Xen devel - Dec 2012 - (updated) test report for xen-unstable tree with upstream QEMU

Hi All,
We did some testing for Xen-unstable tree with upstream QEMU.
We covered basic guest booting up, power management, VT-d, SR-IOV features.
We found 4 new bugs which only exist in upstream QEMU not in qemu-xen.

We followed the below wiki page to use upstream QEMU.
http://wiki.xen.org/wiki/QEMU_Upstream

test tree:
xen-unstable-tree.hg: C/S 26193 (about 20 days ago)
qemu.git: commit e9bff10f8db (about 20 days ago)
Dom0: Linux 3.6.9 release version.

test machine:
Intel Westmere-EP and SandyBridge-EP systems.

new bugs (which don''t exist with qemu-xen-unstable tree):
1. ''maxvcpus=NUM'' item is not supported in upstream QEMU
  http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1837
  -- This blocked vCPU hot-plug for HVM guest.
2. Guest console hangs after save/restore or live-migration when setting
''hpet=0'' in guest config file
  http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1838
3. ''xen_platform_pci=0'' setting cannot make the guest use
emulated PCI devices by default
  http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1839
4. Guest free memory with upstream qemu is 14MB lower than that with
qemu-xen-unstable.git
  http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1836
  -- This might not be a bug; we just curious to know what the additional 14MB
memory is for.

On Mon, 2012-12-24 at 02:24 +0000, Ren, Yongjie wrote:
> Hi All,
> We did some testing for Xen-unstable tree with upstream QEMU.
Thanks.

Out of interest do you also test (or plan to test) the qemu-upstream
stable branch which is built by default when you build Xen?
> We covered basic guest booting up, power management, VT-d, SR-IOV features.
> We found 4 new bugs which only exist in upstream QEMU not in qemu-xen.
> 
> We followed the below wiki page to use upstream QEMU.
> http://wiki.xen.org/wiki/QEMU_Upstream
> 
> test tree:
> xen-unstable-tree.hg: C/S 26193 (about 20 days ago)
> qemu.git: commit e9bff10f8db (about 20 days ago)
> Dom0: Linux 3.6.9 release version.
> 
> test machine:
> Intel Westmere-EP and SandyBridge-EP systems.
> 
> new bugs (which don''t exist with qemu-xen-unstable tree):
> 1. ''maxvcpus=NUM'' item is not supported in upstream QEMU
>   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1837
>   -- This blocked vCPU hot-plug for HVM guest.
> 2. Guest console hangs after save/restore or live-migration when setting
''hpet=0'' in guest config file
>   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1838
> 3. ''xen_platform_pci=0'' setting cannot make the guest use
emulated PCI devices by default
>   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1839
> 4. Guest free memory with upstream qemu is 14MB lower than that with
qemu-xen-unstable.git
>   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1836
>   -- This might not be a bug; we just curious to know what the additional
14MB memory is for.
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

> -----Original Message-----
> From: Ian Campbell [mailto:Ian.Campbell@citrix.com]
> Sent: Thursday, January 03, 2013 12:14 AM
> To: Ren, Yongjie
> Cc: xen-devel@lists.xen.org; Dai, Yan; Xu, YongweiX; Liu, RongrongX; Liu,
> SongtaoX; Zhou, Chao; Stefano Stabellini; Anthony Perard
> Subject: Re: [Xen-devel] (updated) test report for xen-unstable tree with
> upstream QEMU
> 
> On Mon, 2012-12-24 at 02:24 +0000, Ren, Yongjie wrote:
> > Hi All,
> > We did some testing for Xen-unstable tree with upstream QEMU.
> 
> Thanks.
> 
> Out of interest do you also test (or plan to test) the qemu-upstream
> stable branch which is built by default when you build Xen?
> 
No, so far we still use qemu-xen-unstable tree when building Xen in our regular
testing.
We only have enough resource to cover one qemu tree for Xen.
Which qemu do you suggest in our Xen upstream testing? qemu-xen-unstable or pure
qemu upstream ?
BTW, do we have some schedule for officially using qemu upstream tree in Xen by
default in near future (e.g. Xen 4.3)?

> > We covered basic guest booting up, power management, VT-d, SR-IOV
> features.
> > We found 4 new bugs which only exist in upstream QEMU not in
> qemu-xen.
> >
> > We followed the below wiki page to use upstream QEMU.
> > http://wiki.xen.org/wiki/QEMU_Upstream
> >
> > test tree:
> > xen-unstable-tree.hg: C/S 26193 (about 20 days ago)
> > qemu.git: commit e9bff10f8db (about 20 days ago)
> > Dom0: Linux 3.6.9 release version.
> >
> > test machine:
> > Intel Westmere-EP and SandyBridge-EP systems.
> >
> > new bugs (which don''t exist with qemu-xen-unstable tree):
> > 1. ''maxvcpus=NUM'' item is not supported in upstream
QEMU
> >   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1837
> >   -- This blocked vCPU hot-plug for HVM guest.
> > 2. Guest console hangs after save/restore or live-migration when
setting
> ''hpet=0'' in guest config file
> >   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1838
> > 3. ''xen_platform_pci=0'' setting cannot make the
guest use emulated PCI
> devices by default
> >   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1839
> > 4. Guest free memory with upstream qemu is 14MB lower than that with
> qemu-xen-unstable.git
> >   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1836
> >   -- This might not be a bug; we just curious to know what the
additional
> 14MB memory is for.
> >
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xen.org
> > http://lists.xen.org/xen-devel
>

On Fri, 4 Jan 2013, Ren, Yongjie wrote:
> > -----Original Message-----
> > From: Ian Campbell [mailto:Ian.Campbell@citrix.com]
> > Sent: Thursday, January 03, 2013 12:14 AM
> > To: Ren, Yongjie
> > Cc: xen-devel@lists.xen.org; Dai, Yan; Xu, YongweiX; Liu, RongrongX;
Liu,
> > SongtaoX; Zhou, Chao; Stefano Stabellini; Anthony Perard
> > Subject: Re: [Xen-devel] (updated) test report for xen-unstable tree
with
> > upstream QEMU
> > 
> > On Mon, 2012-12-24 at 02:24 +0000, Ren, Yongjie wrote:
> > > Hi All,
> > > We did some testing for Xen-unstable tree with upstream QEMU.
> > 
> > Thanks.
> > 
> > Out of interest do you also test (or plan to test) the qemu-upstream
> > stable branch which is built by default when you build Xen?
> > 
> No, so far we still use qemu-xen-unstable tree when building Xen in our
regular testing.
> We only have enough resource to cover one qemu tree for Xen.
> Which qemu do you suggest in our Xen upstream testing? qemu-xen-unstable or
pure qemu upstream ?
qemu-xen-unstable deviates very little from QEMU upstream nowadays,
however if you test QEMU upstream you risk finding transient QEMU bugs.
I am not sure that you want that, it would be like using the very latest
Linus'' tree as dom0 kernel.
If you want to catch Xen bugs and harden Xen releases it is probably
better to use qemu-xen-unstable.