[deliverable/linux.git] / drivers / md / raid0.c

/*
   raid0.c : Multiple Devices driver for Linux
             Copyright (C) 1994-96 Marc ZYNGIER
	     <zyngier@ufr-info-p7.ibp.fr> or
	     <maz@gloups.fdn.fr>
             Copyright (C) 1999, 2000 Ingo Molnar, Red Hat


   RAID-0 management functions.

   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2, or (at your option)
   any later version.
   
   You should have received a copy of the GNU General Public License
   (for example /usr/src/linux/COPYING); if not, write to the Free
   Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  
*/

#include <linux/blkdev.h>
#include <linux/seq_file.h>
#include "md.h"
#include "raid0.h"

static void raid0_unplug(struct request_queue *q)
{
	mddev_t *mddev = q->queuedata;
	raid0_conf_t *conf = mddev->private;
	mdk_rdev_t **devlist = conf->devlist;
	int i;

	for (i=0; i<mddev->raid_disks; i++) {
		struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev);

		blk_unplug(r_queue);
	}
}

static int raid0_congested(void *data, int bits)
{
	mddev_t *mddev = data;
	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++) {
		struct request_queue *q = bdev_get_queue(devlist[i]->bdev);

		ret |= bdi_congested(&q->backing_dev_info, 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)
{
	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];
	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),
			       (unsigned long long)rdev1->sectors);
			printk(KERN_INFO " with %s(%llu)\n",
			       bdevname(rdev2->bdev,b),
			       (unsigned long long)rdev2->sectors);
			if (rdev2 == rdev1) {
				printk(KERN_INFO "raid0:   END\n");
				break;
			}
			if (rdev2->sectors == rdev1->sectors) {
				/*
				 * Not unique, don't count it as a new
				 * group
				 */
				printk(KERN_INFO "raid0:   EQUAL\n");
				c = 1;
				break;
			}
			printk(KERN_INFO "raid0:   NOT EQUAL\n");
		}
		if (!c) {
			printk(KERN_INFO "raid0:   ==> UNIQUE\n");
			conf->nr_strip_zones++;
			printk(KERN_INFO "raid0: %d zones\n",
				conf->nr_strip_zones);
		}
	}
	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)
		goto abort;
	conf->devlist = kzalloc(sizeof(mdk_rdev_t*)*
				conf->nr_strip_zones*mddev->raid_disks,
				GFP_KERNEL);
	if (!conf->devlist)
		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
	 */
	zone = &conf->strip_zone[0];
	cnt = 0;
	smallest = NULL;
	dev = conf->devlist;
	err = -EINVAL;
	list_for_each_entry(rdev1, &mddev->disks, same_set) {
		int j = rdev1->raid_disk;

		if (j < 0 || j >= mddev->raid_disks) {
			printk(KERN_ERR "raid0: bad disk number %d - "
				"aborting!\n", j);
			goto abort;
		}
		if (dev[j]) {
			printk(KERN_ERR "raid0: multiple devices for %d - "
				"aborting!\n", j);
			goto abort;
		}
		dev[j] = rdev1;

		blk_queue_stack_limits(mddev->queue,
				       rdev1->bdev->bd_disk->queue);
		/* as we don't honour merge_bvec_fn, we must never risk
		 * violating it, so limit ->max_sector to one PAGE, as
		 * a one page request is never in violation.
		 */

		if (rdev1->bdev->bd_disk->queue->merge_bvec_fn &&
		    queue_max_sectors(mddev->queue) > (PAGE_SIZE>>9))
			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);

		if (!smallest || (rdev1->sectors < smallest->sectors))
			smallest = rdev1;
		cnt++;
	}
	if (cnt != mddev->raid_disks) {
		printk(KERN_ERR "raid0: too few disks (%d of %d) - "
			"aborting!\n", cnt, mddev->raid_disks);
		goto abort;
	}
	zone->nb_dev = cnt;
	zone->zone_end = smallest->sectors * cnt;

