ * Note: The parameter @cat must be a valid &u8 value, the parameters @tid,
+ * @iid, and @fun must be either valid &u8 values or %SSAM_ANY_TID,
+ * %SSAM_ANY_IID, or %SSAM_ANY_FUN, respectively. Other non-&u8 values are not
+ * allowed.
+ */
+#define SSAM_VDEV(cat, tid, iid, fun) \
+	SSAM_DEVICE(SSAM_DOMAIN_VIRTUAL, SSAM_VIRTUAL_TC_##cat, tid, iid, fun)
+
+
+/**
+ * SSAM_SDEV() - Initialize a &struct ssam_device_id as physical SSH device
+ * with the given parameters.
+ * @cat: Target category of the device.
+ * @tid: Target ID of the device.
+ * @iid: Instance ID of the device.
+ * @fun: Sub-function of the device.
+ *
+ * Initializes a &struct ssam_device_id with the given parameters in the SSH
+ * domain. See &struct ssam_device_uid for details regarding the parameters.
+ * The special values %SSAM_ANY_TID, %SSAM_ANY_IID, and %SSAM_ANY_FUN can be
+ * used to specify that matching should ignore target ID, instance ID, and/or
+ * sub-function, respectively. This macro initializes the ``match_flags``
+ * field based on the given parameters.
+ *
+ * Note: The parameter @cat must be a valid &u8 value, the parameters @tid,
+ * @iid, and @fun must be either valid &u8 values or %SSAM_ANY_TID,
+ * %SSAM_ANY_IID, or %SSAM_ANY_FUN, respectively. Other non-&u8 values are not
+ * allowed.
+ */
+#define SSAM_SDEV(cat, tid, iid, fun) \
+	SSAM_DEVICE(SSAM_DOMAIN_SERIALHUB, SSAM_SSH_TC_##cat, tid, iid, fun)
+
+
+/**
+ * struct ssam_device - SSAM client device.
+ * @dev:  Driver model representation of the device.
+ * @ctrl: SSAM controller managing this device.
+ * @uid:  UID identifying the device.
+ */
+struct ssam_device {
+	struct device dev;
+	struct ssam_controller *ctrl;
+
+	struct ssam_device_uid uid;
+};
+
+/**
+ * struct ssam_device_driver - SSAM client device driver.
+ * @driver:      Base driver model structure.
+ * @match_table: Match table specifying which devices the driver should bind to.
+ * @probe:       Called when the driver is being bound to a device.
+ * @remove:      Called when the driver is being unbound from the device.
+ */
+struct ssam_device_driver {
+	struct device_driver driver;
+
+	const struct ssam_device_id *match_table;
+
+	int  (*probe)(struct ssam_device *sdev);
+	void (*remove)(struct ssam_device *sdev);
+};
+
+extern struct bus_type ssam_bus_type;
+extern const struct device_type ssam_device_type;
+
+
+/**
+ * is_ssam_device() - Check if the given device is a SSAM client device.
+ * @d: The device to test the type of.
+ *
+ * Return: Returns %true iff the specified device is of type &struct
+ * ssam_device, i.e. the device type points to %ssam_device_type, and %false
+ * otherwise.
+ */
+static inline bool is_ssam_device(struct device *d)
+{
+	return d->type == &ssam_device_type;
+}
+
+/**
+ * to_ssam_device() - Casts the given device to a SSAM client device.
+ * @d: The device to cast.
+ *
+ * Casts the given &struct device to a &struct ssam_device. The caller has to
+ * ensure that the given device is actually enclosed in a &struct ssam_device,
+ * e.g. by calling is_ssam_device().
+ *
+ * Return: Returns a pointer to the &struct ssam_device wrapping the given
+ * device @d.
+ */
+static inline struct ssam_device *to_ssam_device(struct device *d)
+{
+	return container_of(d, struct ssam_device, dev);
+}
+
+/**
+ * to_ssam_device_driver() - Casts the given device driver to a SSAM client
+ * device driver.
+ * @d: The driver to cast.
+ *
+ * Casts the given &struct device_driver to a &struct ssam_device_driver. The
+ * caller has to ensure that the given driver is actually enclosed in a
+ * &struct ssam_device_driver.
+ *
+ * Return: Returns the pointer to the &struct ssam_device_driver wrapping the
+ * given device driver @d.
+ */
+static inline
+struct ssam_device_driver *to_ssam_device_driver(struct device_driver *d)
+{
+	return container_of(d, struct ssam_device_driver, driver);
+}
+
+
+const struct ssam_device_id *ssam_device_id_match(
+		const struct ssam_device_id *table,
+		const struct ssam_device_uid uid);
+
+const struct ssam_device_id *ssam_device_get_match(
+		const struct ssam_device *dev);
+
+const void *ssam_device_get_match_data(const struct ssam_device *dev);
+
+struct ssam_device *ssam_device_alloc(struct ssam_controller *ctrl,
+				      struct ssam_device_uid uid);
+
+int ssam_device_add(struct ssam_device *sdev);
+void ssam_device_remove(struct ssam_device *sdev);
+
+/**
+ * ssam_device_get() - Increment reference count of SSAM client device.
+ * @sdev: The device to increment the reference count of.
+ *
+ * Increments the reference count of the given SSAM client device by
+ * incrementing the reference count of the enclosed &struct device via
+ * get_device().
+ *
+ * See ssam_device_put() for the counter-part of this function.
+ *
+ * Return: Returns the device provided as input.
+ */
+static inline struct ssam_device *ssam_device_get(struct ssam_device *sdev)
+{
+	return sdev ? to_ssam_device(get_device(&sdev->dev)) : NULL;
+}
+
+/**
+ * ssam_device_put() - Decrement reference count of SSAM client device.
+ * @sdev: The device to decrement the reference count of.
+ *
+ * Decrements the reference count of the given SSAM client device by
+ * decrementing the reference count of the enclosed &struct device via
+ * put_device().
+ *
+ * See ssam_device_get() for the counter-part of this function.
+ */
+static inline void ssam_device_put(struct ssam_device *sdev)
+{
+	if (sdev)
+		put_device(&sdev->dev);
+}
+
+/**
+ * ssam_device_get_drvdata() - Get driver-data of SSAM client device.
+ * @sdev: The device to get the driver-data from.
+ *
+ * Return: Returns the driver-data of the given device, previously set via
+ * ssam_device_set_drvdata().
+ */
+static inline void *ssam_device_get_drvdata(struct ssam_device *sdev)
+{
+	return dev_get_drvdata(&sdev->dev);
+}
+
+/**
+ * ssam_device_set_drvdata() - Set driver-data of SSAM client device.
+ * @sdev: The device to set the driver-data of.
+ * @data: The data to set the device's driver-data pointer to.
+ */
+static inline void ssam_device_set_drvdata(struct ssam_device *sdev, void *data)
+{
+	dev_set_drvdata(&sdev->dev, data);
+}
+
+
+int __ssam_device_driver_register(struct ssam_device_driver *d, struct module *o);
+void ssam_device_driver_unregister(struct ssam_device_driver *d);
+
+/**
+ * ssam_device_driver_register() - Register a SSAM client device driver.
+ * @drv: The driver to register.
+ */
+#define ssam_device_driver_register(drv) \
+	__ssam_device_driver_register(drv, THIS_MODULE)
+
+/**
+ * module_ssam_device_driver() - Helper macro for SSAM device driver
+ * registration.
+ * @drv: The driver managed by this module.
+ *
+ * Helper macro to register a SSAM device driver via module_init() and
+ * module_exit(). This macro may only be used once per module and replaces
+ * the afforementioned definitions.
+ */
+#define module_ssam_device_driver(drv)			\
+	module_driver(drv, ssam_device_driver_register,	\
+		      ssam_device_driver_unregister)
+
+
+/* -- Helpers for client-device requests. ----------------------------------- */
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_CL_N() - Define synchronous client-device SAM
+ * request function with neither argument nor return value.
+ * @name: Name of the generated function.
+ * @spec: Specification (&struct ssam_request_spec_md) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request having neither argument nor return value. Device
+ * specifying parameters are not hard-coded, but instead are provided via the
+ * client device, specifically its UID, supplied when calling this function.
+ * The generated function takes care of setting up the request struct, buffer
+ * allocation, as well as execution of the request itself, returning once the
+ * request has been fully completed. The required transport buffer will be
+ * allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_device *sdev)``,
+ * returning the status of the request, which is zero on success and negative
+ * on failure. The ``sdev`` parameter specifies both the target device of the
+ * request and by association the controller via which the request is sent.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_CL_N(name, spec...)			\
+	SSAM_DEFINE_SYNC_REQUEST_MD_N(__raw_##name, spec)		\
+	int name(struct ssam_device *sdev)				\
+	{								\
+		return __raw_##name(sdev->ctrl, sdev->uid.target,	\
+				    sdev->uid.instance);		\
+	}
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_CL_W() - Define synchronous client-device SAM
+ * request function with argument.
+ * @name:  Name of the generated function.
+ * @atype: Type of the request's argument.
+ * @spec:  Specification (&struct ssam_request_spec_md) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request taking an argument of type @atype and having no
+ * return value. Device specifying parameters are not hard-coded, but instead
+ * are provided via the client device, specifically its UID, supplied when
+ * calling this function. The generated function takes care of setting up the
+ * request struct, buffer allocation, as well as execution of the request
+ * itself, returning once the request has been fully completed. The required
+ * transport buffer will be allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_device *sdev,
+ * const atype *arg)``, returning the status of the request, which is zero on
+ * success and negative on failure. The ``sdev`` parameter specifies both the
+ * target device of the request and by association the controller via which
+ * the request is sent. The request's argument is specified via the ``arg``
+ * pointer.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_CL_W(name, atype, spec...)		\
+	SSAM_DEFINE_SYNC_REQUEST_MD_W(__raw_##name, atype, spec)	\
+	int name(struct ssam_device *sdev, const atype *arg)		\
+	{								\
+		return __raw_##name(sdev->ctrl, sdev->uid.target,	\
+				    sdev->uid.instance, arg);		\
+	}
+
+/**
+ * SSAM_DEFINE_SYNC_REQUEST_CL_R() - Define synchronous client-device SAM
+ * request function with return value.
+ * @name:  Name of the generated function.
+ * @rtype: Type of the request's return value.
+ * @spec:  Specification (&struct ssam_request_spec_md) defining the request.
+ *
+ * Defines a function executing the synchronous SAM request specified by
+ * @spec, with the request taking no argument but having a return value of
+ * type @rtype. Device specifying parameters are not hard-coded, but instead
+ * are provided via the client device, specifically its UID, supplied when
+ * calling this function. The generated function takes care of setting up the
+ * request struct, buffer allocation, as well as execution of the request
+ * itself, returning once the request has been fully completed. The required
+ * transport buffer will be allocated on the stack.
+ *
+ * The generated function is defined as ``int name(struct ssam_device *sdev,
+ * rtype *ret)``, returning the status of the request, which is zero on
+ * success and negative on failure. The ``sdev`` parameter specifies both the
+ * target device of the request and by association the controller via which
+ * the request is sent. The request's return value is written to the memory
+ * pointed to by the ``ret`` parameter.
+ *
+ * Refer to ssam_request_sync_onstack() for more details on the behavior of
+ * the generated function.
+ */
+#define SSAM_DEFINE_SYNC_REQUEST_CL_R(name, rtype, spec...)		\
+	SSAM_DEFINE_SYNC_REQUEST_MD_R(__raw_##name, rtype, spec)	\
+	int name(struct ssam_device *sdev, rtype *ret)			\
+	{								\
+		return __raw_##name(sdev->ctrl, sdev->uid.target,	\
+				    sdev->uid.instance, ret);		\
+	}
+
+#endif /* _LINUX_SURFACE_AGGREGATOR_DEVICE_H */
diff --git a/include/linux/surface_aggregator/serial_hub.h b/include/linux/surface_aggregator/serial_hub.h
new file mode 100644
index 000000000000..376313f402b2
--- /dev/null
+++ b/include/linux/surface_aggregator/serial_hub.h
@@ -0,0 +1,659 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Surface Serial Hub (SSH) protocol and communication interface.
+ *
+ * Lower-level communication layers and SSH protocol definitions for the
+ * Surface System Aggregator Module (SSAM). Provides the interface for basic
+ * packet- and request-based communication with the SSAM EC via SSH.
+ *
+ * Copyright (C) 2019-2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _LINUX_SURFACE_AGGREGATOR_SERIAL_HUB_H
+#define _LINUX_SURFACE_AGGREGATOR_SERIAL_HUB_H
+
+#include <linux/crc-ccitt.h>
+#include <linux/kref.h>
+#include <linux/ktime.h>
+#include <linux/list.h>
+#include <linux/types.h>
+
+
+/* -- Data structures for SAM-over-SSH communication. ----------------------- */
+
+/**
+ * enum ssh_frame_type - Frame types for SSH frames.
+ *
+ * @SSH_FRAME_TYPE_DATA_SEQ:
+ *	Indicates a data frame, followed by a payload with the length specified
+ *	in the ``struct ssh_frame.len`` field. This frame is sequenced, meaning
+ *	that an ACK is required.
+ *
+ * @SSH_FRAME_TYPE_DATA_NSQ:
+ *	Same as %SSH_FRAME_TYPE_DATA_SEQ, but unsequenced, meaning that the
+ *	message does not have to be ACKed.
+ *
+ * @SSH_FRAME_TYPE_ACK:
+ *	Indicates an ACK message.
+ *
+ * @SSH_FRAME_TYPE_NAK:
+ *	Indicates an error response for previously sent frame. In general, this
+ *	means that the frame and/or payload is malformed, e.g. a CRC is wrong.
+ *	For command-type payloads, this can also mean that the command is
+ *	invalid.
+ */
+enum ssh_frame_type {
+	SSH_FRAME_TYPE_DATA_SEQ = 0x80,
+	SSH_FRAME_TYPE_DATA_NSQ = 0x00,
+	SSH_FRAME_TYPE_ACK	= 0x40,
+	SSH_FRAME_TYPE_NAK	= 0x04,
+};
+
+/**
+ * struct ssh_frame - SSH communication frame.
+ * @type: The type of the frame. See &enum ssh_frame_type.
+ * @len:  The length of the frame payload directly following the CRC for this
+ *        frame. Does not include the final CRC for that payload.
+ * @seq:  The sequence number for this message/exchange.
+ */
+struct ssh_frame {
+	u8 type;
+	__le16 len;
+	u8 seq;
+} __packed;
+
+static_assert(sizeof(struct ssh_frame) == 4);
+
+/*
+ * SSH_FRAME_MAX_PAYLOAD_SIZE - Maximum SSH frame payload length in bytes.
+ *
+ * This is the physical maximum length of the protocol. Implementations may
+ * set a more constrained limit.
+ */
+#define SSH_FRAME_MAX_PAYLOAD_SIZE	U16_MAX
+
+/**
+ * enum ssh_payload_type - Type indicator for the SSH payload.
+ * @SSH_PLD_TYPE_CMD: The payload is a command structure with optional command
+ *                    payload.
+ */
+enum ssh_payload_type {
+	SSH_PLD_TYPE_CMD = 0x80,
+};
+
+/**
+ * struct ssh_command - Payload of a command-type frame.
+ * @type:    The type of the payload. See &enum ssh_payload_type. Should be
+ *           SSH_PLD_TYPE_CMD for this struct.
+ * @tc:      Command target category.
+ * @tid_out: Output target ID. Should be zero if this an incoming (EC to host)
+ *           message.
+ * @tid_in:  Input target ID. Should be zero if this is an outgoing (host to
+ *           EC) message.
+ * @iid:     Instance ID.
+ * @rqid:    Request ID. Used to match requests with responses and differentiate
+ *           between responses and events.
+ * @cid:     Command ID.
+ */
+struct ssh_command {
+	u8 type;
+	u8 tc;
+	u8 tid_out;
+	u8 tid_in;
+	u8 iid;
+	__le16 rqid;
+	u8 cid;
+} __packed;
+
+static_assert(sizeof(struct ssh_command) == 8);
+
+/*
+ * SSH_COMMAND_MAX_PAYLOAD_SIZE - Maximum SSH command payload length in bytes.
+ *
+ * This is the physical maximum length of the protocol. Implementations may
+ * set a more constrained limit.
+ */
+#define SSH_COMMAND_MAX_PAYLOAD_SIZE \
+	(SSH_FRAME_MAX_PAYLOAD_SIZE - sizeof(struct ssh_command))
+
+/*
+ * SSH_MSG_LEN_BASE - Base-length of a SSH message.
+ *
+ * This is the minimum number of bytes required to form a message. The actual
+ * message length is SSH_MSG_LEN_BASE plus the length of the frame payload.
+ */
+#define SSH_MSG_LEN_BASE	(sizeof(struct ssh_frame) + 3ull * sizeof(u16))
+
+/*
+ * SSH_MSG_LEN_CTRL - Length of a SSH control message.
+ *
+ * This is the length of a SSH control message, which is equal to a SSH
+ * message without any payload.
+ */
+#define SSH_MSG_LEN_CTRL	SSH_MSG_LEN_BASE
+
+/**
+ * SSH_MESSAGE_LENGTH() - Comute length of SSH message.
+ * @payload_size: Length of the payload inside the SSH frame.
+ *
+ * Return: Returns the length of a SSH message with payload of specified size.
+ */
+#define SSH_MESSAGE_LENGTH(payload_size) (SSH_MSG_LEN_BASE + payload_size)
+
+/**
+ * SSH_COMMAND_MESSAGE_LENGTH() - Compute length of SSH command message.
+ * @payload_size: Length of the command payload.
+ *
+ * Return: Returns the length of a SSH command message with command payload of
+ * specified size.
+ */
+#define SSH_COMMAND_MESSAGE_LENGTH(payload_size) \
+	SSH_MESSAGE_LENGTH(sizeof(struct ssh_command) + payload_size)
+
+/**
+ * SSH_MSGOFFSET_FRAME() - Compute offset in SSH message to specified field in
+ * frame.
+ * @field: The field for which the offset should be computed.
+ *
+ * Return: Returns the offset of the specified &struct ssh_frame field in the
+ * raw SSH message data as.
+ */
+#define SSH_MSGOFFSET_FRAME(field) \
+	(sizeof(u16) + offsetof(struct ssh_frame, field))
+
+/**
+ * SSH_MSGOFFSET_COMMAND() - Compute offset in SSH message to specified field
+ * in command.
+ * @field: The field for which the offset should be computed.
+ *
+ * Return: Returns the offset of the specified &struct ssh_command field in
+ * the raw SSH message data.
+ */
+#define SSH_MSGOFFSET_COMMAND(field) \
+	(2ull * sizeof(u16) + sizeof(struct ssh_frame) \
+		+ offsetof(struct ssh_command, field))
+
+/*
+ * SSH_MSG_SYN - SSH message synchronization (SYN) bytes as u16.
+ */
+#define SSH_MSG_SYN		((u16)0x55aa)
+
+/**
+ * ssh_crc() - Compute CRC for SSH messages.
+ * @buf: The pointer pointing to the data for which the CRC should be computed.
+ * @len: The length of the data for which the CRC should be computed.
+ *
+ * Return: Returns the CRC computed on the provided data, as used for SSH
+ * messages.
+ */
+static inline u16 ssh_crc(const u8 *buf, size_t len)
+{
+	return crc_ccitt_false(0xffff, buf, len);
+}
+
+/*
+ * SSH_NUM_EVENTS - The number of reserved event IDs.
+ *
+ * The number of reserved event IDs, used for registering an SSH event
+ * handler. Valid event IDs are numbers below or equal to this value, with
+ * exception of zero, which is not an event ID. Thus, this is also the
+ * absolute maximum number of event handlers that can be registered.
+ */
+#define SSH_NUM_EVENTS		34
+
+/*
+ * SSH_NUM_TARGETS - The number of communication targets used in the protocol.
+ */
+#define SSH_NUM_TARGETS		2
+
+/**
+ * ssh_rqid_next_valid() - Return the next valid request ID.
+ * @rqid: The current request ID.
+ *
+ * Return: Returns the next valid request ID, following the current request ID
+ * provided to this function. This function skips any request IDs reserved for
+ * events.
+ */
+static inline u16 ssh_rqid_next_valid(u16 rqid)
+{
+	return rqid > 0 ? rqid + 1u : rqid + SSH_NUM_EVENTS + 1u;
+}
+
+/**
+ * ssh_rqid_to_event() - Convert request ID to its corresponding event ID.
+ * @rqid: The request ID to convert.
+ */
+static inline u16 ssh_rqid_to_event(u16 rqid)
+{
+	return rqid - 1u;
+}
+
+/**
+ * ssh_rqid_is_event() - Check if given request ID is a valid event ID.
+ * @rqid: The request ID to check.
+ */
+static inline bool ssh_rqid_is_event(u16 rqid)
+{
+	return ssh_rqid_to_event(rqid) < SSH_NUM_EVENTS;
+}
+
+/**
+ * ssh_tc_to_rqid() - Convert target category to its corresponding request ID.
+ * @tc: The target category to convert.
+ */
+static inline u16 ssh_tc_to_rqid(u8 tc)
+{
+	return tc;
+}
+
+/**
+ * ssh_tid_to_index() - Convert target ID to its corresponding target index.
+ * @tid: The target ID to convert.
+ */
+static inline u8 ssh_tid_to_index(u8 tid)
+{
+	return tid - 1u;
+}
+
+/**
+ * ssh_tid_is_valid() - Check if target ID is valid/supported.
+ * @tid: The target ID to check.
+ */
+static inline bool ssh_tid_is_valid(u8 tid)
+{
+	return ssh_tid_to_index(tid) < SSH_NUM_TARGETS;
+}
+
+/**
+ * struct ssam_span - Reference to a buffer region.
+ * @ptr: Pointer to the buffer region.
+ * @len: Length of the buffer region.
+ *
+ * A reference to a (non-owned) buffer segment, consisting of pointer and
+ * length. Use of this struct indicates non-owned data, i.e. data of which the
+ * life-time is managed (i.e. it is allocated/freed) via another pointer.
+ */
+struct ssam_span {
+	u8    *ptr;
+	size_t len;
+};
+
+enum ssam_ssh_tc {
+	/* Known SSH/EC target categories. */
+				// category 0x00 is invalid for EC use
+	SSAM_SSH_TC_SAM = 0x01,	// generic system functionality, real-time clock
+	SSAM_SSH_TC_BAT = 0x02,	// battery/power subsystem
+	SSAM_SSH_TC_TMP = 0x03,	// thermal subsystem
+	SSAM_SSH_TC_PMC = 0x04,
+	SSAM_SSH_TC_FAN = 0x05,
+	SSAM_SSH_TC_PoM = 0x06,
+	SSAM_SSH_TC_DBG = 0x07,
+	SSAM_SSH_TC_KBD = 0x08,	// legacy keyboard (Laptop 1/2)
+	SSAM_SSH_TC_FWU = 0x09,
+	SSAM_SSH_TC_UNI = 0x0a,
+	SSAM_SSH_TC_LPC = 0x0b,
+	SSAM_SSH_TC_TCL = 0x0c,
+	SSAM_SSH_TC_SFL = 0x0d,
+	SSAM_SSH_TC_KIP = 0x0e,
+	SSAM_SSH_TC_EXT = 0x0f,
+	SSAM_SSH_TC_BLD = 0x10,
+	SSAM_SSH_TC_BAS = 0x11,	// detachment system (Surface Book 2/3)
+	SSAM_SSH_TC_SEN = 0x12,
+	SSAM_SSH_TC_SRQ = 0x13,
+	SSAM_SSH_TC_MCU = 0x14,
+	SSAM_SSH_TC_HID = 0x15,	// generic HID input subsystem
+	SSAM_SSH_TC_TCH = 0x16,
+	SSAM_SSH_TC_BKL = 0x17,
+	SSAM_SSH_TC_TAM = 0x18,
+	SSAM_SSH_TC_ACC = 0x19,
+	SSAM_SSH_TC_UFI = 0x1a,
+	SSAM_SSH_TC_USC = 0x1b,
+	SSAM_SSH_TC_PEN = 0x1c,
+	SSAM_SSH_TC_VID = 0x1d,
+	SSAM_SSH_TC_AUD = 0x1e,
+	SSAM_SSH_TC_SMC = 0x1f,
+	SSAM_SSH_TC_KPD = 0x20,
+	SSAM_SSH_TC_REG = 0x21,
+};
+
+
+/* -- Packet transport layer (ptl). ----------------------------------------- */
+
+/**
+ * enum ssh_packet_base_priority - Base priorities for &struct ssh_packet.
+ * @SSH_PACKET_PRIORITY_FLUSH: Base priority for flush packets.
+ * @SSH_PACKET_PRIORITY_DATA:  Base priority for normal data paackets.
+ * @SSH_PACKET_PRIORITY_NAK:   Base priority for NAK packets.
+ * @SSH_PACKET_PRIORITY_ACK:   Base priority for ACK packets.
+ */
+enum ssh_packet_base_priority {
+	SSH_PACKET_PRIORITY_FLUSH = 0,	/* same as DATA to sequence flush */
+	SSH_PACKET_PRIORITY_DATA  = 0,
+	SSH_PACKET_PRIORITY_NAK   = 1,
+	SSH_PACKET_PRIORITY_ACK   = 2,
+};
+
+/*
+ * Same as SSH_PACKET_PRIORITY() below, only with actual values.
+ */
+#define __SSH_PACKET_PRIORITY(base, try) \
+	(((base) << 4) | ((try) & 0x0f))
+
+/**
+ * SSH_PACKET_PRIORITY() - Compute packet priority from base priority and
+ * number of tries.
+ * @base: The base priority as suffix of &enum ssh_packet_base_priority, e.g.
+ *        ``FLUSH``, ``DATA``, ``ACK``, or ``NAK``.
+ * @try:  The number of tries (must be less than 16).
+ *
+ * Compute the combined packet priority. The combined priority is dominated by
+ * the base priority, whereas the number of (re-)tries decides the precedence
+ * of packets with the same base priority, giving higher priority to packets
+ * that already have more tries.
+ *
+ * Return: Returns the computed priority as value fitting inside a &u8. A
+ * higher number means a higher priority.
+ */
+#define SSH_PACKET_PRIORITY(base, try) \
+	__SSH_PACKET_PRIORITY(SSH_PACKET_PRIORITY_##base, (try))
+
+/**
+ * ssh_packet_priority_get_try() - Get number of tries from packet priority.
+ * @priority: The packet priority.
+ *
+ * Return: Returns the number of tries encoded in the specified packet
+ * priority.
+ */
+static inline u8 ssh_packet_priority_get_try(u8 priority)
+{
+	return priority & 0x0f;
+}
+
+/**
+ * ssh_packet_priority_get_base - Get base priority from packet priority.
+ * @priority: The packet priority.
+ *
+ * Return: Returns the base priority encoded in the given packet priority.
+ */
+static inline u8 ssh_packet_priority_get_base(u8 priority)
+{
+	return (priority & 0xf0) >> 4;
+}
+
+
+enum ssh_packet_flags {
+	/* state flags */
+	SSH_PACKET_SF_LOCKED_BIT,
+	SSH_PACKET_SF_QUEUED_BIT,
+	SSH_PACKET_SF_PENDING_BIT,
+	SSH_PACKET_SF_TRANSMITTING_BIT,
+	SSH_PACKET_SF_TRANSMITTED_BIT,
+	SSH_PACKET_SF_ACKED_BIT,
+	SSH_PACKET_SF_CANCELED_BIT,
+	SSH_PACKET_SF_COMPLETED_BIT,
+
+	/* type flags */
+	SSH_PACKET_TY_FLUSH_BIT,
+	SSH_PACKET_TY_SEQUENCED_BIT,
+	SSH_PACKET_TY_BLOCKING_BIT,
+
+	/* mask for state flags */
+	SSH_PACKET_FLAGS_SF_MASK =
+		  BIT(SSH_PACKET_SF_LOCKED_BIT)
+		| BIT(SSH_PACKET_SF_QUEUED_BIT)
+		| BIT(SSH_PACKET_SF_PENDING_BIT)
+		| BIT(SSH_PACKET_SF_TRANSMITTING_BIT)
+		| BIT(SSH_PACKET_SF_TRANSMITTED_BIT)
+		| BIT(SSH_PACKET_SF_ACKED_BIT)
+		| BIT(SSH_PACKET_SF_CANCELED_BIT)
+		| BIT(SSH_PACKET_SF_COMPLETED_BIT),
+
+	/* mask for type flags */
+	SSH_PACKET_FLAGS_TY_MASK =
+		  BIT(SSH_PACKET_TY_FLUSH_BIT)
+		| BIT(SSH_PACKET_TY_SEQUENCED_BIT)
+		| BIT(SSH_PACKET_TY_BLOCKING_BIT),
+};
+
+
+struct ssh_ptl;
+struct ssh_packet;
+
+/**
+ * struct ssh_packet_ops - Callback operations for a SSH packet.
+ * @release:  Function called when the packet reference count reaches zero.
+ *            This callback must be relied upon to ensure that the packet has
+ *            left the transport system(s).
+ * @complete: Function called when the packet is completed, either with
+ *            success or failure. In case of failure, the reason for the
+ *            failure is indicated by the value of the provided status code
+ *            argument. This value will be zero in case of success. Note that
+ *            a call to this callback does not guarantee that the packet is
+ *            not in use by the transport system any more.
+ */
+struct ssh_packet_ops {
+	void (*release)(struct ssh_packet *p);
+	void (*complete)(struct ssh_packet *p, int status);
+};
+
+/**
+ * struct ssh_packet - SSH transport packet.
+ * @ptl:      Pointer to the packet transport layer. May be %NULL if the packet
+ *            (or enclosing request) has not been submitted yet.
+ * @refcnt:   Reference count of the packet.
+ * @priority: Priority of the packet. Must be computed via
+ *            SSH_PACKET_PRIORITY().
+ * @data:     Raw message data.
+ * @data.len: Length of the raw message data.
+ * @data.ptr: Pointer to the raw message data buffer.
+ * @state:    State and type flags describing current packet state (dynamic)
+ *            and type (static). See &enum ssh_packet_flags for possible
+ *            options.
+ * @timestamp: Timestamp specifying when the latest transmission of a
+ *            currently pending packet has been started. May be %KTIME_MAX
+ *            before or in-between transmission attempts. Used for the packet
+ *            timeout implementation.
+ * @queue_node:	The list node for the packet queue.
+ * @pending_node: The list node for the set of pending packets.
+ * @ops:      Packet operations.
+ */
+struct ssh_packet {
+	struct ssh_ptl *ptl;
+	struct kref refcnt;
+
+	u8 priority;
+
+	struct {
+		size_t len;
+		u8 *ptr;
+	} data;
+
+	unsigned long state;
+	ktime_t timestamp;
+
+	struct list_head queue_node;
+	struct list_head pending_node;
+
+	const struct ssh_packet_ops *ops;
+};
+
+struct ssh_packet *ssh_packet_get(struct ssh_packet *p);
+void ssh_packet_put(struct ssh_packet *p);
+
+/**
+ * ssh_packet_set_data() - Set raw message data of packet.
+ * @p:   The packet for which the message data should be set.
+ * @ptr: Pointer to the memory holding the message data.
+ * @len: Length of the message data.
+ *
+ * Sets the raw message data buffer of the packet to the provided memory. The
+ * memory is not copied. Instead, the caller is responsible for management
+ * (i.e. allocation and deallocation) of the memory. The caller must ensure
+ * that the provided memory is valid and contains a valid SSH message,
+ * starting from the time of submission of the packet until the ``release``
+ * callback has been called. During this time, the memory may not be altered
+ * in any way.
