KVM: s390: reset KVM_REQ_MMU_RELOAD if mapping the prefix failed
authorJulius Niedworok <jniedwor@linux.vnet.ibm.com>
Wed, 3 Aug 2016 14:39:55 +0000 (16:39 +0200)
committerChristian Borntraeger <borntraeger@de.ibm.com>
Fri, 12 Aug 2016 07:11:08 +0000 (09:11 +0200)
When triggering KVM_RUN without a user memory region being mapped
(KVM_SET_USER_MEMORY_REGION) a validity intercept occurs. This could
happen, if the user memory region was not mapped initially or if it
was unmapped after the vcpu is initialized. The function
kvm_s390_handle_requests checks for the KVM_REQ_MMU_RELOAD bit. The
check function always clears this bit. If gmap_mprotect_notify
returns an error code, the mapping failed, but the KVM_REQ_MMU_RELOAD
was not set anymore. So the next time kvm_s390_handle_requests is
called, the execution would fall trough the check for
KVM_REQ_MMU_RELOAD. The bit needs to be resetted, if
gmap_mprotect_notify returns an error code. Resetting the bit with
kvm_make_request(KVM_REQ_MMU_RELOAD, vcpu) fixes the bug.

Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Julius Niedworok <jniedwor@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
arch/s390/kvm/kvm-s390.c

index e63f6ed..f142215 100644 (file)
@@ -2362,8 +2362,10 @@ retry:
                rc = gmap_mprotect_notify(vcpu->arch.gmap,
                                          kvm_s390_get_prefix(vcpu),
                                          PAGE_SIZE * 2, PROT_WRITE);
-               if (rc)
+               if (rc) {
+                       kvm_make_request(KVM_REQ_MMU_RELOAD, vcpu);
                        return rc;
+               }
                goto retry;
        }