Re: [XenPPC] Re: [Xen-devel] [PATCH 0/6][TOOLS][XM-TEST] Update xm-test to support new architectures

[Jimi Xenidis <jimix@xxxxxxxxxxxxxx>]

On Oct 1, 2006, at 2:30 AM, Tony Breeds wrote:

> On Fri, Sep 29, 2006 at 08:56:54PM +1100, Tony Breeds wrote:
> > Hi All,
> >         These patches update the xm-test code to be more easily portable
> > to new architecture. This focus od this endevour is PPC but I believe
> > that IA64 also benifits.
> <snip>
>
> By way of a quick status report.
>
> With all thee patches applied to tip (7de08aceff6d) I ran:
>
> I build xm-test:
> cd xen.hg/tools/xm-test
> ./autogen
> ./configure
> INITRD="http://ozlabs.au.ibm.com/~tony/xm-test/"; make existing
>
> I created a grouptest/xenppc file which contains most[1] of the
> potential tests and pushed the whole xm-test dir (yes it's wasteful  
> but
> it's also easy) to my JS20 blade.  Then:
>         ./runtest.sh -d -e tony@xxxxxxxxxxxxxxxxxx -g xenppc xenppc
>
> creates the following report.
> ---
> Xm-test execution summary:
>   PASS:  48
>   FAIL:  17
>   XPASS: 0
>   XFAIL: 0
Tony, this is so great!

> Details:
>
>  FAIL: 01_block_attach_device_pos
>          block-attach failed device did not switch to Connected state
[snip]
>  FAIL: 01_memset_basic_pos
>          Invalid domU meminfo line
>
>  FAIL: 03_memset_random_pos
>          Invalid domU meminfo line
>
> ---
>
> You'd certainly expect the block-* tests to fail.  I don't know why  
> the
> memset tests failed but I'll find out next week when I'm back at work.
xm mem-set does _not_ work because we have yet to implement  
"ballooning".
We have not implemented ballooning YET, for a number reasons, off the  
top of my head:
 - Ballooning works by having a Linux driver allocate memory and  
have Xen release the packing pages
 - We cannot ever balloon the RMA, and Linux does not know how to  
allocate from non-RMA pages
There is also the reason of Domain memory allocations where we  
allocate memory in 16M sizes/aligned extents because Linux tryies to  
use 16M pages for the kernel linear mapping.
> [1] I omitted create/destroy, save/restore, network* and the scheduler
>     tests becuase I know some of them crash the machine, attached  
> is the
>     xenppc file I used.
This is important, nothing should crash, only succeed or fail, these  
need to be attended to.
I'm gonna start using this as my regression test, if anyone else does  
it would be great is we could file bugs and keep:
  http://wiki.xensource.com/xenwiki/XenPPC/XMChecklist

up to date.

-JX


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