Xen devel - Mar 2006 - [PATCH] build in SATA drivers to the -xen kernel for x86_64

This patch adds in the SATA drivers that are supported in the -xen0 kernel. 
Because the default behavior of xen is to reboot on a dom0 crash (i.e. not
finding root filesystem), and the xen0 kernel supported SATA directly, a
number of people (including myself) got caught by the switch over.

Long term, we probably want to modularize more of the kernel, and make
initrd building more of the default, however this should ease people''s
transition from the -xen0 => -xen kernel for testing on x86_64.

	-Sean

-- 
Sean Dague
IBM Linux Technology Center                     email: japh@us.ibm.com
Open Hypervisor Team                           alt: sldague@us.ibm.com



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

On 29 Mar 2006, at 19:07, Sean Dague wrote:
> This patch adds in the SATA drivers that are supported in the -xen0 
> kernel.
> Because the default behavior of xen is to reboot on a dom0 crash (i.e. 
> not
> finding root filesystem), and the xen0 kernel supported SATA directly, 
> a
> number of people (including myself) got caught by the switch over.
>
> Long term, we probably want to modularize more of the kernel, and make
> initrd building more of the default, however this should ease
people''s
> transition from the -xen0 => -xen kernel for testing on x86_64.
If this is done at all, it should be done for i386 as well. I think 
it''s debatable really -- -xen is supposed to be a modular kernel build.
If people want stuff built in they should modify the config or use 
-xen0.

  -- Keir


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

On Wed, Mar 29, 2006 at 07:19:18PM +0100, Keir Fraser
wrote:
> 
> On 29 Mar 2006, at 19:07, Sean Dague wrote:
> 
> >This patch adds in the SATA drivers that are supported in the -xen0 
> >kernel.
> >Because the default behavior of xen is to reboot on a dom0 crash (i.e. 
> >not
> >finding root filesystem), and the xen0 kernel supported SATA directly, 
> >a
> >number of people (including myself) got caught by the switch over.
> >
> >Long term, we probably want to modularize more of the kernel, and make
> >initrd building more of the default, however this should ease
people''s
> >transition from the -xen0 => -xen kernel for testing on x86_64.
> 
> If this is done at all, it should be done for i386 as well. I think 
> it''s debatable really -- -xen is supposed to be a modular kernel
build.
> If people want stuff built in they should modify the config or use 
> -xen0.
Well, SATA is significantly more relevant for x86_64, as a very large
percentage of x86_64 systems are SATA based.  I doubt that even 1/2 of the
IDE controllers that are compiled in by default for the x86_64 kernel have
ever existed in an x86_64 system.  If we''re shooting for truely
modular, we
should start removing the IDE drivers from the -xen kernel.

I would agree that changes here are debateable, however given that xen
default behavior is to reboot on dom0 fail (i.e. no screen log unless on
serial), a large number of users have been tripped up by this process on the
-xen kernels, and will find it hard to debug.  The approach of making the
-xen0 storage drivers also on in the -xen kernel just makes it a bit more
predicable.  (i.e. -xen == -xen0 + lots of other device support).  While it
isn''t that today, I think it would make more sense if it was that way. 
But
then again, that''s just my $0.02. :)

An alternate (or even in addition to any changes to the -xen kernel) would
be to default xen behavior to "noreboot", which would make it a lot
clearer
to users *why* their system doesn''t come up with Xen on the first try.

	-Sean

-- 
Sean Dague
IBM Linux Technology Center                     email: japh@us.ibm.com
Open Hypervisor Team                           alt: sldague@us.ibm.com


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

On 29 Mar 2006, at 20:28, Sean Dague wrote:
> An alternate (or even in addition to any changes to the -xen kernel) 
> would
> be to default xen behavior to "noreboot", which would make it a
lot
> clearer
> to users *why* their system doesn''t come up with Xen on the first
try.
What is native kernel behaviour when it fails to find its rootfs? Does 
it reboot?

  -- Keir


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

* Keir Fraser (Keir.Fraser@cl.cam.ac.uk) wrote:
> What is native kernel behaviour when it fails to find its rootfs? Does 
> it reboot?
Default, no, just panic.  If you add kernel cmdline param it will
reboot, ''panic=<timeout>''

thanks,
-chris

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

On Wed, Mar 29, 2006 at 12:02:48PM -0800, Chris Wright
wrote:
> * Keir Fraser (Keir.Fraser@cl.cam.ac.uk) wrote:
> > What is native kernel behaviour when it fails to find its rootfs? Does
> > it reboot?
> 
> Default, no, just panic.  If you add kernel cmdline param it will
> reboot, ''panic=<timeout>''
Right, but Xen reboots on a dom0 kernel panic by default, which ends up
completely masking why the dom0 kernel paniced, as your screen blips off
immediately (or at least it does for me). :)

Keeping the message on the CRT that dom0 paniced because it couldn''t
find
the root block device would dispense with a lot of questions. 

	-Sean

-- 
Sean Dague
IBM Linux Technology Center                     email: japh@us.ibm.com
Open Hypervisor Team                           alt: sldague@us.ibm.com


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