Bug 4648 - klogd: timestamps are incorrect around suspend
Summary: klogd: timestamps are incorrect around suspend
Status: REJECTED WILL_NOT_FIX
Alias: None
Product: Other
Classification: Unclassified
Component: Other (show other bugs)
Hardware: i386 Linux
: P2 low
Assignee: other_other
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2005-05-21 07:19 UTC by Toralf Förster
Modified: 2005-10-04 03:37 UTC (History)
2 users (show)

See Also:
Kernel Version: 2.6.12
Subsystem:
Regression: ---
Bisected commit-id:


Attachments

Description Toralf Förster 2005-05-21 07:19:23 UTC
Distribution: Gentoo
Hardware Environment: Thinkpad T41
Software Environment: straight Linux installation
Problem Description:During suspend and resume some logging entries got the wrong
timestamp. As you can see in the dump below the lines '... Stopping tasks ...'
and '... Back to C! ...' should have a timestamp around 15:40:01 but they have
the timestamp 16:07:36 of the systems it waked up. BTW I use syslog-ng but I
think it is related to klogd because I observed the same problem using syslog.


Steps to reproduce:

n22 ~ # tail -n 60 /var/log/messages | head -n 8
May 21 15:40:01 n22 cron[16939]: (root) CMD (test -x /usr/sbin/run-crons &&
/usr/sbin/run-crons )
May 21 16:07:36 n22 Stopping tasks:
==========================================================================|
May 21 16:07:36 n22 radeonfb (0000:01:00.0): suspending to state: 3...
May 21 16:07:36 n22 Back to C!
May 21 16:07:36 n22 Warning: CPU frequency is 1700000, cpufreq assumed 600000 kHz.
May 21 16:07:36 n22 PCI: Setting latency timer of device 0000:00:1d.0 to 64
Comment 1 Toralf Förster 2005-09-09 01:34:28 UTC
As described in https://bugs.gentoo.org/show_bug.cgi?id=104829#c1 the logger
should get a chance to log the message bofore the system really shuts off, isn't
it ?
Comment 2 john stultz 2005-10-03 15:51:48 UTC
Pavel, do you have any ideas on this one?

Also, I'd probably change the summary to "timestamps are incorrect around
suspend" for clarity.
Comment 3 Pavel Machek 2005-10-04 03:37:21 UTC
Seems to me as it "works as designed".

dmesg buffer is not timestamped, so klogd adds its own stamps. That means that
timestamp for "Freezing processes" is wrong, because it is done before suspend
but logger can only see it after suspend....

....but that's nothing we can fix.

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