Bug 152022

Summary: kernel sets hardware clock to random past dates during boot on Pegasos
Product: [openSUSE] SUSE Linux 10.1 Reporter: peter czanik <peter>
Component: KernelAssignee: Olaf Hering <ohering>
Status: RESOLVED WORKSFORME QA Contact: E-mail List <qa-bugs>
Severity: Normal    
Priority: P5 - None    
Version: Beta 4   
Target Milestone: ---   
Hardware: PowerPC   
OS: Other   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: output of kernel update

Description peter czanik 2006-02-18 03:03:33 UTC
The kernel sets hardware clock to random past dates during boot on Pegasos. It does not do this always, but there is about a 50% chance. Kernel of 10.0 never did that, but if 10.0 is booted with kernel from 10.1 beta3/4, the problem appears there also. This causes long fsck-s if ext2/3 is used...
Some typical years are 1930, 1933, 1968, 1969, 1970, 1999
The battery is brand new, and there is no such problem on the same machine with 10.0 with it's original 2.6.13 kernel, or Gentoo with 2.6.15.
Comment 1 Chris L Mason 2006-02-18 16:48:13 UTC
Olaf?
Comment 2 Olaf Hering 2006-02-18 16:52:46 UTC
could be related to bug #149895, fix is work in progress.

http://lkml.org/lkml/2006/2/15/5
Comment 3 Olaf Hering 2006-02-20 15:59:02 UTC
Can you test the kernel of the day tomorrow?

ftp://ftp.suse.com/pub/projects/kernel/kotd/ppc/HEAD/

timestamp needs to be at least 20060220092013
Comment 4 peter czanik 2006-02-21 09:18:44 UTC
Created attachment 69482 [details]
output of kernel update

There are some strange messages from a perl module at the end of the log. Probably due that there is no boot manager installed on Pegasos (might change with GRUB2 in the future...).
Comment 5 peter czanik 2006-02-21 09:27:37 UTC
I did about 30 reboots this morning with kernel-default-2.6.16_rc4-20060220183514.ppc.rpm

The situation improved a lot, but still not perfect.

When I boot this KOTD kernel over and over again, there are no more random time changes to the past, as it was the case before.

When I boot this KOTD kernel after the 10.0 kernel, the time change is no more random, but always back to 1970.

If I boot the 10.0 kernel over and over again, time remains correct between the reboots, just as before. Booting 10.0 after the 10.1 KOTD kernel shows correct time also.
Comment 6 peter czanik 2006-02-21 09:53:00 UTC
One more note: booting Gentoo with 2.6.15 after this KOTD has correct date, and booting this KOTD after the above mentioned Gentoo kernel also shows proper date.
Comment 7 Olaf Hering 2006-03-13 15:13:39 UTC
is that still broken? can you try 2.6.16-rc6? there was a related patch, maybe it fixes also this bug.
Comment 8 peter czanik 2006-03-13 17:10:33 UTC
Tested with KOTD, the bug is still there, the same way as in comment #5.
Comment 9 Olaf Hering 2006-03-15 16:15:48 UTC
I dont see it here, btw.  
do you know when it started?
Comment 10 peter czanik 2006-03-16 06:36:38 UTC
I noticed it some time in January, I don't recall when exactly. I also had a real battery problem for forgetting date, that's why I did not report it earlier.
Comment 11 Olaf Hering 2006-03-16 07:51:51 UTC
can you try a selfcompiled 2.6.15.X and see if its aleady broken there?
Comment 12 peter czanik 2006-03-16 08:34:23 UTC
Does a gentoo kernel count as self compiled? It's a 2.6.15 with gentoo patches and compiled by me.
Did a good number of reboots, and here are the results:
- gentoo->suse10.1: correct date
- suse10.0->suse10.1: wrong date
- suse10.0->gentoo: correct date, but booting 10.1 thereafter it's 1970 again.
In summary: if I ever boot 10.0, 10.1 will go back to 1970.
Comment 13 Olaf Hering 2006-03-16 09:39:10 UTC
hmm, can you do a chmod a-x /sbin/hwclock on all these installations? just to check if its the kernel or if some app sets the wrong date.
Comment 14 peter czanik 2006-03-16 10:00:28 UTC
Checked it: dates remain correct.
Comment 15 Ihno Krumreich 2006-04-25 07:43:27 UTC
Status?
Comment 16 Olaf Hering 2006-07-19 09:24:03 UTC
Peter, does this still happen? I guess it does.
Comment 17 peter czanik 2006-07-20 09:03:32 UTC
Yes, it still happens. I think, it's a problem with 10.0 hwclock, as I have not seen it with any other distributions, or any later SUSE versions.
Comment 18 Olaf Hering 2007-09-17 09:59:41 UTC
lets not bother about this anymore.