	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;
		dev = conf->devlist + i * mddev->raid_disks;

		printk(KERN_INFO "raid0: zone %d\n", i);
		zone->dev_start = smallest->sectors;
		smallest = NULL;
		c = 0;

		for (j=0; j<cnt; j++) {
			char b[BDEVNAME_SIZE];
			rdev = conf->devlist[j];
			printk(KERN_INFO "raid0: checking %s ...",
				bdevname(rdev->bdev, b));
			if (rdev->sectors <= zone->dev_start) {
				printk(KERN_INFO " nope.\n");
				continue;
			}
			printk(KERN_INFO " contained as device %d\n", c);
			dev[c] = rdev;
			c++;
			if (!smallest || rdev->sectors < smallest->sectors) {
				smallest = rdev;
				printk(KERN_INFO "  (%llu) is smallest!.\n",
					(unsigned long long)rdev->sectors);
			}
		}

		zone->nb_dev = 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)sectors);

		curr_zone_end += sectors;
		zone->zone_end = curr_zone_end;

		printk(KERN_INFO "raid0: current zone start: %llu\n",
			(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:
	kfree(conf->strip_zone);
	kfree(conf->devlist);
	kfree(conf);
	mddev->private = NULL;
	return err;
}

/**
 *	raid0_mergeable_bvec -- tell bio layer if a two requests can be merged
 *	@q: request queue
 *	@bvm: properties of new bio
 *	@biovec: the request that could be merged to it.
 *
 *	Return amount of bytes we can accept at this offset
 */
static int raid0_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
{
	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_sectors;
	unsigned int bio_sectors = bvm->bi_size >> 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;
	else 
		return max;
}

static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
	sector_t array_sectors = 0;
	mdk_rdev_t *rdev;

	WARN_ONCE(sectors || raid_disks,
		  "%s does not support generic reshape\n", __func__);

	list_for_each_entry(rdev, &mddev->disks, same_set)
		array_sectors += rdev->sectors;

	return array_sectors;
}

static int raid0_run(mddev_t *mddev)
{
	int ret;

	if (mddev->chunk_sectors == 0) {
		printk(KERN_ERR "md/raid0: chunk size must be set.\n");
		return -EINVAL;
	}
	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;

	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);
	/* 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
	 * multiplied by chunk size times 2.
	 * If an individual device has an ra_pages greater than the
	 * chunk size, then we will not drive that device as hard as it
	 * wants.  We consider this a configuration error: a larger
	 * chunksize should be used in that case.
	 */
	{
		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;
}

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;
	return 0;
}

/* 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)
{
	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();
}

/*
 * 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)];
}

/*
 * 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)
{
	mddev_t *mddev = q->queuedata;
	unsigned int chunk_sects;
	sector_t sector_offset;
	struct strip_zone *zone;
	mdk_rdev_t *tmp_dev;
	const int rw = bio_data_dir(bio);
	int cpu;

	if (unlikely(bio_barrier(bio))) {
		bio_endio(bio, -EOPNOTSUPP);
		return 0;
	}

	cpu = part_stat_lock();
	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
		      bio_sectors(bio));
	part_stat_unlock();

	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 ||
		    bio->bi_idx != 0)
			goto bad_map;
		/* This is a one page bio that upper layers
		 * refuse to split for us, so we need to split it.
		 */
		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))
			generic_make_request(&bp->bio2);

		bio_pair_release(bp);
		return 0;
	}

	sector_offset = bio->bi_sector;
	zone =  find_zone(mddev->private, &sector_offset);
	tmp_dev = map_sector(mddev, zone, bio->bi_sector,
			     &sector_offset);
	bio->bi_bdev = tmp_dev->bdev;
	bio->bi_sector = sector_offset + zone->dev_start +
		tmp_dev->data_offset;
	/*
	 * Let the main block layer submit the IO and resolve recursion:
	 */
	return 1;

bad_map:
	printk("raid0_make_request bug: can't convert block across chunks"
		" or bigger than %dk %llu %d\n", chunk_sects / 2,
		(unsigned long long)bio->bi_sector, bio->bi_size >> 10);

	bio_io_error(bio);
	return 0;
}

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->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);
		seq_printf(seq, "=[");
		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
			seq_printf(seq, "%s/", bdevname(
				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_sectors / 2);
	return;
}