+ */
+static inline void ssh_packet_set_data(struct ssh_packet *p, u8 *ptr, size_t len)
+{
+	p->data.ptr = ptr;
+	p->data.len = len;
+}
+
+
+/* -- Request transport layer (rtl). ---------------------------------------- */
+
+enum ssh_request_flags {
+	/* state flags */
+	SSH_REQUEST_SF_LOCKED_BIT,
+	SSH_REQUEST_SF_QUEUED_BIT,
+	SSH_REQUEST_SF_PENDING_BIT,
+	SSH_REQUEST_SF_TRANSMITTING_BIT,
+	SSH_REQUEST_SF_TRANSMITTED_BIT,
+	SSH_REQUEST_SF_RSPRCVD_BIT,
+	SSH_REQUEST_SF_CANCELED_BIT,
+	SSH_REQUEST_SF_COMPLETED_BIT,
+
+	/* type flags */
+	SSH_REQUEST_TY_FLUSH_BIT,
+	SSH_REQUEST_TY_HAS_RESPONSE_BIT,
+
+	/* mask for state flags */
+	SSH_REQUEST_FLAGS_SF_MASK =
+		  BIT(SSH_REQUEST_SF_LOCKED_BIT)
+		| BIT(SSH_REQUEST_SF_QUEUED_BIT)
+		| BIT(SSH_REQUEST_SF_PENDING_BIT)
+		| BIT(SSH_REQUEST_SF_TRANSMITTING_BIT)
+		| BIT(SSH_REQUEST_SF_TRANSMITTED_BIT)
+		| BIT(SSH_REQUEST_SF_RSPRCVD_BIT)
+		| BIT(SSH_REQUEST_SF_CANCELED_BIT)
+		| BIT(SSH_REQUEST_SF_COMPLETED_BIT),
+
+	/* mask for type flags */
+	SSH_REQUEST_FLAGS_TY_MASK =
+		  BIT(SSH_REQUEST_TY_FLUSH_BIT)
+		| BIT(SSH_REQUEST_TY_HAS_RESPONSE_BIT),
+};
+
+
+struct ssh_rtl;
+struct ssh_request;
+
+/**
+ * struct ssh_request_ops - Callback operations for a SSH request.
+ * @release:  Function called when the request's reference count reaches zero.
+ *            This callback must be relied upon to ensure that the request has
+ *            left the transport systems (both, packet an request systems).
+ * @complete: Function called when the request is completed, either with
+ *            success or failure. The command data for the request response
+ *            is provided via the &struct ssh_command parameter (``cmd``),
+ *            the command payload of the request response via the &struct
+ *            ssh_span parameter (``data``).
+ *
+ *            If the request does not have any response or has not been
+ *            completed with success, both ``cmd`` and ``data`` parameters will
+ *            be NULL. If the request response does not have any command
+ *            payload, the ``data`` span will be an empty (zero-length) span.
+ *
+ *            In case of failure, the reason for the failure is indicated by
+ *            the value of the provided status code argument (``status``). This
+ *            value will be zero in case of success.
+ *
+ *            Note that a call to this callback does not guarantee that the
+ *            request is not in use by the transport systems any more.
+ */
+struct ssh_request_ops {
+	void (*release)(struct ssh_request *rqst);
+	void (*complete)(struct ssh_request *rqst,
+			 const struct ssh_command *cmd,
+			 const struct ssam_span *data, int status);
+};
+
+/**
+ * struct ssh_request - SSH transport request.
+ * @packet: The underlying SSH transport packet.
+ * @node:   List node for the request queue and pending set.
+ * @state:  State and type flags describing current request state (dynamic)
+ *          and type (static). See &enum ssh_request_flags for possible
+ *          options.
+ * @timestamp: Timestamp specifying when we start waiting on the respnse of the
+ *          request. This is set once the underlying packet has been completed
+ *          and may be %KTIME_MAX before that, or when the request does not
+ *          expect a response. Used for the request timeout implementation.
+ * @ops:    Request Operations.
+ */
+struct ssh_request {
+	struct ssh_packet packet;
+	struct list_head node;
+
+	unsigned long state;
+	ktime_t timestamp;
+
+	const struct ssh_request_ops *ops;
+};
+
+/**
+ * to_ssh_request() - Cast a SSH packet to its enclosing SSH request.
+ * @p: The packet to cast.
+ *
+ * Casts the given &struct ssh_packet to its enclosing &struct ssh_request.
+ * The caller is responsible for making sure that the packet is actually
+ * wrapped in a &struct ssh_request.
+ *
+ * Return: Returns the &struct ssh_request wrapping the provided packet.
+ */
+static inline struct ssh_request *to_ssh_request(struct ssh_packet *p)
+{
+	return container_of(p, struct ssh_request, packet);
+}
+
+/**
+ * ssh_request_get() - Increment reference count of request.
+ * @r: The request to increment the reference count of.
+ *
+ * Increments the reference count of the given request by incrementing the
+ * reference count of the underlying &struct ssh_packet, enclosed in it.
+ *
+ * See also ssh_request_put(), ssh_packet_get().
+ *
+ * Return: Returns the request provided as input.
+ */
+static inline struct ssh_request *ssh_request_get(struct ssh_request *r)
+{
+	return r ? to_ssh_request(ssh_packet_get(&r->packet)) : NULL;
+}
+
+/**
+ * ssh_request_put() - Decrement reference count of request.
+ * @r: The request to decrement the reference count of.
+ *
+ * Decrements the reference count of the given request by decrementing the
+ * reference count of the underlying &struct ssh_packet, enclosed in it. If
+ * the reference count reaches zero, the ``release`` callback specified in the
+ * request's &struct ssh_request_ops, i.e. ``r->ops->release``, will be
+ * called.
+ *
+ * See also ssh_request_get(), ssh_packet_put().
+ */
+static inline void ssh_request_put(struct ssh_request *r)
+{
+	if (r)
+		ssh_packet_put(&r->packet);
+}
+
+/**
+ * ssh_request_set_data() - Set raw message data of request.
+ * @r:   The request for which the message data should be set.
+ * @ptr: Pointer to the memory holding the message data.
+ * @len: Length of the message data.
+ *
+ * Sets the raw message data buffer of the underlying packet to the specified
+ * buffer. Does not copy the actual message data, just sets the buffer pointer
+ * and length. Refer to ssh_packet_set_data() for more details.
