PT-2026-151683
CVE-2026-99472
·
تاريخ النشر
2026-11-19
·
تاريخ التحديث
2026-11-19
الوصف من
الدرجة
N/A
غير متوفرلا تتوفر حاليًا تقييمات أو مقاييس للخطورة. سنحدّث المعلومات المقابلة في الصفحة فور توفرها.
In the Linux kernel, the following vulnerability has been resolved:
KVM: pfncache: use a dedicated invalidation sequence for cache refresh
The gfn to pfn cache refresh path guards against mmu notifier
invalidations which complete while it has dropped gpc->lock for the
HVA->PFN lookup: hva to pfn retry() samples kvm->mmu invalidate seq
and retries if it changed, or if mn active invalidate count is still
elevated.
That is insufficient for HVA-based caches. mmu invalidate seq is only
advanced by kvm mmu invalidate end() when the invalidated range
overlaps a memslot, and an HVA-based cache (e.g. the Xen shared info
page mapped with KVM XEN ATTR TYPE SHARED INFO HVA) need not be backed
by any memslot at all. An invalidation of the cached HVA which starts
and ends entirely within the lookup window is thus invisible to the
retry check: mn active invalidate count is back to zero and the
sequence never moved. The refresh then publishes a mapping of a page
which has already been freed, and the next reader dereferences it:
BUG: KASAN: use-after-free in kvm xen shared info init+0x3c6/0x440
Read of size 4 at addr ffff8880599c2900 by task syz.2.383/7257
Since gfn to pfn cache invalidate start() deliberately skips caches
which are not currently valid (including one whose refresh is in
progress, as the refresh clears the valid flag before dropping the
lock), the retry check is the only line of defence, and it must fire
for any invalidation, not just those hitting a memslot.
Add a dedicated kvm->gpc invalidate seq, incremented by every
kvm mmu notifier invalidate range end() under mn invalidate lock
before mn active invalidate count is decremented, and check it in
hva to pfn retry() instead of mmu invalidate seq. Incrementing in
range end() in the same critical section as the in-progress count
also closes the variant where the cache is activated with the
contested HVA only after invalidate range start() has run.
The same bug is also reachable through the per-vCPU vcpu info cache
(KVM XEN VCPU ATTR TYPE VCPU INFO HVA), where the stale mapping is
then dereferenced by kvm setup guest pvclock() on the next KVM RUN:
BUG: KASAN: use-after-free in kvm setup guest pvclock+0x5bf/0x660
This intentionally makes refresh retry on unrelated mmu notifier
events; restoring precision (and reworking the GPC locking more
generally) is left for a subsequent series.
Reproducers: https://david.woodhou.se/xen shinfo race.c
https://david.woodhou.se/vcpu info race.c
وجدت خطأً في الوصف؟ لديك ما تضيفه؟ لا تتردد في مراسلتنا 👾