kernel/linux.git/include/linux/hugetlb.h, branch v5.10.258

mm/hugetlb: fix excessive IPI broadcasts when unsharing PMD tables using mmu_gather

2026-04-18T08:31:14+00:00

commit 8ce720d5bd91e9dc16db3604aa4b1bf76770a9a1 upstream. As reported, ever since commit 1013af4f585f ("mm/hugetlb: fix huge_pmd_unshare() vs GUP-fast race") we can end up in some situations where we perform so many IPI broadcasts when unsharing hugetlb PMD page tables that it severely regresses some workloads. In particular, when we fork()+exit(), or when we munmap() a large area backed by many shared PMD tables, we perform one IPI broadcast per unshared PMD table. There are two optimizations to be had: (1) When we process (unshare) multiple such PMD tables, such as during exit(), it is sufficient to send a single IPI broadcast (as long as we respect locking rules) instead of one per PMD table. Locking prevents that any of these PMD tables could get reused before we drop the lock. (2) When we are not the last sharer (> 2 users including us), there is no need to send the IPI broadcast. The shared PMD tables cannot become exclusive (fully unshared) before an IPI will be broadcasted by the last sharer. Concurrent GUP-fast could walk into a PMD table just before we unshared it. It could then succeed in grabbing a page from the shared page table even after munmap() etc succeeded (and supressed an IPI). But there is not difference compared to GUP-fast just sleeping for a while after grabbing the page and re-enabling IRQs. Most importantly, GUP-fast will never walk into page tables that are no-longer shared, because the last sharer will issue an IPI broadcast. (if ever required, checking whether the PUD changed in GUP-fast after grabbing the page like we do in the PTE case could handle this) So let's rework PMD sharing TLB flushing + IPI sync to use the mmu_gather infrastructure so we can implement these optimizations and demystify the code at least a bit. Extend the mmu_gather infrastructure to be able to deal with our special hugetlb PMD table sharing implementation. To make initialization of the mmu_gather easier when working on a single VMA (in particular, when dealing with hugetlb), provide tlb_gather_mmu_vma(). We'll consolidate the handling for (full) unsharing of PMD tables in tlb_unshare_pmd_ptdesc() and tlb_flush_unshared_tables(), and track in "struct mmu_gather" whether we had (full) unsharing of PMD tables. Because locking is very special (concurrent unsharing+reuse must be prevented), we disallow deferring flushing to tlb_finish_mmu() and instead require an explicit earlier call to tlb_flush_unshared_tables(). >From hugetlb code, we call huge_pmd_unshare_flush() where we make sure that the expected lock protecting us from concurrent unsharing+reuse is still held. Check with a VM_WARN_ON_ONCE() in tlb_finish_mmu() that tlb_flush_unshared_tables() was properly called earlier. Document it all properly. Notes about tlb_remove_table_sync_one() interaction with unsharing: There are two fairly tricky things: (1) tlb_remove_table_sync_one() is a NOP on architectures without CONFIG_MMU_GATHER_RCU_TABLE_FREE. Here, the assumption is that the previous TLB flush would send an IPI to all relevant CPUs. Careful: some architectures like x86 only send IPIs to all relevant CPUs when tlb->freed_tables is set. The relevant architectures should be selecting MMU_GATHER_RCU_TABLE_FREE, but x86 might not do that in stable kernels and it might have been problematic before this patch. Also, the arch flushing behavior (independent of IPIs) is different when tlb->freed_tables is set. Do we have to enlighten them to also take care of tlb->unshared_tables? So