On Mon, Sep 21, 2020 at 07:56:20AM -0700, syzbot wrote:

Hello,

syzbot found the following issue on:

HEAD commit:    eb5f95f1 Merge tag 's390-5.9-6' of git://git.kernel.org/pu..
git tree:       upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=13996ad5900000
kernel config:  https://syzkaller.appspot.com/x/.config?x=ffe85b197a57c180
dashboard link: https://syzkaller.appspot.com/bug?extid=577fbac3145a6eb2e7a5
compiler:       gcc (GCC) 10.1.0-syz 20200507

Unfortunately, I don't have any reproducer for this issue yet.

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+577fbac3145a6eb2e7a5@syzkaller.appspotmail.com

==================================================================
BUG: KASAN: stack-out-of-bounds in xfrm_flowi_dport include/net/xfrm.h:877 [inline]
BUG: KASAN: stack-out-of-bounds in __xfrm6_selector_match net/xfrm/xfrm_policy.c:216 [inline]
BUG: KASAN: stack-out-of-bounds in xfrm_selector_match+0xf36/0xf60 net/xfrm/xfrm_policy.c:229
Read of size 2 at addr ffffc9001914f55c by task syz-executor.4/15633

This is yet another ipv4 mapped ipv6 address with IPsec socket policy
combination bug, and I'm sure it is not the last one. We could fix this
one by adding another check to match the address family of the policy
and the SA selector, but maybe it is better to think about how this
should work at all.

We can have only one socket policy for each direction and that
policy accepts either ipv4 or ipv6. We treat this ipv4 mapped ipv6
address as ipv4 and pass it down the ipv4 stack, so this dual usage
will not work with a socket policy. Maybe we can require IPV6_V6ONLY
for sockets with policy attached. Thoughts?