Re: [Qemu-devel] about kvmclock QOM

Re: [Qemu-devel] about kvmclock QOM

[Zhi Yong Wu]

On Tue, Mar 13, 2012 at 9:52 PM, Jan Kiszka <jan.kiszka@siemens.com> wrote:
> On 2012-03-13 14:40, Zhi Yong Wu wrote:
>> On Tue, Mar 13, 2012 at 9:36 PM, Jan Kiszka <jan.kiszka@siemens.com> wrote:
>>> On 2012-03-13 13:57, Zhi Yong Wu wrote:
>>>> HI, Jan
>>>>
>>>> Why can i not see kvmclock device in the output of qom-list? Since it
>>>> has been converted to QOM.
>>>
>>> qom-list, like all QMP/ scripts, are currently broken for me, can't test
>>> therefore. Please post any findings to the list while I'm digging into
>> I can see other devices such emulated nics via qom-list, but didn't
>> see kvmclock device in its output.
>> what is kvmclock's path if it can work normally?
>> My qemu command line is
>>
>> [root@f15 qemu]# x86_64-softmmu/qemu-system-x86_64 -m 512 -drive
>> file=/home/zwu/work/misc/image/rh6.1.img,if=ide,cache=none -enable-kvm
>> -qmp unix:/tmp/server.sock,server,nowait -netdev type=user,id=net
>> -device virtio-net-pci,netdev=net,id=net -net user,vlan=1 -device
>> virtio-net-pci,id=net2,vlan=1 -chardev id=stdio,backend=stdio -device
>> isa-serial,chardev=stdio,id=serial
>
> Ah, now it works again (forgot mode="control" for -mon).
>
>>
>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>> vga/
>> i440fx/
>> peripheral/
>> peripheral-anon/
>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /peripheral
>> serial/
>> net2/
>> net/
>>
>
> kvmclock (like many devices) is not yet linked to some parent device, so
static TypeInfo kvmclock_info = {
    .name          = "kvmclock",
    .parent        = TYPE_SYS_BUS_DEVICE,
It seems to be linked from here
    .instance_size = sizeof(KVMClockState),
    .class_init    = kvmclock_class_init,
};

I also checked other devices which can be seen in the output of
qom-list. I have not seen kvmclock QOM is different from other
devices.
So i am very surprised that why kvmclock device can not be seen in
qom-list output.

> it's not present in that tree. This also happens to auto-created
> isa-serial, e.g. (just leave out -device isa-serial)
Yeah, i can see it in qom-list output.
>
> Again, please post to the list, not privately.
>
> Jan
>
> --
> Siemens AG, Corporate Technology, CT T DE IT 1
> Corporate Competence Center Embedded Linux



-- 
Regards,

Zhi Yong Wu

Re: [Qemu-devel] about kvmclock QOM

[Andreas Färber]

Am 13.03.2012 15:11, schrieb Zhi Yong Wu:
> On Tue, Mar 13, 2012 at 9:52 PM, Jan Kiszka <jan.kiszka@siemens.com> wrote:
>> On 2012-03-13 14:40, Zhi Yong Wu wrote:
>>> On Tue, Mar 13, 2012 at 9:36 PM, Jan Kiszka <jan.kiszka@siemens.com> wrote:
>>>> On 2012-03-13 13:57, Zhi Yong Wu wrote:
>>>>> HI, Jan
>>>>>
>>>>> Why can i not see kvmclock device in the output of qom-list? Since it
>>>>> has been converted to QOM.
>>>>
>>>> qom-list, like all QMP/ scripts, are currently broken for me, can't test
>>>> therefore. Please post any findings to the list while I'm digging into
>>> I can see other devices such emulated nics via qom-list, but didn't
>>> see kvmclock device in its output.
>>> what is kvmclock's path if it can work normally?
>>> My qemu command line is
>>>
>>> [root@f15 qemu]# x86_64-softmmu/qemu-system-x86_64 -m 512 -drive
>>> file=/home/zwu/work/misc/image/rh6.1.img,if=ide,cache=none -enable-kvm
>>> -qmp unix:/tmp/server.sock,server,nowait -netdev type=user,id=net
>>> -device virtio-net-pci,netdev=net,id=net -net user,vlan=1 -device
>>> virtio-net-pci,id=net2,vlan=1 -chardev id=stdio,backend=stdio -device
>>> isa-serial,chardev=stdio,id=serial
>>
>> Ah, now it works again (forgot mode="control" for -mon).
>>
>>>
>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>>> vga/
>>> i440fx/
>>> peripheral/
>>> peripheral-anon/
>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /peripheral
>>> serial/
>>> net2/
>>> net/
>>>
>>
>> kvmclock (like many devices) is not yet linked to some parent device, so
> static TypeInfo kvmclock_info = {
>     .name          = "kvmclock",
>     .parent        = TYPE_SYS_BUS_DEVICE,
> It seems to be linked from here

No. This specifies the parent class (base or super class).

