Re: [Xen-users] Issues with vcpu-set

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
> Sorry for the so late reply.
> I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50

And are you using the version of qemu-xen which ships with those
releases or your own version, perhaps from upstream?

ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
4.3.x but I thought it had been added during the 4.4.x development
cycle. Adding Anthony + xen-devel to confirm.

> Here is the /var/log/xen/
> Waiting for domain centos65.pv (domid 1) to die [pid 8116]
> 
> 
> Here is the output of "xl -vvv vcpu-set"

Is this from 4.3 or 4.4? I think at this point we should focus on the
issue with 4.4.

I also asked for your guest cfg file -- please can you show it to us.

> libxl: debug: libxl_qmp.c:707:libxl__qmp_initialize: connected to
> /var/run/xen/qmp-libxl-1
> 
> libxl: debug: libxl_qmp.c:299:qmp_handle_response: message type: qmp
> 
> libxl: debug: libxl_qmp.c:555:qmp_send_prepare: next qmp command: '{
> 
>     "execute": "qmp_capabilities",
> 
>     "id": 1
> 
> }
> 
> '
> 
> libxl: debug: libxl_qmp.c:299:qmp_handle_response: message type: return
> 
> libxl: debug: libxl_qmp.c:555:qmp_send_prepare: next qmp command: '{
> 
>     "execute": "cpu-add",
> 
>     "id": 2,
> 
>     "arguments": {
> 
>         "id": 0
> 
>     }
> 
> }
> 
> '
> 
> libxl: debug: libxl_qmp.c:299:qmp_handle_response: message type: error
> 
> libxl: error: libxl_qmp.c:289:qmp_handle_error_response: received an
> error message from QMP server: Not supported
> 
> xc: debug: hypercall buffer: total allocations:9 total releases:9
> 
> xc: debug: hypercall buffer: current allocations:0 maximum allocations:2
> 
> xc: debug: hypercall buffer: cache current size:2
> 
> xc: debug: hypercall buffer: cache hits:7 misses:2 toobig:0
> 
> On Tue, Oct 1, 2013 at 3:24 AM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> > On Mon, 2013-09-30 at 10:52 -0600, Yun Wang wrote:
> >> Hi all,
> >>
> >> I tried to use "xl vcpu-set" to change the vCPU number of VMs in Xen
> >> 4.4-unstable and had the following errors.
> >
> > What was the full command line which you used?
> >
> > Which exact version of 4.4-unstable (i.e. git commit) were you using?
> >
> >> libxl: error: libxl_qmp.c:289:qmp_handle_error_response: received an
> >> error message from QMP server: Unable to add CPU: 0, it already
> >> exists.
> >>
> >> CPU hotplug is enabled in the guest OS.
> >
> > Please can you provide your guest cfg file and any relevant logs from
> > under /var/log/xen (especially the ones with the guest name in them).
> >
> > Please also try "xl -vvv vcpu-set ..." and provide the full logs of that
> > attempt.
> >
> > Ian.
> >

Re: [Xen-users] Issues with vcpu-set

[Anthony PERARD]

On Wed, Jan 29, 2014 at 10:11:44AM +0000, Ian Campbell wrote:
> On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
> > Sorry for the so late reply.
> > I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50
> 
> And are you using the version of qemu-xen which ships with those
> releases or your own version, perhaps from upstream?
> 
> ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
> 4.3.x but I thought it had been added during the 4.4.x development
> cycle. Adding Anthony + xen-devel to confirm.

We've added vcpu hotplug to our tree in Xen 4.3.

QEMU upstream is able to do vcpu hotplug with Xen only with the last
release, 1.7. The two previous release (1.5 and 1.6) miss two patches,
and any QEMU release before that those not support cpu hotplug.

> > Here is the /var/log/xen/
> > Waiting for domain centos65.pv (domid 1) to die [pid 8116]
> > 
> > 
> > Here is the output of "xl -vvv vcpu-set"
> 
> Is this from 4.3 or 4.4? I think at this point we should focus on the
> issue with 4.4.
> 
> I also asked for your guest cfg file -- please can you show it to us.
> 

[...]

> > libxl: debug: libxl_qmp.c:555:qmp_send_prepare: next qmp command: '{
> >     "execute": "cpu-add",
> >     "id": 2,
> >     "arguments": {
> >         "id": 0
> >     }
> > }
> > '
> > libxl: debug: libxl_qmp.c:299:qmp_handle_response: message type: error
> > libxl: error: libxl_qmp.c:289:qmp_handle_error_response: received an
> > error message from QMP server: Not supported

This right here means that xl support vcpu-set with qemu-xen, but the
QEMU used does not.

> > On Tue, Oct 1, 2013 at 3:24 AM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> > > On Mon, 2013-09-30 at 10:52 -0600, Yun Wang wrote:
> > >> Hi all,
> > >>
> > >> I tried to use "xl vcpu-set" to change the vCPU number of VMs in Xen
> > >> 4.4-unstable and had the following errors.
> > >
> > > What was the full command line which you used?
> > >
> > > Which exact version of 4.4-unstable (i.e. git commit) were you using?
> > >
> > >> libxl: error: libxl_qmp.c:289:qmp_handle_error_response: received an
> > >> error message from QMP server: Unable to add CPU: 0, it already
> > >> exists.

This is just a message that I don't know how to remove.

Since libxl does not know which cpu are already plugged-in, it asks QEMU
to plug every asked cpu. QEMU just reply back as an error that some CPU
are already plugged. And libxl does print every error that QEMU send.

-- 
Anthony PERARD

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Wed, 2014-01-29 at 12:18 +0000, Anthony PERARD wrote:
> On Wed, Jan 29, 2014 at 10:11:44AM +0000, Ian Campbell wrote:
> > On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
> > > Sorry for the so late reply.
> > > I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50
> > 
> > And are you using the version of qemu-xen which ships with those
> > releases or your own version, perhaps from upstream?
> > 
> > ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
> > 4.3.x but I thought it had been added during the 4.4.x development
> > cycle. Adding Anthony + xen-devel to confirm.
> 
> We've added vcpu hotplug to our tree in Xen 4.3.
> 
> QEMU upstream is able to do vcpu hotplug with Xen only with the last
> release, 1.7. The two previous release (1.5 and 1.6) miss two patches,
> and any QEMU release before that those not support cpu hotplug.

OK. IIRC Xen 4.3 included qemu 1.3 or so?

4.4-rc1 should have had the 1.6 merge in it, but not 1.7, correct? So it
also misses those patches.

Are we planning to backport those two patches for 4.4?

Ian.

Re: [Xen-users] Issues with vcpu-set

[Anthony PERARD]

On Wed, Jan 29, 2014 at 12:22:17PM +0000, Ian Campbell wrote:
> On Wed, 2014-01-29 at 12:18 +0000, Anthony PERARD wrote:
> > On Wed, Jan 29, 2014 at 10:11:44AM +0000, Ian Campbell wrote:
> > > On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
> > > > Sorry for the so late reply.
> > > > I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50
> > > 
> > > And are you using the version of qemu-xen which ships with those
> > > releases or your own version, perhaps from upstream?
> > > 
> > > ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
> > > 4.3.x but I thought it had been added during the 4.4.x development
> > > cycle. Adding Anthony + xen-devel to confirm.
> > 
> > We've added vcpu hotplug to our tree in Xen 4.3.
> > 
> > QEMU upstream is able to do vcpu hotplug with Xen only with the last
> > release, 1.7. The two previous release (1.5 and 1.6) miss two patches,
> > and any QEMU release before that those not support cpu hotplug.
> 
> OK. IIRC Xen 4.3 included qemu 1.3 or so?

That's right, qemu 1.3.

> 4.4-rc1 should have had the 1.6 merge in it, but not 1.7, correct?

That's correct.

> So it also misses those patches.

No, I've put those patches on top of the merge, otherwise, there would
have been regression.

> Are we planning to backport those two patches for 4.4?

No :), it's not needed.

-- 
Anthony PERARD

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Wed, 2014-01-29 at 12:36 +0000, Anthony PERARD wrote:
> On Wed, Jan 29, 2014 at 12:22:17PM +0000, Ian Campbell wrote:
> > On Wed, 2014-01-29 at 12:18 +0000, Anthony PERARD wrote:
> > > On Wed, Jan 29, 2014 at 10:11:44AM +0000, Ian Campbell wrote:
> > > > On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
> > > > > Sorry for the so late reply.
> > > > > I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50
> > > > 
> > > > And are you using the version of qemu-xen which ships with those
> > > > releases or your own version, perhaps from upstream?
> > > > 
> > > > ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
> > > > 4.3.x but I thought it had been added during the 4.4.x development
> > > > cycle. Adding Anthony + xen-devel to confirm.
> > > 
> > > We've added vcpu hotplug to our tree in Xen 4.3.
> > > 
> > > QEMU upstream is able to do vcpu hotplug with Xen only with the last
> > > release, 1.7. The two previous release (1.5 and 1.6) miss two patches,
> > > and any QEMU release before that those not support cpu hotplug.
> > 
> > OK. IIRC Xen 4.3 included qemu 1.3 or so?
> 
> That's right, qemu 1.3.
> 
> > 4.4-rc1 should have had the 1.6 merge in it, but not 1.7, correct?
> 
> That's correct.
> 
> > So it also misses those patches.
> 
> No, I've put those patches on top of the merge, otherwise, there would
> have been regression.

So this functionality should work with xen 4.4.0-rc1 (using whatever
version was referenced by xen.git/Config.mk) or not?

Ian.

Re: [Xen-users] Issues with vcpu-set

[Anthony PERARD]

On Wed, Jan 29, 2014 at 12:39:47PM +0000, Ian Campbell wrote:
> On Wed, 2014-01-29 at 12:36 +0000, Anthony PERARD wrote:
> > On Wed, Jan 29, 2014 at 12:22:17PM +0000, Ian Campbell wrote:
> > > On Wed, 2014-01-29 at 12:18 +0000, Anthony PERARD wrote:
> > > > On Wed, Jan 29, 2014 at 10:11:44AM +0000, Ian Campbell wrote:
> > > > > On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
> > > > > > Sorry for the so late reply.
> > > > > > I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50
> > > > > 
> > > > > And are you using the version of qemu-xen which ships with those
> > > > > releases or your own version, perhaps from upstream?
> > > > > 
> > > > > ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
> > > > > 4.3.x but I thought it had been added during the 4.4.x development
> > > > > cycle. Adding Anthony + xen-devel to confirm.
> > > > 
> > > > We've added vcpu hotplug to our tree in Xen 4.3.
> > > > 
> > > > QEMU upstream is able to do vcpu hotplug with Xen only with the last
> > > > release, 1.7. The two previous release (1.5 and 1.6) miss two patches,
> > > > and any QEMU release before that those not support cpu hotplug.
> > > 
> > > OK. IIRC Xen 4.3 included qemu 1.3 or so?
> > 
> > That's right, qemu 1.3.
> > 
> > > 4.4-rc1 should have had the 1.6 merge in it, but not 1.7, correct?
> > 
> > That's correct.
> > 
> > > So it also misses those patches.
> > 
> > No, I've put those patches on top of the merge, otherwise, there would
> > have been regression.
> 
> So this functionality should work with xen 4.4.0-rc1 (using whatever
> version was referenced by xen.git/Config.mk) or not?

Yes, it should. You can go back to my first reply to this threads and
read the all mail ;-). I reply to what appear to me the "error" that
made someone think that somethings goes wrong.

-- 
Anthony PERARD

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Wed, 2014-01-29 at 13:04 +0000, Anthony PERARD wrote:
> On Wed, Jan 29, 2014 at 12:39:47PM +0000, Ian Campbell wrote:
> > On Wed, 2014-01-29 at 12:36 +0000, Anthony PERARD wrote:
> > > On Wed, Jan 29, 2014 at 12:22:17PM +0000, Ian Campbell wrote:
> > > > On Wed, 2014-01-29 at 12:18 +0000, Anthony PERARD wrote:
> > > > > On Wed, Jan 29, 2014 at 10:11:44AM +0000, Ian Campbell wrote:
> > > > > > On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
> > > > > > > Sorry for the so late reply.
> > > > > > > I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50
> > > > > > 
> > > > > > And are you using the version of qemu-xen which ships with those
> > > > > > releases or your own version, perhaps from upstream?
> > > > > > 
> > > > > > ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
> > > > > > 4.3.x but I thought it had been added during the 4.4.x development
> > > > > > cycle. Adding Anthony + xen-devel to confirm.
> > > > > 
> > > > > We've added vcpu hotplug to our tree in Xen 4.3.
> > > > > 
> > > > > QEMU upstream is able to do vcpu hotplug with Xen only with the last
> > > > > release, 1.7. The two previous release (1.5 and 1.6) miss two patches,
> > > > > and any QEMU release before that those not support cpu hotplug.
> > > > 
> > > > OK. IIRC Xen 4.3 included qemu 1.3 or so?
> > > 
> > > That's right, qemu 1.3.
> > > 
> > > > 4.4-rc1 should have had the 1.6 merge in it, but not 1.7, correct?
> > > 
> > > That's correct.
> > > 
> > > > So it also misses those patches.
> > > 
> > > No, I've put those patches on top of the merge, otherwise, there would
> > > have been regression.
> > 
> > So this functionality should work with xen 4.4.0-rc1 (using whatever
> > version was referenced by xen.git/Config.mk) or not?
> 
> Yes, it should. You can go back to my first reply to this threads and
> read the all mail ;-).

