[0/2] mptcp: a couple of fix - almost

Message

Paolo Abeni Aug. 20, 2021, 5:44 p.m. UTC

This are couple of changes poped-up while investigating
issues/219. They does not address the mentioned issue,
but the first one should fix an actual bug, the other
is more a cleanup, so the target tree is undefined ;)

Paolo Abeni (2):
  mptcp: fix possible divide by zero
  mptcp: make the locking tx schema more readable

 net/mptcp/protocol.c | 73 +++++++++++++++++++++-----------------------
 1 file changed, 35 insertions(+), 38 deletions(-)

Comments

Mat Martineau Aug. 20, 2021, 11:02 p.m. UTC | #1

On Fri, 20 Aug 2021, Paolo Abeni wrote:

> This are couple of changes poped-up while investigating
> issues/219. They does not address the mentioned issue,
> but the first one should fix an actual bug, the other
> is more a cleanup, so the target tree is undefined ;)
>
> Paolo Abeni (2):
>  mptcp: fix possible divide by zero
>  mptcp: make the locking tx schema more readable
>
> net/mptcp/protocol.c | 73 +++++++++++++++++++++-----------------------
> 1 file changed, 35 insertions(+), 38 deletions(-)
>
> -- 
> 2.26.3

So far my syzkaller setup hasn't seen the div-by-0 again, but it was 
typically taking days to show up. I'll leave that running over the 
weekend.

Changes look good to me - mptcp_push_release() is now only called from 
places where info->mss_now is non-zero.

Patch 1 for -net, patch 2 for net-next seems appropriate.

Reviewed-by: Mat Martineau <[e-mail]>

--
Mat Martineau
Intel

Matthieu Baerts Aug. 23, 2021, 7:16 a.m. UTC | #2

Hi Paolo, Florian, Mat,

On 21/08/2021 01:02, Mat Martineau wrote:
> On Fri, 20 Aug 2021, Paolo Abeni wrote:
> 
>> This are couple of changes poped-up while investigating
>> issues/219. They does not address the mentioned issue,
>> but the first one should fix an actual bug, the other
>> is more a cleanup, so the target tree is undefined ;)
> 
> So far my syzkaller setup hasn't seen the div-by-0 again, but it was
> typically taking days to show up. I'll leave that running over the weekend.
> 
> Changes look good to me - mptcp_push_release() is now only called from
> places where info->mss_now is non-zero.
> 
> Patch 1 for -net, patch 2 for net-next seems appropriate.

Thank you for the patches, the suggestions and the reviews!

These two patches are now in our tree (net and net-next) with Mat's RvB tag:

- 5d8855f6a71f: mptcp: fix possible divide by zero
- Results: 84fc11b83a29..653135eedf64
- 33708631c1fd: mptcp: make the locking tx schema more readable
- Results: 653135eedf64..ebac4c6745a4

Builds and tests are now in progress:

https://cirrus-ci.com/github/multipath-tcp/mptcp_net-next/export/20210823T071618
https://github.com/multipath-tcp/mptcp_net-next/actions/workflows/build-validation.yml?query=branch:export/20210823T071618

Cheers,
Matt