0ct0pu5/ladybird
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 2

Kernel: Replace bespoke & rickety NVMeIO synchronization mechanism

Browse source 
Instead of assuming data races won't occur and trying to somehow verify
it with manual un-atomic tracking, we can just use a recursive spinlock
instead of a normal one, to resolve the original deadlock.
This commit is contained in:
Idan Horowitz 2024-02-09 17:08:39 +02:00 • committed by Andreas Kling
parent a957907f4b
commit 38dad2e27f
Notes: sideshowbarker 2024-07-17 18:23:22 +09:00
2 changed files with 5 additions and 18 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

19
Kernel/Devices/Storage/NVMe/NVMeQueue.cpp
View file

@ -111,11 +111,7 @@ void NVMeQueue::complete_current_request(u16 cmdid, u16 status)
auto current_request = request_pdu.request;
AsyncDeviceRequest::RequestResult req_result = AsyncDeviceRequest::Success;
ScopeGuard guard = [req_result, status, &request_pdu, &lock] {
// FIXME: We should unlock at the end of this function to make sure no new requests is inserted
// before we complete the request and calling end_io_handler but that results in a deadlock
// For now this is avoided by asserting the `used` field while inserting.
lock.unlock();
ScopeGuard guard = [req_result, status, &request_pdu] {
if (request_pdu.request)
request_pdu.request->complete(req_result);
if (request_pdu.end_io_handler)
@ -150,10 +146,7 @@ u16 NVMeQueue::submit_sync_sqe(NVMeSubmission& sub)
{
SpinlockLocker req_lock(m_request_lock);
if (m_requests.contains(sub.cmdid) && m_requests.get(sub.cmdid).release_value().used)
VERIFY_NOT_REACHED();
m_requests.set(sub.cmdid, { nullptr, true, [this, &cmd_status](u16 status) mutable { cmd_status = status; m_sync_wait_queue.wake_all(); } });
m_requests.set(sub.cmdid, { nullptr, [this, &cmd_status](u16 status) mutable { cmd_status = status; m_sync_wait_queue.wake_all(); } });
}
submit_sqe(sub);
@ -176,9 +169,7 @@ void NVMeQueue::read(AsyncBlockDeviceRequest& request, u16 nsid, u64 index, u32
{
SpinlockLocker req_lock(m_request_lock);
if (m_requests.contains(sub.cmdid) && m_requests.get(sub.cmdid).release_value().used)
VERIFY_NOT_REACHED();
m_requests.set(sub.cmdid, { request, true, nullptr });
m_requests.set(sub.cmdid, { request, nullptr });
}
full_memory_barrier();
@ -199,9 +190,7 @@ void NVMeQueue::write(AsyncBlockDeviceRequest& request, u16 nsid, u64 index, u32
{
SpinlockLocker req_lock(m_request_lock);
if (m_requests.contains(sub.cmdid) && m_requests.get(sub.cmdid).release_value().used)
VERIFY_NOT_REACHED();
m_requests.set(sub.cmdid, { request, true, nullptr });
m_requests.set(sub.cmdid, { request, nullptr });
}
if (auto result = request.read_from_buffer(request.buffer(), m_rw_dma_region->vaddr().as_ptr(), request.buffer_size()); result.is_error()) {

4
Kernel/Devices/Storage/NVMe/NVMeQueue.h
View file

@ -42,12 +42,10 @@ class AsyncBlockDeviceRequest;
struct NVMeIO {
void clear()
{
used = false;
request = nullptr;
end_io_handler = nullptr;
}
RefPtr<AsyncBlockDeviceRequest> request;
bool used = false;
Function<void(u16 status)> end_io_handler;
};
@ -117,7 +115,7 @@ protected:
Spinlock<LockRank::Interrupts> m_cq_lock {};
HashMap<u16, NVMeIO> m_requests;
NonnullOwnPtr<Memory::Region> m_rw_dma_region;
Spinlock<LockRank::None> m_request_lock {};
RecursiveSpinlock<LockRank::None> m_request_lock {};
private:
u16 m_qid {};