linux-kernel - Strange CPU usage

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>] [day] [month] [year] [list]

Date:	Sat, 7 Jul 2007 15:25:44 +0200
From:	"Loïc Grenié" <loic.grenie@...il.com>
To:	linux-kernel@...r.kernel.org
Subject: Strange CPU usage

    Hello,

  I've observed the following phenomenon: when I stop using
  my computer (leaving it on) for around 20 minutes, the
  CPU seems to be used (temperature increases fifteen
  degrees C) while usage given by top/w/rrdtool seems to
  be next to 0 (below 2%); at the same time the frequency
  (ondemand cpufreq governor) goes up.

    More specifically, here is a typical example:
   - I stop using the computer at 20:00; background  tasks use
     around 5 to 7% cpu (including system overhead).
   - At 20:20, the cpu_user usage goes to 0%, system lowers
     to 2%, frequency and temperature increase.

   A graphic of rrdgraph can be found at:

http://loic.grenie.googlepages.com/stat.png

  which is the snapshot of a real situation. Several examples of
  what I'm trying to describe are visible at end of afternoon,
  18:30, 19:55 and 22:48. (The bars on the bottom are cpu usage:
  system, user and nice; the red line is the CPU frequency: AMD
  Turion X2 with 800MHz and 1.6GHz available; the other lines are
  temperatures as reported by ACPI, the two core probes and the
  hard drive).

    The 20 minutes delay could be related to DPMS kicking in (but I'm
  not positive). The problem seems to be related to X (I've run the
  computer idle this morning for a couple of hours without this
  phenomenon), it seems to be more easily triggered if firefox is
  running (but this is not a necessary condition). The kernel version
  does not seem very important (observed with 2.6.18 (Debian
  version), .21, .22-rc6-mm1, .22-rc7-hrt1, with and without
  suspend2/TuxOnIce).
    I don't know if this is related but my syslog is full of

APIC error on CPU0: 00(40) (once per session)
APIC error on CPU0: 40(40) (a lot of times per session)
(and the same for CPU1)

  at random times.

    What looks very strange to me is the CPU usage while CPU is
  indicated as idle by the kernel. I was wondering whether this
  could be an interrupt firing like mad ?

    Configuration:
Dell Inspiron 1501
AMD Turion 64X2
ATI RS480+SB600
X driver: radeon
lspci attached
One of the .config attached (2.6.21 with 2.2.9 Suspend2)
dmesg attached (running with the abovementioned kernel)
lsmod attached

    I'm wondering: what's up ? Has it been already observed ? Have I
  configured something really wrong ?

    Thank you for any help, thank you for the kernel,

      Loïc

Disclaimer:
Sorry to disturb

View attachment "lspci.txt" of type "text/plain" (1811 bytes)

Download attachment "config-rc7-hrt1-s2" of type "application/octet-stream" (67266 bytes)

View attachment "dmesg.txt" of type "text/plain" (22108 bytes)

View attachment "lsmod.txt" of type "text/plain" (1529 bytes)