What you are looking for is the "parent" property which gets set up by
adding a child property to some object. Many devices are not yet wired
up as such and can only be inspected by info qtree monitor command.

>     .instance_size = sizeof(KVMClockState),
>     .class_init    = kvmclock_class_init,
> };
> 
> I also checked other devices which can be seen in the output of
> qom-list. I have not seen kvmclock QOM is different from other
> devices.
> So i am very surprised that why kvmclock device can not be seen in
> qom-list output.
> 
>> it's not present in that tree. This also happens to auto-created
>> isa-serial, e.g. (just leave out -device isa-serial)
> Yeah, i can see it in qom-list output.

Devices created by -device should be listed in the "/anon-peripheral"
(or so) container. Check qom-list / for its child nodes.

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg

Re: [Qemu-devel] about kvmclock QOM

[Zhi Yong Wu]

On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
> Am 13.03.2012 15:11, schrieb Zhi Yong Wu:
>> On Tue, Mar 13, 2012 at 9:52 PM, Jan Kiszka <jan.kiszka@siemens.com> wrote:
>>> On 2012-03-13 14:40, Zhi Yong Wu wrote:
>>>> On Tue, Mar 13, 2012 at 9:36 PM, Jan Kiszka <jan.kiszka@siemens.com> wrote:
>>>>> On 2012-03-13 13:57, Zhi Yong Wu wrote:
>>>>>> HI, Jan
>>>>>>
>>>>>> Why can i not see kvmclock device in the output of qom-list? Since it
>>>>>> has been converted to QOM.
>>>>>
>>>>> qom-list, like all QMP/ scripts, are currently broken for me, can't test
>>>>> therefore. Please post any findings to the list while I'm digging into
>>>> I can see other devices such emulated nics via qom-list, but didn't
>>>> see kvmclock device in its output.
>>>> what is kvmclock's path if it can work normally?
>>>> My qemu command line is
>>>>
>>>> [root@f15 qemu]# x86_64-softmmu/qemu-system-x86_64 -m 512 -drive
>>>> file=/home/zwu/work/misc/image/rh6.1.img,if=ide,cache=none -enable-kvm
>>>> -qmp unix:/tmp/server.sock,server,nowait -netdev type=user,id=net
>>>> -device virtio-net-pci,netdev=net,id=net -net user,vlan=1 -device
>>>> virtio-net-pci,id=net2,vlan=1 -chardev id=stdio,backend=stdio -device
>>>> isa-serial,chardev=stdio,id=serial
>>>
>>> Ah, now it works again (forgot mode="control" for -mon).
>>>
>>>>
>>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>>>> vga/
>>>> i440fx/
>>>> peripheral/
>>>> peripheral-anon/
>>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /peripheral
>>>> serial/
>>>> net2/
>>>> net/
>>>>
>>>
>>> kvmclock (like many devices) is not yet linked to some parent device, so
>> static TypeInfo kvmclock_info = {
>>     .name          = "kvmclock",
>>     .parent        = TYPE_SYS_BUS_DEVICE,
>> It seems to be linked from here
>
> No. This specifies the parent class (base or super class).
>
> What you are looking for is the "parent" property which gets set up by
You mean that current device object should have one "parent" property?
Its parent device should have one "child" property pointing to it?
But i have not seen this property in qom-list output.
[root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
vga/
i440fx/
peripheral/
peripheral-anon/
[root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /i440fx
piix3/
type
[root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /i440fx/piix3
rtc/
command_serr_enable
legacy-command_serr_enable
multifunction
legacy-multifunction
rombar
romfile
legacy-romfile
addr
legacy-addr
type
[root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /i440fx/piix3/rtc
date
type
lost_tick_policy
base_year
> adding a child property to some object. Many devices are not yet wired
> up as such and can only be inspected by info qtree monitor command.
>
>>     .instance_size = sizeof(KVMClockState),
>>     .class_init    = kvmclock_class_init,
>> };
>>
>> I also checked other devices which can be seen in the output of
>> qom-list. I have not seen kvmclock QOM is different from other
>> devices.
>> So i am very surprised that why kvmclock device can not be seen in
>> qom-list output.
>>
>>> it's not present in that tree. This also happens to auto-created
>>> isa-serial, e.g. (just leave out -device isa-serial)
>> Yeah, i can see it in qom-list output.
>
> Devices created by -device should be listed in the "/anon-peripheral"
It is listed in /peripheral, not /peripheral-anon...
> (or so) container. Check qom-list / for its child nodes.
>
> Andreas
>
> --
> SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
> GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg



-- 
Regards,

Zhi Yong Wu

Re: [Qemu-devel] about kvmclock QOM

[Andreas Färber]

