KVM: s390: ioeventfd: ignore leftmost bits
authorDominik Dingel <dingel@linux.vnet.ibm.com>
Mon, 9 Dec 2013 17:30:01 +0000 (18:30 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 6 Feb 2014 19:34:01 +0000 (11:34 -0800)
commit ff1f3cb4b3ac5d039f02679f34cb1498d110d241 upstream.

The diagnose 500 subcode 3 contains the 32 bit subchannel id in bits 32-63
(counting from the left). As for other I/O instructions, bits 0-31 should be
ignored and thus not be passed to kvm_io_bus_write_cookie().

This fixes a bug where the guest passed non-zero bits 0-31 which the
host tried to interpret, leading to ioeventfd notification failures.

Signed-off-by: Dominik Dingel <dingel@linux.vnet.ibm.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/s390/kvm/diag.c

index 5ff29be..8216c0e 100644 (file)
@@ -121,7 +121,7 @@ static int __diag_virtio_hypercall(struct kvm_vcpu *vcpu)
         * - gpr 4 contains the index on the bus (optionally)
         */
        ret = kvm_io_bus_write_cookie(vcpu->kvm, KVM_VIRTIO_CCW_NOTIFY_BUS,
-                                     vcpu->run->s.regs.gprs[2],
+                                     vcpu->run->s.regs.gprs[2] & 0xffffffff,
                                      8, &vcpu->run->s.regs.gprs[3],
                                      vcpu->run->s.regs.gprs[4]);