Re: [PATCH net] packet: in packet_snd start writing at link layer allocation

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

On Thu, May 24, 2018 at 1:01 PM, Willem de Bruijn
<willemdebruijn.kernel@gmail.com> wrote:
> On Thu, May 24, 2018 at 11:17 AM, Willem de Bruijn
> <willemdebruijn.kernel@gmail.com> wrote:
>> On Thu, May 24, 2018 at 11:07 AM, Tariq Toukan <tariqt@mellanox.com> wrote:
>>>
>>>
>>> On 14/05/2018 3:20 AM, David Miller wrote:
>>>>
>>>> From: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
>>>> Date: Fri, 11 May 2018 13:24:25 -0400
>>>>
>>>>> From: Willem de Bruijn <willemb@google.com>
>>>>>
>>>>> Packet sockets allow construction of packets shorter than
>>>>> dev->hard_header_len to accommodate protocols with variable length
>>>>> link layer headers. These packets are padded to dev->hard_header_len,
>>>>> because some device drivers interpret that as a minimum packet size.
>>>>>
>>>>> packet_snd reserves dev->hard_header_len bytes on allocation.
>>>>> SOCK_DGRAM sockets call skb_push in dev_hard_header() to ensure that
>>>>> link layer headers are stored in the reserved range. SOCK_RAW sockets
>>>>> do the same in tpacket_snd, but not in packet_snd.
>>>>>
>>>>> Syzbot was able to send a zero byte packet to a device with massive
>>>>> 116B link layer header, causing padding to cross over into skb_shinfo.
>>>>> Fix this by writing from the start of the llheader reserved range also
>>>>> in the case of packet_snd/SOCK_RAW.
>>>>>
>>>>> Update skb_set_network_header to the new offset. This also corrects
>>>>> it for SOCK_DGRAM, where it incorrectly double counted reserve due to
>>>>> the skb_push in dev_hard_header.
>>>>>
>>>>> Fixes: 9ed988cd5915 ("packet: validate variable length ll headers")
>>>>> Reported-by: syzbot+71d74a5406d02057d559@syzkaller.appspotmail.com
>>>>> Signed-off-by: Willem de Bruijn <willemb@google.com>
>>>>
>>>>
>>>> Applied and queued up for -stable, thanks Willem.
>>>>
>>>
>>> Hi,
>>>
>>> One of our regression tests started failing. Once this patch is reverted,
>>> test passes.
>>>
>>> The tests add flow steering rules in the receiver side and in the sender
>>> side it send the packet with some RAW socket applications. Then received
>>> side gets completion with error.
>>>
>>> Our verification team compared the packets between the stable and the broken
>>> version, in the broken version we have some extra bytes at the end of the
>>> packet.
>>>
>>> It looks like some bad push to the SKB, maybe the conditional reserved
>>> addition should be more strict?
>>>
>>> Any idea?
>>
>> Thanks for reporting, sorry for the breakage.
>>
>> I think I might. This skb_push moves back the start of skb->data in the
>> same way that tpacket_snd does. But it does not reduce the length
>> passed to skb_put, so this might double count hard_header_len.
>>
>> Let me construct a test.
>
> Indeed.
>
> Still verifying, but this almost certainly has to be
>
>   @@ -2911,7 +2912,7 @@ static int packet_snd(struct socket *sock,
> struct msghdr *msg, size_t len)
>                   if (unlikely(offset < 0))
>                           goto out_free;
>           } else if (reserve) {
>   -               skb_push(skb, reserve);
>   +               skb_reserve(skb, -reserve);
>           }
>
> to move the start of the packet without changing its length.

I sent http://patchwork.ozlabs.org/patch/920126/

Again, thanks a lot for reporting this, Tariq. I'm working on some
packet socket boundary condition tests for tools/testing/selftests/net,
so that I cannot push such a mistake again.