{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-89450","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-11T19:38:34.706Z","datePublished":"2026-09-11T19:43:16.060Z","dateUpdated":"2026-09-13T06:29:30.075Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-09-13T06:29:30.075Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\niommu/tegra241-cmdqv: Reject a vSID wider than the SID_MATCH field\n\ntegra241_vintf_init_vsid() programs the guest-provided vSID into SID_MATCH,\nwhose VIRT_SID field spans bits [20:1] with bit 0 as the match-enable flag.\nThe HW therefore matches only a 20-bit Stream ID.\n\nThe bound check rejects only virt_sid > UINT_MAX, which admits a value far\nwider than the field. The write \"virt_sid << 1 | 0x1\" then drops every bit\nabove 20: a virt_sid of 0x80000000 lands as SID_MATCH = 0x1, a valid match\non vSID 0, so the entry aliases the wrong Stream ID. Because vdev->virt_id\nis guest-controlled, a VMM can trigger it.\n\nValidate virt_sid against the field width with FIELD_MAX(), and program the\nregister with FIELD_PREP() so the value and the field stay consistent."}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H","baseScore":8.8,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - Reached only via local ioctl on /dev/iommu or VFIO-compat /dev/vfio/vfio: iommufd_fops_ioctl() dispatches IOMMU_VDEVICE_ALLOC to iommufd_vdevice_alloc_ioctl(), which calls tegra241_vintf_init_vsid() to program SID_MATCH. There is no network or packet path into CMDQV SID programming.\nAC:L - The attacker fully chooses cmd->virt_id; the old UINT_MAX check accepts values such as 0x80000000, and writel(virt_sid<<1|0x1) deterministically truncates them onto a 20-bit SID_MATCH (aliasing vSID 0). No race or uninfluenced memory layout is required on NVIDIA Grace/CMDQV.\nPR:L - iommufd_fops_open() and the ioctl table have no capable() gate (/dev/iommu 0660, /dev/vfio/vfio 0666). A tenant or unprivileged QEMU user granted VFIO/iommufd access on Grace can bind a CMDQV-backed device, alloc a TEGRA241_CMDQV vIOMMU, and issue VDEVICE_ALLOC without init-namespace root.\nUI:N - The attacker performs the full sequence themselves (open iommufd, bind the assigned device, alloc vIOMMU/vDEVICE with a truncated virt_id, then guest ATC_INV via VCMDQ). No separate victim action such as mounting a filesystem is required.\nS:C - Truncating a guest-controlled vSID into SID_MATCH aliases it onto another Stream ID so CMDQV rewrites ATC_INV/CFGI to the wrong pSID. That is an IOMMU/DMA boundary bypass: missed invalidations leave stale ATS/IOTLB entries and let device DMA reach host memory outside the guest domain.\nC:H - Wrong SID_MATCH/SID_REPLACE pairing leaves stale ATS translations after IOVA unmap, so a guest-controlled passthrough device can DMA-read host pages recycled by the allocator, yielding an arbitrary host-memory disclosure primitive through the IOMMU.\nI:H - The same stale or retargeted translations allow DMA writes into unmapped host pages, and aliased CFGI/ATC_INV operate on the wrong physical Stream ID, corrupting IOMMU translation state and providing an arbitrary write suitable for control-flow hijack.\nA:H - Misdirected invalidation and CFGI cause SMMU translation faults, DMA into recycled kernel pages, and CMDQV GERROR/ISR failures that oops or panic the host; the truncated vSID can be programmed repeatedly on every VDEVICE_ALLOC."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c"],"versions":[{"version":"4dc0d12474f9d4833c3dd96b73d61e406d3f5dc7","lessThan":"d903d99ffd22b0180bd745a43f221c21bcdd8d7c","status":"affected","versionType":"git"},{"version":"4dc0d12474f9d4833c3dd96b73d61e406d3f5dc7","lessThan":"445204550f894ca325ac80a21e3df177ad073798","status":"affected","versionType":"git"},{"version":"4dc0d12474f9d4833c3dd96b73d61e406d3f5dc7","lessThan":"4379610c79bd88ddbea10e7f6c21e16d4b338c6b","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c"],"versions":[{"version":"6.17","status":"affected"},{"version":"0","lessThan":"6.17","status":"unaffected","versionType":"semver"},{"version":"6.18.50","lessThanOrEqual":"6.18.*","status":"unaffected","versionType":"semver"},{"version":"7.2.4","lessThanOrEqual":"7.2.*","status":"unaffected","versionType":"semver"},{"version":"7.3-rc1","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":"6.17","versionEndExcluding":"6.18.50"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.17","versionEndExcluding":"7.2.4"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.17","versionEndExcluding":"7.3-rc1"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/d903d99ffd22b0180bd745a43f221c21bcdd8d7c"},{"url":"https://git.kernel.org/stable/c/445204550f894ca325ac80a21e3df177ad073798"},{"url":"https://git.kernel.org/stable/c/4379610c79bd88ddbea10e7f6c21e16d4b338c6b"}],"title":"iommu/tegra241-cmdqv: Reject a vSID wider than the SID_MATCH field","x_generator":{"engine":"bippy-1.2.0"}}}}