# iommu/vt-d: Block PASID attachment to nested domain with dirty tracking

CVE record · `CVE-2026-53372`

- **Identifiers:** `CVE-2026-53372`
- **State:** Confirmed
- **Severity:** No CVSS score
- **Critical:** No
- **CNA:** Linux
- **Technologies:** [Linux kernel](https://agentprime.io/advisories/technology/linux-kernel.md)
- **Published:** 19 Jul 2026
- **Updated:** 6 Nov 2026
- **Reviewed by a person:** No

Published by the CVE Program (CNA: Linux) at 19 Jul 2026, 09:10 UTC. agentprime has not analysed this yet, so whether it affects any app is not known. This CVE names a vendor and product rather than a package, so the lockfile check doesn’t match it.

## Affected products, as the CNA wrote them

- Vendor: Linux. Product: Linux.
  - affected: `67f6f56b59126b3bf4fc6f4ea564e450fcfcf9f6 to before 3ea9ce757bd3de955b56e7bc5672fc479e40b045` (git)
  - affected: `67f6f56b59126b3bf4fc6f4ea564e450fcfcf9f6 to before 9009c1af5458322469fa9a4371081a4449c5947d` (git)
  - affected: `67f6f56b59126b3bf4fc6f4ea564e450fcfcf9f6 to before cc5bd898ff70710ffc41cd8e5c2741cb64750047` (git)
  - Other versions: unaffected
- Vendor: Linux. Product: Linux.
  - affected: `6.13`
  - unaffected: `before 6.13` (semver)
  - unaffected: `6.18.30 to 6.18.*, inclusive` (semver)
  - unaffected: `7.0.7 to 7.0.*, inclusive` (semver)
  - unaffected: `7.1 to *, inclusive` (original_commit_for_fix)
  - Other versions: affected

## Machine-written summary

Summary written by Claude Haiku from the source advisory. Not reviewed by a person.

> The Linux kernel's Intel VT-d IOMMU code, in version 6.13, can lose dirty pages when a PASID is attached to a nested domain without dirty tracking while its parent domain has it enabled. The source gives few details on attack requirements. Upgrade to 6.18.30 or later, 7.0.7 or later, or 7.1 or later.

Written 17 Nov 2026.

## Not analysed yet

We haven’t published our own analysis of this CVE. If a person reviews it, this page will say what’s affected and what has to be true for it to matter.

## Severity and exploitation

No source has published a CVSS score for this record yet.

- **CISA KEV:** Not listed.
- **EPSS:** 0.0016, the 5th percentile, on 17 Nov 2026: FIRST’s estimate of the chance of exploitation in the next 30 days.

## Timeline

- 19 Jul 2026, 09:10 UTC: [Confirmed by the CVE List](https://www.cve.org/CVERecord?id=CVE-2026-53372)
- 19 Jul 2026, 09:17 UTC: [Confirmed by NVD](https://nvd.nist.gov/vuln/detail/CVE-2026-53372)

## Sources

- [CVE List: CVE-2026-53372](https://www.cve.org/CVERecord?id=CVE-2026-53372)
- [NVD: CVE-2026-53372](https://nvd.nist.gov/vuln/detail/CVE-2026-53372)

### References from the CNA

- <https://git.kernel.org/stable/c/3ea9ce757bd3de955b56e7bc5672fc479e40b045>
- <https://git.kernel.org/stable/c/9009c1af5458322469fa9a4371081a4449c5947d>
- <https://git.kernel.org/stable/c/cc5bd898ff70710ffc41cd8e5c2741cb64750047>

Facts on this page come from the CVE List as the CNA published them, with CVSS from the sources named above, CISA KEV and FIRST EPSS. Vendors, products and versions are shown as the CNA wrote them. We don’t reproduce the CVE’s description.

## Check whether this affects you

This CVE names a vendor and product rather than a package, so the lockfile check doesn’t match it. Compare the affected versions above with what you run.

- With the agentprime MCP server (<https://api.agentprime.io/mcp>), `get_advisory` returns this record and `search_advisories` finds related ones. Setup: <https://agentprime.io/mcp.md>
- As JSON: <https://api.agentprime.io/v1/advisories/CVE-2026-53372>
- Everything that needs attention now: <https://agentprime.io/advisories/critical.md>

---

The web page: <https://agentprime.io/advisories/CVE-2026-53372/>. Everything for agents: <https://agentprime.io/llms.txt>.
