[ENBD] fr1 hangs when trying to access raid device..

[Peter T. Breuer]

"A month of sundays ago [Arve Emil Myr_s] wrote:"
[Charset iso-8859-1 unsupported, filtering to ASCII...]
> >> running to recompile a non-smp kernel now,, my raid controllers didn't  like the nosmp statement...
> >
> >Can you also try it with two small loopback devices as raid components?
> >I haven't tried scsi targets. There may be a difference in the
> >treatment of the first request on the target. If you can't do loop,
> >try ide.
> 
> ok, here goes:(with same smp-kernel)
> 
> #dd if=/dev/zero of=/root/loop0  bs=1024 count=2096
> #cp /root/loop0 /root/loop1
> #losetup /dev/loop0 /root/loop0
> #losetup /dev/loop1 /root/loop1

perfect.

> ..changing sdb1 & sdb2 in raidtab to loop0 & loop1..
> #mkraid --really-force /dev/md0
> ..Everything ok so far... (as expected)
> 
> #mke2fs /dev/md0
> mke2fs 1.28
> Filesystem label=
> OS type=Linux
> Block size=1024 (log=0)
> Fragment size=1024 (log=0)
> 1048 inodes, 4192 blocks
> 209 blocks (4,99%) reserved for Super User
> First data block=1
> 1 block group
> 8192 blocks per group, 8192 fragments pr group
> 1048 inodes per group
> 
> Segmentation fault

Oh, OK! That's an oops. But in any case, can you now repeat WITHOUT
doing a mke2fs? Or if you do do a mke2fs, use a block count (-b 1024
2048, I believe, in this case).

> And I'm back to a frozen system,, got a little bit longer but with the same result in the end..

Stick to loopback now that it's known you can do it with loop.

mke2fs deliberately writes beyond the device bounds. That should
cause an error condition. It may be that I didn't handle the error
condition correctly. Anyway, to confirm, see if you can read and write
the device normally (using dd), aand then see if giving the device size
explicitly to mke2fs helps. That would confirm.

Peter