system does not boot due to device-mapper error

I can reproduce this on 9.04 with the 2.6.31-rc2 kernel from the ubuntu kernel PPA:

When booting dmraid fails to set up the raid devices and dmesg shows "device-mapper: table: 252:0: sda too small for target" and "device-mapper: table: 252:0: sdb too small for target".

I am also affected by this. I have a two disc raid 0 (ICH9R fakeraid, Abit IP35 Pro).

Also happens with the 2.6.31-rc3 kernel from the repositories.

I can confirm this with all 2.6.31 kernels in karmic to date, *and* with the latest kernel from http://kernel.ubuntu.com/~kernel-ppa/mainline/daily/. The 2.6.30 kernel that used to be in the repositories works fine though (using it at the moment).

My system is on two disc on raid0 with intel software raid.

When I try to activate the raid set manually from the busybox shell prompt, after the boot sequence times out, (by typing dmraid -ay), I get the message 'The set isw_ezfgzegerh_zef could not be activated' (or similar).

I can confirm this also, with ICH10R, I filed a bug and I am experiencing the same problem, cannot get activated and cannot boot with dmraid with the same message.

Has anyone filed a bug against the kernel on the kernel bug tracker? If so, can we get the link here?

DMRAID is so bugged that now I'm afraid of just using it and wasting my entire array, someone should definetly post on the kernel bug tracker, we don't need more bugs, with dmraid alone is enough.

For Kanotix I patch my kernels using this patch since I found it, works fine. Ubuntu seems to ignore my hints in the IRC, so here you have it again in your beloved LP:

http://patchwork.kernel.org/patch/33534/

I have tested a kernel with the patch (.31-4) and it worked for me. I was able to boot my system just fine.

For reference, Roman tested http://kernel.ubuntu.com/~dtchen/test-kernels/lp398059/, the source of which is at http://kernel.ubuntu.com/git?p=dtchen/ubuntu-karmic.git;a=shortlog;h=refs/heads/lp398059

It seems that a version of this patch has been mainlined since 2.6.31-rc4 and will be in the next rebase of the Karmic kernel which should occur this week:

  commit 5dea271b6d87bd1d79a59c1d5baac2596a841c37
  Author: Mike Snitzer <email address hidden>
  Date: Thu Jul 23 20:30:42 2009 +0100

    dm table: pass correct dev area size to device_area_is_valid

    Incorrect device area lengths are being passed to device_area_is_valid().

    The regression appeared in 2.6.31-rc1 through commit
    754c5fc7ebb417b23601a6222a6005cc2e7f2913.

    With the dm-stripe target, the size of the target (ti->len) was used
    instead of the stripe_width (ti->len/#stripes). An example of a
    consequent incorrect error message is:

      device-mapper: table: 254:0: sdb too small for target

    Signed-off-by: Mike Snitzer <email address hidden>
    Signed-off-by: Alasdair G Kergon <email address hidden>

Fixed in the -5.24 upload.