netdev - Re: resend: tcp: performance issue with fastopen connections (mss

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAK6E8=cQdCvaXvcNzZYkC7R=vk2SmK7djS9OXNgqcWnO9CG0ug@mail.gmail.com>
Date:   Wed, 18 Jan 2017 10:27:12 -0800
From:   Yuchung Cheng <ycheng@...gle.com>
To:     Eric Dumazet <edumazet@...gle.com>
Cc:     Alexey Kodanev <alexey.kodanev@...cle.com>,
        Eric Dumazet <eric.dumazet@...il.com>,
        David Miller <davem@...emloft.net>,
        netdev <netdev@...r.kernel.org>,
        Vasily Isaenko <vasily.isaenko@...cle.com>,
        Neal Cardwell <ncardwell@...gle.com>
Subject: Re: resend: tcp: performance issue with fastopen connections (mss > window)

Googler-only

Hi Eric: yeah I think the test was just due to the TSO chunking
difference between prod and upstream, which I was able to avoid with
this patch re-suited by Neal in b/34128974. In fact, this patch
enables me to run all recovery tests on upstream kernels for my RACK
patch set.

Neal: can we polish and check that in? super useful.


On Wed, Jan 18, 2017 at 10:16 AM, Eric Dumazet <edumazet@...gle.com> wrote:
> On Wed, Jan 18, 2017 at 10:13 AM, Yuchung Cheng <ycheng@...gle.com> wrote:
>> On Wed, Jan 18, 2017 at 9:35 AM, Eric Dumazet <edumazet@...gle.com> wrote:
>>>
>>> On Wed, Jan 18, 2017 at 9:32 AM, Alexey Kodanev
>>> <alexey.kodanev@...cle.com> wrote:
>>> > Hi Eric,
>>> >
>>> > On 01/13/2017 08:07 PM, Alexey Kodanev wrote:
>>> >
>>>
>>> > Looks like max_window not correctly initialized for tfo sockets.
>>> > On my test machine it has set to '5592320' in tcp_fastopen_create_child().
>>> >
>>> > This diff fixes the issue, the question: is this the right place to do it?
>>> >
>>> > diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
>>> > index 4e777a3..33ed508 100644
>>> > --- a/net/ipv4/tcp_fastopen.c
>>> > +++ b/net/ipv4/tcp_fastopen.c
>>> > @@ -206,6 +206,8 @@ static struct sock *tcp_fastopen_create_child(struct
>>> > sock *sk,
>>> >          */
>>> >         tp->snd_wnd = ntohs(tcp_hdr(skb)->window);
>>> >
>>> > +       tp->max_window = tp->snd_wnd;
>>> > +
>>>
>>> Excellent catch. Let me test our regression tests with this.
>> Indeed nice catch. Thanks for the investigative work!
>>
>
> We do have 2 failures, but tests might have depended on undocumented behavior
>
> (For googlers :
> Ran 211 tests: 209 passing, 0 flaky 2 failing
> Sponge: http://sponge/f1575065-6e1c-4514-bced-9167ce56d2ee
> )
>
> Please Alexey submit an official patch, thanks a lot !