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  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date:	Sun, 26 Jul 2009 02:31:10 +0200
From:	Michal Soltys <soltys@....info>
To:	netdev@...r.kernel.org
Subject: r8169 (+others ?) and note_interrupt performance hit on 2.6.30.x

I've noticed, that on 2.6.30 kernels I took serious performance
hit compared to e.g. 2.6.28.10 (didn't check 2.6.29 ones).

Generally, with trivial nc -u other_host </dev/zero
- so ~ 1KB packets and 1gbit speed I went from:

01:30:25 AM  CPU    %usr   %nice    %sys %iowait    %irq   %soft  %steal  %guest   %idle
01:30:26 AM  all    0.91    0.00    8.68    0.00    2.28    7.31    0.00    0.00   80.82
01:30:26 AM    0    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00  100.00
01:30:26 AM    1    1.68    0.00   15.13    0.00    5.04   13.45    0.00    0.00   64.71

running at full 1gbit speed with ease, to:

01:35:26 AM  CPU    %usr   %nice    %sys %iowait    %irq   %soft  %steal  %guest   %idle
01:35:27 AM  all    0.50    0.00   12.56    0.00   24.12   12.56    0.00    0.00   50.25
01:35:27 AM    0    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00  100.00
01:35:27 AM    1    2.00    0.00   25.00    0.00   48.00   25.00    0.00    0.00    0.00

struggling at ~90MB/s (standard pfifo_fast)

Both netcat and interface interrupt were pinned at the same (2nd) core.

I did some quick oprofile, and noticed one new call -
note_interrupt - which wasn't present in the earlier kernel,
taking majority of the cpu time, relatively to the rest, e.g.:

samples  cum. samples  %        cum. %     symbol name
90984    90984         42.8695  42.8695    note_interrupt


The hardware is regular 32bit x86:

GA-MA74GM-S2H
AMD 4850e (2.5ghz)
onboard r8169

Both kernels with identical configuration, and tests done on idle machine.
--
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

Powered by blists - more mailing lists