RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-05-15 07:34:59 +00:00
Code Issues Activity Actions

Kernel: Remove VLA usage in Ext2FS block traversal code

Browse source 
This was using up to 12KB of kernel stack in the triply indirect case
and looks generally spooky. Let's just allocate a ByteBuffer for now
and take the performance hit (of heap allocation). Longer term we can
reorganize the code to reduce the majority of the heap churn.
This commit is contained in:
Andreas Kling 2021-03-10 16:24:01 +01:00
parent 3dbb9c8448
commit e58a600d52
1 changed files with 4 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

6
Kernel/FileSystem/Ext2FileSystem.cpp
View file

@ -506,9 +506,11 @@ Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list_impl_internal(const e
auto count = min(blocks_remaining, entries_per_block);
if (!count)
return;
u32 array[count];
size_t read_size = count * sizeof(u32);
auto array_storage = ByteBuffer::create_uninitialized(read_size);
auto* array = (u32*)array_storage.data();
auto buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)array);
auto result = fs().read_block(array_block_index, &buffer, sizeof(array), 0);
auto result = fs().read_block(array_block_index, &buffer, read_size, 0);
if (result.is_error()) {
// FIXME: Stop here and propagate this error.
dbgln("Ext2FS: compute_block_list_impl_internal had error: {}", result.error());