raid10_make_request bug: can't convert block across chunks or bigger [...]

raid10_make_request bug: can't convert block across chunks or bigger [...]

[Ask Bjørn Hansen]

Hi,

I have a logical volume I'm (trying to) use for a Xen box.   When  
it's installing the boot loader I get a bunch of errors like the ones  
below.  The kernel is the latest FC6 kernel - 2.6.18-1.2869.fc6xen.

The md device is a raid10 device (obviously) across 4 sata disks.

  Any ideas?

[....]
raid10_make_request bug: can't convert block across chunks or bigger  
than 64k 81830779 4
raid10_make_request bug: can't convert block across chunks or bigger  
than 64k 81847167 4
raid10_make_request bug: can't convert block across chunks or bigger  
than 64k 81879931 4
raid10_make_request bug: can't convert block across chunks or bigger  
than 64k 81896315 4
[...]


  - ask

-- 
http://develooper.com/ - http://askask.com/

raid10_make_request bug: can't convert block across chunks or bigger [...]

[Andy Smith]

[-- Attachment #1: Type: text/plain, Size: 345 bytes --]

I'm now seeing the exact same problem as Ask did in:

http://www.mail-archive.com/linux-raid@vger.kernel.org/msg06762.html

again using md raid10 under LVM, trying to export an LV to a xen
domain.

This is on the latest Debian Etch kernel based on 2.6.18.4.

Ask didn't seem to get any on list replies - can anyone shed any light?

Cheers,
Andy

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: raid10_make_request bug: can't convert block across chunks or bigger [...]

[Andy Smith]

[-- Attachment #1: Type: text/plain, Size: 1114 bytes --]

Sorry for the reply to self, but the last thing I tried has provided
some more info:

On Sat, Apr 07, 2007 at 03:10:27AM +0000, Andy Smith wrote:
> I'm now seeing the exact same problem as Ask did in:
> 
> http://www.mail-archive.com/linux-raid@vger.kernel.org/msg06762.html
> 
> again using md raid10 under LVM, trying to export an LV to a xen
> domain.

This LV was initially created as 512M in size.  I extended it to 1G
and experienced the same thing.  I then created a new LV of size 2G,
copied the contents of the first LV to the second, and used that as
the domain's root filesystem.  I no longer get this problem.

I then deleted the original LV and renamed the new LV to the name of
the original and things still work.  I am able to do:

dd if=/dev/sda1 of=/dev/null

in the Xen domain and not receive any errors so I have to conclude
that all blocks of this second LV are readable.  So is this an issue
of the layout of the LVM extents on disk?  Will I get this problem
again?  Should I be pursuing this on the LVM list, here, or maybe
even with the Xen folks?

Cheers,
Andy

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]