[PATCH] KVM: selftests: Compare the entire XSAVE buffer in the SEV smoke test

[PATCH] KVM: selftests: Compare the entire XSAVE buffer in the SEV smoke test

[Gokul K]

compare_xsave() stops at index 4094, so the last byte of the XSAVE state
is never compared.  Both buffers handed to it are 4096 bytes: the host
copy is a struct kvm_xsave, whose region[] member is 1024 u32s, and the
guest copy is the PAGE_SIZE shared page that guest_code_xsave() runs
XSAVE into.  A VMSA synchronization bug that corrupted only that final
byte would go unreported and the test would still pass.

Bound the loop with sizeof(struct kvm_xsave) rather than an open-coded
length, and make the index unsigned so it does not mix signedness with
sizeof.

Fixes: 8c53183dbaa2 ("selftests: kvm: add test for transferring FPU state into VMSA")
Signed-off-by: Gokul K <gokul02k@gmail.com>
---
 tools/testing/selftests/kvm/x86/sev_smoke_test.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/kvm/x86/sev_smoke_test.c b/tools/testing/selftests/kvm/x86/sev_smoke_test.c
index 6b2cbe2a90b7..c7af27954356 100644
--- a/tools/testing/selftests/kvm/x86/sev_smoke_test.c
+++ b/tools/testing/selftests/kvm/x86/sev_smoke_test.c
@@ -90,9 +90,10 @@ asm("guest_code_xsave:\n"
 
 static void compare_xsave(u8 *from_host, u8 *from_guest)
 {
-	int i;
+	unsigned int i;
 	bool bad = false;
-	for (i = 0; i < 4095; i++) {
+
+	for (i = 0; i < sizeof(struct kvm_xsave); i++) {
 		if (from_host[i] != from_guest[i]) {
 			printf("mismatch at %u | %02hhx %02hhx\n",
 			       i, from_host[i], from_guest[i]);
-- 
2.54.0

Re: [PATCH] KVM: selftests: Compare the entire XSAVE buffer in the SEV smoke test

[Sean Christopherson]

On Fri, Aug 07, 2026, Gokul K wrote:
> compare_xsave() stops at index 4094, so the last byte of the XSAVE state
> is never compared.  Both buffers handed to it are 4096 bytes: the host
> copy is a struct kvm_xsave, whose region[] member is 1024 u32s, and the
> guest copy is the PAGE_SIZE shared page that guest_code_xsave() runs
> XSAVE into.  A VMSA synchronization bug that corrupted only that final
> byte would go unreported and the test would still pass.
> 
> Bound the loop with sizeof(struct kvm_xsave) rather than an open-coded
> length, and make the index unsigned so it does not mix signedness with
> sizeof.
> 
> Fixes: 8c53183dbaa2 ("selftests: kvm: add test for transferring FPU state into VMSA")
> Signed-off-by: Gokul K <gokul02k@gmail.com>
> ---
>  tools/testing/selftests/kvm/x86/sev_smoke_test.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/x86/sev_smoke_test.c b/tools/testing/selftests/kvm/x86/sev_smoke_test.c
> index 6b2cbe2a90b7..c7af27954356 100644
> --- a/tools/testing/selftests/kvm/x86/sev_smoke_test.c
> +++ b/tools/testing/selftests/kvm/x86/sev_smoke_test.c
> @@ -90,9 +90,10 @@ asm("guest_code_xsave:\n"
>  
>  static void compare_xsave(u8 *from_host, u8 *from_guest)
>  {
> -	int i;
> +	unsigned int i;
>  	bool bad = false;
> -	for (i = 0; i < 4095; i++) {

The fix looks right, but this is so bizarrely wrong that I can't help but wonder
if it was somehow intentional.

Paolo?

> +
> +	for (i = 0; i < sizeof(struct kvm_xsave); i++) {
>  		if (from_host[i] != from_guest[i]) {
>  			printf("mismatch at %u | %02hhx %02hhx\n",
>  			       i, from_host[i], from_guest[i]);
> -- 
> 2.54.0
> 
>