far we didn't care, so hopefully we are fine. Of course, we could be setting tlb->freed_tables as well, but that might then unnecessarily flush too much, because the semantics of tlb->freed_tables are a bit fuzzy. This patch changes nothing in this regard. (2) tlb_remove_table_sync_one() is not a NOP on architectures with CONFIG_MMU_GATHER_RCU_TABLE_FREE that actually don't need a sync. Take x86 as an example: in the common case (!pv, !X86_FEATURE_INVLPGB) we still issue IPIs during TLB flushes and don't actually need the second tlb_remove_table_sync_one(). This optimized can be implemented on top of this, by checking e.g., in tlb_remove_table_sync_one() whether we really need IPIs. But as described in (1), it really must honor tlb->freed_tables then to send IPIs to all relevant CPUs. Notes on TLB flushing changes: (1) Flushing for non-shared PMD tables We're converting from flush_hugetlb_tlb_range() to tlb_remove_huge_tlb_entry(). Given that we properly initialize the MMU gather in tlb_gather_mmu_vma() to be hugetlb aware, similar to __unmap_hugepage_range(), that should be fine. (2) Flushing for shared PMD tables We're converting from various things (flush_hugetlb_tlb_range(), tlb_flush_pmd_range(), flush_tlb_range()) to tlb_flush_pmd_range(). tlb_flush_pmd_range() achieves the same that tlb_remove_huge_tlb_entry() would achieve in these scenarios. Note that tlb_remove_huge_tlb_entry() also calls __tlb_remove_tlb_entry(), however that is only implemented on powerpc, which does not support PMD table sharing. Similar to (1), tlb_gather_mmu_vma() should make sure that TLB flushing keeps on working as expected. Further, note that the ptdesc_pmd_pts_dec() in huge_pmd_share() is not a concern, as we are holding the i_mmap_lock the whole time, preventing concurrent unsharing. That ptdesc_pmd_pts_dec() usage will be removed separately as a cleanup later. There are plenty more cleanups to be had, but they have to wait until this is fixed. [david@kernel.org: fix kerneldoc] Link: https://lkml.kernel.org/r/f223dd74-331c-412d-93fc-69e360a5006c@kernel.org Link: https://lkml.kernel.org/r/20251223214037.580860-5-david@kernel.org Fixes: 1013af4f585f ("mm/hugetlb: fix huge_pmd_unshare() vs GUP-fast race") Signed-off-by: David Hildenbrand (Red Hat) Reported-by: "Uschakow, Stanislav" Closes: https://lore.kernel.org/all/4d3878531c76479d9f8ca9789dc6485d@amazon.de/ Tested-by: Laurence Oberman Acked-by: Harry Yoo Reviewed-by: Lorenzo Stoakes Cc: Lance Yang Cc: Liu Shixin Cc: Oscar Salvador Cc: Rik van Riel Cc: Signed-off-by: Andrew Morton (cherry picked from commit 8ce720d5bd91e9dc16db3604aa4b1bf76770a9a1) [ David: We don't have ptdesc and the wrappers, so work directly on page->pt_share_count and pass "struct page" instead of "struct ptdesc". CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING is still called CONFIG_ARCH_WANT_HUGE_PMD_SHARE and is set even without CONFIG_HUGETLB_PAGE. We don't have 550a7d60bd5e ("mm, hugepages: add mremap() support for hugepage backed vma"), so move_hugetlb_page_tables() does not exist. We don't have 40549ba8f8e0 ("hugetlb: use new vma_lock for pmd sharing synchronization") and a98a2f0c8ce1 ("mm/rmap: split migration into its own so changes in mm/rmap.c looks quite different. We don't have 4ddb4d91b82f ("hugetlb: do not update address in huge_pmd_unshare"), so huge_pmd_unshare() still gets a pointer to an address. tlb_gather_mmu() + tlb_finish_mmu() still consume ranges, so also teach tlb_gather_mmu_vma() to forward ranges. Some smaller contextual stuff, in particular, around tlb_gather_mmu_full(). ] Signed-off-by: David Hildenbrand (Arm) Signed-off-by: Sasha Levin

mm/hugetlb: fix hugetlb_pmd_shared()

2026-04-18T08:31:14+00:00

commit ca1a47cd3f5f4c46ca188b1c9a27af87d1ab2216 upstream. Patch series "mm/hugetlb: fixes for PMD table sharing (incl. using mmu_gather)", v3. One functional fix, one performance regression fix, and two related comment fixes. I cleaned up my prototype I recently shared [1] for the performance fix, deferring most of the cleanups I had in the prototype to a later point. While doing that I identified the other things. The goal of this patch set is to be backported to stable trees "fairly" easily. At least patch #1 and #4. Patch #1 fixes hugetlb_pmd_shared() not detecting any sharing Patch #2 + #3 are simple comment fixes that patch #4 interacts with. Patch #4 is a fix for the reported performance regression due to excessive IPI broadcasts during fork()+exit(). The last patch is all about TLB flushes, IPIs and mmu_gather. Read: complicated There are plenty of cleanups in the future to be had + one reasonable optimization on x86. But that's all out of scope for this series. Runtime tested, with a focus on fixing the performance regression using the original reproducer [2] on x86. This patch (of 4): We switched from (wrongly) using the page count to an independent shared count. Now, shared page tables have a refcount of 1 (excluding speculative references) and instead use ptdesc->pt_share_count to identify sharing. We didn't convert hugetlb_pmd_shared(), so right now, we would never detect a shared PMD table as such, because sharing/unsharing no longer touches the refcount of a PMD table. Page migration, like mbind() or migrate_pages() would allow for migrating folios mapped into such shared PMD tables, even though the folios are not exclusive. In smaps we would account them as "private" although they are "shared", and we would be wrongly setting the PM_MMAP_EXCLUSIVE in the pagemap interface. Fix it by properly using ptdesc_pmd_is_shared() in hugetlb_pmd_shared(). Link: https://lkml.kernel.org/r/20251223214037.580860-1-david@kernel.org Link: https://lkml.kernel.org/r/20251223214037.580860-2-david@kernel.org Link: https://lore.kernel.org/all/8cab934d-4a56-44aa-b641-bfd7e23bd673@kernel.org/ [1] Link: https://lore.kernel.org/all/8cab934d-4a56-44aa-b641-bfd7e23bd673@kernel.org/ [2] Fixes: 59d9094df3d7 ("mm: hugetlb: independent PMD page table shared count") Signed-off-by: David Hildenbrand (Red Hat) Reviewed-by: Rik van Riel Reviewed-by: Lance Yang Tested-by: Lance Yang Reviewed-by: Harry Yoo Tested-by: Laurence Oberman Reviewed-by: Lorenzo Stoakes Acked-by: Oscar Salvador Cc: Liu Shixin Cc: "Uschakow, Stanislav" Cc: Signed-off-by: Andrew Morton (cherry picked from commit ca1a47cd3f5f4c46ca188b1c9a27af87d1ab2216) [ David: We don't have ptdesc and the wrappers, so work directly on page->pt_share_count. ] Signed-off-by: David Hildenbrand (Arm) Signed-off-by: Sasha Levin

mm/hugetlb: unshare page tables during VMA split, not before

2025-06-27T10:04:21+00:00

commit 081056dc00a27bccb55ccc3c6f230a3d5fd3f7e0 upstream. Currently, __split_vma() triggers hugetlb page table unsharing through vm_ops->may_split(). This happens before the VMA lock and rmap locks are taken - which is too early, it allows racing VMA-locked page faults in our process and racing rmap walks from other processes to cause page tables to be shared again before we actually perform the split. Fix it by explicitly calling into the hugetlb unshare logic from __split_vma() in the same place where THP splitting also happens. At that point, both the VMA and the rmap(s) are write-locked. An annoying detail is that we can now call into the helper hugetlb_unshare_pmds() from two different locking contexts: 1. from hugetlb_split(), holding: - mmap lock (exclusively) - VMA lock - file rmap lock (exclusively) 2. hugetlb_unshare_all_pmds(), which I think is designed to be able to call us with only the mmap lock held (in shared mode), but currently only runs while holding mmap lock (exclusively) and VMA lock Backporting note: This commit fixes a racy protection that was introduced in commit b30c14cd6102 ("hugetlb: unshare some PMDs when splitting VMAs"); that commit claimed to fix an issue introduced in 5.13, but it should actually also go all the way back. [jannh@google.com: v2] Link: https://lkml.kernel.org/r/20250528-hugetlb-fixes-splitrace-v2-1-1329349bad1a@google.com Link: https://lkml.kernel.org/r/20250528-hugetlb-fixes-splitrace-v2-0-1329349bad1a@google.com Link: https://lkml.kernel.org/r/20250527-hugetlb-fixes-splitrace-v1-1-f4136f5ec58a@google.com Fixes: 39dde65c9940 ("[PATCH] shared page table for hugetlb page") Signed-off-by: Jann Horn Cc: Liam Howlett Reviewed-by: Lorenzo Stoakes Reviewed-by: Oscar Salvador Cc: Lorenzo Stoakes Cc: Vlastimil Babka Cc: [b30c14cd6102: hugetlb: unshare some PMDs when splitting VMAs] Cc: Signed-off-by: Andrew Morton [stable backport: code got moved around, VMA splitting is in __vma_adjust, hugetlb lock wasn't used back then] Signed-off-by: Jann Horn Signed-off-by: Greg Kroah-Hartman

