Xen devel - Dec 2009 - [ PATCH 4/4 ] HVM vcpu add/remove: qemu logic for vcpu add/revmoe

HVM vcpu add/remove: qemu logic for vcpu add/revmoe

-- at qemu side, get vcpu_avail which used for original cpu avail map;
-- setup gpe ioread/iowrite at qmeu;
-- setup vcpu add/remove user interface through monitor;
-- setup SCI logic;

Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>

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

On 13/12/2009 18:05, "Liu, Jinsong" <jinsong.liu@intel.com>
wrote:
> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
> 
> -- at qemu side, get vcpu_avail which used for original cpu avail map;
> -- setup gpe ioread/iowrite at qmeu;
> -- setup vcpu add/remove user interface through monitor;
> -- setup SCI logic;
I''m guessing because this adds a new command-line option that I need
this
checked into the qemu tree before I can apply your first patch (1/4)?
Otherwise that patch will break domain creation as qemu will exit with an
''unrecognised option'' error. So I need Ian Jackson to apply
this one and
send me an updated QEMU_TAG first.

 -- Keir



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

Keir Fraser wrote:
> On 13/12/2009 18:05, "Liu, Jinsong" <jinsong.liu@intel.com>
wrote:
> 
>> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
>> 
>> -- at qemu side, get vcpu_avail which used for original cpu avail
>> map; 
>> -- setup gpe ioread/iowrite at qmeu;
>> -- setup vcpu add/remove user interface through monitor;
>> -- setup SCI logic;
> 
> I''m guessing because this adds a new command-line option that I
need
> this checked into the qemu tree before I can apply your first patch
> (1/4)? Otherwise that patch will break domain creation as qemu will
> exit with an ''unrecognised option'' error. So I need Ian
Jackson to
> apply this one and send me an updated QEMU_TAG first.
> 
>  -- Keir
Yes, only add patch 1/4 will break domain creation, hence need add qemu patch
4/4 first.
Ian, would you please review patch 4/4 and check into qemu tree?

Thanks,
Jinsong
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 14/12/2009 08:04, "Keir Fraser" <keir.fraser@eu.citrix.com>
wrote:
> On 13/12/2009 18:05, "Liu, Jinsong" <jinsong.liu@intel.com>
wrote:
> 
>> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
>> 
>> -- at qemu side, get vcpu_avail which used for original cpu avail map;
>> -- setup gpe ioread/iowrite at qmeu;
>> -- setup vcpu add/remove user interface through monitor;
>> -- setup SCI logic;
> 
> I''m guessing because this adds a new command-line option that I
need this
> checked into the qemu tree before I can apply your first patch (1/4)?
> Otherwise that patch will break domain creation as qemu will exit with an
> ''unrecognised option'' error. So I need Ian Jackson to
apply this one and
> send me an updated QEMU_TAG first.
As of c/s 20640 all your Xen patches are checked in. I modified them a bit
so you may want to take a look. I commented out the one line that actually
sets the new qemu option, until that option is supported by our qemu. I
think there is a question over whether the new qemu option should (a) have a
better name (I called it vcpu_online[] in hvm_info structure); and (b)
should have a more user-friendly format (currently passing a decimal number
interpreted as a bitmap - perhaps should be a list of vcpus instead).

 -- Keir



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

On 14/12/2009 09:25, "Liu, Jinsong" <jinsong.liu@intel.com>
wrote:
>> I''m guessing because this adds a new command-line option that
I need
>> this checked into the qemu tree before I can apply your first patch
>> (1/4)? Otherwise that patch will break domain creation as qemu will
>> exit with an ''unrecognised option'' error. So I need
Ian Jackson to
>> apply this one and send me an updated QEMU_TAG first.
> 
> Yes, only add patch 1/4 will break domain creation, hence need add qemu
patch
> 4/4 first.
> Ian, would you please review patch 4/4 and check into qemu tree?
There''s just a question mark over whether the option should be more
user-
friendly (see my previous email).

