Re: [Xen-devel] linux distribution ?

[Dulloor <dulloor@xxxxxxxxx>]

I hava a sata drive and actually I had taken care of the driver part (afaik). I cross-verified /proc/modules in initramfs. 
When I tried restarting udevd (on initramfs shell), it warns about deprecated sysfs.

It does make sense for even me to downgrade back to 8.10. 
But, to figure this out, can you please send over your .config ?

-dulloor

On Thu, May 21, 2009 at 7:47 PM, Patrick Colp <pjcolp@xxxxxxxxx> wrote:

Yes, I was able to do it. I only down-graded my test box to 8.10 because
I didn't want to deal with python stuff (and to make sure that errors I
got when running my code was because of my code and not some sort of
weird incompatibility).

Sometimes the 2.6.18 driver gets confused with sata stuff. If both the
regular (pata) and sata drivers are installed, it will pick the wrong
one. I forget which way around you want to make it... I think you want
to have the sata ones only, but I'll check in the morning.

If you're not sure which driver you're needing, do an lspci and it'll
say which IDE system you have (or just remove all pata/ide drivers in
the kernel config and try again, but you will probably have to remove
/lib/modules/2.6.18.8-xen first, to make sure the old one doesn't stick
around by accident).


Patrick

Dulloor wrote:
>   Is anyone able to boot 2.6.18-xen on 9.04 ubuntu (pvops is fine).
>
> - mountroot in my initramfs fails, cause udev doesn't set up any block
> /dev/<...>.
> - All required modules seem to be loaded (checked both in scripts and at
> initramfs shell), but obviously something is missing in 2.6.18
>
> -dulloor
>
> On Thu, May 21, 2009 at 6:42 AM, Tim Post <echo@xxxxxxxxxxxx> wrote:
>
>> Sorry for top posting, another case in point:
>>
>> Depending on the distro, 'awk' might not work for configuring Linux. A
>> warning could be generated telling the user that installing 'gawk' might
>> be a good idea.
>>
>> Moreover, we warn about the absence of hg, git, bcc, etc.
>>
>> The idea is, if the 'check' script spews stderr to some file, that file
>> becomes a list of stuff that the user would (probably) want to install.
>>
>> Again, I am in NO way suggesting ./configure --enable-kernels
>> --enable-ocaml-stubdom , etc, nor am I suggesting that the script create
>> Makefiles. I could do that, I would NOT want to maintain it, neither
>> would anyone else on this list.
>>
>> Said script could also 'better advertise' other targets in the Makefile,
>> i.e. if Mercurial is not installed. At the least, as a 'general
>> failure', make help could be advertised.
>>
>> Cheers,
>> --Tim
>>
>> On Thu, 2009-05-21 at 18:27 +0800, Tim Post wrote:
>>> On Thu, 2009-05-21 at 05:53 -0400, Dulloor wrote:
>>>> Keir et al -
>>>>
>>>> I am on ubuntu and every time I upgrade my distro (dom-0), I end up
>>>> spending half-a-day getting xen working again, like this time on
>>>> moving to jaunty/karmic (problem booting 2.6.18 based xen and then
>>>> python version).
>>>>
>>>> Which distro do the xensource guys use for their development ? All I
>>>> am interested in is xen development/test environment.
>>>>
>>>> -dulloor
>>> You might consider just building Xen from source (tools and hypervisor),
>>> which takes it completely out of the scope of your package manager.
>>>
>>> I know that is taboo in some circles, however it gives you greater
>>> flexibility when upgrading, while also giving you the ability to test
>>> experimental patches.
>>>
>>> The problem is, doing this often violates enterprise warranties. 99.9%
>>> of the time, I'd rather just trust my distro when it comes to packages.
>>>
>>> When it comes to Xen, I usually recommend (and install) the latest
>>> faithful official release. The one and only time I just used distro
>>> packages was with Ubuntu Hardy (LTS) .. and that was chaotic (time going
>>> backwards, etc).
>>>
>>> There was once a universal installer script .. can that be resurrected
>>> and possibly rely on m4 being present for developers? Using that, the
>>> user knows with no uncertainty exactly what they are missing (and what
>>> version is needed).
>>>
>>> For instance, a dependency on 32 bit stubs when building on x86_64.
>>>
>>> It does not have to be named ./configure, it does not have to create
>>> makefiles and I am happy to maintain it. The drawback is 6k+ lines of
>>> generated shell code that has to be tracked in the hg.
>>>
>>> It could be ... scripts/checkbuildconfig .. or whatever. It would not be
>>> a configuration tool, just a diagnostic tool that offers hints on what
>>> is needed to build.
>>>
>>> Why clutter the Makefile needlessly? A script would be more portable,
>>> anyway.
>>>
>>> This approach has solved this exact problem for decades.
>>>
>>> Cheers,
>>> --Tim
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Xen-devel mailing list
>>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>>> http://lists.xensource.com/xen-devel
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>> http://lists.xensource.com/xen-devel
>>
>
>

> ------------------------------------------------------------------------

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

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