GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel IMA: integer underflow in xattr_verify() causes out-of-bounds read on truncated security.ima

CVE-2026-74671Kernel, userspace & hypervisorcurated

Impact

A truncated security.ima extended attribute makes the digest-length check underflow to SIZE_MAX, so the following memcmp() reads past the end of the buffer allocated for the xattr value. On a GPU node this matters where IMA appraisal is part of the attested boot chain used to convince tenants the host is what it claims to be: the out-of-bounds read is an integrity-measurement code path, and the record does not establish anything beyond an out-of-bounds read (information disclosure or oops). Nodes with IMA appraisal disabled are not affected at all. Where it is enabled, an attacker able to write a short security.ima value on a file that gets measured reaches the flawed check directly.

Who can reach it

Local, and only on kernels built and booted with IMA appraisal active. Requires the ability to set a truncated security.ima xattr on a file that IMA later verifies, which normally means CAP_SYS_ADMIN on the filesystem in question. Not reachable remotely and not reachable from an unprivileged tenant container without that capability.

What to do

Take the stable-tree fix for your kernel series - it is a pure arithmetic-signedness change in security/integrity/ima/ima_appraise.c with no ABI impact. Applying it means installing a patched kernel and rebooting each node, which for a GPU fleet means draining tenant workloads first; there is no livepatch published in the record. Fleets that do not enable IMA appraisal can defer to the next scheduled kernel roll.

References

Related entries

All Kernel, userspace & hypervisor entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.