{
  "schema_version": "1.6.1",
  "id": "RPI-CVE-2024-50115",
  "modified": "2026-10-09T11:39:44Z",
  "upstream": [
    "CVE-2024-50115"
  ],
  "published": "2024-11-05T17:11:07Z",
  "summary": "KVM: nSVM: Ignore nCR3[4:0] when loading PDPTEs from memory",
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nKVM: nSVM: Ignore nCR3[4:0] when loading PDPTEs from memory\n\nIgnore nCR3[4:0] when loading PDPTEs from memory for nested SVM, as bits\n4:0 of CR3 are ignored when PAE paging is used, and thus VMRUN doesn't\nenforce 32-byte alignment of nCR3.\n\nIn the absolute worst case scenario, failure to ignore bits 4:0 can result\nin an out-of-bounds read, e.g. if the target page is at the end of a\nmemslot, and the VMM isn't using guard pages.\n\nPer the APM:\n\n  The CR3 register points to the base address of the page-directory-pointer\n  table. The page-directory-pointer table is aligned on a 32-byte boundary,\n  with the low 5 address bits 4:0 assumed to be 0.\n\nAnd the SDM's much more explicit:\n\n  4:0    Ignored\n\nNote, KVM gets this right when loading PDPTRs, it's only the nSVM flow\nthat is broken.",
  "severity": [
    {
      "type": "CVSS_V3",
      "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:N/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": "6.1.115",
            "compiled_filter": false
          }
        }
      }
    }
  ],
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://www.cve.org/CVERecord?id=CVE-2024-50115"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/76ce386feb14ec9a460784fcd495d8432acce7a5"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/58cb697d80e669c56197f703e188867c8c54c494"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/6876793907cbe19d42e9edc8c3315a21e06c32ae"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/2c4adc9b192a0815fe58a62bc0709449416cc884"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/426682afec71ea3f889b972d038238807b9443e4"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/f559b2e9c5c5308850544ab59396b7d53cfc67bd"
    }
  ],
  "database_specific": {
    "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git",
    "upstream_modified": "2026-08-05T11:52:36Z",
    "severity": {
      "source": "kernel CNA",
      "cvss_v3_score": 8.4,
      "cvss_v3_severity": "High",
      "scenario": "AV:L - The attacker must execute code on the host machine, specifically CPL0 inside an L1 guest, issuing VMRUN with an attacker-chosen nCR3; there is no network protocol involved. Guest-to-host attacks against KVM are scored as local access to the host.\nAC:L - Every precondition is under the attacker's control \u2014 L1 chooses 32-bit PAE mode, sets the NPT enable bit, and sets nCR3[4:0]; nested SVM and NPT are enabled by default in kvm_amd, and nCR3 receives no alignment validation, so the out-of-bounds read triggers deterministically on the first VMRUN. The guest knows its own e820 map and can aim at a memslot-terminating gfn and retry against every memslot boundary.\nPR:L - The attacker needs kernel-level (CPL0) control of a guest VM whose CPUID exposes SVM \u2014 i.e. an ordinary tenant VM on a nested-virtualization-enabled AMD host, which is a low privilege level relative to the host kernel being attacked. No host-side root or /dev/kvm privileges beyond running one's own VM are required.\nUI:N - Exploitation is entirely driven by instructions the attacker executes in its own guest; no host administrator or other user needs to take any action.\nS:C - The vulnerable component is KVM in the host kernel, but the impacted resource is host VMM process memory lying outside the guest's memslot, so the attack crosses the guest/host security boundary. This is the classic KVM guest-to-host scope change.\nC:H - Up to 31 bytes past the end of every memslot in the host VMM's address space can be read, and the value is recoverable by the guest through the shadow-NPT root oracle in mmu_alloc_shadow_roots(), disclosing host heap contents and pointers that defeat host ASLR. It is repeatable at each memslot boundary, so it is not strictly bounded to a few bytes at one fixed location.\nI:N - The defect is purely a read; the resulting PDPTE only influences translations that remain gated by kvm_vcpu_is_visible_gfn() and normal memslot lookup, and any accessed/dirty writes land in the attacker's own L1 memory. No host memory is corrupted and no control-flow primitive is produced.\nA:H - In affected v6.12-rc1..rc4 kernels the same guest-controlled offset trips the WARN_ON_ONCE(offset + len > PAGE_SIZE) hardening check in __kvm_read_guest_page(), giving a guest-triggerable host kernel WARN that panics the entire host \u2014 and all co-resident VMs \u2014 on the widely used panic_on_warn=1 configuration."
    }
  }
}