syzbot reported shift-out-of-bounds in nl802154_new_interface.
The problem was in signed representation of enum nl802154_iftype
enum nl802154_iftype {
/* for backwards compatibility TODO */
NL802154_IFTYPE_UNSPEC = -1,
...
Since, enum has negative value in it, objects of this type
will be represented as signed integer.
type = nla_get_u32(info->attrs[NL802154_ATTR_IFTYPE]);
u32 will be casted to signed, which can cause negative value type.
Reported-by: [email protected]
Signed-off-by: Pavel Skripkin <[email protected]>
---
net/ieee802154/nl802154.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/ieee802154/nl802154.c b/net/ieee802154/nl802154.c
index 7c5a1aa5adb4..6cce045e3d40 100644
--- a/net/ieee802154/nl802154.c
+++ b/net/ieee802154/nl802154.c
@@ -910,7 +910,7 @@ static int nl802154_new_interface(struct sk_buff *skb, struct genl_info *info)
if (info->attrs[NL802154_ATTR_IFTYPE]) {
type = nla_get_u32(info->attrs[NL802154_ATTR_IFTYPE]);
- if (type > NL802154_IFTYPE_MAX ||
+ if (type > NL802154_IFTYPE_MAX || type < 0 ||
!(rdev->wpan_phy.supported.iftypes & BIT(type)))
return -EINVAL;
}
--
2.30.2
Hi,
On Mon, 5 Apr 2021 at 15:58, Pavel Skripkin <[email protected]> wrote:
>
> syzbot reported shift-out-of-bounds in nl802154_new_interface.
> The problem was in signed representation of enum nl802154_iftype
>
> enum nl802154_iftype {
> /* for backwards compatibility TODO */
> NL802154_IFTYPE_UNSPEC = -1,
> ...
>
> Since, enum has negative value in it, objects of this type
> will be represented as signed integer.
>
> type = nla_get_u32(info->attrs[NL802154_ATTR_IFTYPE]);
>
> u32 will be casted to signed, which can cause negative value type.
>
> Reported-by: [email protected]
> Signed-off-by: Pavel Skripkin <[email protected]>
Yes, this patch will fix the issue but we discussed that the problem
is deeper than such a fix. The real problem is that we are using a -1
value which doesn't fit into the u32 netlink value and it gets
converted back and forward which we should avoid.
- Alex
On Tue, 2021-04-06 at 08:21 -0400, Alexander Aring wrote:
> Hi,
>
> On Mon, 5 Apr 2021 at 15:58, Pavel Skripkin <[email protected]>
> wrote:
> >
> > syzbot reported shift-out-of-bounds in nl802154_new_interface.
> > The problem was in signed representation of enum nl802154_iftype
> >
> > enum nl802154_iftype {
> > /* for backwards compatibility TODO */
> > NL802154_IFTYPE_UNSPEC = -1,
> > ...
> >
> > Since, enum has negative value in it, objects of this type
> > will be represented as signed integer.
> >
> > type = nla_get_u32(info->attrs[NL802154_ATTR_IFTYPE]);
> >
> > u32 will be casted to signed, which can cause negative value type.
> >
> > Reported-by: [email protected]
> > Signed-off-by: Pavel Skripkin <[email protected]>
>
> Yes, this patch will fix the issue but we discussed that the problem
> is deeper than such a fix. The real problem is that we are using a -1
> value which doesn't fit into the u32 netlink value and it gets
> converted back and forward which we should avoid.
>
OK, thanks for feedback!
>
> - Alex
With regards,
Pavel Skripkin