Modify

Opened 8 years ago

Closed 7 years ago

Last modified 4 years ago

#6054 closed defect (obsolete)

Via C3-based board crashes after 10h38m

Reported by: pfolk-openwrt@… Owned by: developers
Priority: normal Milestone: Barrier Breaker 14.07
Component: kernel Version: Trunk
Keywords: Cc: steve@…

Description

We have a reliable problem whereby the system hangs after about 38270 seconds of uptime (as calculated by subtracting the btime listed in /proc/stat from 'time' in perl), when running on Via C3-based ECS-PV6VEM3 motherboards. The problem occurs over a wide range of OpenWRT versions and kernel versions, BUT it does not occur if we use an ubuntu kernel.

The problem occurs even if we have no modules loaded, with svn head and kernel 2.6.31.1.

The system WAKES UP if we press "enter" on the serial console. When it resumes, it has an uptime of 38271s (the next second), regardless of how long it was hung. It immediately starts responding to pings and altogether behaves normally from that point on.

We are currently working backwards from the openwrt kernel to the ubuntu kernel to find where the change occurred, but having trouble getting other kernel versions to build in the kamikaze system. Help appreciated.

Attachments (0)

Change History (6)

comment:1 Changed 8 years ago by steve@…

I've cross-posted to the Ubuntu kernel team mailing list, in hopes that they might remember fixing this issue:

https://lists.ubuntu.com/archives/kernel-team/2009-November/007785.html

What I know at this point is:

1) The problem exists when I've stripped away all of the OpenWRT kernel patches, and with a very minimal set of kernel and OpenWRT modules enabled.

2) The problem exists when I build the Ubuntu kernel sources (karmic/9.10) with the OpenWRT toolchain.

3) The problem disappears when I use an Ubuntu kernel from one of their installation CDs (i.e. the version running the installation, not the version to-be-installed).

Number 2 here is the new piece of information here, which makes me think the problem might be in the OpenWRT toolchain, but it could still be something special about how the Ubuntu installation kernels are built.

-Steve

comment:2 Changed 8 years ago by steve@…

I have reason to believe it's a problem with gcc 4.1. I'm trying to rebuild with a more recent version of gcc. Building with CONFIG_GCC_VERSION="4.4.1" or "4.3.4" fails with "Building GCC requires GMP 4.1+ and MPFR 2.3.0+", despite more recent versions being listed in tools/{gmp,mpfr}/Makefile. A build with 4.2.4 is progressing, at least. I'll report the results here.

This lkml thread might be related:

http://linux.derkeiler.com/Mailing-Lists/Kernel/2007-01/msg00421.html

comment:3 Changed 8 years ago by florian

If you need to test a new toolchain version, make sure you remove your .config before and re-run make menuconfig.

comment:4 Changed 8 years ago by steve@…

I've finally gotten it to build successfully with gcc 4.4.1, and that seems to have eliminated the hang.

comment:5 Changed 7 years ago by nbd

  • Resolution set to obsolete
  • Status changed from new to closed

comment:6 Changed 4 years ago by jow

  • Milestone changed from Attitude Adjustment 12.09 to Barrier Breaker 14.07

Milestone Attitude Adjustment 12.09 deleted

Add Comment

Modify Ticket

Action
as closed .
The resolution will be deleted. Next status will be 'reopened'.
Author


E-mail address and user name can be saved in the Preferences.

 
Note: See TracTickets for help on using tickets.