Messages in this thread | | | Date | Wed, 30 Aug 2000 18:59:22 +0200 | From | Jens Axboe <> | Subject | Re: [PATCH] Fix disk statistic reporting to include all disks |
| |
On Wed, Aug 30 2000, Bill Wilson wrote: > Won't we loose some important info by not looking at minors? hda and > hdb are minors of IDE0_MAJOR, etc. Is this data distinguishable in the > queues?
hda and hdb each have their own queue, so we don't loose anything there.
> > No big deal, for now I just have the proc code doing blk_get_queue for > > every major. That code is hardly performance critical ;-) > > True for the kernel, but there are some user land apps that read /proc/stat > quite frequently (10 per second) so some performance tuning might be > worth it. There could be a linked list field in the blk_dev[] array > which links all entries that have had statistics collected. Then the /proc > code does not have to blk_get_queue on every major.
Right now the code in my tree looks something like this:
for (i = 0; i < MAX_BLKDEV; i++) { request_queue_t *q = blk_get_queue(MKDEV(i, 0)); if (!q) continue; if (!active) continue;
/* print stats */ }
So it's not really that bad. I suppose things could be made quicker by just browsing a list of active queues, but I'm not sure it's worth the trouble (but hey, proove me wrong). Note that linking blk_dev does not make much sense, since not all queues are available from there. Drivers are free to use their own queue, even several if they want.
-- * Jens Axboe <axboe@suse.de> * SuSE Labs - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/
| |