Ubuntu
linux package

Bug #1073433
Comment #105

Comment 105 for bug 1073433

Revision history for this message

Alexander (lxandr) wrote on 2013-04-15:

#105

1) Problem appears on different hardware. I mean totally different (except for my usb hdd, from which the system boots).
2) I run memtest about a year ago... But I don't see any oops-es, hangs or kernel panics. So the problem is not related with (broken) hardware.
3) How I've removed plymouth? I read about plymouth in comments to #556372. People suggest to install plymouth-dummy package (which has no mountall dependency) so after that plymouth can be easily removed. I have neither plymouth nor network-manager nor modemmanager nor wicd installed. Just plain /etc/network/inerfaces. +guessnet.
4) I'm bringing network interface up manually from /etc/rc.local. Because 'ifup -a' called from init-scrips doesn't work correctly (bug).
5) The harder I'm trying to debug this bug, the less it appears (he knows that I'm hunting on him! haha).
6) I can speak russian, если что.
7) currently I can't set HALT=halt in /etc/default/halt. Because most of the time shutdown is scheduled and done automatically (and I can be far away from computer at that moment). But I'll do that soon.
8) Problem doesn't appear 100% of the time (at least for me). And also not related with any of my debug scripts.
9) Both root and swap partitions are on my usb harddrive. So, maybe it's somehow related to usb? Or swap?
I don't know is it important... but root and swap are mounted from fstab by label:
LABEL=root / ext4 errors=remount-ro 0 1
LABEL=swap none swap sw 0 0
10) I'm also using autofs daemon. Maybe it's somehow related to it?
11) Now I've removed all my debug parts from umountroot. And added two scripts:

/etc/rc0.d/S59-before-umountroot:
#!/bin/bash
/sbin/sysctl kernel.printk="7 4 1 7"
/bin/dmesg > /upstart-messages-1-before-ro-`date +%s`.log

/etc/rc0.d/S75-before-halt:
#!/bin/bash
/bin/dmesg > /upstart-messages-2-after-ro-`date +%s`.log

(also added symlinks to rc6.d)

But, I've got nothing interesting:
...
[22482.298335] init: smbd main process (1057) killed by TERM signal
[22482.299101] init: tty4 main process (1246) killed by TERM signal
[22482.299659] init: tty5 main process (1255) killed by TERM signal
[22482.300259] init: tty2 main process (1293) killed by TERM signal
[22482.304610] init: tty3 main process (1294) killed by TERM signal
[22482.305310] init: tty6 main process (1299) killed by TERM signal
[22482.315235] init: irqbalance main process (1327) killed by TERM signal
[22482.315828] init: cron main process (1339) killed by TERM signal
[22482.318599] init: tty1 main process (2165) killed by TERM signal
[22484.047088] init: Disconnected from system bus
[22484.190023] tg3 0000:10:00.0: wake-up capability enabled by ACPI

Both logs are the same.

12) cat /proc/cmdline
root=LABEL=root ro ipv6.disable=1

1) Problem appears on different hardware. I mean totally different (except for my usb hdd, from which the system boots).
2) I run memtest about a year ago... But I don't see any oops-es, hangs or kernel panics. So the problem is not related with (broken) hardware.
3) How I've removed plymouth? I read about plymouth in comments to #556372. People suggest to install plymouth-dummy package (which has no mountall dependency) so after that plymouth can be easily removed. I have neither plymouth nor network-manager nor modemmanager nor wicd installed. Just plain /etc/network/inerfaces. +guessnet. 
4) I'm bringing network interface up manually from /etc/rc.local.  Because 'ifup -a' called from init-scrips doesn't work correctly (bug).
5) The harder I'm trying to debug this bug, the less it appears (he knows that I'm hunting on him! haha). 
6) I can speak russian, если что.
7) currently I can't set HALT=halt in /etc/default/halt. Because most of the time shutdown is scheduled and done automatically (and I can be far away from computer at that moment). But I'll do that soon.
8) Problem doesn't appear 100% of the time (at least for me). And also not related with any of my debug scripts.
9) Both root and swap partitions are on my usb harddrive. So, maybe it's somehow related to usb? Or swap?
I don't know is it important... but root and swap are mounted from fstab by label:
LABEL=root      /               ext4    errors=remount-ro 0     1
LABEL=swap      none            swap    sw              0       0
10) I'm also using autofs daemon. Maybe it's somehow related to it?
11) Now I've removed all my debug parts from umountroot. And added two scripts: