mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
synced 2026-10-26 10:57:38 +08:00
Linus Torvalds
7a78f7ba9f
Pull slab updates from Harry Yoo: - Performance improvement of the sheaf prefill API, by allowing the barn to hold a partial sheaf and refilling sheaves from the barn. This improves will-it-scale mmap1 throughput by 24% on the maple_node cache (Hao Li) - Restore memcg accounting of ipc msg_msg allocations, lost with e.g. CONFIG_SLAB_BUCKETS=n where the allocations fall back to the general kmalloc caches (Kees Cook) - Fix an old race between kmem_cache_destroy() and slab_late_init(), where a cache destroyed late in the boot process tries to release a kobject that was never initialized (Imre Kaloz) - Remove slab_folio(), the last use of struct folio in slab code (Matthew Wilcox) - Reject kmalloc() requests above KMALLOC_MAX_SIZE explicitly with a warning and NULL return instead of relying on the page allocator's order check (Zi Yan) - Various minor fixes and cleanups (Hao Li, Karl Mehltretter, Kees Cook, Longlong Xia) * tag 'slab-for-7.4' of git://git.kernel.org/pub/scm/linux/kernel/git/mm/slab: mm/slub: initialize the free pointer of the bootstrap kmem_cache_node mm/slab: sample slab_state once in kmem_cache_destroy() ipc, msg: Account msg_msg allocations with GFP_KERNEL_ACCOUNT again mm/slab: Mark the kmem_buckets_create() context as a Context: section slab: Remove slab_folio() mm/slub: refill prefilled sheaves from the barn mm/slub: make the case handling in __slab_free() easier to follow mm/slub: unify slub stat definitions to simplify metric additions mm/slab_common: reject zero object_size before calculate_alignment mm/slab: reject unsupported kmalloc() sizes