netdev - Re: Is 3.18 patch "The tcp: TSO packets automatic sizing" will cause to Troughput drop when enable NO_HZ_IDEL or HIGH_RES

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: <CANHbGWsqBZEw0-6A0zE6PrSUF1osbN86mcYxjo9Z4V6WWyXTww@mail.gmail.com>
Date:	Wed, 7 May 2014 23:51:29 +0800
From:	V JobNickname <workofv@...il.com>
To:	Neal Cardwell <ncardwell@...gle.com>
Cc:	Netdev <netdev@...r.kernel.org>
Subject: Re: Is 3.18 patch "The tcp: TSO packets automatic sizing" will cause
 to Troughput drop when enable NO_HZ_IDEL or HIGH_RES_TIMERS ?

2014-05-07 0:48 GMT+08:00 Neal Cardwell <ncardwell@...gle.com>:
> On Tue, May 6, 2014 at 12:17 PM, V JobNickname <workofv@...il.com> wrote:
>> More detail trace, the first commit of 3.10.18, tcp: TSO packets
>> automatic sizing, cause to this issue.
>> https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/diff/?id=5e25ba5003ee5de0ba2be56bfd54d16d4b1b028d
>
> To provide another data point, it would be interesting to know what
> performance you see with v3.15-rc4. Linux 3.15 has the following
> patch:
>
>  740b0f1841f6  tcp: switch rtt estimations to usec resolution
>
> That patch switches tcp_update_pacing_rate() to use microsecond RTTs
> for its calculation. That may help things in your case. (As Eric says,
> more details would help.)
>
>> I don't know why this TSO patch affect my network since the network
>> driver doesn't enable NETIF_F_TSO.
>
> Keep in mind that anytime the Linux TCP code or commit descriptions
> talk about "TSO", the same applies to GSO. Is GSO enabled for your
> device?
>
> neal

Porting to the v3.15-rc4 from main stream, the performance is well when
 enabled NO_HZ_IDEL or HIGH_RES_TIMERS.

I also try the version before commit 740b0f1841f6  tcp: switch rtt
estimations to usec resolution, i.e
363ec392352 net: add skb_mstamp infrastructure,
also not cause to performance drop. So I think my case is regardless
to use jiffie or nesc in tcp_update_pacing_rate().

Then I try more eraly version and found that in 3.14-rc2 thes case was
hit but solved at 3.14-rc3.
(I use "git checkout -b 3.14-rc2 v3.14-rc2" and "git checkout -b
3.14-rc3 v3.14-rc3"  to get taged version,
and use "git checkout 3.14-rc2/3.14-rc3 to switch between". Hope my
usage are correct, since I'm not familer with git.)

I think some commit between 3.14-rc2 and 3.14-rc3 can help find the
root cause of my case.

ps.
But I don't know how to use git to update from rc2 to rc3 by each
commit(or maybe binary search commit log), and find that one commit.

For example, I would like to following logs start from 3.14-rc2 tag in
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/log/?h=linux-3.14.y&ofs=1650

Can I,

git checkout 3.14-rc2 -- switch to 3.14-rc2
git checout 07b0e5b10258b48e5edfb6c8ac156f05510eb775 (ASoC: da9055:
Fix device registration of PMIC and CODEC devices)
will the commit between this commit and 3.14-rc2 also updated?
if this commit solve my case, I can try some one commit between it and 3.14-rc2.
such as
git checkout a49f56eec54d864ba0fda838e4c8bf5c72f3eb08 (microblaze: Fix
a typo when disabling stack protection)
should I revert to 3.14-rc2 first, or the git will done it?

Thanks
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html