diff options
Diffstat (limited to 'drivers/md/raid5.c')
| -rw-r--r-- | drivers/md/raid5.c | 50 |
1 files changed, 45 insertions, 5 deletions
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 7b820b81d8c2..4739ed891e75 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -6079,6 +6079,38 @@ out_release: return ret; } +/* + * If the bio covers multiple data disks, find sector within the bio that has + * the lowest chunk offset in the first chunk. + */ +static sector_t raid5_bio_lowest_chunk_sector(struct r5conf *conf, + struct bio *bi) +{ + int sectors_per_chunk = conf->chunk_sectors; + int raid_disks = conf->raid_disks; + int dd_idx; + struct stripe_head sh; + unsigned int chunk_offset; + sector_t r_sector = bi->bi_iter.bi_sector & ~((sector_t)RAID5_STRIPE_SECTORS(conf)-1); + sector_t sector; + + /* We pass in fake stripe_head to get back parity disk numbers */ + sector = raid5_compute_sector(conf, r_sector, 0, &dd_idx, &sh); + chunk_offset = sector_div(sector, sectors_per_chunk); + if (sectors_per_chunk - chunk_offset >= bio_sectors(bi)) + return r_sector; + /* + * Bio crosses to the next data disk. Check whether it's in the same + * chunk. + */ + dd_idx++; + while (dd_idx == sh.pd_idx || dd_idx == sh.qd_idx) + dd_idx++; + if (dd_idx >= raid_disks) + return r_sector; + return r_sector + sectors_per_chunk - chunk_offset; +} + static bool raid5_make_request(struct mddev *mddev, struct bio * bi) { DEFINE_WAIT_FUNC(wait, woken_wake_function); @@ -6150,6 +6182,17 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi) } md_account_bio(mddev, &bi); + /* + * Lets start with the stripe with the lowest chunk offset in the first + * chunk. That has the best chances of creating IOs adjacent to + * previous IOs in case of sequential IO and thus creates the most + * sequential IO pattern. We don't bother with the optimization when + * reshaping as the performance benefit is not worth the complexity. + */ + if (likely(conf->reshape_progress == MaxSector)) + logical_sector = raid5_bio_lowest_chunk_sector(conf, bi); + s = (logical_sector - ctx.first_sector) >> RAID5_STRIPE_SHIFT(conf); + add_wait_queue(&conf->wait_for_overlap, &wait); while (1) { res = make_stripe_request(mddev, conf, &ctx, logical_sector, @@ -6178,7 +6221,7 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi) continue; } - s = find_first_bit(ctx.sectors_to_do, stripe_cnt); + s = find_next_bit_wrap(ctx.sectors_to_do, stripe_cnt, s); if (s == stripe_cnt) break; @@ -7716,7 +7759,6 @@ static void raid5_set_io_opt(struct r5conf *conf) static int raid5_run(struct mddev *mddev) { struct r5conf *conf; - int working_disks = 0; int dirty_parity_disks = 0; struct md_rdev *rdev; struct md_rdev *journal_dev = NULL; @@ -7912,10 +7954,8 @@ static int raid5_run(struct mddev *mddev) pr_warn("md: cannot handle concurrent replacement and reshape.\n"); goto abort; } - if (test_bit(In_sync, &rdev->flags)) { - working_disks++; + if (test_bit(In_sync, &rdev->flags)) continue; - } /* This disc is not fully in-sync. However if it * just stored parity (beyond the recovery_offset), * when we don't need to be concerned about the |