On Wed, 6 May 2015 23:34:17 -0700 Ming Lin <m...@kernel.org> wrote:
> If a read request fits entirely in a chunk, it will be passed directly to the
> underlying device (providing it hasn't failed of course). If it doesn't fit,
> the slightly less efficient path that uses the stripe_cache is used.
> Requests that get to the stripe cache are always completely split up as
> necessary.
>
> So with RAID5, ripping out the merge_bvec_fn doesn't cause it to stop work,
> but could cause it to take the less efficient path more often.
>
> All that is needed to manage this is for 'chunk_aligned_read' do some bio
> splitting, much like the RAID0 code does.
>
> Cc: Neil Brown <ne...@suse.de>
> Cc: linux-r...@vger.kernel.org
> Signed-off-by: Ming Lin <m...@kernel.org>
> ---
> drivers/md/raid5.c | 42 +++++++++++++++++++++++++++++++++++++-----
> 1 file changed, 37 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
> index 7f4a717..b18f548 100644
> --- a/drivers/md/raid5.c
> +++ b/drivers/md/raid5.c
> @@ -4738,7 +4738,7 @@ static void raid5_align_endio(struct bio *bi, int error)
> add_bio_to_retry(raid_bi, conf);
> }
>
> -static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
> +static int raid5_read_one_chunk(struct mddev *mddev, struct bio *raid_bio)
> {
> struct r5conf *conf = mddev->private;
> int dd_idx;
> @@ -4747,7 +4747,7 @@ static int chunk_aligned_read(struct mddev *mddev,
> struct bio * raid_bio)
> sector_t end_sector;
>
> if (!in_chunk_boundary(mddev, raid_bio)) {
> - pr_debug("chunk_aligned_read : non aligned\n");
> + pr_debug("%s: non aligned\n", __func__);
> return 0;
> }
> /*
> @@ -4822,6 +4822,36 @@ static int chunk_aligned_read(struct mddev *mddev,
> struct bio * raid_bio)
> }
> }
>
> +static struct bio *chunk_aligned_read(struct mddev *mddev, struct bio
> *raid_bio)
> +{
> + struct bio *split;
> +
> + do {
> + sector_t sector = raid_bio->bi_iter.bi_sector;
> + unsigned chunk_sects = mddev->chunk_sectors;
> + unsigned sectors;
> +
> + if (likely(is_power_of_2(chunk_sects)))
> + sectors = chunk_sects - (sector & (chunk_sects-1));
> + else
> + sectors = chunk_sects - sector_div(sector, chunk_sects);
RAID5 doesn't currently allow non-power-of-2 chunks. So this test is
pointless, but not really harmful. Maybe someday we will.
I'm equally happy for it to stay or go.
Acked-by: NeilBrown <ne...@suse.de>
Thanks,
NeilBrown
> +
> + if (sectors < bio_sectors(raid_bio)) {
> + split = bio_split(raid_bio, sectors, GFP_NOIO,
> fs_bio_set);
> + bio_chain(split, raid_bio);
> + } else
> + split = raid_bio;
> +
> + if (!raid5_read_one_chunk(mddev, split)) {
> + if (split != raid_bio)
> + generic_make_request(raid_bio);
> + return split;
> + }
> + } while (split != raid_bio);
> +
> + return NULL;
> +}
> +
> /* __get_priority_stripe - get the next stripe to process
> *
> * Full stripe writes are allowed to pass preread active stripes up until
> @@ -5099,9 +5129,11 @@ static void make_request(struct mddev *mddev, struct
> bio * bi)
> * data on failed drives.
> */
> if (rw == READ && mddev->degraded == 0 &&
> - mddev->reshape_position == MaxSector &&
> - chunk_aligned_read(mddev,bi))
> - return;
> + mddev->reshape_position == MaxSector) {
> + bi = chunk_aligned_read(mddev, bi);
> + if (!bi)
> + return;
> + }
>
> if (unlikely(bi->bi_rw & REQ_DISCARD)) {
> make_discard_request(mddev, bi);
pgprw2fVOBeD3.pgp
Description: OpenPGP digital signature
