KVM: s390: prevent buffer overrun on memory hotplug during migration

We must not go beyond the pre-allocated buffer. This can happen when
a new memory slot is added during migration.

Reported-by: David Hildenbrand <[email protected]>
Signed-off-by: Christian Borntraeger <[email protected]>
Cc: [email protected] # 4.13+
Fixes: 190df4a212a7 (KVM: s390: CMMA tracking, ESSA emulation, migration mode)
Reviewed-by: Cornelia Huck <[email protected]>
Reviewed-by: David Hildenbrand <[email protected]>

diff --git a/arch/s390/kvm/priv.c b/arch/s390/kvm/priv.c
index 572496c688cc0c647bd220310bfdc6e9635d4723..0714bfa56da0f54cae66b26bd5329b18beb77b0b 100644 (file)
--- a/arch/s390/kvm/priv.c
+++ b/arch/s390/kvm/priv.c
@@ -1006,7 +1006,7 @@ static inline int do_essa(struct kvm_vcpu *vcpu, const int orc)
                cbrlo[entries] = gfn << PAGE_SHIFT;
        }
 
                cbrlo[entries] = gfn << PAGE_SHIFT;
        }
 

-       if (orc) {
+       if (orc && gfn < ms->bitmap_size) {

                /* increment only if we are really flipping the bit to 1 */
                if (!test_and_set_bit(gfn, ms->pgste_bitmap))
                        atomic64_inc(&ms->dirty_pages);
                /* increment only if we are really flipping the bit to 1 */
                if (!test_and_set_bit(gfn, ms->pgste_bitmap))
                        atomic64_inc(&ms->dirty_pages);