ladybird/Kernel/Memory/AnonymousVMObject.cpp
Andreas Kling 6cd3695761 Kernel: Stop taking MM lock while using regular quickmaps
You're still required to disable interrupts though, as the mappings are
per-CPU. This exposed the fact that our CR3 lookup map is insufficiently
protected (but we'll address that in a separate commit.)
2022-08-22 17:56:03 +02:00

402 lines
15 KiB
C++

/*
* Copyright (c) 2018-2022, Andreas Kling <kling@serenityos.org>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <Kernel/Arch/SafeMem.h>
#include <Kernel/Arch/SmapDisabler.h>
#include <Kernel/Debug.h>
#include <Kernel/Memory/AnonymousVMObject.h>
#include <Kernel/Memory/MemoryManager.h>
#include <Kernel/Memory/PhysicalPage.h>
#include <Kernel/Process.h>
namespace Kernel::Memory {
ErrorOr<NonnullLockRefPtr<VMObject>> AnonymousVMObject::try_clone()
{
// We need to acquire our lock so we copy a sane state
SpinlockLocker lock(m_lock);
if (is_purgeable() && is_volatile()) {
// If this object is purgeable+volatile, create a new zero-filled purgeable+volatile
// object, effectively "pre-purging" it in the child process.
auto clone = TRY(try_create_purgeable_with_size(size(), AllocationStrategy::None));
clone->m_volatile = true;
return clone;
}
// We're the parent. Since we're about to become COW we need to
// commit the number of pages that we need to potentially allocate
// so that the parent is still guaranteed to be able to have all
// non-volatile memory available.
size_t new_cow_pages_needed = 0;
for (auto const& page : m_physical_pages) {
if (!page->is_shared_zero_page())
++new_cow_pages_needed;
}
if (new_cow_pages_needed == 0)
return TRY(try_create_with_size(size(), AllocationStrategy::None));
dbgln_if(COMMIT_DEBUG, "Cloning {:p}, need {} committed cow pages", this, new_cow_pages_needed);
auto committed_pages = TRY(MM.commit_physical_pages(new_cow_pages_needed));
// Create or replace the committed cow pages. When cloning a previously
// cloned vmobject, we want to essentially "fork", leaving us and the
// new clone with one set of shared committed cow pages, and the original
// one would keep the one it still has. This ensures that the original
// one and this one, as well as the clone have sufficient resources
// to cow all pages as needed
auto new_shared_committed_cow_pages = TRY(adopt_nonnull_lock_ref_or_enomem(new (nothrow) SharedCommittedCowPages(move(committed_pages))));
auto new_physical_pages = TRY(this->try_clone_physical_pages());
auto clone = TRY(try_create_with_shared_cow(*this, *new_shared_committed_cow_pages, move(new_physical_pages)));
// Both original and clone become COW. So create a COW map for ourselves
// or reset all pages to be copied again if we were previously cloned
TRY(ensure_or_reset_cow_map());
m_shared_committed_cow_pages = move(new_shared_committed_cow_pages);
if (m_unused_committed_pages.has_value() && !m_unused_committed_pages->is_empty()) {
// The parent vmobject didn't use up all committed pages. When
// cloning (fork) we will overcommit. For this purpose we drop all
// lazy-commit references and replace them with shared zero pages.
for (size_t i = 0; i < page_count(); i++) {
auto& page = clone->m_physical_pages[i];
if (page && page->is_lazy_committed_page()) {
page = MM.shared_zero_page();
}
}
}
return clone;
}
ErrorOr<NonnullLockRefPtr<AnonymousVMObject>> AnonymousVMObject::try_create_with_size(size_t size, AllocationStrategy strategy)
{
Optional<CommittedPhysicalPageSet> committed_pages;
if (strategy == AllocationStrategy::Reserve || strategy == AllocationStrategy::AllocateNow) {
committed_pages = TRY(MM.commit_physical_pages(ceil_div(size, static_cast<size_t>(PAGE_SIZE))));
}
auto new_physical_pages = TRY(VMObject::try_create_physical_pages(size));
return adopt_nonnull_lock_ref_or_enomem(new (nothrow) AnonymousVMObject(move(new_physical_pages), strategy, move(committed_pages)));
}
ErrorOr<NonnullLockRefPtr<AnonymousVMObject>> AnonymousVMObject::try_create_physically_contiguous_with_size(size_t size)
{
auto contiguous_physical_pages = TRY(MM.allocate_contiguous_physical_pages(size));
auto new_physical_pages = TRY(FixedArray<LockRefPtr<PhysicalPage>>::try_create(contiguous_physical_pages.span()));
return adopt_nonnull_lock_ref_or_enomem(new (nothrow) AnonymousVMObject(move(new_physical_pages)));
}
ErrorOr<NonnullLockRefPtr<AnonymousVMObject>> AnonymousVMObject::try_create_purgeable_with_size(size_t size, AllocationStrategy strategy)
{
Optional<CommittedPhysicalPageSet> committed_pages;
if (strategy == AllocationStrategy::Reserve || strategy == AllocationStrategy::AllocateNow) {
committed_pages = TRY(MM.commit_physical_pages(ceil_div(size, static_cast<size_t>(PAGE_SIZE))));
}
auto new_physical_pages = TRY(VMObject::try_create_physical_pages(size));
auto vmobject = TRY(adopt_nonnull_lock_ref_or_enomem(new (nothrow) AnonymousVMObject(move(new_physical_pages), strategy, move(committed_pages))));
vmobject->m_purgeable = true;
return vmobject;
}
ErrorOr<NonnullLockRefPtr<AnonymousVMObject>> AnonymousVMObject::try_create_with_physical_pages(Span<NonnullLockRefPtr<PhysicalPage>> physical_pages)
{
auto new_physical_pages = TRY(FixedArray<LockRefPtr<PhysicalPage>>::try_create(physical_pages));
return adopt_nonnull_lock_ref_or_enomem(new (nothrow) AnonymousVMObject(move(new_physical_pages)));
}
ErrorOr<NonnullLockRefPtr<AnonymousVMObject>> AnonymousVMObject::try_create_for_physical_range(PhysicalAddress paddr, size_t size)
{
if (paddr.offset(size) < paddr) {
dbgln("Shenanigans! try_create_for_physical_range({}, {}) would wrap around", paddr, size);
// Since we can't wrap around yet, let's pretend to OOM.
return ENOMEM;
}
auto new_physical_pages = TRY(VMObject::try_create_physical_pages(size));
return adopt_nonnull_lock_ref_or_enomem(new (nothrow) AnonymousVMObject(paddr, move(new_physical_pages)));
}
ErrorOr<NonnullLockRefPtr<AnonymousVMObject>> AnonymousVMObject::try_create_with_shared_cow(AnonymousVMObject const& other, NonnullLockRefPtr<SharedCommittedCowPages> shared_committed_cow_pages, FixedArray<LockRefPtr<PhysicalPage>>&& new_physical_pages)
{
auto weak_parent = TRY(other.try_make_weak_ptr<AnonymousVMObject>());
auto vmobject = TRY(adopt_nonnull_lock_ref_or_enomem(new (nothrow) AnonymousVMObject(move(weak_parent), move(shared_committed_cow_pages), move(new_physical_pages))));
TRY(vmobject->ensure_cow_map());
return vmobject;
}
AnonymousVMObject::AnonymousVMObject(FixedArray<LockRefPtr<PhysicalPage>>&& new_physical_pages, AllocationStrategy strategy, Optional<CommittedPhysicalPageSet> committed_pages)
: VMObject(move(new_physical_pages))
, m_unused_committed_pages(move(committed_pages))
{
if (strategy == AllocationStrategy::AllocateNow) {
// Allocate all pages right now. We know we can get all because we committed the amount needed
for (size_t i = 0; i < page_count(); ++i)
physical_pages()[i] = m_unused_committed_pages->take_one();
} else {
auto& initial_page = (strategy == AllocationStrategy::Reserve) ? MM.lazy_committed_page() : MM.shared_zero_page();
for (size_t i = 0; i < page_count(); ++i)
physical_pages()[i] = initial_page;
}
}
AnonymousVMObject::AnonymousVMObject(PhysicalAddress paddr, FixedArray<LockRefPtr<PhysicalPage>>&& new_physical_pages)
: VMObject(move(new_physical_pages))
{
VERIFY(paddr.page_base() == paddr);
for (size_t i = 0; i < page_count(); ++i)
physical_pages()[i] = PhysicalPage::create(paddr.offset(i * PAGE_SIZE), MayReturnToFreeList::No);
}
AnonymousVMObject::AnonymousVMObject(FixedArray<LockRefPtr<PhysicalPage>>&& new_physical_pages)
: VMObject(move(new_physical_pages))
{
}
AnonymousVMObject::AnonymousVMObject(LockWeakPtr<AnonymousVMObject> other, NonnullLockRefPtr<SharedCommittedCowPages> shared_committed_cow_pages, FixedArray<LockRefPtr<PhysicalPage>>&& new_physical_pages)
: VMObject(move(new_physical_pages))
, m_cow_parent(move(other))
, m_shared_committed_cow_pages(move(shared_committed_cow_pages))
, m_purgeable(m_cow_parent.strong_ref()->m_purgeable)
{
}
AnonymousVMObject::~AnonymousVMObject()
{
if (!m_shared_committed_cow_pages || m_shared_committed_cow_pages->is_empty())
return;
auto cow_parent = m_cow_parent.strong_ref();
if (!cow_parent)
return;
SpinlockLocker lock(cow_parent->m_lock);
if (cow_parent->m_shared_committed_cow_pages == m_shared_committed_cow_pages)
cow_parent->m_shared_committed_cow_pages.clear();
}
size_t AnonymousVMObject::purge()
{
SpinlockLocker lock(m_lock);
if (!is_purgeable() || !is_volatile())
return 0;
size_t total_pages_purged = 0;
for (auto& page : m_physical_pages) {
VERIFY(page);
if (page->is_shared_zero_page())
continue;
page = MM.shared_zero_page();
++total_pages_purged;
}
m_was_purged = true;
for_each_region([](Region& region) {
region.remap();
});
return total_pages_purged;
}
ErrorOr<void> AnonymousVMObject::set_volatile(bool is_volatile, bool& was_purged)
{
VERIFY(is_purgeable());
SpinlockLocker locker(m_lock);
was_purged = m_was_purged;
if (m_volatile == is_volatile)
return {};
if (is_volatile) {
// When a VMObject is made volatile, it gives up all of its committed memory.
// Any physical pages already allocated remain in the VMObject for now, but the kernel is free to take them at any moment.
for (auto& page : m_physical_pages) {
if (page && page->is_lazy_committed_page())
page = MM.shared_zero_page();
}
m_unused_committed_pages = {};
m_shared_committed_cow_pages = nullptr;
if (!m_cow_map.is_null())
m_cow_map = {};
m_volatile = true;
m_was_purged = false;
for_each_region([&](auto& region) { region.remap(); });
return {};
}
// When a VMObject is made non-volatile, we try to commit however many pages are not currently available.
// If that fails, we return false to indicate that memory allocation failed.
size_t committed_pages_needed = 0;
for (auto& page : m_physical_pages) {
VERIFY(page);
if (page->is_shared_zero_page())
++committed_pages_needed;
}
if (!committed_pages_needed) {
m_volatile = false;
return {};
}
m_unused_committed_pages = TRY(MM.commit_physical_pages(committed_pages_needed));
for (auto& page : m_physical_pages) {
if (page->is_shared_zero_page())
page = MM.lazy_committed_page();
}
m_volatile = false;
m_was_purged = false;
for_each_region([&](auto& region) { region.remap(); });
return {};
}
NonnullLockRefPtr<PhysicalPage> AnonymousVMObject::allocate_committed_page(Badge<Region>)
{
return m_unused_committed_pages->take_one();
}
ErrorOr<void> AnonymousVMObject::ensure_cow_map()
{
if (m_cow_map.is_null())
m_cow_map = TRY(Bitmap::try_create(page_count(), true));
return {};
}
ErrorOr<void> AnonymousVMObject::ensure_or_reset_cow_map()
{
if (m_cow_map.is_null())
TRY(ensure_cow_map());
else
m_cow_map.fill(true);
return {};
}
bool AnonymousVMObject::should_cow(size_t page_index, bool is_shared) const
{
auto const& page = physical_pages()[page_index];
if (page && (page->is_shared_zero_page() || page->is_lazy_committed_page()))
return true;
if (is_shared)
return false;
return !m_cow_map.is_null() && m_cow_map.get(page_index);
}
ErrorOr<void> AnonymousVMObject::set_should_cow(size_t page_index, bool cow)
{
TRY(ensure_cow_map());
m_cow_map.set(page_index, cow);
return {};
}
size_t AnonymousVMObject::cow_pages() const
{
if (m_cow_map.is_null())
return 0;
return m_cow_map.count_slow(true);
}
PageFaultResponse AnonymousVMObject::handle_cow_fault(size_t page_index, VirtualAddress vaddr)
{
SpinlockLocker lock(m_lock);
if (is_volatile()) {
// A COW fault in a volatile region? Userspace is writing to volatile memory, this is a bug. Crash.
