Re: [PATCH] IDE Power Management, try 2

[Jens Axboe]

On Thu, Jun 05 2003, Bartlomiej Zolnierkiewicz wrote:
> > Jens, Bart, what do you think ? Should I add pm_step & pm_state to
> > struct request ? Do the "extended taskfile structure" thing ? Or just
> > keep things like they are in this new patch and forget about carrying
> > the PM state value ?
> 
> I think extending struct request is the way to go,
> pm_step & pm_state or even pointer to rq_pm_struct.

Agree

> > I also added another rq->flags bit for requests forced at the head of
> > the queue with ide_preempt. This is typically for sense requests done
> > by ide-cd (though I also spotted a user in the tcq stuff). I need that
> > to make sure that if such a request ever happens to be pushed in front
> > of the current PM request (with the drive->blocked flag already set),
> > we don't enter an endless loop, fetching that new request and dropping
> > it right away because we only accept PM requests from the queue once
> > the drive is suspended.
> 
> Jens, I think generic version of ide_do_drive_cmd() would be useful for
> other block devices, what do you think?

Yes very much so, scsi_ioctl also basically implements part of this
functionality.

-- 
Jens Axboe