xm-test thoughts

xm-test thoughts

[Eric Schwartz]

I have created (by hand; the normal build process does not work on ia64) 
a ramdisk for use with xm-test on ia64.  What's the best way to get this 
included on http://xm-test.xensource.com/ramdisks/ so that others might 
be able to use it?  It will require the arbitrary ramdisk size patch I 
previously posted, as it's 8MB, but other than that, I've been able to 
run xm-test successfully with it.

-=Eric

RE: xm-test thoughts

[Ian Pratt]

> I have created (by hand; the normal build process does not work on
> ia64)
> a ramdisk for use with xm-test on ia64.  What's the best way to get
> this
> included on http://xm-test.xensource.com/ramdisks/ so that others
might
> be able to use it?  It will require the arbitrary ramdisk size patch I
> previously posted, as it's 8MB, but other than that, I've been able to
> run xm-test successfully with it.

Post a URL I can pull it from and I'll make sure it happens.

Ian

Re: xm-test thoughts

[Eric Schwartz]

Ian Pratt wrote:
> Post a URL I can pull it from and I'll make sure it happens.
There's a (hopefully quick) legal sign-off that needs to happen first; 
as soon as I get clearance, I'll let you know. This brings in a topic 
I'd intended to include in the original email, and decided not to on 
grounds of clarity: how xm-test images are created.

Right now, the process involves buildroot and uclibc, which works fine 
for ppc and x86, but is fairly broken for ia64, and is in any event a 
project designed for embedded systems. I assume uclibc was chosen 
because it's relatively small compared to glibc, but with the patch in 
my earlier email, any size ramdisk can be used. Since glibc is going to 
be available, almost by definition, on any system on which Xen will be 
built, would there be any interest in using that instead of buildroot 
and uclibc?

Honestly, buildroot is kinda overkill for the purpose; it's trivial to 
compile busybox and hping2 from scratch, and installing them into a 
chroot shouldn't be particularly difficult. The only tricky thing is 
glibc, since it is reasonably heavily-patched by most distributions, but 
I'd argue even that isn't particularly hard; pick one distro's patches 
(any distro would do; I used the Debian libc, but any distro's libc 
should be fine) and go with that.

Another possibility, depending on how people feel about this, is to 
simply require the binaries to exist on the system the ramdisk is built 
from, and use a combination of ldd and python (or perl, or whatever, I 
don't care) to copy all the required libraries into the image. This is 
the method I used to create the image; I just didn't have an automated 
process around it.

Is there any interest in either of these alternatives? Do most people 
using xm-test just download a pre-generated image, or would efforts to 
make images easier and faster to build be worth it? If nobody cares, I 
don't want to waste my time, but I think making this process easier 
would certainly help with Xen's portability.

-=Eric

Re: xm-test thoughts

[Tony Breeds]

On Thu, May 03, 2007 at 01:18:52PM -0600, Eric Schwartz wrote:
> I have created (by hand; the normal build process does not work on ia64) 
> a ramdisk for use with xm-test on ia64.  What's the best way to get this 
> included on http://xm-test.xensource.com/ramdisks/ so that others might 
> be able to use it?  It will require the arbitrary ramdisk size patch I 
> previously posted, as it's 8MB, but other than that, I've been able to 
> run xm-test successfully with it.

Asside from this patch, Where and how does the build system fail for
ia64?  It wasn't too hard to bolt in powerpc support.  I thought I but
all the arch hooks in place to allow it to work most places.

Yours Tony

  linux.conf.au        http://linux.conf.au/ || http://lca2008.linux.org.au/
  Jan 28 - Feb 02 2008 The Australian Linux Technical Conference!

Re: xm-test thoughts

[Eric Schwartz]

Tony Breeds wrote:
> On Thu, May 03, 2007 at 01:18:52PM -0600, Eric Schwartz wrote:
>   
>> I have created (by hand; the normal build process does not work on ia64) 
>> a ramdisk for use with xm-test on ia64.  What's the best way to get this 
>> included on http://xm-test.xensource.com/ramdisks/ so that others might 
>> be able to use it?  It will require the arbitrary ramdisk size patch I 
>> previously posted, as it's 8MB, but other than that, I've been able to 
>> run xm-test successfully with it.
>>     
>
> Asside from this patch, Where and how does the build system fail for
> ia64?  It wasn't too hard to bolt in powerpc support.  I thought I but
> all the arch hooks in place to allow it to work most places.
>   
uClibc is pretty badly broken for ia64; there's no ldso support, though 
a co-worker, Al Stone, has been trying to patch that in. buildroot 
requires uClibc (in fact, I read someone describe buildroot as a testbed 
for uClibc), so that's a pretty fundamental failure.

-=ERic