diff options
Diffstat (limited to 'drivers/md/raid0.c')
-rw-r--r-- | drivers/md/raid0.c | 405 |
1 files changed, 212 insertions, 193 deletions
diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index c08d755..ab4a489 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -26,8 +26,8 @@ static void raid0_unplug(struct request_queue *q) { mddev_t *mddev = q->queuedata; - raid0_conf_t *conf = mddev_to_conf(mddev); - mdk_rdev_t **devlist = conf->strip_zone[0].dev; + raid0_conf_t *conf = mddev->private; + mdk_rdev_t **devlist = conf->devlist; int i; for (i=0; i<mddev->raid_disks; i++) { @@ -40,8 +40,8 @@ static void raid0_unplug(struct request_queue *q) static int raid0_congested(void *data, int bits) { mddev_t *mddev = data; - raid0_conf_t *conf = mddev_to_conf(mddev); - mdk_rdev_t **devlist = conf->strip_zone[0].dev; + raid0_conf_t *conf = mddev->private; + mdk_rdev_t **devlist = conf->devlist; int i, ret = 0; for (i = 0; i < mddev->raid_disks && !ret ; i++) { @@ -52,27 +52,60 @@ static int raid0_congested(void *data, int bits) return ret; } +/* + * inform the user of the raid configuration +*/ +static void dump_zones(mddev_t *mddev) +{ + int j, k, h; + sector_t zone_size = 0; + sector_t zone_start = 0; + char b[BDEVNAME_SIZE]; + raid0_conf_t *conf = mddev->private; + printk(KERN_INFO "******* %s configuration *********\n", + mdname(mddev)); + h = 0; + for (j = 0; j < conf->nr_strip_zones; j++) { + printk(KERN_INFO "zone%d=[", j); + for (k = 0; k < conf->strip_zone[j].nb_dev; k++) + printk("%s/", + bdevname(conf->devlist[j*mddev->raid_disks + + k]->bdev, b)); + printk("]\n"); + + zone_size = conf->strip_zone[j].zone_end - zone_start; + printk(KERN_INFO " zone offset=%llukb " + "device offset=%llukb size=%llukb\n", + (unsigned long long)zone_start>>1, + (unsigned long long)conf->strip_zone[j].dev_start>>1, + (unsigned long long)zone_size>>1); + zone_start = conf->strip_zone[j].zone_end; + } + printk(KERN_INFO "**********************************\n\n"); +} -static int create_strip_zones (mddev_t *mddev) +static int create_strip_zones(mddev_t *mddev) { - int i, c, j; - sector_t current_start, curr_zone_start; - sector_t min_spacing; - raid0_conf_t *conf = mddev_to_conf(mddev); - mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev; + int i, c, j, err; + sector_t curr_zone_end, sectors; + mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev, **dev; struct strip_zone *zone; int cnt; char b[BDEVNAME_SIZE]; - - /* - * The number of 'same size groups' - */ - conf->nr_strip_zones = 0; - + raid0_conf_t *conf = kzalloc(sizeof(*conf), GFP_KERNEL); + + if (!conf) + return -ENOMEM; list_for_each_entry(rdev1, &mddev->disks, same_set) { printk(KERN_INFO "raid0: looking at %s\n", bdevname(rdev1->bdev,b)); c = 0; + + /* round size to chunk_size */ + sectors = rdev1->sectors; + sector_div(sectors, mddev->chunk_sectors); + rdev1->sectors = sectors * mddev->chunk_sectors; + list_for_each_entry(rdev2, &mddev->disks, same_set) { printk(KERN_INFO "raid0: comparing %s(%llu)", bdevname(rdev1->bdev,b), @@ -103,16 +136,16 @@ static int create_strip_zones (mddev_t *mddev) } } printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones); - + err = -ENOMEM; conf->strip_zone = kzalloc(sizeof(struct strip_zone)* conf->nr_strip_zones, GFP_KERNEL); if (!conf->strip_zone) - return 1; + goto abort; conf->devlist = kzalloc(sizeof(mdk_rdev_t*)* conf->nr_strip_zones*mddev->raid_disks, GFP_KERNEL); if (!conf->devlist) - return 1; + goto abort; /* The first zone must contain all devices, so here we check that * there is a proper alignment of slots to devices and find them all @@ -120,7 +153,8 @@ static int create_strip_zones (mddev_t *mddev) zone = &conf->strip_zone[0]; cnt = 0; smallest = NULL; - zone->dev = conf->devlist; + dev = conf->devlist; + err = -EINVAL; list_for_each_entry(rdev1, &mddev->disks, same_set) { int j = rdev1->raid_disk; @@ -129,12 +163,12 @@ static int create_strip_zones (mddev_t *mddev) "aborting!