Odd code in sd_init()

[Alexander Viro]

sd_init() contains the following loop:

        for (k = 0; k < N_USED_SD_MAJORS; k++) {
                request_queue_t *q = blk_get_queue(mk_kdev(SD_MAJOR(k), 0));
                blk_queue_hardsect_size(q, 512);
        }

... which is a damn interesting thing to do, seeing that it's called when
we do scsi_register_device(&sd_template) and before we get any chance to
set ->queue, so we actually end up setting sector size for hell knows
what (well, for default queues of our majors - i.e. stuff that won't be
used).

Either I'm missing something, or that code is bogus.  Jens, Linus?