mm/hugetlb: change hugetlb_reserve_pages() to type bool

2024-03-15T14:48:22+00:00

[ Upstream commit 33b8f84a4ee78491a8f4f9e4c5520c9da4a10983 ] While reviewing a bug in hugetlb_reserve_pages, it was noticed that all callers ignore the return value. Any failure is considered an ENOMEM error by the callers. Change the function to be of type bool. The function will return true if the reservation was successful, false otherwise. Callers currently assume a zero return code indicates success. Change the callers to look for true to indicate success. No functional change, only code cleanup. Link: https://lkml.kernel.org/r/20201221192542.15732-1-mike.kravetz@oracle.com Signed-off-by: Mike Kravetz Reviewed-by: Matthew Wilcox (Oracle) Cc: David Hildenbrand Cc: Dan Carpenter Cc: Michal Hocko Cc: Davidlohr Bueso Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds Stable-dep-of: e656c7a9e596 ("mm: hugetlb pages should not be reserved by shmat() if SHM_NORESERVE") Signed-off-by: Sasha Levin

hugetlb: check for undefined shift on 32 bit architectures

2023-02-22T11:55:56+00:00

commit ec4288fe63966b26d53907212ecd05dfa81dd2cc upstream. Users can specify the hugetlb page size in the mmap, shmget and memfd_create system calls. This is done by using 6 bits within the flags argument to encode the base-2 logarithm of the desired page size. The routine hstate_sizelog() uses the log2 value to find the corresponding hugetlb hstate structure. Converting the log2 value (page_size_log) to potential hugetlb page size is the simple statement: 1UL << page_size_log Because only 6 bits are used for page_size_log, the left shift can not be greater than 63. This is fine on 64 bit architectures where a long is 64 bits. However, if a value greater than 31 is passed on a 32 bit architecture (where long is 32 bits) the shift will result in undefined behavior. This was generally not an issue as the result of the undefined shift had to exactly match hugetlb page size to proceed. Recent improvements in runtime checking have resulted in this undefined behavior throwing errors such as reported below. Fix by comparing page_size_log to BITS_PER_LONG before doing shift. Link: https://lkml.kernel.org/r/20230216013542.138708-1-mike.kravetz@oracle.com Link: https://lore.kernel.org/lkml/CA+G9fYuei_Tr-vN9GS7SfFyU1y9hNysnf=PB7kT0=yv4MiPgVg@mail.gmail.com/ Fixes: 42d7395feb56 ("mm: support more pagesizes for MAP_HUGETLB/SHM_HUGETLB") Signed-off-by: Mike Kravetz Reported-by: Naresh Kamboju Reviewed-by: Jesper Juhl Acked-by: Muchun Song Tested-by: Linux Kernel Functional Testing Tested-by: Naresh Kamboju Cc: Anders Roxell Cc: Andi Kleen Cc: Sasha Levin Cc: Signed-off-by: Andrew Morton Signed-off-by: Greg Kroah-Hartman

