Bug 397816

Created attachment 220616 [details]
additional partitioning details

I tried modprobe  dm-multipath prior to kpartx -a /dev/sda, but that did not change result.

I am a-865 on freenode's #opensuse-factory.

There was a bug in kpartx which prevented it from creating the correct dm devices for extended partitions. Please verify that the kpartx changelog contains:

* Wed May 28 2008 hare@suse.de
- Calculate correct partition offset in kpartx (bnc#394658)

(In reply to comment #3 from Hannes Reinecke)
> Please verify that the kpartx changelog contains:
 
# changelog kpartx
-bash: changelog: not found
# kpartx --changelog
kpartx: invalid option
# kpartx -v
(usage)
# rpm -qa | grep kpartx
kpartx-0.4.7-127
#

I have no idea how to find this changelog info. 0.4.7-127 is what is now on the mirrors.

rpm -qi --changelog kpartx
would give you the answer. But seeing the 0.4.7-128 contains the fix, chances are that -127 does not.

-127 has a 30 May timestamp and shows
* Wed May 28 2008 hare@suse.de
- Calculate correct partition offset in kpartx (bnc#394658)

Apparently -127 is what I used for filing this bug, which still exists now with -128 not yet available.

Would have been too easy. Hmm.

Oh for crying out loud.
The DFSee tool is _so_ broken. If you ever come across the author please let him know that there are systems with more than 8 disks.

Sigh.

Sorry, no luck here.

I managed to re-create the partition table here locally:

yarrow:/tmp/dfsee/linux # kpartx -a -l /dev/sdb
sdb1 : 0 208782 /dev/sdb 63
sdb2 : 0 16065 /dev/sdb 208845
sdb3 : 0 514080 /dev/sdb 224910
sdb4 : 0 311805585 /dev/sdb 755055
sdb5 : 0 417627 /dev/dm-3 63
sdb6 : 0 13108977 /dev/dm-3 433818
sdb7 : 0 514017 /dev/dm-3 13542858
sdb8 : 0 2104452 /dev/dm-3 14073003
sdb9 : 0 9831717 /dev/dm-3 16177518
sdb10 : 0 9831717 /dev/dm-3 26009298
sdb11 : 0 9831717 /dev/dm-3 35841078
sdb12 : 0 3277197 /dev/dm-3 45672858
sdb13 : 0 14345982 /dev/dm-3 48950118
sdb14 : 0 481887 /dev/dm-3 63296163
sdb15 : 0 2249037 /dev/dm-3 63778113
sdb16 : 0 9831717 /dev/dm-3 233151408
sdb17 : 0 9831717 /dev/dm-3 242983188
sdb18 : 0 9831717 /dev/dm-3 252814968
sdb19 : 0 9831717 /dev/dm-3 262646748
sdb20 : 0 9831717 /dev/dm-3 272478528
sdb21 : 0 9831717 /dev/dm-3 282310308
sdb22 : 0 9831717 /dev/dm-3 292142088
sdb23 : 0 9831717 /dev/dm-3 301973868
yarrow:/tmp/dfsee/linux # kpartx -a /dev/sdb
yarrow:/tmp/dfsee/linux # dmsetup table
sdb14: 0 481887 linear 253:3 63296163
sdb3: 0 514080 linear 8:16 224910
sdb13: 0 14345982 linear 253:3 48950118
sdb2: 0 16065 linear 8:16 208845
sdb12: 0 3277197 linear 253:3 45672858
sdb1: 0 208782 linear 8:16 63
sdb11: 0 9831717 linear 253:3 35841078
sdb10: 0 9831717 linear 253:3 26009298
sdb23: 0 9831717 linear 253:3 301973868
sdb22: 0 9831717 linear 253:3 292142088
sdb9: 0 9831717 linear 253:3 16177518
sdb19: 0 9831717 linear 253:3 262646748
sdb21: 0 9831717 linear 253:3 282310308
sdb8: 0 2104452 linear 253:3 14073003
sdb18: 0 9831717 linear 253:3 252814968
sdb20: 0 9831717 linear 253:3 272478528
sdb7: 0 514017 linear 253:3 13542858
sdb17: 0 9831717 linear 253:3 242983188
sdb6: 0 13108977 linear 253:3 433818
sdb16: 0 9831717 linear 253:3 233151408
sdb5: 0 417627 linear 253:3 63
sdb15: 0 2249037 linear 253:3 63778113
sdb4: 0 311805585 linear 8:16 755055

Can you attach /var/log/messages?
And 'kpartx -v -a /dev/sda' might be helpful, too.

Does it work with OpenSUSE 11.0?

(In reply to comment #8 from Hannes Reinecke)
> Oh for crying out loud.
> The DFSee tool is _so_ broken. If you ever come across the author please let
> him know that there are systems with more than 8 disks.

He will normally reply to bug reports within 24 hours, but this week and next he is on holiday. I don't know what the "broken" is, so you'll need to report what makes you cry out loud.
http://tech.groups.yahoo.com/group/dfsee-support/?yguid=12054984
http://www.dfsee.com/dfsee/

Created attachment 221461 [details]
/var/log/messages tail

Can you try to switch off smartd?
There might be a chance that smartd is keeping the device busy and device is refusing to do something with it.

(In reply to comment #10 from Hannes Reinecke)
> Does it work with OpenSUSE 11.0?
 
This particular install reports itself as 11.0 and was updated after report on mailing list of Factory lockdown for 11.0 release. When I try to access gwdg or mirrors.kernel.org ftp 11.0 dirs I get access denied messages.

(In reply to comment #13 from Hannes Reinecke)
> Can you try to switch off smartd?
> There might be a chance that smartd is keeping the device busy and device is
> refusing to do something with it.
 
I turned it off and rebooted but get the same result.

