languages/cpython
languages/cpython
1
0
Fork 0
mirror of https://github.com/python/cpython synced 2024-10-14 17:18:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

gh-117755: Fix mimalloc for huge allocation on s390x (#117809)

Browse source 
Fix mimalloc allocator for huge memory allocation (around
8,589,934,592 GiB) on s390x.

Abort allocation early in mimalloc if the number of slices doesn't
fit into uint32_t, to prevent a integer overflow (cast 64-bit
size_t to uint32_t).
This commit is contained in:
Victor Stinner 2024-04-16 22:34:12 +02:00 committed by GitHub
parent e05d202ebf
commit 3fe03ccea6
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 8 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
Misc/NEWS.d/next/Core and Builtins/2024-04-12-12-28-49.gh-issue-117755.6ct8kU.rst Normal file
View file

@ -0,0 +1,2 @@
Fix mimalloc allocator for huge memory allocation (around 8,589,934,592 GiB) on
s390x. Patch by Victor Stinner.

6
Objects/mimalloc/segment.c
View file

@ -814,6 +814,9 @@ static mi_segment_t* mi_segment_os_alloc( size_t required, size_t page_alignment
const size_t extra = align_offset - info_size;
// recalculate due to potential guard pages
*psegment_slices = mi_segment_calculate_slices(required + extra, ppre_size, pinfo_slices);
// mi_page_t.slice_count type is uint32_t
if (*psegment_slices > (size_t)UINT32_MAX) return NULL;
}
const size_t segment_size = (*psegment_slices) * MI_SEGMENT_SLICE_SIZE;
@ -865,6 +868,9 @@ static mi_segment_t* mi_segment_alloc(size_t required, size_t page_alignment, mi
size_t pre_size;
size_t segment_slices = mi_segment_calculate_slices(required, &pre_size, &info_slices);
// mi_page_t.slice_count type is uint32_t
if (segment_slices > (size_t)UINT32_MAX) return NULL;
// Commit eagerly only if not the first N lazy segments (to reduce impact of many threads that allocate just a little)
const bool eager_delay = (// !_mi_os_has_overcommit() && // never delay on overcommit systems
_mi_current_thread_count() > 1 && // do not delay for the first N threads