[PATCH] Fix another race condition in xm destroy

[PATCH] Fix another race condition in xm destroy

[Anthony Liguori]

[-- Attachment #1: Type: text/plain, Size: 617 bytes --]

There's a race condition in xm destroy that gets triggered when 
xc_domain_destroy results in a domain disappearing before we invoke the 
cleanup routines.  Specifically, the saveToDB routines will through a No 
such process error.

Christian, I'd appreciate if you could sanity check this before 
committing.  I think it's okay because starting the cleanup shouldn't 
have any adverse effects.  Having another pair of eyes confirm it would 
make me feel better though :-)

BTW, this is not the fix for the xm destroy && xm create race 
condition.  Just another one I found in the process.

Regards,

Anthony Liguori

[-- Attachment #2: 6884-xm-destroy-race.diff --]
[-- Type: text/x-patch, Size: 914 bytes --]

# HG changeset patch
# User anthony@localhost.localdomain
# Node ID 2790025adf8eae599d026c9ad6f0b3deacda9b9f
# Parent  fd19e760932d095b23d38e67eaec24dd02ba3aba
Fix a race in xm destroy.  Once we call xc_domain_destroy, there's no guarentee
that the domain will exist (even immediately after the call).

Starting the cleanup before we call destroy will keep us from getting hosed.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

diff -r fd19e760932d -r 2790025adf8e tools/python/xen/xend/XendDomainInfo.py
--- a/tools/python/xen/xend/XendDomainInfo.py	Thu Sep 15 19:46:14 2005
+++ b/tools/python/xen/xend/XendDomainInfo.py	Fri Sep 16 04:35:26 2005
@@ -771,9 +771,9 @@
     def destroy(self):
         """Clenup vm and destroy domain.
         """
-        self.destroy_domain()
         self.cleanup()
         self.saveToDB()
+        self.destroy_domain()
         return 0
 
     def is_terminated(self):

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

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

Re: [PATCH] Fix another race condition in xm destroy

[Christian Limpach]

On 9/16/05, Anthony Liguori <aliguori@us.ibm.com> wrote:
> There's a race condition in xm destroy that gets triggered when
> xc_domain_destroy results in a domain disappearing before we invoke the
> cleanup routines.  Specifically, the saveToDB routines will through a No
> such process error.
> 
> Christian, I'd appreciate if you could sanity check this before
> committing.  I think it's okay because starting the cleanup shouldn't
> have any adverse effects.  Having another pair of eyes confirm it would
> make me feel better though :-)

How does it fail for you?  I fixed one issue where we were clearing
the store event channel and writing that back to the store was
failing.

I'd prefer to keep doing the cleanup after destroy_domain.

     christian

Re: [PATCH] Fix another race condition in xm destroy

[Anthony Liguori]

Christian Limpach wrote:

>On 9/16/05, Anthony Liguori <aliguori@us.ibm.com> wrote:
>  
>
>>There's a race condition in xm destroy that gets triggered when
>>xc_domain_destroy results in a domain disappearing before we invoke the
>>cleanup routines.  Specifically, the saveToDB routines will through a No
>>such process error.
>>
>>Christian, I'd appreciate if you could sanity check this before
>>committing.  I think it's okay because starting the cleanup shouldn't
>>have any adverse effects.  Having another pair of eyes confirm it would
>>make me feel better though :-)
>>    
>>
>
>How does it fail for you?  I fixed one issue where we were clearing
>the store event channel and writing that back to the store was
>failing.
>  
>
You're fix solved the problem.  That's why I made sure to ask, I wasn't 
confident my solution was correct :-)

Thanks,

Anthony Liguori

>I'd prefer to keep doing the cleanup after destroy_domain.
>
>     christian
>
>  
>