xen-netback: fix input validation in xenvif_set_hash_mapping()
authorJan Beulich <JBeulich@suse.com>
Tue, 25 Sep 2018 08:12:30 +0000 (02:12 -0600)
committerDavid S. Miller <davem@davemloft.net>
Tue, 25 Sep 2018 17:39:34 +0000 (10:39 -0700)
Both len and off are frontend specified values, so we need to make
sure there's no overflow when adding the two for the bounds check. We
also want to avoid undefined behavior and hence use off to index into
->hash.mapping[] only after bounds checking. This at the same time
allows to take care of not applying off twice for the bounds checking
against vif->num_queues.

It is also insufficient to bounds check copy_op.len, as this is len
truncated to 16 bits.

This is XSA-270 / CVE-2018-15471.

Reported-by: Felix Wilhelm <fwilhelm@google.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Tested-by: Paul Durrant <paul.durrant@citrix.com>
Cc: stable@vger.kernel.org [4.7 onwards]
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/xen-netback/hash.c

index 3c4c58b9fe76edfbf3d27fb5b6dbd0184ba706c0..3b6fb5b3bdb23ce6ab65b9947ba95cecbd8b59b1 100644 (file)
@@ -332,20 +332,22 @@ u32 xenvif_set_hash_mapping_size(struct xenvif *vif, u32 size)
 u32 xenvif_set_hash_mapping(struct xenvif *vif, u32 gref, u32 len,
                            u32 off)
 {
-       u32 *mapping = &vif->hash.mapping[off];
+       u32 *mapping = vif->hash.mapping;
        struct gnttab_copy copy_op = {
                .source.u.ref = gref,
                .source.domid = vif->domid,
-               .dest.u.gmfn = virt_to_gfn(mapping),
                .dest.domid = DOMID_SELF,
-               .dest.offset = xen_offset_in_page(mapping),
-               .len = len * sizeof(u32),
+               .len = len * sizeof(*mapping),
                .flags = GNTCOPY_source_gref
        };
 
-       if ((off + len > vif->hash.size) || copy_op.len > XEN_PAGE_SIZE)
+       if ((off + len < off) || (off + len > vif->hash.size) ||
+           len > XEN_PAGE_SIZE / sizeof(*mapping))
                return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;
 
+       copy_op.dest.u.gmfn = virt_to_gfn(mapping + off);
+       copy_op.dest.offset = xen_offset_in_page(mapping + off);
+
        while (len-- != 0)
                if (mapping[off++] >= vif->num_queues)
                        return XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER;