PT-2026-151515
CVE-2026-99334
·
Published
2026-11-19
·
Updated
2026-11-19
Description by
Score
N/A
Not availableNo severity ratings or metrics are available. When they are, we'll update the corresponding info on the page.
In the Linux kernel, the following vulnerability has been resolved:
KVM: PPC: Book3S HV: Add preempt disable() around virtual-mode HPTE bit-lock users
kvmppc hv find lock hpte() requires virtual-mode callers to run with
preemption disabled, because it can return with HPTE V HVLOCK still held
until the caller later unlocks the HPTE. Existing virtual-mode callers
in book3s 64 mmu hv.c already follow that rule.
Two virtual-mode call paths do not:
-
kvmppc handle exit hv() calls kvmppc hpte hv fault() for hash-mode data-side and instruction-side faults after guest exit, with preemption already enabled. kvmppc hpte hv fault() calls kvmppc hv find lock hpte() directly.
-
kvmppc pseries do hcall() executes virtual-mode handlers for HPT hcalls via kvmppc pseries do hpt hcall() with preemption enabled. The handlers for H ENTER, H REMOVE, H READ, H CLEAR MOD, H CLEAR REF, H PROTECT, and H BULK REMOVE spin on try lock hpte() or lock rmap(), and H ENTER also reaches kvmppc do h enter(), which uses arch spin lock() on kvm->mmu lock. That raw lock choice is intentional because kvmppc do h enter() is also used by real-mode callers, so the correct fix is to establish the proper preemption context at the virtual-mode caller boundary.
If a vCPU thread is preempted while holding HPTE V HVLOCK, any other
thread on the same CPU trying to acquire the same bit-lock can spin
indefinitely. The same applies to lock rmap() users and to the H ENTER
path while holding the raw mmu lock.
Fix this by adding preempt disable()/preempt enable() pairs around the
two kvmppc hpte hv fault() call sites in kvmppc handle exit hv(), and
around the kvmppc pseries do hpt hcall() invocation in
kvmppc pseries do hcall(). Targeting only the HPT hcall helper avoids
wrapping non-HPT hcalls (such as H RTAS, H CONFER, H REGISTER VPA) that
perform page faults, blocking memory allocations, or context switches.
Found an issue in the description? Have something to add? Feel free to write us 👾