\n", j); goto abort; } - if (zone->dev[j]) { + if (dev[j]) { printk(KERN_ERR "raid0: multiple devices for %d - " "aborting!\n", j); goto abort; } - zone->dev[j] = rdev1; + dev[j] = rdev1; blk_queue_stack_limits(mddev->queue, rdev1->bdev->bd_disk->queue); @@ -144,7 +178,7 @@ static int create_strip_zones (mddev_t *mddev) */ if (rdev1->bdev->bd_disk->queue->merge_bvec_fn && - mddev->queue->max_sectors > (PAGE_SIZE>>9)) + queue_max_sectors(mddev->queue) > (PAGE_SIZE>>9)) blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9); if (!smallest || (rdev1->sectors < smallest->sectors)) @@ -157,34 +191,32 @@ static int create_strip_zones (mddev_t *mddev) goto abort; } zone->nb_dev = cnt; - zone->sectors = smallest->sectors * cnt; - zone->zone_start = 0; + zone->zone_end = smallest->sectors * cnt; - current_start = smallest->sectors; - curr_zone_start = zone->sectors; + curr_zone_end = zone->zone_end; /* now do the other zones */ for (i = 1; i < conf->nr_strip_zones; i++) { zone = conf->strip_zone + i; - zone->dev = conf->strip_zone[i-1].dev + mddev->raid_disks; + dev = conf->devlist + i * mddev->raid_disks; printk(KERN_INFO "raid0: zone %d\n", i); - zone->dev_start = current_start; + zone->dev_start = smallest->sectors; smallest = NULL; c = 0; for (j=0; j<cnt; j++) { char b[BDEVNAME_SIZE]; - rdev = conf->strip_zone[0].dev[j]; + rdev = conf->devlist[j]; printk(KERN_INFO "raid0: checking %s ...", bdevname(rdev->bdev, b)); - if (rdev->sectors <= current_start) { + if (rdev->sectors <= zone->dev_start) { printk(KERN_INFO " nope.\n"); continue; } printk(KERN_INFO " contained as device %d\n", c); - zone->dev[c] = rdev; + dev[c] = rdev; c++; if (!smallest || rdev->sectors < smallest->sectors) { smallest = rdev; @@ -194,47 +226,39 @@ static int create_strip_zones (mddev_t *mddev) } zone->nb_dev = c; - zone->sectors = (smallest->sectors - current_start) * c; + sectors = (smallest->sectors - zone->dev_start) * c; printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n", - zone->nb_dev, (unsigned long long)zone->sectors); + zone->nb_dev, (unsigned long long)sectors); - zone->zone_start = curr_zone_start; - curr_zone_start += zone->sectors; + curr_zone_end += sectors; + zone->zone_end = curr_zone_end; - current_start = smallest->sectors; printk(KERN_INFO "raid0: current zone start: %llu\n", - (unsigned long long)current_start); - } - - /* Now find appropriate hash spacing. - * We want a number which causes most hash entries to cover - * at most two strips, but the hash table must be at most - * 1 PAGE. We choose the smallest strip, or contiguous collection - * of strips, that has big enough size. We never consider the last - * strip though as it's size has no bearing on the efficacy of the hash - * table. - */ - conf->spacing = curr_zone_start; - min_spacing = curr_zone_start; - sector_div(min_spacing, PAGE_SIZE/sizeof(struct strip_zone*)); - for (i=0; i < conf->nr_strip_zones-1; i++) { - sector_t s = 0; - for (j = i; j < conf->nr_strip_zones - 1 && - s < min_spacing; j++) - s += conf->strip_zone[j].sectors; - if (s >= min_spacing && s < conf->spacing) - conf->spacing = s; + (unsigned long long)smallest->sectors); } - mddev->queue->unplug_fn = raid0_unplug; - mddev->queue->backing_dev_info.congested_fn = raid0_congested; mddev->queue->backing_dev_info.congested_data = mddev; + /* + * now since we have the hard sector sizes, we can make sure + * chunk size is a multiple of that sector size + */ + if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) { + printk(KERN_ERR "%s chunk_size of %d not valid\n", + mdname(mddev), + mddev->chunk_sectors << 9); + goto abort; + } printk(KERN_INFO "raid0: done.\n"); + mddev->private = conf; return 0; - abort: - return 1; +abort: + kfree(conf->strip_zone); + kfree(conf->devlist); + kfree(conf); + mddev->private = NULL; + return err; } /** @@ -252,10 +276,15 @@ static int raid0_mergeable_bvec(struct request_queue *q, mddev_t *mddev = q->queuedata; sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); int max; - unsigned int chunk_sectors = mddev->chunk_size >> 9; + unsigned int chunk_sectors = mddev->chunk_sectors; unsigned int bio_sectors = bvm->bi_size >> 9; - max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9; + if (is_power_of_2(chunk_sectors)) + max = (chunk_sectors - ((sector & (chunk_sectors-1)) + + bio_sectors)) << 9; + else + max = (chunk_sectors - (sector_div(sector, chunk_sectors) + + bio_sectors)) << 9; if (max < 0) max = 0; /* bio_add cannot handle a negative return */ if (max <= biovec->bv_len && bio_sectors == 0) return biovec->bv_len; @@ -277,84 +306,28 @@ static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks) return array_sectors; } -static int raid0_run (mddev_t *mddev) +static int raid0_run(mddev_t *mddev) { - unsigned cur=0, i=0, nb_zone; - s64 sectors; - raid0_conf_t *conf; + int ret; - if (mddev->chunk_size == 0) { - printk(KERN_ERR "md/raid0: non-zero chunk size required.\n"); + if (mddev->chunk_sectors == 0) { + printk(KERN_ERR "md/raid0: chunk size must be set.\n"); return -EINVAL; } - printk(KERN_INFO "%s: setting max_sectors to %d, segment boundary to %d\n", - mdname(mddev), - mddev->chunk_size >> 9, - (mddev->chunk_size>>1)-1); - blk_queue_max_sectors(mddev->queue, mddev->chunk_size >> 9); - blk_queue_segment_boundary(mddev->queue, (mddev->chunk_size>>1) - 1); + if (md_check_no_bitmap(mddev)) + return -EINVAL; + blk_queue_max_sectors(mddev->queue, mddev->chunk_sectors); mddev->queue->queue_lock = &mddev->queue->__queue_lock; - conf = kmalloc(sizeof (raid0_conf_t), GFP_KERNEL); - if (!conf) - goto out; - mddev->private = (void *)conf; - - conf->strip_zone = NULL; - conf->devlist = NULL; - if (create_strip_zones (mddev)) - goto out_free_conf; + ret = create_strip_zones(mddev); + if (ret < 0) + return ret; /* calculate array device size */ md_set_array_sectors(mddev, raid0_size(mddev, 0, 0)); printk(KERN_INFO "raid0 : md_size is %llu sectors.\n", (unsigned long long)mddev->array_sectors); - printk(KERN_INFO "raid0 : conf->spacing is %llu sectors.\n", - (unsigned long long)conf->spacing); - { - sector_t s = raid0_size(mddev, 0, 0); - sector_t space = conf->spacing; - int round; - conf->sector_shift = 0; - if (sizeof(sector_t) > sizeof(u32)) { - /*shift down space and s so that sector_div will work */ - while (space > (sector_t) (~(u32)0)) { - s >>= 1; - space >>= 1; - s += 1; /* force round-up */ - conf->sector_shift++; - } - } - round = sector_div(s, (u32)space) ? 1 : 0; - nb_zone = s + round; - } - printk(KERN_INFO "raid0 : nb_zone is %d.\n", nb_zone); - - printk(KERN_INFO "raid0 : Allocating %zu bytes for hash.