Kernel Bug Tracker – Bug 13017
ATA bus errors on resume
Last modified: 2012-05-30 15:19:23 UTC
Subject : 2.6.29 regression: ATA bus errors on resume
Submitter : Niel Lambrechts <firstname.lastname@example.org>
Date : 2009-03-25 5:19
References : http://marc.info/?l=linux-kernel&m=123795841615989&w=4
Handled-By : Tejun Heo <email@example.com>
Patch : http://marc.info/?l=linux-kernel&m=123865517107421&w=4
This entry is being used for tracking a regression from 2.6.28. Please don't
close it until the problem is fixed in the mainline.
Ignore-Patch : http://marc.info/?l=linux-kernel&m=123865517107421&w=4
No reporter response, closing.
EXT4 file-system errors resurfaced in 2.6.30-rc6. Was able to trigger the bug with Tejun Heo's debug patch. Problem under discussion.
Created attachment 21548 [details]
EXT4 errors, captured with Tejun's debug patch
Created attachment 21549 [details]
No EXT4 errors, using Tejun's debug patch
Created attachment 21550 [details]
messages (126.96.36.199 - no EXT4 errors)
Created attachment 21551 [details]
messages (188.8.131.52 - no EXT4 errors)
Description of above /var/log/messages attachments:
I did a hibernation/resume cycle using console, then X for each kernel I have back to 2.6.28. I originally noticed EXT4 errors in 2.6.29 and now again in 2.6.30-rc6 (perhaps rc5 as well), but these still only happen intermittently.
I attached 2 x /var/log/messages output for 2.6.30-rc6. Both running Tejun's debug patch, but one of them has output when the EXT4 file-system errors triggered.
The other older kernel version messages I attached for the sake of comparison. They also seem to contain ATA resets.
Tested Tejun's readahead patch, which did not trigger the bug.
Bug seen again in latest 2.6.30-rc7 git kernel, with a resume after hibernating leading to EXT4 errors.
Also, it seems 184.108.40.206 is good, I took some time to try and trigger it there, but it did not even when hibernating with a reasonably active system.
I am going to try and bisect this again, it is driving me up the wall.
Is this bug still present in current mainline kernels?