CVE-2026-74439

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> iommu/vt-d: Clear Present bit before tearing down scalable-mode context entry<br /> <br /> device_pasid_table_teardown() zeroes the 128-bit scalable-mode context<br /> entry with context_clear_entry() while the Present bit is still set. This<br /> creates a window where the hardware can fetch a torn entry, with some<br /> fields already zeroed while Present is still set, leading to unpredictable<br /> behavior or spurious faults. The context-cache invalidation is issued only<br /> after the entry has been zeroed, and intel_pasid_free_table() then frees<br /> the PASID directory pages, so the IOMMU can keep walking a stale Present=1<br /> entry that points at freed memory.<br /> <br /> While x86 provides strong write ordering, the compiler may reorder the two<br /> 64-bit writes to the entry, and the hardware fetch is not guaranteed to be<br /> atomic with respect to multiple CPU writes.<br /> <br /> Commit c1e4f1dccbe9d ("iommu/vt-d: Clear Present bit before tearing down<br /> context entry") fixed this exact pattern in domain_context_clear_one() and<br /> the copied-context path, but device_pasid_table_teardown() was not<br /> converted.<br /> <br /> Align it with the "Guidance to Software for Invalidations" in the VT-d<br /> spec, Section 6.5.3.3, using the same ownership handshake as the sibling<br /> fix: clear only the Present bit, flush it to the IOMMU, perform the<br /> context-cache invalidation, and only then zero the rest of the entry.