Bug 4982 - Usenet gateway crashes under heavy traffic (survives with 2.6.12-mm1)
Usenet gateway crashes under heavy traffic (survives with 2.6.12-mm1)
Status: REJECTED INSUFFICIENT_DATA
Product: Drivers
Classification: Unclassified
Component: Network
i386 Linux
: P2 normal
Assigned To: Jeff Garzik
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2005-08-01 22:27 UTC by Alexey Dobriyan
Modified: 2007-02-22 09:42 UTC (History)
1 user (show)

See Also:
Kernel Version: 2.6.13-rc4-git2
Tree: Mainline
Regression: ---


Attachments
uname, free, loadavg, /proc/interrupts, ... (915 bytes, text/plain)
2005-08-01 22:31 UTC, Alexey Dobriyan
Details
2.6.12-mm1 .config (27.45 KB, text/plain)
2005-08-01 22:32 UTC, Alexey Dobriyan
Details
2.6.13-rc4-git2 .config (25.65 KB, text/plain)
2005-08-01 22:33 UTC, Alexey Dobriyan
Details
2.6.12-mm1 log (27.10 KB, text/plain)
2005-08-01 22:35 UTC, Alexey Dobriyan
Details
2.6.13-rc4-git2 log (114.92 KB, text/plain)
2005-08-01 22:36 UTC, Alexey Dobriyan
Details

Description Alexey Dobriyan 2005-08-01 22:27:08 UTC
From: Danny ter Haar <dth@picard.cistron.nl>
http://marc.theaimsgroup.com/?t=112280339800001&r=1&w=2

A tyan AMD64 opteron machine functioning as a usenet gateway really
pumps some traffic a day (http://newsgate.newsserver.nl)
Incoming traffic comes through a optical gig-E card (acenic)
and local traffic is fed to our spool boxes through cupper gig-E
(tigon3). Machine uses adaptec onboard scsi disks.
At first i thought i had a hardware problem since no kernel would
survive longer than 30 hours. Ofcourse i ran memtest for a couple
of days. Than i compiled 2.6.12-mm1 and this kernel surviced 18 days
without a problem.

reboot   system boot  2.6.12-mm1       Sun Jul 31 09:47          (01:48)
reboot   system boot  2.6.13-rc4-git2  Sat Jul 30 18:29          (17:07)
reboot   system boot  2.6.12-mm1       Sat Jul 30 14:12          (04:14)
reboot   system boot  2.6.13-rc4       Fri Jul 29 14:16        (1+04:10)
reboot   system boot  2.6.13-rc3-mm3   Fri Jul 29 12:17          (01:50)
reboot   system boot  2.6.12-mm1       Thu Jul 28 00:06        (1+12:09)
reboot   system boot  2.6.13-rc3-mm2   Wed Jul 27 22:27          (01:36)
reboot   system boot  2.6.13-rc3-mm1   Wed Jul 27 11:22          (12:41)
reboot   system boot  2.6.12-mm1       Sun Jul 17 15:51        (9+19:29)

Machine does have serial console (and remote powerboot) but no logging
possibility (portmaster1). When it crashes i think most of the times it
has something to do with IRQ.
2.6.13-rc4-git2 stopped working with the following notice:

Jul 31 03:28:18 newsgate kernel: hw tcp v4 csum failed
Jul 31 05:56:59 newsgate kernel: NETDEV WATCHDOG: eth3: transmit timed out
Jul 31 05:56:59 newsgate kernel: tg3: eth3: transmit timed out, resetting

Serial console kept spitting those messages but it gave no prompt
anymore. remote powercycle was needed to get it back.

More info/config can be found at: http://newsgate.newsserver.nl/kernel/
Comment 1 Alexey Dobriyan 2005-08-01 22:31:07 UTC
Created attachment 5470 [details]
uname, free, loadavg, /proc/interrupts, ...
Comment 2 Alexey Dobriyan 2005-08-01 22:32:38 UTC
Created attachment 5471 [details]
2.6.12-mm1 .config
Comment 3 Alexey Dobriyan 2005-08-01 22:33:30 UTC
Created attachment 5472 [details]
2.6.13-rc4-git2 .config
Comment 4 Alexey Dobriyan 2005-08-01 22:35:57 UTC
Created attachment 5473 [details]
2.6.12-mm1 log
Comment 5 Andrew Morton 2005-08-01 22:36:35 UTC
bugme-daemon@kernel-bugs.osdl.org wrote:
>
> http://bugzilla.kernel.org/show_bug.cgi?id=4982

Do we not have the oops output for this crash?

Comment 6 Alexey Dobriyan 2005-08-01 22:36:39 UTC
Created attachment 5474 [details]
2.6.13-rc4-git2 log
Comment 7 Alexey Dobriyan 2005-09-27 02:17:06 UTC
From: Danny ter Haar <dth@picard.cistron.nl>
Assorted posts to linux-kernel

2.6.13-rc7
2.6.14-rc2-git3
2.6.14-rc2-git5 unstable
Comment 8 Adrian Bunk 2006-12-18 17:43:07 UTC
Is this issue still present in kernel 2.6.19?
Comment 9 Adrian Bunk 2007-02-22 09:42:19 UTC
Please reopen this bug if it's still present with kernel 2.6.20.

Note You need to log in before you can comment on or make changes to this bug.