On Sun, Jan 14, 2007 at 05:19:38PM +0900, YAMAMOTO Takashi wrote:

> > On Sat, Jan 13, 2007 at 09:11:06PM +0900, YAMAMOTO Takashi wrote:
> > 
> > > > I'm thinking something like:
> > > > 
> > > > 	callout_init(&c);
> > > > 	callout_setwork(&c, func, arg);
> > > > 	callout_schedule(&c, timeout);
> > > 
> > > i implemented it.  see the attached patch.
> > > 
> > > once callout_setwork() is called on a callout,
> > > callout_schedule() is only legal operation on the callout.
> > > (i'm not a fan of making primitives have rich functionalities...)
> > 
> > I would prefer not to have a seperate callout_setwork(), but instead
> > a flag to callout_init that can be ignored if we later do interrupts
> > as threads:
> > 
> > 	callout_init(&foo, CALLOUT_THREAD);
> > 
> > We will need to add a CALLOUT_MPSAFE flag at some point.
> > 
> > Andrew
> 
> what will CALLOUT_MPSAFE do, if softclock() is called with kernel_lock held?

We will need to change it so that the kernel lock isn't taken until
softclock() - which also means having a flag for softintr_establish().

Andrew