{
  "affected": [
    {
      "ranges": [
        {
          "events": [
            {
              "introduced": "6375e95f381e3dc85065b6f74263a61522736203"
            },
            {
              "fixed": "8011e987026b65f59ea6b6082c777807a008b227"
            },
            {
              "fixed": "b57cfd4da476a4636bf1ed69dee176cc266e5a06"
            },
            {
              "fixed": "243cf1f67b365668389832463fc31cbb9fec47af"
            }
          ],
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "type": "GIT"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Linux",
        "name": "Kernel"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "6.14.0"
            },
            {
              "fixed": "6.18.64"
            }
          ],
          "type": "ECOSYSTEM"
        },
        {
          "events": [
            {
              "introduced": "6.19.0"
            },
            {
              "fixed": "7.3.5"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "database_specific": {
    "cna_assigner": "Linux",
    "osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/99xxx/CVE-2026-99705.json"
  },
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nmm: memory: fix truncation removing mapped folios\n\nAyush reported an issue that shmem/tmpfs folios are still mapped when\ntruncating them on production:\n\"\nBUG: Bad page cache in process ...  pfn:...\npage dumped because: still mapped when deleted\n\"\n\nI reproduced the issue with Ayush's reproducer [1] with the latest\nmm-unstable branch, and I got the following crash info.  From the dump\nmessage, we can see that truncate_inode_folio() is really trying to remove\nmapped folios, which is incorrect.\n\n\"\n[  189.171977] page: refcount:3 mapcount:1 mapping:00000000fe168ee1\nindex:0x3880 pfn:0x191647\n[  189.171995] memcg:ffff0000cc073d40\n[  189.171997] aops:shmem_aops ino:3c01 dentry name(?):\"memfd:runsc-memory\"\n[  189.172006] flags: 0x17fffef0002022d(locked|referenced|uptodate|lru|\nworkingset|swapbacked|node=0|zone=2|lastcpupid=0x3ffff)\n.......\n[  189.172017] page dumped because: VM_BUG_ON_FOLIO(folio_mapped(folio))\n[  189.172026] ------------[ cut here ]------------\n[  189.172027] kernel BUG at mm/filemap.c:155!\n[  189.172057] Internal error: Oops - BUG: 00000000f2000800 [#1]  SMP\n.......\n[  189.177721] CPU: 3 UID: 0 PID: 5521 Comm: repro Kdump: loaded Tainted:\n7.3.0-rc4+ #277 PREEMPT(lazy)\n.......\n[  189.183986] Call trace:\n[  189.184124]  filemap_unaccount_folio+0xf0/0x1e8 (P)\n[  189.184391]  __filemap_remove_folio+0x34/0x160\n[  189.184633]  filemap_remove_folio+0x4c/0xb0\n[  189.184859]  truncate_inode_folio+0x34/0x58\n[  189.185087]  shmem_undo_range+0x220/0x658\n[  189.185308]  shmem_fallocate+0x2f0/0x470\n[  189.185527]  vfs_fallocate+0x128/0x328\n......\n[  189.187078]  el0t_64_sync+0x184/0x188\n\"\n\nAfter analysis, I believe the race exists between truncation and\nMADV_DONTNEED, and shmem's -\u003emap_pages() merely makes the issue easier to\nreproduce.  Since MADV_DONTNEED synchronously releases the pagetable page\nbefore calling tlb_flush_rmaps(), this could cause another thread's\ntruncation to skip zap_pte_range() (due to pmd is none) but still observe\nthe folio's mapcount as non-zero.  A possible race scenario is as follows:\n\nCPU 0                           CPU 1\nshmem_fallocate                 madvise_dontneed_single_vma\n  unmap_mapping_range            .......\n  ......\n  (filemap_map_pages() remap folios)\n\t\t\t\t  zap_pte_range\n  truncate_inode_folio\t\t    zap_empty_pte_table(pmd clear)\n    unmap_mapping_folio\n    ......\n      zap_pmd_range (saw pmd none, skip)\n\n  filemap_remove_folio\n    BUG_ON(folio_mapped)\n                                    tlb_flush_rmaps(remove rmaps)\n\nTo fix this issue, we should move the pmd clear operation (via\nzap_empty_pte_table()) to after tlb_flush_rmaps() and use\nsync_with_folio_pmd_zap() to sync with concurrent zapping, so that when we\nfind a pmd_none() while unmapping a folio without holding the PTL, the\nfolio's mappings are guaranteed to have been removed via\ntlb_flush_rmaps().\n\nMeanwhile move the second pmd_none() check to where it belongs: if\nsplitting zapped the PMD.",
  "id": "CVE-2026-99705",
  "modified": "2026-11-20T06:45:21.108030693Z",
  "published": "2026-11-20T05:50:46.263Z",
  "references": [
    {
      "type": "PACKAGE",
      "url": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/243cf1f67b365668389832463fc31cbb9fec47af"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/8011e987026b65f59ea6b6082c777807a008b227"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/b57cfd4da476a4636bf1ed69dee176cc266e5a06"
    },
    {
      "type": "ADVISORY",
      "url": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/99xxx/CVE-2026-99705.json"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-99705"
    }
  ],
  "schema_version": "1.9.2",
  "summary": "mm: memory: fix truncation removing mapped folios"
}