[lkml]   [2004]   [Aug]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subjectoops with 2.6.8-rc4 (ipv6 / nfs / 3c59x related?)
Hash: SHA1

hi list,

yesterday i encounterd several oopses with my "server". the subject of
this mail is just plain saying that it is all "2.6.8-rc4's fault". the
reality however turned out to be much more complex and confusing so i
try to recapitulate in a chronologically way:

- - the box (debian/unstable, i368 [PentiumIII] ) was running for 28 days
with 2.6.7-ck5 (no other patches but the -ck5 [1] )

- - ipv6 transition interface was down again, i had to start the freenet6
client again, to setup sit1. then i tried from a nfs-client to use this
interface by issuing "tracepath6" comamnds when i noticed my nfs-server
(the pIII-box) was dead. i could not catch the oops, had no digicam at
hand / was too lazy to write, i'll tell later on that.

- - rebooted same kernel, tried again locally (setup sit1, tracepath6 to
some ipv6 host) -> the box oopsed and locked up. (SYSRQ-s|u did not
succeed, only SYSRQ-b...rebooting)

- - ok, 2.6.8-rc4 is out anyway, time to update. compiled, booted, setup
sit1, tracepath6....-> the box oopsed and locked up.

- - am i too dumb to boot my self-built kernels? perhaps, so i used
kernel-image-2.6.7-1-686 from debian. again: i set up sit1, tracepath6
to some host -> ha! it did not lock up instantly, but tracepath6 gave
not results, so i CTRL-C'ed it and received a message:

KERNEL: assertion (!atomic_read(&sk->sk_wmem_alloc)) failed at
net/ipv4/af_inet.c (155)

everytime i CTRL-C'ed i received this message. and 2.6.7-1-686 was still
alive....for a few minutes. when i used my nfs-client to issue
tracepath6 commands, the box oopsed but this time the oops mad it to the
disk [2]. then it locked up, and the screen filled with traces.

- - today i got a camera so i was able to catch the oops [3] + [4]. now
the confusing part: i rememberd the oopses from the 2.6.8-rc4: they were
the same as [3] is showing, so it seemed to be related to the 3c59x
module. i even suspected (sudden) harware problems so i changed the NIC
to another 3c59x card. hm, the 3c59x oopses were gone indeed (somehow i
still doubt the card is damaged), but the nfs related oopses occured
with 2.6.7-1-i686 *and* 2.6.8-rc4 [4] + [5].

- - so, the nfs-related oopses were 100% reproducable when i used the
nfs-shares on the client (both 2.6.7-1-686 and 2.6.8-rc4)

- - somehow ipv6 was always involved, i could not trigger the oops when
ipv6 was not loaded/sit1 was not setup.

- - vanilla 2.6.7 seems not affected, 2.6.8-rc[1..3] are also ok
(2.6.8-rc3 is running atm).

i hope i did not mix something up here. yes, maybe this report has to
split up into one ipv6 and one nfs report. i hope [3], [4], [5] are
helpful to find out what the oops was all about. (the "Tainted" flags in
the 2.6.7-1-686 kernels came from loop-aes modules:
"loop: no version for "struct_module" found: kernel tainted.")

sorry for the crossposting (please delete rcpts if not applicable)
Thank you for your commments.





- --
BOFH excuse #78:

Yes, yes, its called a design limitation
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird -

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 14:05    [W:0.027 / U:4.908 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site