PERFORCE change 126835 for review

[John Baldwin]

On Thursday 27 September 2007 03:18:10 pm Kip Macy wrote:
> It is in there, it just doesn't seem to do the trick :-/.

Note that there was a recent bug fix in HEAD that fixed a race where if you 
were using a callout that rescheduled itself and didn't do an explicit 
callout_stop() while holding the lock but just did a callout_drain(), then 
the callout could remain scheduled.  The latest commit to kern_timeout.c 
should fix this though.

>      -Kip
> 
> On 9/26/07, John Baldwin <jhb at freebsd.org> wrote:
> > On Wednesday 26 September 2007 04:40:08 pm Kip Macy wrote:
> > > http://perforce.freebsd.org/chv.cgi?CH=126835
> > >
> > > Change 126835 by kmacy at kmacy_home:ethng on 2007/09/26 20:40:03
> > >
> > >       increase delay in shutdown to avoid destroying the port lock
> > >       before the last call to the tick handler
> > >
> > > Affected files ...
> > >
> > > .. //depot/projects/ethng/src/sys/dev/cxgb/cxgb_main.c#21 edit
> > >
> > > Differences ...
> > >
> > > ==== //depot/projects/ethng/src/sys/dev/cxgb/cxgb_main.c#21 (text+ko) 
====
> > >
> > > @@ -1023,7 +1023,7 @@
> > >
> > >       ether_ifdetach(p->ifp);
> > >       printf("waiting for callout to stop ...");
> > > -     DELAY(10000);
> > > +     DELAY(100000);
> > >       printf("done\n");
> >
> > Umm, what about callout_drain()?  That's its purpose.
> >
> > --
> > John Baldwin
> >
> 



-- 
John Baldwin