Re: [PATCH] 2.5.67-ac2 direct-IO for IDE taskfile ioctl (0/4)

[Bartlomiej Zolnierkiewicz]

On Wed, 23 Apr 2003, Andrew Morton wrote:

> Andries Brouwer <[e-mail]> wrote:
> >
> > On Wed, Apr 23, 2003 at 03:35:00PM -0700, Andrew Morton wrote:
> >
> > > What is special about the IDE ioctl approach?
> >
> > Usually one wants to use the standard commands for I/O.
> > But if the purpose is to talk to the drive (set password,
> > set native max, eject, change ZIP drive from big floppy
> > mode to removable disk mode, etc. etc.) then one needs
> > a means to execute IDE commands "by hand".
>
> Yes, but none of these are performance-critical and they don't involve
> large amnounts of data.  A copy is OK.

low-level benchmarking is somehow performance critical :-)

note that direct-IO here is _only_ bonus, main goal is to
use the same code for _all_ ioctl and fs generated IDE IO
(remember that all fs IDE IO is rq->bio based)

and this is work-in-progress...

> If all the rework against bio_map_user() and friends is needed for other
> reasons then fine.  But it doesn't seem to be needed for the IDE taskfile
> ioctl.

Rework of bio_map_user() and co. is minimal and not needed but otherwise
I will have to duplicate same code in 4 places. bio_map_user() now does
allocated bio checking and grabs extra reference to bio which all users of
old bio_map_user() have to do anyway (and they will probably forgot to).