RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-05-20 13:45:06 +00:00
Code Issues Activity Actions

Make font loading use mmap().

Browse source 
This exposed a serious race condition in page_in_from_inode().
Reordered the logic and added a paging lock to VMObject.
Now, only one process can page in from a VMObject at a time.
There are definitely ways to optimize this, for instance by making
the locking be per-page instead. It's not something that I'm going
to worry about right now though.
This commit is contained in:
Andreas Kling 2019-02-03 01:36:25 +01:00
parent 7f91aec25c
commit abe3f515b1
4 changed files with 65 additions and 39 deletions
Download patch file Download diff file Expand all files Collapse all files

51
Kernel/MemoryManager.cpp
View file

@ -301,9 +301,41 @@ bool MemoryManager::page_in_from_inode(Region& region, unsigned page_index_in_re
auto& vmo = region.vmo();
ASSERT(!vmo.is_anonymous());
ASSERT(vmo.inode());
auto& inode = *vmo.inode();
auto& vmo_page = vmo.physical_pages()[region.first_page_index() + page_index_in_region];
ASSERT(vmo_page.is_null());
bool interrupts_were_enabled = are_interrupts_enabled();
if (!interrupts_were_enabled)
sti();
LOCKER(vmo.m_paging_lock);
if (!interrupts_were_enabled)
cli();
if (!vmo_page.is_null()) {
kprintf("MM: page_in_from_inode was served by someone else while lock was held\n");
remap_region_page(region, page_index_in_region, true);
return true;
}
#ifdef MM_DEBUG
dbgprintf("MM: page_in_from_inode ready to read from inode, will write to L%x!\n", dest_ptr);
#endif
sti(); // Oh god here we go...
byte page_buffer[PAGE_SIZE];
auto& inode = *vmo.inode();
auto nread = inode.read_bytes(vmo.inode_offset() + ((region.first_page_index() + page_index_in_region) * PAGE_SIZE), PAGE_SIZE, page_buffer, nullptr);
if (nread < 0) {
kprintf("MM: page_in_from_inode had error (%d) while reading!\n", nread);
return false;
}
if (nread < PAGE_SIZE) {
// If we read less than a page, zero out the rest to avoid leaking uninitialized data.
memset(page_buffer + nread, 0, PAGE_SIZE - nread);
}
cli();
vmo_page = allocate_physical_page(ShouldZeroFill::No);
if (vmo_page.is_null()) {
kprintf("MM: page_in_from_inode was unable to allocate a physical page\n");
@ -311,20 +343,7 @@ bool MemoryManager::page_in_from_inode(Region& region, unsigned page_index_in_re
}
remap_region_page(region, page_index_in_region, true);
byte* dest_ptr = region.laddr().offset(page_index_in_region * PAGE_SIZE).as_ptr();
#ifdef MM_DEBUG
dbgprintf("MM: page_in_from_inode ready to read from inode, will write to L%x!\n", dest_ptr);
#endif
sti(); // Oh god here we go...
auto nread = inode.read_bytes(vmo.inode_offset() + ((region.first_page_index() + page_index_in_region) * PAGE_SIZE), PAGE_SIZE, dest_ptr, nullptr);
if (nread < 0) {
kprintf("MM: page_in_from_inode had error (%d) while reading!\n", nread);
return false;
}
if (nread < PAGE_SIZE) {
// If we read less than a page, zero out the rest to avoid leaking uninitialized data.
memset(dest_ptr + nread, 0, PAGE_SIZE - nread);
}
cli();
memcpy(dest_ptr, page_buffer, PAGE_SIZE);
return true;
}