Sorry, but this wasn't unambiguously clear to me from your initial
reply, which spoke about broken releases and a fix which was in a
version we don't include yet.

>  I reply to what appear to me the "error" that
> made someone think that somethings goes wrong.

So it sounds like Yun Wang is using an upstream qemu and not our branch
and is therefore missing some fixes?

Ian.

Re: [Xen-users] Issues with vcpu-set

[Yun Wang]

So to fix the problem, I need to update the qemu version to version
1.7 or later?
BTW. I had this problem in both pvhvm and pv guest.
Does pv guest rely on qemu also?

Thanks for all the reply!

On Wed, Jan 29, 2014 at 6:14 AM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> On Wed, 2014-01-29 at 13:04 +0000, Anthony PERARD wrote:
>> On Wed, Jan 29, 2014 at 12:39:47PM +0000, Ian Campbell wrote:
>> > On Wed, 2014-01-29 at 12:36 +0000, Anthony PERARD wrote:
>> > > On Wed, Jan 29, 2014 at 12:22:17PM +0000, Ian Campbell wrote:
>> > > > On Wed, 2014-01-29 at 12:18 +0000, Anthony PERARD wrote:
>> > > > > On Wed, Jan 29, 2014 at 10:11:44AM +0000, Ian Campbell wrote:
>> > > > > > On Tue, 2014-01-28 at 18:44 -0700, Yun Wang wrote:
>> > > > > > > Sorry for the so late reply.
>> > > > > > > I had this issue in Xen-4.3.0 (official release) and Xen-4.4.0-rc1-25-g9a80d50
>> > > > > >
>> > > > > > And are you using the version of qemu-xen which ships with those
>> > > > > > releases or your own version, perhaps from upstream?
>> > > > > >
>> > > > > > ISTR that vcpu hotplug for HVM guests was missing from qemu-xen in Xen
>> > > > > > 4.3.x but I thought it had been added during the 4.4.x development
>> > > > > > cycle. Adding Anthony + xen-devel to confirm.
>> > > > >
>> > > > > We've added vcpu hotplug to our tree in Xen 4.3.
>> > > > >
>> > > > > QEMU upstream is able to do vcpu hotplug with Xen only with the last
>> > > > > release, 1.7. The two previous release (1.5 and 1.6) miss two patches,
>> > > > > and any QEMU release before that those not support cpu hotplug.
>> > > >
>> > > > OK. IIRC Xen 4.3 included qemu 1.3 or so?
>> > >
>> > > That's right, qemu 1.3.
>> > >
>> > > > 4.4-rc1 should have had the 1.6 merge in it, but not 1.7, correct?
>> > >
>> > > That's correct.
>> > >
>> > > > So it also misses those patches.
>> > >
>> > > No, I've put those patches on top of the merge, otherwise, there would
>> > > have been regression.
>> >
>> > So this functionality should work with xen 4.4.0-rc1 (using whatever
>> > version was referenced by xen.git/Config.mk) or not?
>>
>> Yes, it should. You can go back to my first reply to this threads and
>> read the all mail ;-).
>
> Sorry, but this wasn't unambiguously clear to me from your initial
> reply, which spoke about broken releases and a fix which was in a
> version we don't include yet.
>
>>  I reply to what appear to me the "error" that
>> made someone think that somethings goes wrong.
>
> So it sounds like Yun Wang is using an upstream qemu and not our branch
> and is therefore missing some fixes?
>
> Ian.
>

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Wed, 2014-01-29 at 07:38 -0700, Yun Wang wrote:
> So to fix the problem, I need to update the qemu version to version
> 1.7 or later?

Yes. Or, AIUI, you can use the version of Qemu which is bundled with the
Xen releases.

> BTW. I had this problem in both pvhvm and pv guest.
> Does pv guest rely on qemu also?

It does not.

If you are having an issue with a PV guest vcpu hotplug then it is not
the same underlying issue. Please report it separately with full logs,
config info etc.

Ian.

Re: [Xen-users] Issues with vcpu-set

[Yun Wang]

Here is the PV guest config
#########################
name = "centos65.pv"
bootloader = "/usr/local/bin/pygrub"
extra = "root=/dev/xvda"
memory = 4096
vcpus = 2
vfb=[ "type=vnc, vncpass=123456" ]
vif = [ 'mac=00:16:3e:54:02:01, bridge=xenbr0' ]
disk = [ 'file:/vms/centos65_pv.img, xvda, w']
######################################


The output of "xl -vvv vcpu-set"
########################

libxl: debug: libxl_qmp.c:696:libxl__qmp_initialize: connected to
/var/run/xen/qmp-libxl-38

libxl: debug: libxl_qmp.c:296:qmp_handle_response: message type: qmp

libxl: debug: libxl_qmp.c:546:qmp_send_prepare: next qmp command: '{

    "execute": "qmp_capabilities",

    "id": 1

}

'

libxl: debug: libxl_qmp.c:296:qmp_handle_response: message type: return

libxl: debug: libxl_qmp.c:546:qmp_send_prepare: next qmp command: '{

    "execute": "cpu-add",

    "id": 2,

    "arguments": {

        "id": 0

    }

}

'

libxl: debug: libxl_qmp.c:296:qmp_handle_response: message type: error

libxl: error: libxl_qmp.c:287:qmp_handle_error_response: received an
error message from QMP server: Not supported

xc: debug: hypercall buffer: total allocations:9 total releases:9

xc: debug: hypercall buffer: current allocations:0 maximum allocations:2

xc: debug: hypercall buffer: cache current size:2

xc: debug: hypercall buffer: cache hits:7 misses:2 toobig:0
######################################
Again, this issue exists both on Xen-4.3.0 (official release) and
Xen-4.4.0-rc1-25-g9a80d50

On Wed, Jan 29, 2014 at 7:45 AM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> On Wed, 2014-01-29 at 07:38 -0700, Yun Wang wrote:
>> So to fix the problem, I need to update the qemu version to version
>> 1.7 or later?
>
> Yes. Or, AIUI, you can use the version of Qemu which is bundled with the
> Xen releases.
>
>> BTW. I had this problem in both pvhvm and pv guest.
>> Does pv guest rely on qemu also?
>
> It does not.
>
> If you are having an issue with a PV guest vcpu hotplug then it is not
> the same underlying issue. Please report it separately with full logs,
> config info etc.
>
> Ian.
>

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Wed, 2014-01-29 at 09:18 -0700, Yun Wang wrote:
> Here is the PV guest config
> #########################
> name = "centos65.pv"
> bootloader = "/usr/local/bin/pygrub"
> extra = "root=/dev/xvda"
> memory = 4096
> vcpus = 2
> vfb=[ "type=vnc, vncpass=123456" ]
> vif = [ 'mac=00:16:3e:54:02:01, bridge=xenbr0' ]
> disk = [ 'file:/vms/centos65_pv.img, xvda, w']

This will most likely use a qdisk backend, so this guest will have a
qemu instance running even though it is PV.

> ######################################
> 
> 
> The output of "xl -vvv vcpu-set"
> ########################
> 
> libxl: debug: libxl_qmp.c:696:libxl__qmp_initialize: connected to
> /var/run/xen/qmp-libxl-38

But we certainly shouldn't be doing this for such a guest, regardless of
whether a qemu is running for it.

Anthony -- looks like you forgot about the PV case in 62fe11fb --
libxl_set_vcpuonline shoulod use the xenstored version for such guests.
Can you send a fix please.

Ian.