Created attachment 221467 [details]
boot track

Created attachment 221468 [details]
display output of kpartx -v -a /dev/sda

(In reply to comment #10 from Hannes Reinecke)
> And 'kpartx -v -a /dev/sda' might be helpful, too.
.

Created attachment 221471 [details]
files created by dfsdisk script

DFSee can use these to precisely replicate the partitioning of my ST3160023A 8.01 /dev/sda. http://www.dfsee.com/dfsee/dfsee9xx_linux.tgz

(In reply to comment #13 from Hannes Reinecke)
> There might be a chance that ... keeping the device busy and device is
> refusing to do something with it.
 
Could this problem have anything to do with the fact that the system is a single disk system?

/etc/mtab:
/dev/sda10 / ext3 rw,noatime,nodiratime,acl 0 0
/dev/sda3 /disks/C msdos rw,noexec,nosuid,nodev,dmask=0000,fmask=0111 0 0
/dev/sda5 /disks/hda/boot ext2 rw,noatime,nodiratime,noacl 0 0
/dev/sda6 /disks/D vfat rw,noexec,nosuid,nodev,dmask=0000,fmask=0111,utf8=true 0 0
/dev/sda7 /disks/E msdos rw,noexec,nosuid,nodev,dmask=0000,fmask=0111 0 0
/dev/sda12 /home ext3 rw,noatime,nodiratime,acl 0 0
/dev/sda13 /pub ext3 rw,noatime,nodiratime,acl 0 0
/dev/sda14 /srv ext3 rw,noatime,nodiratime,noacl 0 0
/dev/sda15 /usr/local ext3 rw,noatime,nodiratime,noacl 0 0

I tried running kpartx -v -a /dev/sda on Mandriva Cooker and get similar messages, e.g. this tail:
device-mapper: table: 252:0: linear: dm-linear: Device lookup failed
device-mapper: ioctl: error adding target to table
device-mapper: reload ioctl failed: Invalid argument
add map sda23: 0 9831717 linear 0:0 302728923

Pretty much so. device-mapper refuses to create dm-devices if the underlying devices are busy, eg mounted.

So I fear we need to do some initrd hackery here to get the required result.

What you need to do is to create a initrd script /lib/mkinitrd/scripts/boot-dm_linear.sh which just creates a linear table on top of your existing sda device. mkinitrd already has the required infrastructure in
place, so whenever a dm-device is specified as a root device it'll do the right
thing, like running kpartx etc.

Then re-run mkinitrd and everything should be nice an dandy.

Let's see how the mkinitrd script could look like ...

(In reply to comment #20 from Hannes Reinecke)
> What you need to do is to create a initrd script

Literally? I'm no script writer.

> Let's see how the mkinitrd script could look like ...
 
I can only guess one line, same as subject, 'kpartx -a /dev/sda', which would only be good if sda is the only disk you want it done on.

Oh well. It all comes back to me.

There is a script /sbin/activate_dm_linear, which takes the device name as an argument. That will create the required udev rules.
And you need to set 'DM_BLOCK=yes' in /etc/sysconfig/kernel.

Then re-run mkinitrd, reboot and you should be set.

Created attachment 221792 [details]
output of mkinitrd -v

Created attachment 221793 [details]
/var/log/boot.msg from using new initrd

(In reply to comment #22 from Hannes Reinecke)
> There is a script /sbin/activate_dm_linear, which takes the device name as an
> argument. That will create the required udev rules.

So all /lib/mkinitrd/scripts/boot-dm_linear.sh needs is? ->

/sbin/activate_dm_linear /dev/sda

> And you need to set 'DM_BLOCK=yes' in /etc/sysconfig/kernel.

With, or without, quotes aroung yes? I tried with.
 
>Then re-run mkinitrd, reboot and you should be set.

It booted, but paused waiting for root to appear, then gave: 
Waiting for device /dev/root to appear: ..............................Could not find /dev/root.
Want me to fall back to /dev/disk/by-id/ata-ST3160023A_4JS0L8FE-part10? (Y/n) 
Waiting for device /dev/disk/by-id/ata-ST3160023A_4JS0L8FE-part10 to appear:  ok

I don't see anything new in the /dev tree. /dev/mapper only has control. Nothing in /dev/disk beyond 15. Only 3 /dev/d* files.

Created attachment 227827 [details]
activate_dm_linear

Updated activate_dm_linear script.

Please use the above script to update the udev rules and re-run mkinitrd.
Ensure that 'kpartx' is listed in the features of the mkinitrd output.
The system should now boot up running on a device-mapper disk with all partitions visible (as device-mapper devices).

Created attachment 228274 [details]
mkintrd, fdisk -l, mount, /dev/md*, /dev/dm*, /var/log/boot.msg

I don't think it worked.

It must have worked at some point. Now after some more updates without paying any mind to this, and using 2.6.26-2, mount shows:

/dev/dm-22 on / type ext3 (rw,noatime,nodiratime,acl)
/dev/dm-12 on /disks/hda/boot type ext2 (rw,noatime,nodiratime,noacl)
/dev/dm-26 on /home type ext3 (rw,noatime,nodiratime,acl)
/dev/dm-28 on /pub type ext3 (rw,noatime,nodiratime,acl)
/dev/dm-30 on /srv type ext3 (rw,noatime,nodiratime,noacl)
/dev/dm-32 on /usr/local type ext3 (rw,noatime,nodiratime,noacl)
/dev/dm-24 on /disks/hda/fedora type ext3 (rw,noatime,nodiratime,noacl)

Is it time to try a fresh install and see if the installer offers up all my partitions? If so, do I need to manually intervene with kpartx at installation startup?

Closing as 11.0 is now out of maintenance. Please reopen if you encounter this bug in any maintained openSUSE release. Thanks.