Bug 134471

Summary:	Problem with mke2fs during install
Product:	[openSUSE] SUSE Linux 10.1	Reporter:	Tony Jones <tonyj>
Component:	Installation	Assignee:	Hendrik Vogelsang <hvogel>
Status:	RESOLVED FIXED	QA Contact:	Klaus Kämpf <kkaempf>
Severity:	Normal
Priority:	P5 - None
Version:	Alpha 3
Target Milestone:	---
Hardware:	IA64
OS:	Other
Whiteboard:
Found By:	Other	Services Priority:
Business Priority:		Blocker:	---
Marketing QA Status:	---	IT Deployment:	---
Attachments:	Y2log of mk2efs strace of mke2fs (run from vcons/2)

Description Tony Jones 2005-11-18 20:33:56 UTC

Following partition table exists:

 Disk geometry for /dev/sdb: 0kB - 36GB
 Disk label type: gpt
 Number  Start   End     Size    File system  Name                  Flags
 1       17kB    200MB   200MB   fat32                              
 2       200MB   36GB    36GB    ext3     

Reusing these partitions but instructing Yast to reformat them results in mke2fs appearing to hang for 5 minutes.

ps shows the following:

F S   UID   PID  PPID  C PRI  NI ADDR SZ WCHAN  TTY          TIME CMD
0 D     0  2972  2101  1  78   0 -   346 get_re tty1     00:00:02 mke2fs

Totally reproducable when run directly from vcons (HP zx6000 2x36GB U320 disk).
It gets as far as "Writing superblocks and filesystem accounting information:" and then goes on vacation for 5 minutes (no disk activity).

See attached y2log for more details.  It eventually recovers and proceeeds.

Comment 1 Tony Jones 2005-11-18 20:34:25 UTC

Created attachment 57758 [details]
Y2log of mk2efs

Comment 4 Arvin Schnell 2005-11-22 09:24:55 UTC

Since it also hangs when run on the vcon this really looks like
a problem within mke2fs.

Comment 5 Hendrik Vogelsang 2005-11-22 09:32:19 UTC

Tony please try mke2fs calls in the running system. You can use the -n switch that does dry runs...

Comment 6 Tony Jones 2005-11-22 18:59:04 UTC

-n from the INSTALLER vcons works just fine.

I don't have a spare disk to try actually running it from the booted SL10.1

Suggest you try it from the installation DVD.
The default install is reiser3, so it's possible this just hasn't been tried before, or maybe it's specific to my hardware.

Tue Nov 22 10:35:53 PST 2005
0 S root      2707   712  1  80   0 -   218 wait   10:32 tty2     00:00:02 strace -tt mke2fs /dev/sdb2
0 D root      2708  2707  1  78   0 -   456 get_re 10:32 tty2     00:00:03 mke2fs /dev/sdb2
0 S root      2745  2742  0  79   0 -   321 pipe_w 10:35 tty6     00:00:00 grep mke2fs
WIDE_WCHAN_COLUMN   PID
get_request_wait   2708

Attaching strace of what it's doing.

For comparison, on Debian, 'mke2fs /dev/sdb2" (same target disk) takes 12.18 seconds elapsed time to complete.

Comment 7 Tony Jones 2005-11-22 19:02:04 UTC

Created attachment 58130 [details]
strace of mke2fs (run from vcons/2)

Actually not that useful, just shows a 4 minute "black hole".

Comment 8 Hendrik Vogelsang 2005-11-28 10:48:38 UTC

is there anything else in the logs? Maybe harddisk errors in the kernel buffer? (dmesg)

Comment 9 Tony Jones 2005-12-23 20:09:47 UTC

The machine got wiped. Sorry for the delay.

I was waiting for an alpha4 iso to retest but it appears there won't be one for ia64.

The problem does not occur when installing SLES10p3.  This is mke2fs 1.38.    I imagine both SL10.1 and SLES are using the same stable package?

Anyways, clearly the disk isn't at fault.

Could be an issue with the installation kernel. I didn't pay attention to what the versioning was for this.

Let me know how you want to proceed.

Comment 10 Tony Jones 2006-01-25 18:49:16 UTC

This doesn't happen any longer in SLES10p4.  Recommend close.

Comment 11 Hendrik Vogelsang 2006-01-29 13:56:54 UTC

ok