{
  "schema_version": "1.6.1",
  "id": "RPI-CVE-2023-53810",
  "modified": "2026-10-09T11:39:44Z",
  "upstream": [
    "CVE-2023-53810"
  ],
  "published": "2025-12-09T00:01:55Z",
  "summary": "blk-mq: release crypto keyslot before reporting I/O complete",
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nblk-mq: release crypto keyslot before reporting I/O complete\n\nOnce all I/O using a blk_crypto_key has completed, filesystems can call\nblk_crypto_evict_key().  However, the block layer currently doesn't call\nblk_crypto_put_keyslot() until the request is being freed, which happens\nafter upper layers have been told (via bio_endio()) the I/O has\ncompleted.  This causes a race condition where blk_crypto_evict_key()\ncan see 'slot_refs != 0' without there being an actual bug.\n\nThis makes __blk_crypto_evict_key() hit the\n'WARN_ON_ONCE(atomic_read(&slot->slot_refs) != 0)' and return without\ndoing anything, eventually causing a use-after-free in\nblk_crypto_reprogram_all_keys().  (This is a very rare bug and has only\nbeen seen when per-file keys are being used with fscrypt.)\n\nThere are two options to fix this: either release the keyslot before\nbio_endio() is called on the request's last bio, or make\n__blk_crypto_evict_key() ignore slot_refs.  Let's go with the first\nsolution, since it preserves the ability to report bugs (via\nWARN_ON_ONCE) where a key is evicted while still in-use.",
  "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": [
        "1:6.1.21-1+rpt1"
      ],
      "ecosystem_specific": {
        "binaries": [
          {
            "binary_name": "linux-image-6.1.0-rpi1-rpi-v6",
            "binary_version": "1:6.1.21-1+rpt1"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi1-rpi-v7",
            "binary_version": "1:6.1.21-1+rpt1"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi1-rpi-v7l",
            "binary_version": "1:6.1.21-1+rpt1"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi1-rpi-v8",
            "binary_version": "1:6.1.21-1+rpt1"
          }
        ]
      },
      "database_specific": {
        "per_version": {
          "1:6.1.21-1+rpt1": {
            "linux_commit": "0afb5e98488aed7017b9bf321b575d0177feb7ed",
            "upstream_version": "6.1.21",
            "assessment": "commit ancestry (strak rule)",
            "upstream_fixed": "6.1.28"
          }
        }
      }
    }
  ],
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://www.cve.org/CVERecord?id=CVE-2023-53810"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/874bdf43b4a7dc5463c31508f62b3e42eb237b08"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/d206f79d9cd658665b37ce8134c6ec849ac7af0c"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/7d206ec7a04e8545828191b6ea8b49d3ea61391f"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/b278570e2c59d538216f8b656e97680188a8fba4"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/92d5d233b9ff531cf9cc36ab4251779e07adb633"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/9cd1e566676bbcb8a126acd921e4e194e6339603"
    }
  ],
  "database_specific": {
    "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git",
    "upstream_modified": "2026-08-05T11:24:39Z",
    "severity": {
      "source": "kernel CNA",
      "cvss_v3_score": 7.8,
      "cvss_v3_severity": "High",
      "scenario": "AV:L - Triggering requires local filesystem I/O on an fscrypt-encrypted file plus locally-driven inode/key eviction (unlink/close or FS_IOC_REMOVE_ENCRYPTION_KEY), all via ordinary syscalls on the local system. No network protocol handler is involved.\nAC:L - The attacker controls both sides of the race \u2014 submitting encrypted I/O and forcing key eviction \u2014 and can retry the loop indefinitely across threads/CPUs until the window between bio_endio() and blk_mq_free_request() is hit. The subsequent reprogram_all_keys() trigger occurs routinely on suspend/resume and UFS error recovery, so no attacker-uncontrollable precondition gates exploitation.\nPR:L - Only an ordinary unprivileged local account (or an Android app) with write access to an fscrypt-encrypted directory is needed; per-file keys are the default fscrypt mode and every app's data directory qualifies. No capabilities or root are required.\nUI:N - The attacker performs every step \u2014 the I/O, the key eviction, and the heap spray \u2014 with no action from any other user. Device reset/resume that consumes the dangling pointer happens automatically or can be induced by the attacker.\nS:U - The use-after-free and the resulting memory corruption occur within the kernel's own security authority; no VM, IOMMU, or sandbox boundary is crossed.\nC:H - The freed blk_crypto_key is re-read and its attacker-controlled crypto_mode is used as an unbounded array index in ufshcd_crypto_keyslot_program(), yielding an arbitrary out-of-bounds kernel read oracle; the stale keyslot also causes stored data to be decrypted under the wrong key, compromising fscrypt's confidentiality guarantees.\nI:H - The same UAF gives an attacker-controlled key->size to a memcpy into a 72-byte stack union, an unbounded stack buffer overflow with attacker-supplied bytes suitable for control-flow hijacking; additionally, attacker-chosen key material can be programmed into hardware keyslots and on-disk data silently encrypted under the wrong key.\nA:H - The immediate effect is a WARN_ON_ONCE (a panic under panic_on_warn) and a use-after-free that dereferences freed/wild pointers during UFS reset or resume, causing kernel oops or panic; corrupting the hardware keyslot state also breaks storage encryption and can render the filesystem unusable."
    }
  }
}