CVE-2026-62434

When switching from Grant Table v2 to v1 in Xen, code holding the grant table lock but temporarily dropping and re-acquiring it wrongly assumes that table properties won't change during the window without the lock. This leads to issues with status frames that disappear when switching from v2 to v1.

When switching from Grant Table v1 to v2 in Xen, code holding the grant table lock but temporarily dropping and re-acquiring it wrongly assumes that table properties won't change during the window without the lock. This leads to a reduction in the number of valid grant references.

The EVTCHNOP_expand_array hypercall checks for FIFO event channels being enabled, but without holding the correct lock. It can race with EVTCHNOP_reset, resulting in dereferencing a NULL pointer.

The logic to handle periodic Viridian STIMERs performs a division with an unchecked user-controlled divisor value, that can be set to zero to cause a #DE fault.

Accesses to CMOS memory contents are done using an indirect IO port pair. Xen needs to cache the guest chosen index, and one of the usages of the index didn't take the necessary locking to avoid concurrent changes, allowing a guest to change the index after it being checked, causing an out-of-bound read access.

Accessing vNUMA configuration data of a guest is still possible when domain destruction has already started. The cleanup of that configuration information is not synchronized with its retrieval by a device model controlling the guest.

When grant-copy operations are processed, permission checks may be carried out on a page different from the one involved in the copy when the grant is already pinned, leading to inconsistency.

Platform-op operations use a system-wide lock, but the lock acquisition lacks fairness, and with XSM/Flask enabled, the lock is acquired before permission checks. This is the platform-op issue.

Sysctl operations use a system-wide lock, but the lock acquisition lacks fairness, and with XSM/Flask enabled, the lock is acquired before permission checks. This is the sysctl issue.

Addressing certain issues related to long-running operations has proven overly costly. Since alternatives (HVM/PVH: HAP, PV: shim) are commonly available, the decision was to deprecate the functionality, while still retaining it for use at own risk. Memory-wise small guests may still be okay to run.