2.2.15pre5: gfp called by non-running task..

From: Philippe Strauss (philou@philou.ch)
Date: Sat Jan 29 2000 - 11:18:33 EST

  • Next message: Graham Murray: "2.3.41 - tcp lo interface very slow"

    Hi there,

    following a 2.2.14 crash on one of our test webcache server (running squid
    to test raid-0.90-19990824 stability), I compiled a new kernel:

    - 2.2.15pre5
    - raid-0.90-19990824 for 2.2.11 (I had to backout the diff to raid0.c in 2.2.13 -> 2.2.14
    patch).
    - reiserfs 3.5.16 for 2.2.14

    After reboot and launching squid, kernel msg scrolled continuously on my pty:

    'gfp called by non-running (1) task from da49e960!'

    (~ 20 per second)

    stopping squid stop also the messages.

    At first I thought it could be either raid-0.90 or reiserfs related, so I've
    reformated squid spooling area to a plain ext2fs on a normal partition (no raid)
    but the get_free_pages msg keep scrolling at high rate when I start squid.

    My squid is compiled with threads support, allowing squid disk i/o to be
    scheduled outside the main select loop of squid.
    Lot's of squid thread are sitting idle, waiting for the main squid
    to dispatch it's disk i/o.

    Context: PII 450 512MB SDRAM, Compaq proliant 1850R, Debian 2.1r4,
    kernel and squid compiled with gcc-2.7.2.3 (part of debian).
    2 other such server running vanilla 2.2.14 and the same squid
    without a glitch.

    Squid 2.2S5 with Henrik Nordström's patch: 2.2.STABLE5 snapshot 2000-01-16 

    -- 
Philippe Strauss, Telecom Consulting
home email address: <philou@philou.ch>

    Be Here Now. If you do not live in the future, there is nothing to fear.
If you do not live in the past, there is nothing to regret.
--

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

    This archive was generated by hypermail 2b29 : Sat Jan 29 2000 - 17:38:29 EST