Since commit b1f202060afe ("mm: remap unused subpages to shared zeropage when splitting isolated thp"), splitting an anonymous THP remaps all zero-filled subpages to the shared zeropage via TTU_USE_SHARED_ZEROPAGE. This flag is set unconditionally for every anonymous folio split, including splits triggered by KSM. When KSM is enabled with THP=always, this causes two regressions: 1. use_zero_pages=1: KSM calls try_to_merge_one_page() which triggers split_huge_page(). The split remaps all 512 zero-filled subpages to the shared zeropage at once, freeing the entire 2MB THP when KSM only intended to process a single 4KB page. This bypasses KSM's pages_to_scan rate limiting, causing ~1GB to be freed almost instantly. 2. use_zero_pages=0: The same split side-effect occurs through the stable/unstable tree merge paths. Each pages_to_scan iteration triggers an expensive split_huge_page() that silently frees 2MB, while the scanner wastes cycles on tree searches for zero-filled pages that were already freed as a side-effect. Introduce current_is_ksmd(), to check if the current task is the KSM thread. Use it in try_to_map_unused_to_zeropage() to skip zero-page remapping when the split is triggered by ksmd. Allow remapping during reclaim running in ksmd's context. Fixes: b1f202060afe ("mm: remap unused subpages to shared zeropage when splitting isolated thp") Signed-off-by: Nico Pache (Red Hat) --- This patch fixes a bug in KSM where the pages_to_scan limit is bypassed and the policy (use_zero_pages) is ignored. Commit b1f202060afe ("mm: remap unused subpages to shared zeropage when splitting isolated thp") added zero-page remapping when an anonymous folio is split, including splits triggered by KSM. Skip zero-page remapping when ksmd triggers the split, while allowing remapping during reclaim running in ksmd's context. Changes in v3: - Drop the patch gating zero-page remapping on shrink_underused. - Allow zero-page remapping during reclaim running in ksmd's context - Add a comment explaining the ksmd check (David Hildenbrand). V2: https://lore.kernel.org/all/20261025-b4-ksm_fixv2-v2-0-8b61ad0491f1@linux.dev V1: https://lore.kernel.org/all/20260609114619.144416-1-npache@redhat.com/ RFC: https://lore.kernel.org/all/20260508170509.640851-1-npache@redhat.com/ --- include/linux/ksm.h | 6 ++++++ mm/ksm.c | 5 +++++ mm/migrate.c | 8 ++++++++ 3 files changed, 19 insertions(+) diff --git a/include/linux/ksm.h b/include/linux/ksm.h index a0d358a1feb6..23c8e9f60118 100644 --- a/include/linux/ksm.h +++ b/include/linux/ksm.h @@ -17,6 +17,7 @@ #ifdef CONFIG_KSM int ksm_madvise(struct vm_area_struct *vma, unsigned long start, unsigned long end, int advice, vm_flags_t *vm_flags); +bool current_is_ksmd(void); vma_flags_t ksm_vma_flags(struct mm_struct *mm, const struct file *file, vma_flags_t vma_flags); int ksm_enable_merge_any(struct mm_struct *mm); @@ -144,6 +145,11 @@ static inline int ksm_madvise(struct vm_area_struct *vma, unsigned long start, return 0; } +static inline bool current_is_ksmd(void) +{ + return false; +} + static inline struct folio *ksm_might_need_to_copy(struct folio *folio, struct page *page, struct vm_area_struct *vma, unsigned long addr) { diff --git a/mm/ksm.c b/mm/ksm.c index d24b129932bf..60656b05419a 100644 --- a/mm/ksm.c +++ b/mm/ksm.c @@ -2873,6 +2873,11 @@ static int ksm_scan_thread(void *nothing) return 0; } +bool current_is_ksmd(void) +{ + return kthread_func(current) == ksm_scan_thread; +} + static bool __ksm_should_add_vma(const struct file *file, vma_flags_t vma_flags) { if (vma_flags_test(&vma_flags, VMA_MERGEABLE_BIT)) diff --git a/mm/migrate.c b/mm/migrate.c index 65869b8acec5..b0949f43ed5c 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -305,6 +305,14 @@ static bool try_to_map_unused_to_zeropage(struct page_vma_mapped_walk *pvmw, if (PageCompound(page) || PageHWPoison(page)) return false; + /* + * When a split is triggered by KSM, let KSM take care of scanning for + * and deduplicating the shared zeropage. Reclaim in the ksmd context + * should still be able to reclaim zero-filled pages. + */ + if (current_is_ksmd() && !current->reclaim_state) + return false; + VM_BUG_ON_PAGE(!PageAnon(page), page); VM_BUG_ON_PAGE(!PageLocked(page), page); VM_BUG_ON_PAGE(pte_present(old_pte), page); --- base-commit: 2dc60bd9975fc8666a13d58410499e06e6799ba9 change-id: 20261025-b4-ksm_fixv2-2bf062e92946 Best regards, -- Nico Pache (Red Hat)