On Mon, Jan 05 2004, Bartlomiej Zolnierkiewicz wrote:

On Monday 05 of January 2004 17:49, Jens Axboe wrote:

quoted

On Mon, Jan 05 2004, Bartlomiej Zolnierkiewicz wrote:

quoted

quoted

calling end_request with a null sector count, ide_end_request will then
take hard_nr_sectors which will end the whole request even if only one
bio was finished, huh? Am I missing something here?

No, it is used mainly to fail requests.

This hack should be later removed with care
(there is some strange comment about locking).

IIRC, it's due to it not always being safe to inspect rq state outside
of ide_lock. So that makes 0 a magic value that just means 'end the
first chunk' for ide_end_request().

Why/when it is not safe to do?

You would need to read hwgroup->rq.

-- 
Jens Axboe