[PATCH] KVM: x86: Bypass cpuid check for empty arch-lbr leaf

From: Yang Weijiang
Date: Mon May 30 2022 - 11:43:14 EST


On arch-lbr capable platforms, cpuid(0x1c,0) returns meaningful
arch-lbr supported values, in this case, eax[7:0] = lbr depth mask.
Whereas on legacy platforms(non-arch-lbr), cpuid(0x1c,0) returns
with eax/ebx/ecx/edx zeroed out.

On legacy platforms, during selftests app startup, it first gets
supported cpuids by KVM_GET_SUPPORTED_CPUID then sets the returned
data with KVM_SET_CPUID2, this leads to empty cpuid leaf(0x1c,0)
written to KVM and makes the check fail, app finally ends up with
below error message when run selftest:

KVM_SET_CPUID2 failed, rc: -1 errno: 22

So check the validity of the leaf(0x1c,0) before validate lbr
depth value.

QEMU filters out empty CPUID leaves before calls KVM_SET_CPUID2,
so this is not a problem.

Fixes: 4b73207592: ("KVM: x86/cpuid: Advertise Arch LBR feature in CPUID")
Signed-off-by: Yang Weijiang <weijiang.yang@xxxxxxxxx>
---
arch/x86/kvm/cpuid.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index 9c107b5cc88f..c2eab1a73aab 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -103,13 +103,14 @@ static int kvm_check_cpuid(struct kvm_vcpu *vcpu,
return -EINVAL;
}
best = cpuid_entry2_find(entries, nent, 0x1c, 0);
- if (best) {
+ if (best && best->eax) {
unsigned int eax, ebx, ecx, edx;

/* Reject user-space CPUID if depth is different from host's.*/
cpuid_count(0x1c, 0, &eax, &ebx, &ecx, &edx);

- if ((best->eax & 0xff) != BIT(fls(eax & 0xff) - 1))
+ if ((eax & 0xff) &&
+ (best->eax & 0xff) != BIT(fls(eax & 0xff) - 1))
return -EINVAL;
}

--
2.27.0