Am 13.03.2012 15:52, schrieb Zhi Yong Wu:
> On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
>> What you are looking for is the "parent" property which gets set up by
> You mean that current device object should have one "parent" property?
> Its parent device should have one "child" property pointing to it?
> But i have not seen this property in qom-list output.
> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
> vga/
> i440fx/
> peripheral/
> peripheral-anon/

Every one of these with */ is a child, in this case of the root "/". :)

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg

Re: [Qemu-devel] about kvmclock QOM

[Zhi Yong Wu]

On Tue, Mar 13, 2012 at 11:10 PM, Andreas Färber <afaerber@suse.de> wrote:
> Am 13.03.2012 15:52, schrieb Zhi Yong Wu:
>> On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
>>> What you are looking for is the "parent" property which gets set up by
>> You mean that current device object should have one "parent" property?
>> Its parent device should have one "child" property pointing to it?
>> But i have not seen this property in qom-list output.
>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>> vga/
>> i440fx/
>> peripheral/
>> peripheral-anon/
>
> Every one of these with */ is a child, in this case of the root "/". :)
I see. but i want to know how it use object to represent this relationship.
>
> Andreas
>
> --
> SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
> GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg



-- 
Regards,

Zhi Yong Wu

Re: [Qemu-devel] about kvmclock QOM

[Andreas Färber]

Am 13.03.2012 16:17, schrieb Zhi Yong Wu:
> On Tue, Mar 13, 2012 at 11:10 PM, Andreas Färber <afaerber@suse.de> wrote:
>> Am 13.03.2012 15:52, schrieb Zhi Yong Wu:
>>> On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
>>>> What you are looking for is the "parent" property which gets set up by
>>> You mean that current device object should have one "parent" property?
>>> Its parent device should have one "child" property pointing to it?
>>> But i have not seen this property in qom-list output.
>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>>> vga/
>>> i440fx/
>>> peripheral/
>>> peripheral-anon/
>>
>> Every one of these with */ is a child, in this case of the root "/". :)
> I see. but i want to know how it use object to represent this relationship.

See object_property_add_child() in qom/object.c.

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg

Re: [Qemu-devel] about kvmclock QOM

[Zhi Yong Wu]

On Tue, Mar 13, 2012 at 11:10 PM, Andreas Färber <afaerber@suse.de> wrote:
> Am 13.03.2012 15:52, schrieb Zhi Yong Wu:
>> On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
>>> What you are looking for is the "parent" property which gets set up by
>> You mean that current device object should have one "parent" property?
>> Its parent device should have one "child" property pointing to it?
>> But i have not seen this property in qom-list output.
>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>> vga/
>> i440fx/
>> peripheral/
>> peripheral-anon/
>
> Every one of these with */ is a child, in this case of the root "/". :)
Moreover, why can isa-serial device be liste in qom-list output but
kvmclock can't? where is the difference between their QOM?
>
> Andreas
>
> --
> SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
> GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg



-- 
Regards,

Zhi Yong Wu

Re: [Qemu-devel] about kvmclock QOM

[Zhi Yong Wu]

On Tue, Mar 13, 2012 at 11:18 PM, Andreas Färber <afaerber@suse.de> wrote:
> Am 13.03.2012 16:17, schrieb Zhi Yong Wu:
>> On Tue, Mar 13, 2012 at 11:10 PM, Andreas Färber <afaerber@suse.de> wrote:
>>> Am 13.03.2012 15:52, schrieb Zhi Yong Wu:
>>>> On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
>>>>> What you are looking for is the "parent" property which gets set up by
>>>> You mean that current device object should have one "parent" property?
>>>> Its parent device should have one "child" property pointing to it?
>>>> But i have not seen this property in qom-list output.
>>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>>>> vga/
>>>> i440fx/
>>>> peripheral/
>>>> peripheral-anon/
>>>
>>> Every one of these with */ is a child, in this case of the root "/". :)
>> I see. but i want to know how it use object to represent this relationship.
>
> See object_property_add_child() in qom/object.c.
great, thanks.
>
> Andreas
>
> --
> SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
> GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg



-- 
Regards,

Zhi Yong Wu

Re: [Qemu-devel] about kvmclock QOM

[Andreas Färber]

Am 13.03.2012 16:21, schrieb Zhi Yong Wu:
> On Tue, Mar 13, 2012 at 11:10 PM, Andreas Färber <afaerber@suse.de> wrote:
>> Am 13.03.2012 15:52, schrieb Zhi Yong Wu:
>>> On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
>>>> What you are looking for is the "parent" property which gets set up by
[...adding a child property to some object. Many devices are not yet
wired up as such...]
>>> You mean that current device object should have one "parent" property?
>>> Its parent device should have one "child" property pointing to it?
>>> But i have not seen this property in qom-list output.
>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>>> vga/
>>> i440fx/
>>> peripheral/
>>> peripheral-anon/
>>
>> Every one of these with */ is a child, in this case of the root "/". :)
> Moreover, why can isa-serial device be liste in qom-list output but
> kvmclock can't? where is the difference between their QOM?

Because someone must give the child a name and assign a parent to it by
calling said object_property_add_child(). ("wired up" above)

Anthony had an RFC series refactoring i440fx, for instance, that sparked
some controversy and was not followed-up yet.

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg

Re: [Qemu-devel] about kvmclock QOM

[Zhi Yong Wu]

On Wed, Mar 14, 2012 at 12:28 AM, Andreas Färber <afaerber@suse.de> wrote:
> Am 13.03.2012 16:21, schrieb Zhi Yong Wu:
>> On Tue, Mar 13, 2012 at 11:10 PM, Andreas Färber <afaerber@suse.de> wrote:
>>> Am 13.03.2012 15:52, schrieb Zhi Yong Wu:
>>>> On Tue, Mar 13, 2012 at 10:36 PM, Andreas Färber <afaerber@suse.de> wrote:
>>>>> What you are looking for is the "parent" property which gets set up by
> [...adding a child property to some object. Many devices are not yet
> wired up as such...]
>>>> You mean that current device object should have one "parent" property?
>>>> Its parent device should have one "child" property pointing to it?
>>>> But i have not seen this property in qom-list output.
>>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>>>> vga/
>>>> i440fx/
>>>> peripheral/
>>>> peripheral-anon/
>>>
>>> Every one of these with */ is a child, in this case of the root "/". :)
>> Moreover, why can isa-serial device be liste in qom-list output but
>> kvmclock can't? where is the difference between their QOM?
>
> Because someone must give the child a name and assign a parent to it by
> calling said object_property_add_child(). ("wired up" above)
>
> Anthony had an RFC series refactoring i440fx, for instance, that sparked
> some controversy and was not followed-up yet.
thanks.
>
> Andreas
>
> --
> SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
> GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg



-- 
Regards,

Zhi Yong Wu

Re: [Qemu-devel] about kvmclock QOM

[Gavin Shan]

> >>>>> What you are looking for is the "parent" property which gets set up by
> > [...adding a child property to some object. Many devices are not yet
> > wired up as such...]
> >>>> You mean that current device object should have one "parent" property?
> >>>> Its parent device should have one "child" property pointing to it?
> >>>> But i have not seen this property in qom-list output.
> >>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
> >>>> vga/
> >>>> i440fx/
> >>>> peripheral/
> >>>> peripheral-anon/
> >>>
> >>> Every one of these with */ is a child, in this case of the root "/". :)
> >> Moreover, why can isa-serial device be liste in qom-list output but
> >> kvmclock can't? where is the difference between their QOM?
> >
> > Because someone must give the child a name and assign a parent to it by
> > calling said object_property_add_child(). ("wired up" above)
> >
> > Anthony had an RFC series refactoring i440fx, for instance, that sparked
> > some controversy and was not followed-up yet.
> thanks.

I'm a little curious that QOM will become more dependent than ever. Originally,
qemu has qdev prototype to trace device/bus staff. Is QOM going to replace qdev/qbus
staff in future?

Thanks,
Gavin

Re: [Qemu-devel] about kvmclock QOM

[Zhi Yong Wu]

On Wed, Mar 14, 2012 at 12:27 PM, Gavin Shan <shangw@linux.vnet.ibm.com> wrote:
>> >>>>> What you are looking for is the "parent" property which gets set up by
>> > [...adding a child property to some object. Many devices are not yet
>> > wired up as such...]
>> >>>> You mean that current device object should have one "parent" property?
>> >>>> Its parent device should have one "child" property pointing to it?
>> >>>> But i have not seen this property in qom-list output.
>> >>>> [root@f15 qemu]# QMP/qom-list -s /tmp/server.sock /
>> >>>> vga/
>> >>>> i440fx/
>> >>>> peripheral/
>> >>>> peripheral-anon/
>> >>>
>> >>> Every one of these with */ is a child, in this case of the root "/". :)
>> >> Moreover, why can isa-serial device be liste in qom-list output but
>> >> kvmclock can't? where is the difference between their QOM?
>> >
>> > Because someone must give the child a name and assign a parent to it by
>> > calling said object_property_add_child(). ("wired up" above)
>> >
>> > Anthony had an RFC series refactoring i440fx, for instance, that sparked
>> > some controversy and was not followed-up yet.
>> thanks.
>
> I'm a little curious that QOM will become more dependent than ever. Originally,
It has better expandability than qdev.
> qemu has qdev prototype to trace device/bus staff. Is QOM going to replace qdev/qbus
Yeah
> staff in future?
>
> Thanks,
> Gavin
>



-- 
Regards,

Zhi Yong Wu