Bug#409875: cannot setup device-mapper mapping ontop of /dev/md* device

Bug#409875: cannot setup device-mapper mapping ontop of /dev/md* device

[David Härdeman]

On Thu, Feb 08, 2007 at 10:51:50PM +0100, Johannes Schlumberger wrote:
>I built the packages and installed them.
>The attached file contains the generated stderr stream (which looks a bit
>strange to me). Just to make shure, here is also the output generated from the
>command without the redirection.

The stderr stream is fine. When you run it without redirection, bash 
will helpfully mask the null characters that are printed (which is not 
done when they are redirected to a file).

However, everything looks ok after studying that file, the first two 
ioctl's are executed (i.e. the device-mapper version is checked and the 
"root" mapping name is reserved).

Then when it is time to pass the actual parameters for the mapping using 
the "reload" ioctl, everything goes south. I'm quickly running out of 
ideas what might cause this or what the best way to debug it would be.

So I'm CC:ing this mail to some different people with the hope that 
someone other than me can spot the error.

Quoting from previous mails:
> [04:11:35][root@asso:/home/spjsschl]# cryptsetup -c aes -s 256 -d /dev/stdin
> create root /dev/md0 < /key
> Command failed: device-mapper: reload ioctl failed: Invalid argument
> 
> [04:12:57][root@asso:/home/spjsschl]# dmesg | tail -n 3
> device-mapper: table: 253:2: crypt: Device lookup failed
> device-mapper: ioctl: error adding target to table
> device-mapper: ioctl: device doesn't appear to be in the dev hash table.
> 
> [22:28:31][root@asso:/tmp]# echo '0 10000 crypt aes-cbc-essiv:sha256 \
> 0123456789abcdef0123456789abcdef 0 /dev/md0 0' > map
> [22:28:46][root@asso:/tmp]# dmsetup create root map
> device-mapper: reload ioctl failed: Invalid argument Command failed

Executive summary:
dmsetup / cryptsetup both fail to create a mapping on top of a raid 
device (/dev/md0 in this case).

For much more details, see the bug report, any help appreciated:
http://bugs.debian.org/409875

-- 
David Härdeman

Bug#409875: cannot setup device-mapper mapping ontop of /dev/md* device

[Loïc Minier]

On Thu, Feb 08, 2007, David Härdeman wrote:
> dmsetup / cryptsetup both fail to create a mapping on top of a raid 
> device (/dev/md0 in this case).

 I don't know where the error lies, but I created two loop devices loop0
 and loop1, added them to a new RAID 1 md0 device, and ran cryptsetup on
 the resulting md0 successfully.

 bee:~# dmsetup  ls
 ...
 crypt0  (253, 6)

 bee:~# dmsetup table crypt0
 0 65408 crypt aes-cbc-plain
 42cfa211018ea492fdee45ac637b7972a0ad687380f9a6d0484900ad8cca2d8f 0 9:0
 0

 bee:~# mdadm --detail /dev/md0
 ...
    Number   Major   Minor   RaidDevice State
       0       7        0        0      active sync   /dev/loop0
       1       7        1        1      active sync   /dev/loop1

-- 
Loïc Minier <lool@dooz.org>

Bug#409875: [Pkg-cryptsetup-devel] Bug#409875: cannot setup device-mapper mapping ontop of /dev/md* device

[David Härdeman]

On Fri, February 9, 2007 11:45, Loic Minier said:
> On Thu, Feb 08, 2007, David Härdeman wrote:
>> dmsetup / cryptsetup both fail to create a mapping on top of a raid
>> device (/dev/md0 in this case).
>
>  I don't know where the error lies, but I created two loop devices loop0
>  and loop1, added them to a new RAID 1 md0 device, and ran cryptsetup on
>  the resulting md0 successfully.

Ditto. I've succesfully setup a crypto mapping on a raid1 device under
qemu, which leads me to believe that it is something specific to Johannes
setup.

So I'm hoping that someone will come up with some suggestions on how to
troubleshoot it further on his machine.

(This is my last cross-post on this subject, I promise, please keep
further discussions to the bug report address only -
409875@bugs.debian.org)

-- 
David Härdeman



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Bug#409875: [Pkg-cryptsetup-devel] Bug#409875: cannot setup device-mapper mapping ontop of /dev/md* device

[Henrique de Moraes Holschuh]

On Fri, 09 Feb 2007, David Härdeman wrote:
> > On Thu, Feb 08, 2007, David Härdeman wrote:
> >> dmsetup / cryptsetup both fail to create a mapping on top of a raid
> >> device (/dev/md0 in this case).
> >
> >  I don't know where the error lies, but I created two loop devices loop0
> >  and loop1, added them to a new RAID 1 md0 device, and ran cryptsetup on
> >  the resulting md0 successfully.
> 
> Ditto. I've succesfully setup a crypto mapping on a raid1 device under
> qemu, which leads me to believe that it is something specific to Johannes
> setup.

Warning: It is not safe on all kernels out there, due to kernel bugs.  See
recent flow of patches from the md maintainer in LKML.  I don't know if they
all landed on 2.6.16.y and 2.6.19.y and 2.6.20 yet.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

Bug#409875: [Pkg-cryptsetup-devel] Bug#409875: bug #409875

[David Härdeman]

reassign 409875 libdevmapper1.02
retitle 409875 libdevmapper should provide more helpful error messages
severity 409875 minor
thanks

On Fri, Feb 09, 2007 at 09:44:32PM +0100, Johannes Schlumberger wrote:
>> Thanks, this looks mighty suspicious...it looks like /dev/md0 is mounted 
>> as your root partition by the initramfs? Please provide me with "ls -al 
>> /dev/root" and "cat /proc/cmdline" as well as the contents of your grub 
>> or lilo config file.
>
>I am so sorry for stealing yours and other persons time like this. I had
>root=/dev/md0 in my menu.lst and / as /dev/hdd6 in my fstab. I have to say I am
>strongly embarassed.

No problem, it happends to all of us. There still is a bug here 
though...the error messages from libdevmapper are obtuse to say the 
least.

libdevmapper explained that /dev/md0 is in use with the message:
"Command failed: device-mapper: reload ioctl failed: Invalid argument"

That is a bug in my book.

I'll reassign this bug to libdevmapper (and retitle it accordingly).

I'll also CC all people that I've bugged earlier so that they know that 
this issue is no more.

>Thank you a lot for figuring this out and all your patience.

That's ok, you just owe me a beer.

-- 
David Härdeman