Xen devel - Mar 2006 - patch to test vbd loopback sharing issue

Hi Ping and Rick,

Could you both test this patch to see if the loopback sharing issues go
away? Ping was right, Xend - or the /etc/xen/scripts/block script -
checks to make sure if a device is already in use. I''m not sure why we
didn''t hit this before, but we are now.

This patch just changes the mode for HVM disk.img mount to
''w!'' to force
it to not check for sharing. We need the sharing currently. 

Thanks to Dan Smith for the suggested fix.

Thanks,

Dan




_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Daniel,

Thanks a lot for your patch. In fact, I have another solution for this problem,
that is, we can force mode "w" to "r". I have tested this
patch and it works quite well in today''s xm-test report.
Please help me review this patch. :-)
>-----Original Message-----
>From: Daniel Stekloff [mailto:dsteklof@us.ibm.com]
>Sent: 2006年3月17日 8:51
>To: Ricardo C Gonzalez; Yu, Ping Y
>Cc: Xen Mailing List
>Subject: patch to test vbd loopback sharing issue
>
>
>Hi Ping and Rick,
>
>Could you both test this patch to see if the loopback sharing issues go
>away? Ping was right, Xend - or the /etc/xen/scripts/block script -
>checks to make sure if a device is already in use. I''m not sure why
we
>didn''t hit this before, but we are now.
>
>This patch just changes the mode for HVM disk.img mount to
''w!'' to force
>it to not check for sharing. We need the sharing currently.
>
>Thanks to Dan Smith for the suggested fix.
>
>Thanks,
>
>Dan
>


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

PY> Daniel, Thanks a lot for your patch. In fact, I have another
PY> solution for this problem, that is, we can force mode "w" to
PY> "r". I have tested this patch and it works quite well in
today''s
PY> xm-test report.  Please help me review this patch. :-)

AFAIK, qemu-dm does not honor the read-only flag from xend.  This has
a side effect of making the concurrent use checks go away, but it also
would indicate to someone that the disk image is read-only, which
would not be true.  So, since ''r'' is not honored,
''w!'' is more
appropriate, as it describes what is actually happening.

So, IMHO, Dan''s patch should go into the tree.

-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms@us.ibm.com


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Thu, Mar 16, 2006 at 04:51:04PM -0800, Daniel Stekloff wrote:
> # HG changeset patch
> # User stekloff@elm3b49.beaverton.ibm.com
> # Node ID 66c247ce736523fa20dd142232bf3f5c25c65813
> # Parent  00aba64e034e70a262cbe3b729b3601ce8142eda
> 
> Make sure the check to see if device is in use doesn''t happen. We
need
> to share disk.imgs. Qemu doesn''t honor the ''r''
read-only value.
> 
> Signed-off-by: Daniel Stekloff <dsteklof@us.ibm.com>
Applied, thanks.

Ewan.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Fri, Mar 17, 2006 at 09:48:28AM +0800, Yu, Ping Y wrote:
> Daniel,
> 
> Thanks a lot for your patch. In fact, I have another solution for this
problem, that is, we can force mode "w" to "r". I have
tested this patch and it works quite well in today''s xm-test report.
Following Dan Smith''s argument that "r" is misleading, given
that qemu-dm does
not honour the read-only flag, I have taken Dan Stekloff''s patch, using
"w!"
to avoid the multiple writer check, instead.

Ewan.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel