s390/nmi: handle guarded storage validity failures for KVM guests

machine check validity bits reflect the state of the machine check. If a
guest does not make use of guarded storage, the validity bit might be
off. We can not use the host CR bit to decide if the validity bit must
be on. So ignore "invalid" guarded storage controls for KVM guests in
the host and rely on the machine check being forwarded to the guest.  If
no other errors happen from a host perspective everything is fine and no
process must be killed and the host can continue to run.

Cc: stable@vger.kernel.org
Fixes: c929500d7a5a ("s390/nmi: s390: New low level handling for machine check happening in guest")
Reported-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Tested-by: Carsten Otte <cotte@de.ibm.com>
Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

diff --git a/arch/s390/kernel/nmi.c b/arch/s390/kernel/nmi.c
index 0c9e894913dccd614002bbc6e562683386220788..147c0d5fd9b459630ec166d6de69a79b156fe7d6 100644 (file)
--- a/arch/s390/kernel/nmi.c
+++ b/arch/s390/kernel/nmi.c
@@ -307,11 +307,21 @@ static int notrace s390_validate_registers(union mci mci, int umode)
        if (cr2.gse) {
                if (!mci.gs) {
                        /*
-                        * Guarded storage register can't be restored and
-                        * the current processes uses guarded storage.
-                        * It has to be terminated.
+                        * 2 cases:
+                        * - machine check in kernel or userspace
+                        * - machine check while running SIE (KVM guest)
+                        * For kernel or userspace the userspace values of
+                        * guarded storage control can not be recreated, the
+                        * process must be terminated.
+                        * For SIE the guest values of guarded storage can not
+                        * be recreated. This is either due to a bug or due to
+                        * GS being disabled in the guest. The guest will be
+                        * notified by KVM code and the guests machine check
+                        * handling must take care of this.  The host values
+                        * are saved by KVM and are not affected.
                         */
-                       kill_task = 1;
+                       if (!test_cpu_flag(CIF_MCCK_GUEST))
+                               kill_task = 1;
                } else {
                        load_gs_cb((struct gs_cb *)mcesa->guarded_storage_save_area);
                }