mm/migration: return errno when isolate_huge_page failed

2023-02-15T16:22:21+00:00

[ Upstream commit 7ce82f4c3f3ead13a9d9498768e3b1a79975c4d8 ] We might fail to isolate huge page due to e.g. the page is under migration which cleared HPageMigratable. We should return errno in this case rather than always return 1 which could confuse the user, i.e. the caller might think all of the memory is migrated while the hugetlb page is left behind. We make the prototype of isolate_huge_page consistent with isolate_lru_page as suggested by Huang Ying and rename isolate_huge_page to isolate_hugetlb as suggested by Muchun to improve the readability. Link: https://lkml.kernel.org/r/20220530113016.16663-4-linmiaohe@huawei.com Fixes: e8db67eb0ded ("mm: migrate: move_pages() supports thp migration") Signed-off-by: Miaohe Lin Suggested-by: Huang Ying Reported-by: kernel test robot (build error) Cc: Alistair Popple Cc: Christoph Hellwig Cc: Christoph Lameter Cc: David Hildenbrand Cc: David Howells Cc: Mike Kravetz Cc: Muchun Song Cc: Oscar Salvador Cc: Peter Xu Signed-off-by: Andrew Morton Stable-dep-of: 73bdf65ea748 ("migrate: hugetlb: check for hugetlb shared PMD in node migration") Signed-off-by: Sasha Levin

mm: hugetlb: proc: check for hugetlb shared PMD in /proc/PID/smaps

2023-02-15T16:22:19+00:00

commit 3489dbb696d25602aea8c3e669a6d43b76bd5358 upstream. Patch series "Fixes for hugetlb mapcount at most 1 for shared PMDs". This issue of mapcount in hugetlb pages referenced by shared PMDs was discussed in [1]. The following two patches address user visible behavior caused by this issue. [1] https://lore.kernel.org/linux-mm/Y9BF+OCdWnCSilEu@monkey/ This patch (of 2): A hugetlb page will have a mapcount of 1 if mapped by multiple processes via a shared PMD. This is because only the first process increases the map count, and subsequent processes just add the shared PMD page to their page table. page_mapcount is being used to decide if a hugetlb page is shared or private in /proc/PID/smaps. Pages referenced via a shared PMD were incorrectly being counted as private. To fix, check for a shared PMD if mapcount is 1. If a shared PMD is found count the hugetlb page as shared. A new helper to check for a shared PMD is added. [akpm@linux-foundation.org: simplification, per David] [akpm@linux-foundation.org: hugetlb.h: include page_ref.h for page_count()] Link: https://lkml.kernel.org/r/20230126222721.222195-2-mike.kravetz@oracle.com Fixes: 25ee01a2fca0 ("mm: hugetlb: proc: add hugetlb-related fields to /proc/PID/smaps") Signed-off-by: Mike Kravetz Acked-by: Peter Xu Cc: David Hildenbrand Cc: James Houghton Cc: Matthew Wilcox Cc: Michal Hocko Cc: Muchun Song Cc: Naoya Horiguchi Cc: Vishal Moola (Oracle) Cc: Yang Shi Cc: Signed-off-by: Andrew Morton Signed-off-by: Greg Kroah-Hartman

