Bug 104350

Summary:	scsi errorhandler from usb storage device not removed
Product:	[openSUSE] SUSE LINUX 10.0	Reporter:	Danny Al-Gaaf <dalgaaf>
Component:	Kernel	Assignee:	Jens Axboe <axboe>
Status:	RESOLVED INVALID	QA Contact:	E-mail List <qa-bugs>
Severity:	Normal
Priority:	P5 - None
Version:	Beta 1
Target Milestone:	---
Hardware:	Other
OS:	All
Whiteboard:
Found By:	Other	Services Priority:
Business Priority:		Blocker:	---
Marketing QA Status:	---	IT Deployment:	---

Description Danny Al-Gaaf 2005-08-12 10:00:20 UTC

I added and removed a USB Harddisk several times. In this case the [scsi_eh_*] 
not full removed. Always if I plug in the USB storage device again the 
[scsi_eh_*] is increased. 

Because of this I can't unload the sd_mod module because it is 4 or more time 
used.

Also a problem: one partition of the harddisk is xfs. There are after remove 
several xfs related processes.

----------------------
root      6040  0.0  0.0      0     0 ?        S    13:32   0:00 [kjournald]
root      6052  0.0  0.0      0     0 ?        S<   13:32   0:00 [xfslogd/0]
root      6054  0.0  0.0      0     0 ?        S<   13:32   0:00 [xfsdatad/0]
root      6055  0.0  0.0      0     0 ?        S    13:32   0:00 [xfsbufd]
root      6158  0.0  0.0      0     0 ?        S    13:41   0:00 [scsi_eh_1]
root      6217  0.0  0.0      0     0 ?        S    13:41   0:00 [kjournald]
----------------------
and from an other machine:

root     17423  0.0  0.0      0     0 ?        S    11:41   0:00 [scsi_eh_2]
root     17490  0.0  0.0      0     0 ?        S<   11:41   0:00 [xfslogd/0]
root     17492  0.0  0.0      0     0 ?        S<   11:41   0:00 [xfsdatad/0]
root     17493  0.0  0.0      0     0 ?        S    11:41   0:00 [xfsbufd]
root     17495  0.0  0.0      0     0 ?        S    11:41   0:00 [xfssyncd]
root     17608  0.0  0.0      0     0 ?        S    11:41   0:00 [scsi_eh_6]
root     17671  0.0  0.0      0     0 ?        S    11:41   0:00 [kjournald]

Comment 1 Hubert Mantel 2005-08-12 10:03:01 UTC

Not being able to unload the sd_mod module does not qualify as a critical bug.

Comment 2 Danny Al-Gaaf 2005-08-12 10:17:09 UTC

sorry ... the problem with sd_mod is solved: Because of fast add and remove the 
device several times not all partitions are unmounted. But the problem with 
[scsi_eh_*] and xfs processes exists anymore also if I umount the related 
partitions

Comment 3 Olaf Kirch 2005-09-01 11:03:41 UTC

Can we close this report? Can you reproduce the problem on beta4?

Comment 4 Danny Al-Gaaf 2005-09-01 11:55:49 UTC

Can't reproduce this at the moment. I close the bug and reopen if this happen 
again

Comment 5 Danny Al-Gaaf 2005-09-01 13:33:06 UTC

While tests with firewire and usb with a harddisk with 4 partitions the same 
happens again. The device is removed and all mountpoints are umounted and there 
are now again two [scsi_eh_*] left

Comment 6 Olaf Kirch 2005-09-05 19:33:47 UTC

Chris, can someone from your team look into this?

Comment 7 Chris L Mason 2005-09-05 20:28:53 UTC

Jens, does this sound like a known bug?

Comment 8 Olaf Hering 2005-09-05 20:39:32 UTC

asked once about that, but it happens also with usb, right?

<olaf>  jejb: sbp2 drives will get a new scsi_eh_N after each replugin. should I
be concerned? 
<olaf>  do you know if thats fixed in your trees, or by some patches?
<jejb>  olaf, I presume they remove and re-add the host?
<olaf>  I cant remember how it behaved in older kernels
<jejb>  olaf, if they remove and re-add the host then its expected
<olaf>  how large can N grow?
<jejb>  boundless
<olaf>  ok, guess they will fix it someday
<hch>   well, 32bits
<hch>   and then it starts from 0 and causes bad trouble
<hch>   we should maybe switch it to an idr allocator to reallocate them as soon
as the old hose has been completely released
<jejb>  hch, yes, I was wondering that ... it would stop Ben Collins whining too
... 
<olaf>  I think the physical connector will die before I fill an uint 


does the left over kernel thread cause problems?

Comment 9 Bernhard Kaindl 2006-02-22 16:59:52 UTC

Since the last question was not answered since more than 5 months,
I do not assume that the xfs processes cause any problems.

It seems that this is normal, I tested mounting and unmounting xfs
with the current 10.1 kernel and also with this newer xfs, two xfs
processes stay, but repeated mounting/unmounting works without problems
and the processes are cleaned up when unloading the xfs module.

xfs modules staying ore being removed is not a function of
the usb storage driver, so I'd say that has also nothing to
do with unplugging an usb disk (if it's unmounted before).

Just in case, the very latest kernels can be found in
ftp://ftp.suse.com/pub/projects/kernel/kotd/