[linux-cifs-client] Re: huge variation in wait_event called by wait_for_response. Potential bug? Help please

[Jeff Layton]

On Wed, 12 Nov 2008 17:41:13 -0800
"wasrshi nimara" <warshinimara at gmail.com> wrote:

> Hi,
> 
> Figured this one out. That was a bad question!
> 
> What i should have said:
> 
> Is it ok to replace the wait_event with wait_event_timeout. Looks like the
> denotify thread
> periodically wakes up and cleans this queue. So the wait time is always
> delta 0 - 15 seconds.
> 
> I am working on an app that needs strict guarantees on this timeout, hence
> the question.
> 
> thanks
> warshi
> 
> 

It sounds reasonable to me. I have some questions about what the
denotify thread is for anyway. It doesn't seem like having a separate
kernel thread to wake up the waitqueues every 15 seconds is a good
design, but maybe I'm missing something.

Wasrshi, do you have a patch? It's easier to discuss ideas like this
with some concrete code as a starting point.

Thanks,
-- 
Jeff Layton <jlayton at redhat.com>