Re: [Xen-devel] [PATCH 2/2] libxl: fix stale timeout event callback race

[Jim Fehlig <jfehlig@xxxxxxxx>]

Ian Campbell wrote:
> On Wed, 2012-12-12 at 18:01 +0000, Jim Fehlig wrote:
>   
>> Ian Jackson wrote:
>>     
>
>   
>>>   
>>>       
>>>>  After again reading
>>>> libxl_event.h, I'm considering the below patch in the libvirt libxl
>>>> driver.  The change is primarily inspired by this comment for
>>>> libxl_osevent_occurred_timeout:
>>>>     
>>>>         
>>> ...
>>>   
>>>       
>>>> /* Implicitly, on entry to this function the timeout has been
>>>>  * deregistered.  If _occurred_timeout is called, libxl will not
>>>>  * call timeout_deregister; if it wants to requeue the timeout it
>>>>  * will call timeout_register again.
>>>>  */
>>>>     
>>>>         
>>> Well your patch is only correct when used with the new libxl, with my
>>> patches.
>>>   
>>>       
>> Hmm, it is not clear to me how to make the libxl driver work correctly
>> with libxl pre and post your patches :-/.
>>     
>
> Ideally we will find a way to make this work without changes on the
> application side.
>   

I think Ian J. is right about applications still working, *if* they have
the callbacks coded correctly.  There are some bugs on the libvirt side
as well :).

> But if that turns out to be impossible and applications are going to
> need patching anyway then I think we should consider just fixing the API
> rather than playing tricks like the "modify to 0" thing to try and keep
> it compatible.
>
> One option is to add new hooks which libxl can call to take/release the
> application's event loop lock + a LIBXL_HAVE_EVENT_LOOP_LOCK define so
> the application can conditionally provide them.

libvirt's event loop lock is private to the event impl and not exposed
to its numerous users.

Regards,
Jim



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel