Xen devel - Sep 2005 - [PATCH][VT] Fix vnc configure problem for vmx guest

Fix vnc configure problem for vmx guest induced in 6984,
Signed-off-by: Xiaofeng Ling <xiaofeng.ling@intel.com>

diff -r 10759a44ce3b tools/python/xen/xend/image.py
--- a/tools/python/xen/xend/image.py    Thu Sep 22 15:12:14 2005
+++ b/tools/python/xen/xend/image.py    Fri Sep 23 00:33:50 2005
@@ -283,14 +283,12 @@
         self.pid = 0
         self.memmap_value = []

-        self.dmargs += self.configVNC(imageConfig)
-
-
     def createImage(self):
         """Create a VM for the VMX environment.
         """
         self.parseMemmap()
         self.createDomain()
+        self.dmargs += self.configVNC(self.vm.info[''device''])

     def buildDomain(self):
         # Create an event channel
-------------------
Ling Xiaofeng(Daniel)
Open Source Technology Center
Intel China Software Center
iNet: 8-752-1243
8621-52574545-1243(O)
xfling@users.sourceforge.net

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

On Fri, Sep 23, 2005 at 08:31:39AM +0800, Ling, Xiaofeng wrote:
> Fix vnc configure problem for vmx guest induced in 6984,
Hi Xiofeng,

Sorry for breaking your code yesterday, but could you please explain to me
how it broke?  I don''t understand what was wrong, because as far as I
can see,
configVNC only depends on the config it receives, so moving it
shouldn''t have
made any difference.  Obviously it did, so could you explain it to me please?

I specifically don''t want to apply your patch because I want remove the
dependency from image.py to self.info, and your patch puts that dependency
back.

Thanks,

Ewan.


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

>> Fix vnc configure problem for vmx guest induced in 6984,
>
>Hi Xiofeng,
>
>Sorry for breaking your code yesterday, but could you please 
>explain to me
>how it broke?  I don''t understand what was wrong, because as 
>far as I can see,
>configVNC only depends on the config it receives, so moving it 
>shouldn''t have
>made any difference.  Obviously it did, so could you explain 
>it to me please?
>
>I specifically don''t want to apply your patch because I want remove
the
>dependency from image.py to self.info, and your patch puts 
>that dependency
>back.
>
Hi Ewan,
configVNC will use domid plus a constant as listening port number to
VNCserver, however at the time it is called, the domain has not been
created yet...
-Xin

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

see attach mail. That is why configVNC is seperated.
I know this fix is not seems so ideal, 
I''d like to know if you have good idea for the problem.

Ewan Mellor <> wrote:
> On Fri, Sep 23, 2005 at 08:31:39AM +0800, Ling, Xiaofeng wrote:
> 
>> Fix vnc configure problem for vmx guest induced in 6984,
> 
> Hi Xiofeng,
> 
> Sorry for breaking your code yesterday, but could you please explain
> to me how it broke?  I don''t understand what was wrong, because as
> far as I can see, configVNC only depends on the config it receives,
> so moving it shouldn''t have made any difference.  Obviously it
did,
> so could you explain it to me please?    
> 
> I specifically don''t want to apply your patch because I want
remove
> the dependency from image.py to self.info, and your patch puts that
> dependency back.  
> 
> Thanks,
> 
> Ewan.
> 
> 
> _______________________________________________
> 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 Fri, Sep 23, 2005 at 11:12:26PM +0800, Ling, Xiaofeng wrote:
> see attach mail. That is why configVNC is seperated.
> I know this fix is not seems so ideal, 
> I''d like to know if you have good idea for the problem.
Thanks for that.  I''ve had a look, and I don''t see any reason
for ImageHandler
to be created before the domain itself, so I''ve moved the
ImageHandler.create
call so that it is after the xc.domain_create call.  This means that you have
the domid available inside configVNC, as necessary.

The change will be pushed to the public server in the next few minutes.  Let
me know how you get on.

Thanks,

Ewan.

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

Ewan Mellor wrote:
> On Fri, Sep 23, 2005 at 11:12:26PM +0800, Ling, Xiaofeng wrote:
> 
> 
>>see attach mail. That is why configVNC is seperated.
>>I know this fix is not seems so ideal, 
>>I''d like to know if you have good idea for the problem.
> 
> 
> Thanks for that.  I''ve had a look, and I don''t see any
reason for ImageHandler
> to be created before the domain itself, so I''ve moved the
ImageHandler.create
> call so that it is after the xc.domain_create call.  This means that you
have
> the domid available inside configVNC, as necessary.
> 
> The change will be pushed to the public server in the next few minutes. 
Let
> me know how you get on.
Yes, moving ImageGandler.create after domain create resolves the 
problem. but I guess the original purpose of putting ImageHandler.create 
after domain creating maybe if ImageHanler.ceate fails(some 
configuration wrong), there''ll be no useless domain struct exists.
So I seperates the configVNC and put it after the domain creating.
Maybe better way is have a pre-config and post-config?


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

On Mon, Sep 26, 2005 at 09:51:17AM +0800, Xiaofeng Ling wrote:
> Yes, moving ImageGandler.create after domain create resolves the 
> problem. but I guess the original purpose of putting ImageHandler.create 
> after domain creating maybe if ImageHanler.ceate fails(some 
> configuration wrong), there''ll be no useless domain struct exists.
> So I seperates the configVNC and put it after the domain creating.
> Maybe better way is have a pre-config and post-config?
Well, if the post-config fails, we still have to be able to clean up, so
there''s no extra effort in cleaning up after ImageHandler.create as
well.
That said, there''s no problem having an explicit phase of
post-configuration
that sees the image- and device-configuration structures, just as you suggest.

If it''s working for you now, I''ll leave it alone for this
week, but I''ll
revisit it later and if it looks neater to have a post-config phase,
I''ll do
that.

Thanks for all your help,

Ewan.

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

Ewan Mellor <> wrote:
> On Mon, Sep 26, 2005 at 09:51:17AM +0800, Xiaofeng Ling wrote:
> 
>> Yes, moving ImageGandler.create after domain create resolves the
>> problem. but I guess the original purpose of putting
>> ImageHandler.create after domain creating maybe if ImageHanler.ceate
>> fails(some configuration wrong), there''ll be no useless domain
>> struct exists. So I seperates the configVNC and put it after the
>> domain creating. 
>> Maybe better way is have a pre-config and post-config?
> 
> Well, if the post-config fails, we still have to be able to clean up,
> so there''s no extra effort in cleaning up after
ImageHandler.create
> as well.  
> That said, there''s no problem having an explicit phase of
> post-configuration that sees the image- and device-configuration
> structures, just as you suggest.  
> 
> If it''s working for you now, I''ll leave it alone for this
week, but
> I''ll revisit it later and if it looks neater to have a post-config
> phase, I''ll do that.  
ok.

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