s390/nmi: handle vector validity failures for KVM guests

commit f094a39c6ba168f2df1edfd1731cca377af5f442 upstream.

The machine check validity bit tells about the context. If a KVM guest
was running the bit tells about the guest validity and the host state is
not affected. As a guest can disable the guest validity this might
result in unwanted host errors on machine checks.

Cc: stable@vger.kernel.org
Fixes: c929500d7a5a ("s390/nmi: s390: New low level handling for machine check happening in guest")
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/arch/s390/kernel/nmi.c b/arch/s390/kernel/nmi.c
index dffc65d..a50f2ff 100644 (file)
--- a/arch/s390/kernel/nmi.c
+++ b/arch/s390/kernel/nmi.c
@@ -273,7 +273,14 @@ static int notrace s390_validate_registers(union mci mci, int umode)
                /* Validate vector registers */
                union ctlreg0 cr0;
 
-               if (!mci.vr) {
+               /*
+                * The vector validity must only be checked if not running a
+                * KVM guest. For KVM guests the machine check is forwarded by
+                * KVM and it is the responsibility of the guest to take
+                * appropriate actions. The host vector or FPU values have been
+                * saved by KVM and will be restored by KVM.
+                */
+               if (!mci.vr && !test_cpu_flag(CIF_MCCK_GUEST)) {
                        /*
                         * Vector registers can't be restored. If the kernel
                         * currently uses vector registers the system is