[Qemu-devel] Re: [PATCH v6 18/18] virtio-serial-bus: wake up iothread upon guest read notification

[Anthony Liguori <anthony@codemonkey.ws>]

On 04/27/2010 12:58 PM, Marcelo Tosatti wrote:
> On Tue, Apr 27, 2010 at 12:41:27PM -0500, Anthony Liguori wrote:
>    
>> On 04/27/2010 07:34 AM, Amit Shah wrote:
>>      
>>> From: Marcelo Tosatti<mtosatti@redhat.com>
>>>
>>> Wake up iothread when buffers are consumed.
>>>
>>> Signed-off-by: Marcelo Tosatti<mtosatti@redhat.com>
>>> Signed-off-by: Amit Shah<amit.shah@redhat.com>
>>>        
>> What's the race here?  This looks very odd to me.
>>      
> We discussed this on the following thread:
>
> http://www.mail-archive.com/kvm@vger.kernel.org/msg29249.html
>    

I don't think there was ever an adequate explanation of exactly what was 
happening.

My suspicion is that the real problem is due to the can_read() handlers 
whereas it's necessary to invoke the main loop for can_read() to be 
executed again to cause the event to be polled.  The better solution 
though is to not use the can_read() handler and instead explicitly 
register and deregister the read callbacks.

qemu_set_fd_handler() would require a notify call but that makes sense.

I think this is a good justification for removing uses of 
qemu_set_fd_handler2().

Regards,

Anthony Liguori