{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-99728","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-25T10:25:14.431Z","datePublished":"2026-11-20T05:51:03.661Z","dateUpdated":"2026-11-22T11:09:54.039Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-11-22T11:09:54.039Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nmm/huge_memory: fix list corruption when deferred_split_scan() races with deferred_split_folio()\n\nSince the deferred split queue became a list_lru, the isolate callback\nmoves the folios it took a reference on to an on-stack list, linked\nthrough folio->_deferred_list, and deferred_split_scan() takes them off\nthat list with list_del_init() without holding the list_lru lock.\n\ndeferred_split_folio() does not know about the on-stack list. It only\nholds the list_lru lock when it looks at the same list_head, and it can\nrun for such a folio: a THP is queued when it is mapped, as a candidate\nfor the underused check, and is queued \"again\" when a part of it gets\nunmapped:\n\n  CPU0 deferred_split_scan()          CPU1 deferred_split_folio()\n  list_del_init(&folio->_deferred_list)\n    next->prev = prev\n    prev->next = next\n    entry->next = entry\n                                      list_lru_lock_irqsave()\n                                      folio_set_partially_mapped()\n                                      __list_lru_add()\n                                        list_empty(item)  -> true\n                                        list_add_tail(item, &l->list)\n                                          item->next = &l->list\n                                          item->prev = tail\n    entry->prev = entry\n\nThe folio now sits on the list_lru with ->prev pointing to itself. The\nnext removal leaves the neighbours pointing to it, and they still do so\nafter the folio has been freed. With CONFIG_DEBUG_LIST:\n\n  list_del corruption. prev->next should be ffd4000000270090, but was\n  ffd4000000278090. (prev=ffd4000000270090)\n\nFix it by grabbing the lock around list_del_init() so\ndeferred_split_folio() cannot race with it."}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","baseScore":7.8,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - deferred_split_scan() runs from __alloc_pages_direct_reclaim() via try_to_free_pages() and shrink_slab(); the racing deferred_split_folio(true) is reached from madvise(MADV_DONTNEED) through zap_vma_range_batched(), zap_pmd_range(), and folio_remove_rmap_ptes(). Both sides act on the caller's own anonymous THP, and no remote protocol message carries the triggering state.\nAC:L - The caller faults an anonymous THP so map_anon_folio_pmd_nopf() queues it with deferred_split_folio(folio, false) (split_underused_thp defaults to true), allocates until __alloc_pages_direct_reclaim() runs list_del_init() in deferred_split_scan(), and concurrently issues MADV_DONTNEED on a sub-PMD range so __folio_remove_rmap() calls deferred_split_folio(folio, true). The same threads produce both sides.\nPR:L - do_huge_pmd_anonymous_page(), madvise(MADV_DONTNEED), and the allocation that enters __alloc_pages_direct_reclaim() are available to an unprivileged process. No capability is checked; MADV_HUGEPAGE covers sysfs \"madvise\" mode, and TRANSPARENT_HUGEPAGE_ALWAYS is the Kconfig default.\nUI:N - The attacker triggers the race with their own mmap, anonymous page fault, madvise(MADV_DONTNEED), and allocations.\nS:U - The dangling pointer is folio->_deferred_list on deferred_split_lru, kernel metadata of the same kernel the local process is running in.\nC:H - The raced list_del_init() leaves the deferred_split_lru predecessor pointing at folio->_deferred_list after __folio_put() calls free_frozen_pages(). The next deferred_split_isolate() container_of()s that dangling list_head and folio_try_get()s it, reading the recycled struct page (the _deferred_list slot overlaps lru.prev and mapping).\nI:H - Once deferred_split_isolate() walks that freed _deferred_list, __list_del() writes next->prev = prev and prev->next = next using pointers loaded from the recycled struct page, corrupting the kernel objects those lru and mapping words address.\nA:H - The predecessor link left by the raced list_del_init() still names the folio after it is freed, so the next deferred_split_scan() dereferences it. CONFIG_DEBUG_LIST reports this as list_del corruption and BUGs; otherwise the wild pointer in deferred_split_isolate() oopses the kernel."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["mm/huge_memory.c"],"versions":[{"version":"fafaeceb89a5e2e856ff04c2cacb6cae4a2ecb67","lessThan":"877a9fbabf760714f510f36b310a986c4be2c2d8","status":"affected","versionType":"git"},{"version":"fafaeceb89a5e2e856ff04c2cacb6cae4a2ecb67","lessThan":"f09171b67ea06a7fe425db33e8b74a92de1426b9","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["mm/huge_memory.c"],"versions":[{"version":"7.2","status":"affected"},{"version":"0","lessThan":"7.2","status":"unaffected","versionType":"semver"},{"version":"7.3.5","lessThanOrEqual":"7.3.*","status":"unaffected","versionType":"semver"},{"version":"7.4-rc2","lessThanOrEqual":"*","status":"unaffected","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"operator":"OR","negate":false,"cpeMatch":[{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"7.2","versionEndExcluding":"7.3.5"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"7.2","versionEndExcluding":"7.4-rc2"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/877a9fbabf760714f510f36b310a986c4be2c2d8"},{"url":"https://git.kernel.org/stable/c/f09171b67ea06a7fe425db33e8b74a92de1426b9"}],"title":"mm/huge_memory: fix list corruption when deferred_split_scan() races with deferred_split_folio()","x_generator":{"engine":"bippy-1.2.0"}}}}