{
  "schema_version": "1.6.1",
  "id": "RPI-CVE-2026-74482",
  "modified": "2026-10-09T11:39:44Z",
  "upstream": [
    "CVE-2026-74482"
  ],
  "published": "2026-08-15T12:28:42Z",
  "summary": "mm/huge_memory: unlock i_mmap_rwsem before releasing after-split folios",
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nmm/huge_memory: unlock i_mmap_rwsem before releasing after-split folios\n\n__folio_split() keeps dereferencing the mapping after the split:\nshmem_uncharge(mapping->host) and remap_page() while the folios are still\nfrozen/locked, and i_mmap_unlock_read(mapping) at the very end, after the\nafter-split folios have been unlocked and freed.\n\nNothing holds an inode reference across that.  The split relies on @folio\n-- which the beyond-EOF drop loop never removes, as it starts at\nfolio_next(folio) -- staying locked and in the page cache to hold off\neviction.  But the unlock loop unlocks @folio before i_mmap_unlock_read()\nruns.  If the caller's @lock_at is a tail beyond EOF, as memory_failure()\npasses when splitting a poisoned tail of a shmem THP that reaches past\ni_size during truncation, it too is gone from the page cache; so once\n@folio is unlocked no locked, in-cache folio pins the inode, and a\nconcurrent final iput() can evict and RCU-free it before\ni_mmap_unlock_read() touches i_mmap_rwsem:\n\n  BUG: KASAN: slab-use-after-free in __up_read+0x634/0x790\n   i_mmap_unlock_read include/linux/fs.h:537 [inline]\n   __folio_split+0x732/0x1640 mm/huge_memory.c:4100\n   try_to_split_thp_page+0xab/0x390 mm/memory-failure.c:1675\n   memory_failure+0x1394/0x26e0 mm/memory-failure.c:2470\n\n  Freed by task 4601:\n   shmem_free_in_core_inode+0x54/0xb0 mm/shmem.c:5177\n   evict+0x57f/0xac0 fs/inode.c:870\n\nDo every mapping dereference while @folio still pins the inode: drop\ni_mmap_rwsem right after remap_page(), before the loop that unlocks and\nfrees the after-split folios, and clear @mapping so the exit path does not\nunlock it again.  shmem_uncharge() and remap_page() already run before\nthat point, so after this nothing past the unlock loop touches the inode\nor the mapping.\n\nThis is now a rule the split depends on, alongside keeping @folio frozen\nuntil the page cache is updated: no inode or mapping dereference once the\nafter-split folios start being unlocked.",
  "severity": [
    {
      "type": "CVSS_V3",
      "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H"
    }
  ],
  "affected": [
    {
      "package": {
        "ecosystem": "Raspberry Pi OS:12",
        "name": "linux"
      },
      "versions": [
        "1:6.1.47-1+rpt4"
      ],
      "ecosystem_specific": {
        "binaries": [
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-2712",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v6",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v7",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v7l",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v8",
            "binary_version": "1:6.1.47-1+rpt4"
          }
        ]
      },
      "database_specific": {
        "per_version": {
          "1:6.1.47-1+rpt4": {
            "linux_commit": "655fc658a15ae7a6f37103754adb39ba52a9a14e",
            "upstream_version": "6.1.47",
            "assessment": "commit ancestry (strak rule)",
            "upstream_fixed": "6.1.183",
            "compiled_filter": false
          }
        }
      }
    }
  ],
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://www.cve.org/CVERecord?id=CVE-2026-74482"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/bc2f5eabaaf60ec18da70b619a8fba1bfb7dea3a"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/f87c08060818ebb19bafed37c38244538da25097"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/6f5c272d71845a669e4c8ee5c72b376e29a0e6e5"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/be106f7855f03d3128ed0ce70ba74b484a90b473"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/e3dd774dbfd0b5bc2dbd0995221751b1234f8205"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/10065fb891651d9541e7a5a2db84c1e656ece4f9"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/d640efe94d86d3be893d4c19220362546a637e90"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/e923bd21058ea02fd0dcd3549d151d143fd036e5"
    }
  ],
  "database_specific": {
    "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git",
    "upstream_modified": "2026-08-19T16:47:12Z",
    "severity": {
      "source": "kernel CNA",
      "cvss_v3_score": 7.8,
      "cvss_v3_severity": "High",
      "scenario": "AV:L - The UAF is in __folio_split() on the memory_failure()->try_to_split_thp_page() path while splitting file-backed shmem THPs; reachability is via local syscalls (memfd/mmap/madvise/ftruncate) or admin hwpoison injection, not network or physical interfaces.\nAC:L - An attacker can deterministically create a shmem THP with tail pages beyond EOF and race MADV_HWPOISON-driven memory_failure() against concurrent truncate/close(iput) using attacker-controlled threads; both sides of the race are attacker-driven.\nPR:L - Reliable exploitation uses MADV_HWPOISON to invoke memory_failure() on a chosen tail page; CAP_SYS_ADMIN is available to user-namespace/container root, and the shmem THP setup needs only ordinary local mapping and truncation privileges.\nUI:N - No separate victim action is required; the attacker creates, maps, poisons, and truncates their own memfd/shmem/tmpfs object to reach the vulnerable split path.\nS:U - Impact is a kernel slab use-after-free on inode/mapping metadata in core MM within the same kernel security domain, not a VM escape, sandbox breakout, or other cross-authority boundary violation.\nC:H - KASAN reports slab-use-after-free in i_mmap_unlock_read() after shmem_free_in_core_inode() RCU-frees the inode; this inode/mapping UAF can be heap-sprayed into an arbitrary kernel memory read primitive.\nI:H - The same UAF corrupts freed inode slab objects and i_mmap_rwsem state, enabling heap grooming for arbitrary kernel writes and potential privilege escalation via control-flow hijacking per kernel UAF guidance.\nA:H - The use-after-free provokes a KASAN BUG/oops during i_mmap_unlock_read() and corrupts core MM inode state, capable of kernel panic or hang when triggered on production shmem/tmpfs workloads."
    }
  }
}