+ */
+static inline void ssh_request_set_data(struct ssh_request *r, u8 *ptr, size_t len)
+{
+	ssh_packet_set_data(&r->packet, ptr, len);
+}
+
+#endif /* _LINUX_SURFACE_AGGREGATOR_SERIAL_HUB_H */
diff --git a/include/uapi/linux/surface_aggregator/cdev.h b/include/uapi/linux/surface_aggregator/cdev.h
new file mode 100644
index 000000000000..1a8bc0249f8e
--- /dev/null
+++ b/include/uapi/linux/surface_aggregator/cdev.h
@@ -0,0 +1,58 @@
+/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
+/*
+ * Surface System Aggregator Module (SSAM) user-space EC interface.
+ *
+ * Definitions, structs, and IOCTLs for the /dev/surface/aggregator misc
+ * device. This device provides direct user-space access to the SSAM EC.
+ * Intended for debugging and development.
+ *
+ * Copyright (C) 2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _UAPI_LINUX_SURFACE_AGGREGATOR_CDEV_H
+#define _UAPI_LINUX_SURFACE_AGGREGATOR_CDEV_H
+
+#include <linux/ioctl.h>
+#include <linux/types.h>
+
+/**
+ * struct ssam_cdev_request - Controller request IOCTL argument.
+ * @target_category: Target category of the SAM request.
+ * @target_id:       Target ID of the SAM request.
+ * @command_id:      Command ID of the SAM request.
+ * @instance_id:     Instance ID of the SAM request.
+ * @flags:           SAM Request flags.
+ * @status:          Request status (output).
+ * @payload:         Request payload (input data).
+ * @payload.data:    Pointer to request payload data.
+ * @payload.length:  Length of request payload data (in bytes).
+ * @response:        Request response (output data).
+ * @response.data:   Pointer to response buffer.
+ * @response.length: On input: Capacity of response buffer (in bytes).
+ *                   On output: Length of request response (number of bytes
+ *                   in the buffer that are actually used).
+ */
+struct ssam_cdev_request {
+	__u8 target_category;
+	__u8 target_id;
+	__u8 command_id;
+	__u8 instance_id;
+	__u16 flags;
+	__s16 status;
+
+	struct {
+		__u64 data;
+		__u16 length;
+		__u8 __pad[6];
+	} payload;
+
+	struct {
+		__u64 data;
+		__u16 length;
+		__u8 __pad[6];
+	} response;
+} __attribute__((__packed__));
+
+#define SSAM_CDEV_REQUEST	_IOWR(0xA5, 1, struct ssam_cdev_request)
+
+#endif /* _UAPI_LINUX_SURFACE_AGGREGATOR_CDEV_H */
diff --git a/include/uapi/linux/surface_aggregator/dtx.h b/include/uapi/linux/surface_aggregator/dtx.h
new file mode 100644
index 000000000000..fb6c49e254af
--- /dev/null
+++ b/include/uapi/linux/surface_aggregator/dtx.h
@@ -0,0 +1,150 @@
+/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
+/*
+ * Surface DTX (clipboard detachment system driver) user-space interface.
+ *
+ * Definitions, structs, and IOCTLs for the /dev/surface/dtx misc device. This
+ * device allows user-space to control the clipboard detachment process on
+ * Surface Book series devices.
+ *
+ * Copyright (C) 2020 Maximilian Luz <luzmaximilian@gmail.com>
+ */
+
+#ifndef _UAPI_LINUX_SURFACE_AGGREGATOR_DTX_H
+#define _UAPI_LINUX_SURFACE_AGGREGATOR_DTX_H
+
+#include <linux/ioctl.h>
+#include <linux/types.h>
+
+
+/* Status/error categories */
+#define SDTX_CATEGORY_STATUS		0x0000
+#define SDTX_CATEGORY_RUNTIME_ERROR	0x1000
+#define SDTX_CATEGORY_HARDWARE_ERROR	0x2000
+#define SDTX_CATEGORY_UNKNOWN		0xf000
+
+#define SDTX_CATEGORY_MASK		0xf000
+#define SDTX_CATEGORY(value)		((value) & SDTX_CATEGORY_MASK)
+
+#define SDTX_STATUS(code)		((code) | SDTX_CATEGORY_STATUS)
+#define SDTX_ERR_RT(code)		((code) | SDTX_CATEGORY_RUNTIME_ERROR)
+#define SDTX_ERR_HW(code)		((code) | SDTX_CATEGORY_HARDWARE_ERROR)
+#define SDTX_UNKNOWN(code)		((code) | SDTX_CATEGORY_UNKNOWN)
+
+#define SDTX_SUCCESS(value)	(SDTX_CATEGORY(value) == SDTX_CATEGORY_STATUS)
+
+/* Latch status values */
+#define SDTX_LATCH_CLOSED		SDTX_STATUS(0x00)
+#define SDTX_LATCH_OPENED		SDTX_STATUS(0x01)
+
+/* Base state values */
+#define SDTX_BASE_DETACHED		SDTX_STATUS(0x00)
+#define SDTX_BASE_ATTACHED		SDTX_STATUS(0x01)
+
+/* Runtime errors (non-critical) */
+#define SDTX_DETACH_NOT_FEASIBLE	SDTX_ERR_RT(0x01)
+#define SDTX_DETACH_TIMEDOUT		SDTX_ERR_RT(0x02)
+
+/* Hardware errors (critical) */
+#define SDTX_ERR_FAILED_TO_OPEN		SDTX_ERR_HW(0x01)
+#define SDTX_ERR_FAILED_TO_REMAIN_OPEN	SDTX_ERR_HW(0x02)
+#define SDTX_ERR_FAILED_TO_CLOSE	SDTX_ERR_HW(0x03)
+
+
+/* Base types */
+#define SDTX_DEVICE_TYPE_HID		0x0100
+#define SDTX_DEVICE_TYPE_SSH		0x0200
+
+#define SDTX_DEVICE_TYPE_MASK		0x0f00
+#define SDTX_DEVICE_TYPE(value)		((value) & SDTX_DEVICE_TYPE_MASK)
+
+#define SDTX_BASE_TYPE_HID(id)		((id) | SDTX_DEVICE_TYPE_HID)
+#define SDTX_BASE_TYPE_SSH(id)		((id) | SDTX_DEVICE_TYPE_SSH)
+
+
+/**
+ * enum sdtx_device_mode - Mode describing how (and if) the clipboard is
+ * attached to the base of the device.
+ * @SDTX_DEVICE_MODE_TABLET: The clipboard is detached from the base and the
+ *                           device operates as tablet.
+ * @SDTX_DEVICE_MODE_LAPTOP: The clipboard is attached normally to the base
+ *                           and the device operates as laptop.