> libxl: debug: libxl_qmp.c:296:qmp_handle_response: message type: qmp
> 
> libxl: debug: libxl_qmp.c:546:qmp_send_prepare: next qmp command: '{
> 
>     "execute": "qmp_capabilities",
> 
>     "id": 1
> 
> }
> 
> '
> 
> libxl: debug: libxl_qmp.c:296:qmp_handle_response: message type: return
> 
> libxl: debug: libxl_qmp.c:546:qmp_send_prepare: next qmp command: '{
> 
>     "execute": "cpu-add",
> 
>     "id": 2,
> 
>     "arguments": {
> 
>         "id": 0
> 
>     }
> 
> }
> 
> '
> 
> libxl: debug: libxl_qmp.c:296:qmp_handle_response: message type: error
> 
> libxl: error: libxl_qmp.c:287:qmp_handle_error_response: received an
> error message from QMP server: Not supported
> 
> xc: debug: hypercall buffer: total allocations:9 total releases:9
> 
> xc: debug: hypercall buffer: current allocations:0 maximum allocations:2
> 
> xc: debug: hypercall buffer: cache current size:2
> 
> xc: debug: hypercall buffer: cache hits:7 misses:2 toobig:0
> ######################################
> Again, this issue exists both on Xen-4.3.0 (official release) and
> Xen-4.4.0-rc1-25-g9a80d50
> 
> On Wed, Jan 29, 2014 at 7:45 AM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> > On Wed, 2014-01-29 at 07:38 -0700, Yun Wang wrote:
> >> So to fix the problem, I need to update the qemu version to version
> >> 1.7 or later?
> >
> > Yes. Or, AIUI, you can use the version of Qemu which is bundled with the
> > Xen releases.
> >
> >> BTW. I had this problem in both pvhvm and pv guest.
> >> Does pv guest rely on qemu also?
> >
> > It does not.
> >
> > If you are having an issue with a PV guest vcpu hotplug then it is not
> > the same underlying issue. Please report it separately with full logs,
> > config info etc.
> >
> > Ian.
> >

Re: [Xen-users] Issues with vcpu-set

[Yun Wang]

Ian,

What disk option should I use for a PV guest? tap2:aio seems to have a
similar issue with "xl vcpu-set"
Thanks,
Yun.

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Thu, 2014-01-30 at 09:04 -0700, Yun Wang wrote:
> What disk option should I use for a PV guest? tap2:aio seems to have a
> similar issue with "xl vcpu-set"

To workaround the issue you would need phy: I think, perhaps by setting
up a loopback device on the raw image by hand.

Hopefully the underlying issue can get fixed though.

Ian.

Re: [Xen-users] Issues with vcpu-set

[Yun Wang]

See if I am doing it correctly.

1. setup a loopback device
losetup /dev/loop0 /vms/centos65_pv.img

2. check the output of "losetup -a"
/dev/loop0: [fb00]:926576 (/vms/centos65_pv.img)

3. use the loopback device in the PV config
disk = [ 'phy:/dev/loop0, xvda, w']

However, similar error still exists.

On Thu, Jan 30, 2014 at 9:12 AM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> On Thu, 2014-01-30 at 09:04 -0700, Yun Wang wrote:
>> What disk option should I use for a PV guest? tap2:aio seems to have a
>> similar issue with "xl vcpu-set"
>
> To workaround the issue you would need phy: I think, perhaps by setting
> up a loopback device on the raw image by hand.
>
> Hopefully the underlying issue can get fixed though.
>
> Ian.
>

Re: [Xen-users] Issues with vcpu-set

[Ian Campbell]

On Thu, 2014-01-30 at 09:35 -0700, Yun Wang wrote:
> See if I am doing it correctly.
> 
> 1. setup a loopback device
> losetup /dev/loop0 /vms/centos65_pv.img
> 
> 2. check the output of "losetup -a"
> /dev/loop0: [fb00]:926576 (/vms/centos65_pv.img)
> 
> 3. use the loopback device in the PV config
> disk = [ 'phy:/dev/loop0, xvda, w']
> 
> However, similar error still exists.

Oh, you'll also need to nix your vfb I'm afraid...

If you are comfortable patching C code then you could try modifying
tools/libxl/libxl.c:libxl_set_vcpuonline() to call libxl__domain_type
and arrange to call libxl__set_vcpuonline_xenstore for PV guests and the
existing logic for HVM guests to choose between _xenstore and _qmp.

If you are comfortable doing that then please submit a patch.
http://wiki.xen.org/wiki/Submitting_Xen_Patches has some general
guidance.

Ian.

> 
> On Thu, Jan 30, 2014 at 9:12 AM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> > On Thu, 2014-01-30 at 09:04 -0700, Yun Wang wrote:
> >> What disk option should I use for a PV guest? tap2:aio seems to have a
> >> similar issue with "xl vcpu-set"
> >
> > To workaround the issue you would need phy: I think, perhaps by setting
> > up a loopback device on the raw image by hand.
> >
> > Hopefully the underlying issue can get fixed though.
> >
> > Ian.
> >