KVM: s390: sthyi: remove invalid guest write access

handle_sthyi() always writes to guest memory if the sthyi function
code is zero in order to fault in the page that later is written to.

However a function code of zero does not necessarily mean that a write
to guest memory happens: if the KVM host is running as a second level
guest under z/VM 6.2 the sthyi instruction is indicated to be
available to the KVM host, however if the instruction is executed it
will always return with a return code that indicates "unsupported
function code".

In such a case handle_sthyi() must not write to guest memory. This
means that the prior write access to fault in the guest page may
result in invalid guest exceptions, and/or invalid data modification.

In order to be architecture compliant simply remove the write_guest()
call.

Given that the guest assumed a write access anyway, this fix does not
qualify for -stable. This just makes sure the sthyi handler is
architecture compliant.

Fixes: 95ca2cb57985 ("KVM: s390: Add sthyi emulation")
Reviewed-by: Janosch Frank <frankja@linux.vnet.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>

diff --git a/arch/s390/kvm/sthyi.c b/arch/s390/kvm/sthyi.c
index a2e5c24..395926b 100644 (file)
--- a/arch/s390/kvm/sthyi.c
+++ b/arch/s390/kvm/sthyi.c
@@ -436,14 +436,6 @@ int handle_sthyi(struct kvm_vcpu *vcpu)
        if (addr & ~PAGE_MASK)
                return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 
-       /*
-        * If the page has not yet been faulted in, we want to do that
-        * now and not after all the expensive calculations.
-        */
-       r = write_guest(vcpu, addr, reg2, &cc, 1);
-       if (r)
-               return kvm_s390_inject_prog_cond(vcpu, r);
-
        sctns = (void *)get_zeroed_page(GFP_KERNEL);
        if (!sctns)
                return -ENOMEM;