mm/hugetlb: fix races when looking up a CONT-PTE/PMD size hugetlb page

2022-12-14T10:31:59+00:00

commit fac35ba763ed07ba93154c95ffc0c4a55023707f upstream. On some architectures (like ARM64), it can support CONT-PTE/PMD size hugetlb, which means it can support not only PMD/PUD size hugetlb (2M and 1G), but also CONT-PTE/PMD size(64K and 32M) if a 4K page size specified. So when looking up a CONT-PTE size hugetlb page by follow_page(), it will use pte_offset_map_lock() to get the pte entry lock for the CONT-PTE size hugetlb in follow_page_pte(). However this pte entry lock is incorrect for the CONT-PTE size hugetlb, since we should use huge_pte_lock() to get the correct lock, which is mm->page_table_lock. That means the pte entry of the CONT-PTE size hugetlb under current pte lock is unstable in follow_page_pte(), we can continue to migrate or poison the pte entry of the CONT-PTE size hugetlb, which can cause some potential race issues, even though they are under the 'pte lock'. For example, suppose thread A is trying to look up a CONT-PTE size hugetlb page by move_pages() syscall under the lock, however antoher thread B can migrate the CONT-PTE hugetlb page at the same time, which will cause thread A to get an incorrect page, if thread A also wants to do page migration, then data inconsistency error occurs. Moreover we have the same issue for CONT-PMD size hugetlb in follow_huge_pmd(). To fix above issues, rename the follow_huge_pmd() as follow_huge_pmd_pte() to handle PMD and PTE level size hugetlb, which uses huge_pte_lock() to get the correct pte entry lock to make the pte entry stable. Mike said: Support for CONT_PMD/_PTE was added with bb9dd3df8ee9 ("arm64: hugetlb: refactor find_num_contig()"). Patch series "Support for contiguous pte hugepages", v4. However, I do not believe these code paths were executed until migration support was added with 5480280d3f2d ("arm64/mm: enable HugeTLB migration for contiguous bit HugeTLB pages") I would go with 5480280d3f2d for the Fixes: targe. Link: https://lkml.kernel.org/r/635f43bdd85ac2615a58405da82b4d33c6e5eb05.1662017562.git.baolin.wang@linux.alibaba.com Fixes: 5480280d3f2d ("arm64/mm: enable HugeTLB migration for contiguous bit HugeTLB pages") Signed-off-by: Baolin Wang Suggested-by: Mike Kravetz Reviewed-by: Mike Kravetz Cc: David Hildenbrand Cc: Muchun Song Cc: Signed-off-by: Andrew Morton Signed-off-by: Sasha Levin Signed-off-by: Samuel Mendoza-Jonas Signed-off-by: Greg Kroah-Hartman

mm/hugetlb: initialize hugetlb_usage in mm_init

2021-09-18T11:40:36+00:00

commit 13db8c50477d83ad3e3b9b0ae247e5cd833a7ae4 upstream. After fork, the child process will get incorrect (2x) hugetlb_usage. If a process uses 5 2MB hugetlb pages in an anonymous mapping, HugetlbPages: 10240 kB and then forks, the child will show, HugetlbPages: 20480 kB The reason for double the amount is because hugetlb_usage will be copied from the parent and then increased when we copy page tables from parent to child. Child will have 2x actual usage. Fix this by adding hugetlb_count_init in mm_init. Link: https://lkml.kernel.org/r/20210826071742.877-1-liuzixian4@huawei.com Fixes: 5d317b2b6536 ("mm: hugetlb: proc: add HugetlbPages field to /proc/PID/status") Signed-off-by: Liu Zixian Reviewed-by: Naoya Horiguchi Reviewed-by: Mike Kravetz Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds Signed-off-by: Greg Kroah-Hartman

mm, futex: fix shared futex pgoff on shmem huge page

2021-06-30T12:47:29+00:00

commit fe19bd3dae3d15d2fbfdb3de8839a6ea0fe94264 upstream. If more than one futex is placed on a shmem huge page, it can happen that waking the second wakes the first instead, and leaves the second waiting: the key's shared.pgoff is wrong. When 3.11 commit 13d60f4b6ab5 ("futex: Take hugepages into account when generating futex_key"), the only shared huge pages came from hugetlbfs, and the code added to deal with its exceptional page->index was put into hugetlb source. Then that was missed when 4.8 added shmem huge pages. page_to_pgoff() is what others use for this nowadays: except that, as currently written, it gives the right answer on hugetlbfs head, but nonsense on hugetlbfs tails. Fix that by calling hugetlbfs-specific hugetlb_basepage_index() on PageHuge tails as well as on head. Yes, it's unconventional to declare hugetlb_basepage_index() there in pagemap.h, rather than in hugetlb.h; but I do not expect anything but page_to_pgoff() ever to need it. [akpm@linux-foundation.org: give hugetlb_basepage_index() prototype the correct scope] Link: https://lkml.kernel.org/r/b17d946b-d09-326e-b42a-52884c36df32@google.com Fixes: 800d8c63b2e9 ("shmem: add huge pages support") Reported-by: Neel Natu Signed-off-by: Hugh Dickins Reviewed-by: Matthew Wilcox (Oracle) Acked-by: Thomas Gleixner Cc: "Kirill A. Shutemov" Cc: Zhang Yi Cc: Mel Gorman Cc: Mike Kravetz Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Darren Hart Cc: Davidlohr Bueso Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds Signed-off-by: Greg Kroah-Hartman