dbgln("COW fault in volatile region, will crash.");
return PageFaultResponse::ShouldCrash;
}
auto& page_slot = physical_pages()[page_index];
// If we were sharing committed COW pages with another process, and the other process
// has exhausted the supply, we can stop counting the shared pages.
if (m_shared_committed_cow_pages && m_shared_committed_cow_pages->is_empty())
m_shared_committed_cow_pages = nullptr;
if (page_slot->ref_count() == 1) {
dbgln_if(PAGE_FAULT_DEBUG, " >> It's a COW page but nobody is sharing it anymore. Remap r/w");
MUST(set_should_cow(page_index, false)); // If we received a COW fault, we already have a cow map allocated, so this is infallible
if (m_shared_committed_cow_pages) {
m_shared_committed_cow_pages->uncommit_one();
if (m_shared_committed_cow_pages->is_empty())
m_shared_committed_cow_pages = nullptr;
}
return PageFaultResponse::Continue;
}
LockRefPtr<PhysicalPage> page;
if (m_shared_committed_cow_pages) {
dbgln_if(PAGE_FAULT_DEBUG, " >> It's a committed COW page and it's time to COW!");
page = m_shared_committed_cow_pages->take_one();
} else {
dbgln_if(PAGE_FAULT_DEBUG, " >> It's a COW page and it's time to COW!");
auto page_or_error = MM.allocate_physical_page(MemoryManager::ShouldZeroFill::No);
if (page_or_error.is_error()) {
dmesgln("MM: handle_cow_fault was unable to allocate a physical page");
return PageFaultResponse::OutOfMemory;
}
page = page_or_error.release_value();
}
dbgln_if(PAGE_FAULT_DEBUG, " >> COW {} <- {}", page->paddr(), page_slot->paddr());
{
u8* dest_ptr = MM.quickmap_page(*page);
SmapDisabler disabler;
void* fault_at;
if (!safe_memcpy(dest_ptr, vaddr.as_ptr(), PAGE_SIZE, fault_at)) {
if ((u8*)fault_at >= dest_ptr && (u8*)fault_at <= dest_ptr + PAGE_SIZE)
dbgln(" >> COW: error copying page {}/{} to {}/{}: failed to write to page at {}",
page_slot->paddr(), vaddr, page->paddr(), VirtualAddress(dest_ptr), VirtualAddress(fault_at));
else if ((u8*)fault_at >= vaddr.as_ptr() && (u8*)fault_at <= vaddr.as_ptr() + PAGE_SIZE)
dbgln(" >> COW: error copying page {}/{} to {}/{}: failed to read from page at {}",
page_slot->paddr(), vaddr, page->paddr(), VirtualAddress(dest_ptr), VirtualAddress(fault_at));
else
VERIFY_NOT_REACHED();
}
MM.unquickmap_page();
}
page_slot = move(page);
MUST(set_should_cow(page_index, false)); // If we received a COW fault, we already have a cow map allocated, so this is infallible
return PageFaultResponse::Continue;
}
AnonymousVMObject::SharedCommittedCowPages::SharedCommittedCowPages(CommittedPhysicalPageSet&& committed_pages)
: m_committed_pages(move(committed_pages))
{
}
AnonymousVMObject::SharedCommittedCowPages::~SharedCommittedCowPages() = default;
NonnullLockRefPtr<PhysicalPage> AnonymousVMObject::SharedCommittedCowPages::take_one()
{
SpinlockLocker locker(m_lock);
return m_committed_pages.take_one();
}
void AnonymousVMObject::SharedCommittedCowPages::uncommit_one()
{
SpinlockLocker locker(m_lock);
m_committed_pages.uncommit_one();
}
}