Re: [PATCH] IDE Power Management, try 2

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Thu, 2003-06-05 at 16:27, Benjamin Herrenschmidt wrote:
> > > I had to add yet another rq->flags bit for that, and I think that sucks
> > 
> > You don't have if you use additional, default pm_state (on == 0).
> > This sucks too, but a bit less.
> 
> Can you elaborate ? I'm not sure I understand what you meant

Forget it, my brain finally got a clue ;) Though I don't like the
solution. Adding pm_step & pm_state to struct request or a ptr to
rq_pm_struct seem the way to go to me, though I'm not sure which
of these 2 solution is the best, struct request is already a can
of worms imho... ;) If we ever need more PM fields in there, then
the pointer may be the best solution, but right know, I can't think
of any reason to add more stuffs

> > I think extending struct request is the way to go,
> > pm_step & pm_state or even pointer to rq_pm_struct.
> 
> Ok. I'll wait for Jens ack and go that way if he agrees.
> 
> Ben.
-- 
Benjamin Herrenschmidt <benh@kernel.crashing.org>