pdflush uses all cpu-time with 2.6.7 and slow media

From: Georg Chini (georg.chini_at_triaton-webhosting.com)
Date: 06/30/04

  • Next message: Sean Neakums: "Re: malloc overlap?"
    Date:	Wed, 30 Jun 2004 11:55:46 +0200
    To: linux-kernel@vger.kernel.org
    
    

    Hello,

    there seems to be a problem with pdflush and
    slow media in 2.6.7. I'm using the packet-writing patch
    maintained by Peter Osterlund to write to dvd+rw.
    When I copy large files to the dvd, pdflush
    starts using all cpu-time up to a point where
    the system hangs completely. I found a posting here
    with a similar problem concerning nfs and pdflush
    (28.04.2004, Brent Cook). This thread mentions,
    that the problem is not observable in 2.6.5.
    So I tried 2.6.5 and everything works fine if
    I set dirty_ratio in /proc/sys/vm to 15. With
    the default value of 40 there are still some problems,
    but it is way better than 2.6.7-bk12. BTW my machine
    is a dual PIII.
    Any idea what might cause the problem? Any more
    information I should give?
    As I'm not on the list, please CC to me in your replies.

    Thanks in advance
                      Georg Chini

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/


  • Next message: Sean Neakums: "Re: malloc overlap?"

    Relevant Pages

    • Re: Complete I/O starvation with 3ware raid on 2.6
      ... >> kernel somewhere in D state. ... > They're pdflush and kjournald. ... > I've noticed the problem does not occur when the raid can absorb data ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: 2.6.5-rc1-mm2 and direct_read_under and wb
      ... as opposed to for a regular dirty memory flush". ... > many queues and cannot accord to get stuck on any one queue. ... > It's only OK to let pdflush avoid blocking on the buffer because pdflush ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: pdflush and dm-crypt
      ... I got feedback from a person who said that renicing ... So it looks like the CPU scheduler doesn't want to ... > schedule pdflush away. ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: 2.6.6-rc1-mm1 (and earlier): pdflush taking 100% cpu time (profile, .config etc. provided)
      ... pdflush is taking over my system: ... there's some logic error in there. ... If I could reproduce it, ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: pdflush question...
      ... > Is it worth having a kernel config option to vary the number of 'pdflush' ... number of busy spindles, really. ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)