Re: END GAME (Re: Linux 2.5.3-pre1-aia1)

Jens Axboe (axboe@suse.de)
Wed, 23 Jan 2002 09:55:15 +0100


On Tue, Jan 22 2002, Andre Hedrick wrote:
>
> Linus, Jens,
>
> I need a function that performs the kmapping to return a pointer with all
> the data needed for that transaction of the data phase, and will cross
> pages correctly, and may cross more than 2 pages at a time in PIO.
> I do not care how you do.
>
> char * majic_voodoo_mapping (
> struct request *rq,
> int nsect,
> unsigned long *flags)
> {
> char * buffer_walk = ide_map_rq(rq, &flags);
> nsect -= ide_rq_offset(rq);
> do {
> buffer_walk += get_some_more(rq, nsect);
> } while (nsect)
> return buffer_walk;
> }
>
> This should solve all the problems in the data-phases and let the driver
> run correctly. The result is on each "get_some_more" will all BLOCK/BIO to
> return the partial competions of at least one page
>
> The function would behave like ide_end_request but only to adjust the
> buffer in process, and make block/bio deal with munging it back to the top
> layers on the partial completions, it will not stop the data IO process of
> the ATOMIC command in process.

That _is_ what ide_end_request is doing, it is not stopping any data I/O
in progress. It is also atomic. What exactly do you think is missing?

> Better yet, I will shut my PIEHOLE!

Oh, that'll be the day. Lets just say I'm not running to get my
calendar and the big fat marker. :-)

-- 
Jens Axboe

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/