\n", - nb_zone*sizeof(struct strip_zone*)); - conf->hash_table = kmalloc (sizeof (struct strip_zone *)*nb_zone, GFP_KERNEL); - if (!conf->hash_table) - goto out_free_conf; - sectors = conf->strip_zone[cur].sectors; - - conf->hash_table[0] = conf->strip_zone + cur; - for (i=1; i< nb_zone; i++) { - while (sectors <= conf->spacing) { - cur++; - sectors += conf->strip_zone[cur].sectors; - } - sectors -= conf->spacing; - conf->hash_table[i] = conf->strip_zone + cur; - } - if (conf->sector_shift) { - conf->spacing >>= conf->sector_shift; - /* round spacing up so when we divide by it, we - * err on the side of too-low, which is safest - */ - conf->spacing++; - } - /* calculate the max read-ahead size. * For read-ahead of large files to be effective, we need to * readahead at least twice a whole stripe. i.e. number of devices @@ -365,48 +338,107 @@ static int raid0_run (mddev_t *mddev) * chunksize should be used in that case. */ { - int stripe = mddev->raid_disks * mddev->chunk_size / PAGE_SIZE; + int stripe = mddev->raid_disks * + (mddev->chunk_sectors << 9) / PAGE_SIZE; if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) mddev->queue->backing_dev_info.ra_pages = 2* stripe; } - blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec); + dump_zones(mddev); return 0; +} -out_free_conf: +static int raid0_stop(mddev_t *mddev) +{ + raid0_conf_t *conf = mddev->private; + + blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ kfree(conf->strip_zone); kfree(conf->devlist); kfree(conf); mddev->private = NULL; -out: - return -ENOMEM; + return 0; } -static int raid0_stop (mddev_t *mddev) +/* Find the zone which holds a particular offset + * Update *sectorp to be an offset in that zone + */ +static struct strip_zone *find_zone(struct raid0_private_data *conf, + sector_t *sectorp) { - raid0_conf_t *conf = mddev_to_conf(mddev); + int i; + struct strip_zone *z = conf->strip_zone; + sector_t sector = *sectorp; + + for (i = 0; i < conf->nr_strip_zones; i++) + if (sector < z[i].zone_end) { + if (i) + *sectorp = sector - z[i-1].zone_end; + return z + i; + } + BUG(); +} - blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ - kfree(conf->hash_table); - conf->hash_table = NULL; - kfree(conf->strip_zone); - conf->strip_zone = NULL; - kfree(conf); - mddev->private = NULL; +/* + * remaps the bio to the target device. we separate two flows. + * power 2 flow and a general flow for the sake of perfromance +*/ +static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone, + sector_t sector, sector_t *sector_offset) +{ + unsigned int sect_in_chunk; + sector_t chunk; + raid0_conf_t *conf = mddev->private; + unsigned int chunk_sects = mddev->chunk_sectors; + + if (is_power_of_2(chunk_sects)) { + int chunksect_bits = ffz(~chunk_sects); + /* find the sector offset inside the chunk */ + sect_in_chunk = sector & (chunk_sects - 1); + sector >>= chunksect_bits; + /* chunk in zone */ + chunk = *sector_offset; + /* quotient is the chunk in real device*/ + sector_div(chunk, zone->nb_dev << chunksect_bits); + } else{ + sect_in_chunk = sector_div(sector, chunk_sects); + chunk = *sector_offset; + sector_div(chunk, chunk_sects * zone->nb_dev); + } + /* + * position the bio over the real device + * real sector = chunk in device + starting of zone + * + the position in the chunk + */ + *sector_offset = (chunk * chunk_sects) + sect_in_chunk; + return conf->devlist[(zone - conf->strip_zone)*mddev->raid_disks + + sector_div(sector, zone->nb_dev)]; +} - return 0; +/* + * Is io distribute over 1 or more chunks ? +*/ +static inline int is_io_in_chunk_boundary(mddev_t *mddev, + unsigned int chunk_sects, struct bio *bio) +{ + if (likely(is_power_of_2(chunk_sects))) { + return chunk_sects >= ((bio->bi_sector & (chunk_sects-1)) + + (bio->bi_size >> 9)); + } else{ + sector_t sector = bio->bi_sector; + return chunk_sects >= (sector_div(sector, chunk_sects) + + (bio->bi_size >> 9)); + } } -static int raid0_make_request (struct request_queue *q, struct bio *bio) +static int raid0_make_request(struct request_queue *q, struct bio *bio) { mddev_t *mddev = q->queuedata; - unsigned int sect_in_chunk, chunksect_bits, chunk_sects; - raid0_conf_t *conf = mddev_to_conf(mddev); + unsigned int chunk_sects; + sector_t sector_offset; struct strip_zone *zone; mdk_rdev_t *tmp_dev; - sector_t chunk; - sector_t sector, rsect; const int rw = bio_data_dir(bio); int cpu; @@ -421,11 +453,9 @@ static int raid0_make_request (struct request_queue *q, struct bio *bio) bio_sectors(bio)); part_stat_unlock(); - chunk_sects = mddev->chunk_size >> 9; - chunksect_bits = ffz(~chunk_sects); - sector = bio->bi_sector; - - if (unlikely(chunk_sects < (bio->bi_sector & (chunk_sects - 1)) + (bio->bi_size >> 9))) { + chunk_sects = mddev->chunk_sectors; + if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) { + sector_t sector = bio->bi_sector; struct bio_pair *bp; /* Sanity check -- queue functions should prevent this happening */ if (bio->bi_vcnt != 1 || @@ -434,7 +464,12 @@ static int raid0_make_request (struct request_queue *q, struct bio *bio) /* This is a one page bio that upper layers * refuse to split for us, so we need to split it. */ - bp = bio_split(bio, chunk_sects - (bio->bi_sector & (chunk_sects - 1))); + if (likely(is_power_of_2(chunk_sects))) + bp = bio_split(bio, chunk_sects - (sector & + (chunk_sects-1))); + else + bp = bio_split(bio, chunk_sects - + sector_div(sector, chunk_sects)); if (raid0_make_request(q, &bp->bio1)) generic_make_request(&bp->bio1); if (raid0_make_request(q, &bp->bio2)) @@ -443,34 +478,14 @@ static int raid0_make_request (struct request_queue *q, struct bio *bio) bio_pair_release(bp); return 0; } - - - { - sector_t x = sector >> conf->sector_shift; - sector_div(x, (u32)conf->spacing); - zone = conf->hash_table[x]; - } - while (sector >= zone->zone_start + zone->sectors) - zone++; - - sect_in_chunk = bio->bi_sector & (chunk_sects - 1); - - - { - sector_t x = (sector - zone->zone_start) >> chunksect_bits; - - sector_div(x, zone->nb_dev); - chunk = x; - - x = sector >> chunksect_bits; - tmp_dev = zone->dev[sector_div(x, zone->nb_dev)]; - } - rsect = (chunk << chunksect_bits) + zone->dev_start + sect_in_chunk; - + sector_offset = bio->bi_sector; + zone = find_zone(mddev->private, §or_offset); + tmp_dev = map_sector(mddev, zone, bio->bi_sector, + §or_offset); bio->bi_bdev = tmp_dev->bdev; - bio->bi_sector = rsect + tmp_dev->data_offset; - + bio->bi_sector = sector_offset + zone->dev_start + + tmp_dev->data_offset; /* * Let the main block layer submit the IO and resolve recursion: */ @@ -485,31 +500,35 @@ bad_map: return 0; } -static void raid0_status (struct seq_file *seq, mddev_t *mddev) +static void raid0_status(struct seq_file *seq, mddev_t *mddev) { #undef MD_DEBUG #ifdef MD_DEBUG int j, k, h; char b[BDEVNAME_SIZE]; - raid0_conf_t *conf = mddev_to_conf(mddev); + raid0_conf_t *conf = mddev->private; + sector_t zone_size; + sector_t zone_start = 0; h = 0; + for (j = 0; j < conf->nr_strip_zones; j++) { seq_printf(seq, " z%d", j); - if (conf->hash_table[h] == conf->strip_zone+j) - seq_printf(seq, "(h%d)", h++); seq_printf(seq, "=["); for (k = 0; k < conf->strip_zone[j].nb_dev; k++) seq_printf(seq, "%s/", bdevname( - conf->strip_zone[j].dev[k]->bdev,b)); - - seq_printf(seq, "] zs=%d ds=%d s=%d\n", - conf->strip_zone[j].zone_start, - conf->strip_zone[j].dev_start, - conf->strip_zone[j].sectors); + conf->devlist[j*mddev->raid_disks + k] + ->bdev, b)); + + zone_size = conf->strip_zone[j].zone_end - zone_start; + seq_printf(seq, "] ze=%lld ds=%lld s=%lld\n", + (unsigned long long)zone_start>>1, + (unsigned long long)conf->strip_zone[j].dev_start>>1, + (unsigned long long)zone_size>>1); + zone_start = conf->strip_zone[j].zone_end; } #endif - seq_printf(seq, " %dk chunks", mddev->chunk_size/1024); + seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2); return; } |