 -- Keir



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

Keir Fraser wrote:
> On 14/12/2009 08:04, "Keir Fraser"
<keir.fraser@eu.citrix.com> wrote:
> 
>> On 13/12/2009 18:05, "Liu, Jinsong"
<jinsong.liu@intel.com> wrote:
>> 
>>> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
>>> 
>>> -- at qemu side, get vcpu_avail which used for original cpu avail
>>> map; 
>>> -- setup gpe ioread/iowrite at qmeu;
>>> -- setup vcpu add/remove user interface through monitor;
>>> -- setup SCI logic;
>> 
>> I''m guessing because this adds a new command-line option that
I need
>> this checked into the qemu tree before I can apply your first patch
>> (1/4)? Otherwise that patch will break domain creation as qemu will
>> exit with an ''unrecognised option'' error. So I need
Ian Jackson to
>> apply this one and send me an updated QEMU_TAG first.
> 
> As of c/s 20640 all your Xen patches are checked in. I modified them
> a bit so you may want to take a look. I commented out the one line
> that actually sets the new qemu option, until that option is
> supported by our qemu. I think there is a question over whether the
> new qemu option should (a) have a better name (I called it
> vcpu_online[] in hvm_info structure); and (b) should have a more
> user-friendly format (currently passing a decimal number interpreted
> as a bitmap - perhaps should be a list of vcpus instead). 
> 
>  -- Keir
Thanks!

Currently at xm level, HVM config keep compatible with PV config (patch 20495,
20502), they both set maxvcpus/ avail vcpus at config file as
maxvcpus = xxx
vcpus = yyy
and both HVM and PV can dynamic add/remove vcpus now.

One question is, patch 20384/ 20386/ 20389 and qemu patch
3140780e451d3919ef2c81f91ae0ebe3f286eb06 extend HVM vcpus max to 128, however,
current xm and xend python logic seems only support max 64 since xm/xend now
interpret vcpu bitmap to a ''long''.
I agree that the bitmap would better be replaced by a list of vcpus so that
vcpus number will not be limited in the future.

Regards,
Jinsong
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 12/14/2009 10:54 AM, Liu, Jinsong wrote:
> Keir Fraser wrote:
>    
>> On 14/12/2009 08:04, "Keir
Fraser"<keir.fraser@eu.citrix.com>  wrote:
>>
>>      
>>> On 13/12/2009 18:05, "Liu,
Jinsong"<jinsong.liu@intel.com>  wrote:
>>>
>>>        
>>>> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
>>>>
>>>> -- at qemu side, get vcpu_avail which used for original cpu
avail
>>>> map;
>>>> -- setup gpe ioread/iowrite at qmeu;
>>>> -- setup vcpu add/remove user interface through monitor;
>>>> -- setup SCI logic;
>>>>          
>>> I''m guessing because this adds a new command-line option
that I need
>>> this checked into the qemu tree before I can apply your first patch
>>> (1/4)? Otherwise that patch will break domain creation as qemu will
>>> exit with an ''unrecognised option'' error. So I
need Ian Jackson to
>>> apply this one and send me an updated QEMU_TAG first.
>>>        
>> As of c/s 20640 all your Xen patches are checked in. I modified them
>> a bit so you may want to take a look. I commented out the one line
>> that actually sets the new qemu option, until that option is
>> supported by our qemu. I think there is a question over whether the
>> new qemu option should (a) have a better name (I called it
>> vcpu_online[] in hvm_info structure); and (b) should have a more
>> user-friendly format (currently passing a decimal number interpreted
>> as a bitmap - perhaps should be a list of vcpus instead).
>>
>>   -- Keir
>>      
> Thanks!
>
> Currently at xm level, HVM config keep compatible with PV config (patch
20495, 20502), they both set maxvcpus/ avail vcpus at config file as
> maxvcpus = xxx
> vcpus = yyy
> and both HVM and PV can dynamic add/remove vcpus now.
>
> One question is, patch 20384/ 20386/ 20389 and qemu patch
3140780e451d3919ef2c81f91ae0ebe3f286eb06 extend HVM vcpus max to 128, however,
current xm and xend python logic seems only support max 64 since xm/xend now
interpret vcpu bitmap to a ''long''.
> I agree that the bitmap would better be replaced by a list of vcpus so that
vcpus number will not be limited in the future.
>    
Well, actually I did the patches 20495 and 20502 and I used the existing 
infrastructure of vcpus and vcpu_avail here so there were no changes in 
current infrastructure since I wanted to preserve it. That''s the reason
why my patch was done this way. Changing existing infrastructure could 
introduce some problems with that so that''s why maxvcpus values is
going
to be vcpus in xend python as well as vcpus is set the bitmask of 
vcpu_avail.