+ * @SDTX_DEVICE_MODE_STUDIO: The clipboard is attached to the base in reverse.
+ *                           The device operates as tablet with keyboard and
+ *                           touchpad deactivated, however, the base battery
+ *                           and, if present in the specific device model, dGPU
+ *                           are available to the system.
+ */
+enum sdtx_device_mode {
+	SDTX_DEVICE_MODE_TABLET		= 0x00,
+	SDTX_DEVICE_MODE_LAPTOP		= 0x01,
+	SDTX_DEVICE_MODE_STUDIO		= 0x02,
+};
+
+/**
+ * struct sdtx_event - Event provided by reading from the DTX device file.
+ * @length: Length of the event payload, in bytes.
+ * @code:   Event code, detailing what type of event this is.
+ * @data:   Payload of the event, containing @length bytes.
+ *
+ * See &enum sdtx_event_code for currently valid event codes.
+ */
+struct sdtx_event {
+	__u16 length;
+	__u16 code;
+	__u8 data[];
+} __attribute__((__packed__));
+
+/**
+ * enum sdtx_event_code - Code describing the type of an event.
+ * @SDTX_EVENT_REQUEST:         Detachment request event type.
+ * @SDTX_EVENT_CANCEL:          Cancel detachment process event type.
+ * @SDTX_EVENT_BASE_CONNECTION: Base/clipboard connection change event type.
+ * @SDTX_EVENT_LATCH_STATUS:    Latch status change event type.
+ * @SDTX_EVENT_DEVICE_MODE:     Device mode change event type.
+ *
+ * Used in @struct sdtx_event to describe the type of the event. Further event
+ * codes are reserved for future use. Any event parser should be able to
+ * gracefully handle unknown events, i.e. by simply skipping them.
+ *
+ * Consult the DTX user-space interface documentation for details regarding
+ * the individual event types.
+ */
+enum sdtx_event_code {
+	SDTX_EVENT_REQUEST		= 1,
+	SDTX_EVENT_CANCEL		= 2,
+	SDTX_EVENT_BASE_CONNECTION	= 3,
+	SDTX_EVENT_LATCH_STATUS		= 4,
+	SDTX_EVENT_DEVICE_MODE		= 5,
+};
+
+/**
+ * struct sdtx_base_info - Describes if and what type of base is connected.
+ * @state:   The state of the connection. Valid values are %SDTX_BASE_DETACHED,
+ *           %SDTX_BASE_ATTACHED, and %SDTX_DETACH_NOT_FEASIBLE (in case a base
+ *           is attached but low clipboard battery prevents detachment). Other
+ *           values are currently reserved.
+ * @base_id: The type of base connected. Zero if no base is connected.
+ */
+struct sdtx_base_info {
+	__u16 state;
+	__u16 base_id;
+} __attribute__((__packed__));
+
+
+/* IOCTLs */
+#define SDTX_IOCTL_EVENTS_ENABLE	_IO(0xa5, 0x21)
+#define SDTX_IOCTL_EVENTS_DISABLE	_IO(0xa5, 0x22)
+
+#define SDTX_IOCTL_LATCH_LOCK		_IO(0xa5, 0x23)
+#define SDTX_IOCTL_LATCH_UNLOCK		_IO(0xa5, 0x24)
+
+#define SDTX_IOCTL_LATCH_REQUEST	_IO(0xa5, 0x25)
+#define SDTX_IOCTL_LATCH_CONFIRM	_IO(0xa5, 0x26)
+#define SDTX_IOCTL_LATCH_HEARTBEAT	_IO(0xa5, 0x27)
+#define SDTX_IOCTL_LATCH_CANCEL		_IO(0xa5, 0x28)
+
+#define SDTX_IOCTL_GET_BASE_INFO	_IOR(0xa5, 0x29, struct sdtx_base_info)
+#define SDTX_IOCTL_GET_DEVICE_MODE	_IOR(0xa5, 0x2a, __u16)
+#define SDTX_IOCTL_GET_LATCH_STATUS	_IOR(0xa5, 0x2b, __u16)
+
+#endif /* _UAPI_LINUX_SURFACE_AGGREGATOR_DTX_H */
diff --git a/scripts/mod/devicetable-offsets.c b/scripts/mod/devicetable-offsets.c
index 35b9899b0b0f..6b761c5b3ddd 100644
--- a/scripts/mod/devicetable-offsets.c
+++ b/scripts/mod/devicetable-offsets.c
@@ -243,8 +243,9 @@ int main(void)
 
 	DEVID(ssam_device_id);
 	DEVID_FIELD(ssam_device_id, match_flags);
+	DEVID_FIELD(ssam_device_id, domain);
 	DEVID_FIELD(ssam_device_id, category);
-	DEVID_FIELD(ssam_device_id, channel);
+	DEVID_FIELD(ssam_device_id, target);
 	DEVID_FIELD(ssam_device_id, instance);
 	DEVID_FIELD(ssam_device_id, function);
 
diff --git a/scripts/mod/file2alias.c b/scripts/mod/file2alias.c
index 079672e0747a..23d9d368fde9 100644
--- a/scripts/mod/file2alias.c
+++ b/scripts/mod/file2alias.c
@@ -1364,20 +1364,22 @@ static int do_mhi_entry(const char *filename, void *symval, char *alias)
 	return 1;
 }
 
-/* Looks like: ssam:cNtNiNfN
+/*
+ * Looks like: ssam:dNcNtNiNfN
  *
  * N is exactly 2 digits, where each is an upper-case hex digit.
  */
 static int do_ssam_entry(const char *filename, void *symval, char *alias)
 {
 	DEF_FIELD(symval, ssam_device_id, match_flags);
+	DEF_FIELD(symval, ssam_device_id, domain);
 	DEF_FIELD(symval, ssam_device_id, category);
-	DEF_FIELD(symval, ssam_device_id, channel);
+	DEF_FIELD(symval, ssam_device_id, target);
 	DEF_FIELD(symval, ssam_device_id, instance);
 	DEF_FIELD(symval, ssam_device_id, function);
 
-	sprintf(alias, "ssam:c%02X", category);
-	ADD(alias, "t", match_flags & SSAM_MATCH_CHANNEL, channel);
+	sprintf(alias, "ssam:d%02Xc%02X", domain, category);
+	ADD(alias, "t", match_flags & SSAM_MATCH_TARGET, target);
 	ADD(alias, "i", match_flags & SSAM_MATCH_INSTANCE, instance);
 	ADD(alias, "f", match_flags & SSAM_MATCH_FUNCTION, function);
 
-- 
2.29.2