Re: [RFC][PATCH] Ingo's smptimers patch experiment

[Dipankar Sarma <dipankar@in.ibm.com>]

On Wed, Feb 06, 2002 at 06:27:46PM +0100, Christoph Hellwig wrote:
> In article <20020206211925.A8720@in.ibm.com> you wrote:
> > Hi Ingo,
> >
> > I ported your smptimers patch to 2.5.3 and experimented with 
> > it a little bit. Basically I am curious about why we
> > we need to call run_all_timers() (which runs timers for all
> > CPUs) through the timer bh if locking fails in run_local_timers(). 
> 
> Some driver do ugly things with TIMER_BH, and Ingo's 2.4 patched
> tried to stayed source compatible with 2.4 drivers.
> 
> For 2.5 I'd really like to see TIMER_BH (all BH's in fact) to gone.

I can see that net driver relies on being able to disable all timers 
by doing in net/core/dev.c -

        tasklet_disable(bh_task_vec+TIMER_BH);

But this doesn't completely disable timers in Ingo's patch 
since timers can also be fired through run_local_timers() if 
locking succeeds, no TIMER_BH in that case.

There are only a few places where TIMER_BH is used. I will see if
I can make another smptimers patch that gets rid of them.

Thanks
Dipankar
-- 
Dipankar Sarma  <dipankar@in.ibm.com> http://lse.sourceforge.net
Linux Technology Center, IBM Software Lab, Bangalore, India.