Regards,
Michal
> Regards,
> Jinsong
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
>    

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

On Mon, Dec 14, 2009 at 02:05:18AM +0800, Liu, Jinsong
wrote:
> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
> 
> -- at qemu side, get vcpu_avail which used for original cpu avail map;
> -- setup gpe ioread/iowrite at qmeu;
> -- setup vcpu add/remove user interface through monitor;
> -- setup SCI logic;
> 
> Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
It says in the patch:

/* Please update qemu-doc.texi when adding or changing commands */^M
 static const term_cmd_t term_cmds[] = {^M
.. snip..
+    { "cpu_set", "is", do_cpu_set_nr,^M

Is there a forthcomming patch to update the qemu-doc.texi?

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

Keir Fraser wrote:
> On 14/12/2009 08:04, "Keir Fraser"
<keir.fraser@eu.citrix.com> wrote:
> 
>> On 13/12/2009 18:05, "Liu, Jinsong"
<jinsong.liu@intel.com> wrote:
>> 
>>> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
>>> 
>>> -- at qemu side, get vcpu_avail which used for original cpu avail
>>> map; 
>>> -- setup gpe ioread/iowrite at qmeu;
>>> -- setup vcpu add/remove user interface through monitor;
>>> -- setup SCI logic;
>> 
>> I''m guessing because this adds a new command-line option that
I need
>> this checked into the qemu tree before I can apply your first patch
>> (1/4)? Otherwise that patch will break domain creation as qemu will
>> exit with an ''unrecognised option'' error. So I need
Ian Jackson to
>> apply this one and send me an updated QEMU_TAG first.
> 
> As of c/s 20640 all your Xen patches are checked in. I modified them
> a bit so you may want to take a look. I commented out the one line
> that actually sets the new qemu option, until that option is
> supported by our qemu. I think there is a question over whether the
> new qemu option should (a) have a better name (I called it
> vcpu_online[] in hvm_info structure); and (b) should have a more
> user-friendly format (currently passing a decimal number interpreted
> as a bitmap - perhaps should be a list of vcpus instead). 
> 
>  -- Keir
Just go through your modification, bios_info is really good, I don''t
know its usage before, thanks!

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

Ian,

Would you please review this patch and check into qemu-xen tree?

We recently enable HVM vcpu add/remove feature, total 4 patches, among them 3
patches is at Xen upstream tree now (c/s 20638, 20639, 20640),
and without this patch at qemu side, HVM vcpu add/remove feature cannot be
enabled.

Thanks,
Jinsong

Liu, Jinsong wrote:
> Keir Fraser wrote:
>> On 13/12/2009 18:05, "Liu, Jinsong"
<jinsong.liu@intel.com> wrote:
>> 
>>> HVM vcpu add/remove: qemu logic for vcpu add/revmoe
>>> 
>>> -- at qemu side, get vcpu_avail which used for original cpu avail
>>> map; 
>>> -- setup gpe ioread/iowrite at qmeu;
>>> -- setup vcpu add/remove user interface through monitor;
>>> -- setup SCI logic;
>> 
>> I''m guessing because this adds a new command-line option that
I need
>> this checked into the qemu tree before I can apply your first patch
>> (1/4)? Otherwise that patch will break domain creation as qemu will
>> exit with an ''unrecognised option'' error. So I need
Ian Jackson to
>> apply this one and send me an updated QEMU_TAG first.
>> 
>>  -- Keir
> 
> Yes, only add patch 1/4 will break domain creation, hence need add
> qemu patch 4/4 first. Ian, would you please review patch 4/4 and
> check into qemu tree? 
> 
> Thanks,
> Jinsong
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel


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