Re: [PATCH-RFC 2/2] eventfd: EFD_STATE flag

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Aug 04, 2009 at 12:17:44PM +0300, Avi Kivity wrote:
> On 08/04/2009 11:54 AM, Michael S. Tsirkin wrote:
>> On Tue, Aug 04, 2009 at 11:53:03AM +0300, Avi Kivity wrote:
>>    
>>> On 08/03/2009 07:57 PM, Michael S. Tsirkin wrote:
>>>      
>>>>> Why not do it at the point of the write?
>>>>>
>>>>>       if (value != ctx->count) {
>>>>>           ctx->count = value;
>>>>>           wake_things_up();
>>>>>       }
>>>>>
>>>>>          
>>>> What if write comes before read?
>>>>
>>>>        
>>> The read will get the new value.
>>>      
>>
>> Yes :) But how does read know it should not block?
>>    
>
> If a different read comes after the write but after our read, it will  
> have transferred the value, resulting in the same situation.

Not the same: one reader wakes up, others sleep.

Multiple reads from the same fd behave this way for any file I can think
of. Consider regular eventfd, or a pipe, a socket ...  But if we want to
support blocking reads, we probably should not require the readers to
sync with writers.


> I think reads should never block with a state based mechanism.

Yes, with no support for blocking reads, we don't need the state.
In that case, we probably want to error on open out unless O_NONBLOCK is
specified.  But why is this a good idea?


> -- 
> error compiling committee.c: too many arguments to function