static struct mdk_personality raid0_personality=
{
	.name		= "raid0",
	.level		= 0,
	.owner		= THIS_MODULE,
	.make_request	= raid0_make_request,
	.run		= raid0_run,
	.stop		= raid0_stop,
	.status		= raid0_status,
	.size		= raid0_size,
};

static int __init raid0_init (void)
{
	return register_md_personality (&raid0_personality);
}

static void raid0_exit (void)
{
	unregister_md_personality (&raid0_personality);
}

module_init(raid0_init);
module_exit(raid0_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-2"); /* RAID0 */
MODULE_ALIAS("md-raid0");
MODULE_ALIAS("md-level-0");
Commit	Line	Data
1da177e4 LT	1	/*
	2	raid0.c : Multiple Devices driver for Linux
	3	Copyright (C) 1994-96 Marc ZYNGIER
	4	<zyngier@ufr-info-p7.ibp.fr> or
	5	<maz@gloups.fdn.fr>
	6	Copyright (C) 1999, 2000 Ingo Molnar, Red Hat
	7
	8
	9	RAID-0 management functions.
	10
	11	This program is free software; you can redistribute it and/or modify
	12	it under the terms of the GNU General Public License as published by
	13	the Free Software Foundation; either version 2, or (at your option)
	14	any later version.
	15
	16	You should have received a copy of the GNU General Public License
	17	(for example /usr/src/linux/COPYING); if not, write to the Free
	18	Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
	19	*/
	20
bff61975	21	#include <linux/blkdev.h>
bff61975	22	#include <linux/seq_file.h>
43b2e5d8	23	#include "md.h"
ef740c37	24	#include "raid0.h"
1da177e4	25
165125e1	26	static void raid0_unplug(struct request_queue *q)
1da177e4 LT	27	{
1da177e4 LT	28	mddev_t *mddev = q->queuedata;
070ec55d	29	raid0_conf_t *conf = mddev->private;
b414579f	30	mdk_rdev_t **devlist = conf->devlist;
1da177e4 LT	31	int i;
	32
	33	for (i=0; i<mddev->raid_disks; i++) {
165125e1	34	struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev);
1da177e4	35
2ad8b1ef	36	blk_unplug(r_queue);
1da177e4 LT	37	}
	38	}
	39
26be34dc N	40	static int raid0_congested(void *data, int bits)
	41	{
	42	mddev_t *mddev = data;
070ec55d	43	raid0_conf_t *conf = mddev->private;
b414579f	44	mdk_rdev_t **devlist = conf->devlist;
26be34dc N	45	int i, ret = 0;
	46
	47	for (i = 0; i < mddev->raid_disks && !ret ; i++) {
165125e1	48	struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
26be34dc N	49
	50	ret \|= bdi_congested(&q->backing_dev_info, bits);
	51	}
	52	return ret;
	53	}
	54
46994191	55	/*
	56	* inform the user of the raid configuration
	57	*/
	58	static void dump_zones(mddev_t *mddev)
	59	{
	60	int j, k, h;
	61	sector_t zone_size = 0;
	62	sector_t zone_start = 0;
	63	char b[BDEVNAME_SIZE];
	64	raid0_conf_t *conf = mddev->private;
	65	printk(KERN_INFO "***** %s configuration *******\n",
	66	mdname(mddev));
	67	h = 0;
	68	for (j = 0; j < conf->nr_strip_zones; j++) {
	69	printk(KERN_INFO "zone%d=[", j);
	70	for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
	71	printk("%s/",
	72	bdevname(conf->devlist[j*mddev->raid_disks
	73	+ k]->bdev, b));
	74	printk("]\n");
	75
	76	zone_size = conf->strip_zone[j].zone_end - zone_start;
	77	printk(KERN_INFO " zone offset=%llukb "
	78	"device offset=%llukb size=%llukb\n",
	79	(unsigned long long)zone_start>>1,
	80	(unsigned long long)conf->strip_zone[j].dev_start>>1,
	81	(unsigned long long)zone_size>>1);
	82	zone_start = conf->strip_zone[j].zone_end;
	83	}
	84	printk(KERN_INFO "**********************************\n\n");
	85	}
	86
ed7b0038	87	static int create_strip_zones(mddev_t *mddev)
1da177e4	88	{
ed7b0038	89	int i, c, j, err;
49f357a2	90	sector_t curr_zone_end, sectors;
b414579f	91	mdk_rdev_t smallest, rdev1, rdev2, rdev, **dev;
1da177e4 LT	92	struct strip_zone *zone;
	93	int cnt;
	94	char b[BDEVNAME_SIZE];
ed7b0038 AN	95	raid0_conf_t conf = kzalloc(sizeof(conf), GFP_KERNEL);
	96
	97	if (!conf)
	98	return -ENOMEM;
159ec1fc	99	list_for_each_entry(rdev1, &mddev->disks, same_set) {
0825b87a	100	printk(KERN_INFO "raid0: looking at %s\n",
1da177e4 LT	101	bdevname(rdev1->bdev,b));
1da177e4 LT	102	c = 0;
13f2682b N	103
	104	/* round size to chunk_size */
	105	sectors = rdev1->sectors;
	106	sector_div(sectors, mddev->chunk_sectors);
	107	rdev1->sectors = sectors * mddev->chunk_sectors;
	108
159ec1fc	109	list_for_each_entry(rdev2, &mddev->disks, same_set) {
0825b87a	110	printk(KERN_INFO "raid0: comparing %s(%llu)",
1da177e4	111	bdevname(rdev1->bdev,b),
dd8ac336	112	(unsigned long long)rdev1->sectors);
0825b87a	113	printk(KERN_INFO " with %s(%llu)\n",
1da177e4	114	bdevname(rdev2->bdev,b),
dd8ac336	115	(unsigned long long)rdev2->sectors);
1da177e4	116	if (rdev2 == rdev1) {
0825b87a	117	printk(KERN_INFO "raid0: END\n");
1da177e4 LT	118	break;
1da177e4 LT	119	}
dd8ac336	120	if (rdev2->sectors == rdev1->sectors) {
1da177e4 LT	121	/*
	122	* Not unique, don't count it as a new
	123	* group
	124	*/
0825b87a	125	printk(KERN_INFO "raid0: EQUAL\n");
1da177e4 LT	126	c = 1;
	127	break;
	128	}
0825b87a	129	printk(KERN_INFO "raid0: NOT EQUAL\n");
1da177e4 LT	130	}
1da177e4 LT	131	if (!c) {
0825b87a	132	printk(KERN_INFO "raid0: ==> UNIQUE\n");
1da177e4	133	conf->nr_strip_zones++;
0825b87a AN	134	printk(KERN_INFO "raid0: %d zones\n",
0825b87a AN	135	conf->nr_strip_zones);
1da177e4 LT	136	}
1da177e4 LT	137	}
0825b87a	138	printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones);
ed7b0038	139	err = -ENOMEM;
9ffae0cf	140	conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
1da177e4 LT	141	conf->nr_strip_zones, GFP_KERNEL);
1da177e4 LT	142	if (!conf->strip_zone)
ed7b0038	143	goto abort;
9ffae0cf	144	conf->devlist = kzalloc(sizeof(mdk_rdev_t)
1da177e4 LT	145	conf->nr_strip_zones*mddev->raid_disks,
	146	GFP_KERNEL);
	147	if (!conf->devlist)
ed7b0038	148	goto abort;
1da177e4	149
1da177e4 LT	150	/* The first zone must contain all devices, so here we check that
	151	* there is a proper alignment of slots to devices and find them all
	152	*/
	153	zone = &conf->strip_zone[0];
	154	cnt = 0;
	155	smallest = NULL;
b414579f	156	dev = conf->devlist;
ed7b0038	157	err = -EINVAL;
159ec1fc	158	list_for_each_entry(rdev1, &mddev->disks, same_set) {
1da177e4 LT	159	int j = rdev1->raid_disk;
	160
	161	if (j < 0 \|\| j >= mddev->raid_disks) {
0825b87a AN	162	printk(KERN_ERR "raid0: bad disk number %d - "
0825b87a AN	163	"aborting!\n", j);
1da177e4 LT	164	goto abort;
1da177e4 LT	165	}
b414579f	166	if (dev[j]) {
0825b87a AN	167	printk(KERN_ERR "raid0: multiple devices for %d - "
0825b87a AN	168	"aborting!\n", j);
1da177e4 LT	169	goto abort;
1da177e4 LT	170	}
b414579f	171	dev[j] = rdev1;
1da177e4 LT	172
	173	blk_queue_stack_limits(mddev->queue,
	174	rdev1->bdev->bd_disk->queue);
	175	/* as we don't honour merge_bvec_fn, we must never risk
	176	* violating it, so limit ->max_sector to one PAGE, as
	177	* a one page request is never in violation.
	178	*/
	179
	180	if (rdev1->bdev->bd_disk->queue->merge_bvec_fn &&
ae03bf63	181	queue_max_sectors(mddev->queue) > (PAGE_SIZE>>9))
1da177e4 LT	182	blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
1da177e4 LT	183
dd8ac336	184	if (!smallest \|\| (rdev1->sectors < smallest->sectors))
1da177e4 LT	185	smallest = rdev1;
	186	cnt++;
	187	}
	188	if (cnt != mddev->raid_disks) {
0825b87a AN	189	printk(KERN_ERR "raid0: too few disks (%d of %d) - "
0825b87a AN	190	"aborting!\n", cnt, mddev->raid_disks);
1da177e4 LT	191	goto abort;
	192	}
	193	zone->nb_dev = cnt;
49f357a2	194	zone->zone_end = smallest->sectors * cnt;
1da177e4	195
49f357a2	196	curr_zone_end = zone->zone_end;
1da177e4 LT	197
	198	/* now do the other zones */
	199	for (i = 1; i < conf->nr_strip_zones; i++)
	200	{
	201	zone = conf->strip_zone + i;
b414579f	202	dev = conf->devlist + i * mddev->raid_disks;
1da177e4	203
0825b87a	204	printk(KERN_INFO "raid0: zone %d\n", i);
d27a43ab	205	zone->dev_start = smallest->sectors;
1da177e4 LT	206	smallest = NULL;
	207	c = 0;
	208
	209	for (j=0; j<cnt; j++) {
	210	char b[BDEVNAME_SIZE];
b414579f	211	rdev = conf->devlist[j];
0825b87a AN	212	printk(KERN_INFO "raid0: checking %s ...",
0825b87a AN	213	bdevname(rdev->bdev, b));
d27a43ab	214	if (rdev->sectors <= zone->dev_start) {
0825b87a	215	printk(KERN_INFO " nope.\n");
dd8ac336 AN	216	continue;
	217	}
	218	printk(KERN_INFO " contained as device %d\n", c);
b414579f	219	dev[c] = rdev;
dd8ac336 AN	220	c++;
	221	if (!smallest \|\| rdev->sectors < smallest->sectors) {
	222	smallest = rdev;
	223	printk(KERN_INFO " (%llu) is smallest!.\n",
	224	(unsigned long long)rdev->sectors);
	225	}
1da177e4 LT	226	}
	227
	228	zone->nb_dev = c;
49f357a2	229	sectors = (smallest->sectors - zone->dev_start) * c;
83838ed8	230	printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n",
49f357a2	231	zone->nb_dev, (unsigned long long)sectors);
1da177e4	232
49f357a2	233	curr_zone_end += sectors;
d27a43ab	234	zone->zone_end = curr_zone_end;
1da177e4	235
6b8796cc	236	printk(KERN_INFO "raid0: current zone start: %llu\n",
d27a43ab	237	(unsigned long long)smallest->sectors);
1da177e4	238	}
1da177e4	239	mddev->queue->unplug_fn = raid0_unplug;
26be34dc N	240	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
26be34dc N	241	mddev->queue->backing_dev_info.congested_data = mddev;
1da177e4	242
92e59b6b	243	/*
	244	* now since we have the hard sector sizes, we can make sure
	245	* chunk size is a multiple of that sector size
	246	*/
9d8f0363	247	if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) {
92e59b6b	248	printk(KERN_ERR "%s chunk_size of %d not valid\n",
92e59b6b	249	mdname(mddev),
9d8f0363	250	mddev->chunk_sectors << 9);
92e59b6b	251	goto abort;
92e59b6b	252	}
0825b87a	253	printk(KERN_INFO "raid0: done.\n");
ed7b0038	254	mddev->private = conf;
1da177e4	255	return 0;
5568a603	256	abort:
ed7b0038 AN	257	kfree(conf->strip_zone);
	258	kfree(conf->devlist);
	259	kfree(conf);
	260	mddev->private = NULL;
	261	return err;
1da177e4 LT	262	}
	263
	264	/**
	265	* raid0_mergeable_bvec -- tell bio layer if a two requests can be merged
	266	* @q: request queue
cc371e66	267	* @bvm: properties of new bio
1da177e4 LT	268	* @biovec: the request that could be merged to it.
	269	*
	270	* Return amount of bytes we can accept at this offset
	271	*/
cc371e66 AK	272	static int raid0_mergeable_bvec(struct request_queue *q,
	273	struct bvec_merge_data *bvm,
	274	struct bio_vec *biovec)
1da177e4 LT	275	{
1da177e4 LT	276	mddev_t *mddev = q->queuedata;
cc371e66	277	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
1da177e4	278	int max;
9d8f0363	279	unsigned int chunk_sectors = mddev->chunk_sectors;
cc371e66	280	unsigned int bio_sectors = bvm->bi_size >> 9;
1da177e4	281
d6e412ea	282	if (is_power_of_2(chunk_sectors))
fbb704ef	283	max = (chunk_sectors - ((sector & (chunk_sectors-1))
	284	+ bio_sectors)) << 9;
	285	else
	286	max = (chunk_sectors - (sector_div(sector, chunk_sectors)
	287	+ bio_sectors)) << 9;
1da177e4 LT	288	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
	289	if (max <= biovec->bv_len && bio_sectors == 0)
	290	return biovec->bv_len;
	291	else
	292	return max;
	293	}
	294
80c3a6ce DW	295	static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks)
	296	{
	297	sector_t array_sectors = 0;
	298	mdk_rdev_t *rdev;
	299
	300	WARN_ONCE(sectors \|\| raid_disks,
	301	"%s does not support generic reshape\n", __func__);
	302
	303	list_for_each_entry(rdev, &mddev->disks, same_set)
	304	array_sectors += rdev->sectors;
	305
	306	return array_sectors;
	307	}
	308
8f79cfcd	309	static int raid0_run(mddev_t *mddev)
1da177e4	310	{
5568a603	311	int ret;
1da177e4	312
9d8f0363	313	if (mddev->chunk_sectors == 0) {
fbb704ef	314	printk(KERN_ERR "md/raid0: chunk size must be set.\n");
2604b703 N	315	return -EINVAL;
2604b703 N	316	}
0894cc30 AN	317	if (md_check_no_bitmap(mddev))
0894cc30 AN	318	return -EINVAL;
9d8f0363	319	blk_queue_max_sectors(mddev->queue, mddev->chunk_sectors);
e7e72bf6	320	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
1da177e4	321
5568a603 AN	322	ret = create_strip_zones(mddev);
5568a603 AN	323	if (ret < 0)
ed7b0038	324	return ret;
1da177e4 LT	325
1da177e4 LT	326	/* calculate array device size */
1f403624	327	md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
1da177e4	328
ccacc7d2 AN	329	printk(KERN_INFO "raid0 : md_size is %llu sectors.\n",
ccacc7d2 AN	330	(unsigned long long)mddev->array_sectors);
1da177e4 LT	331	/* calculate the max read-ahead size.
	332	* For read-ahead of large files to be effective, we need to
	333	* readahead at least twice a whole stripe. i.e. number of devices
	334	* multiplied by chunk size times 2.
	335	* If an individual device has an ra_pages greater than the
	336	* chunk size, then we will not drive that device as hard as it
	337	* wants. We consider this a configuration error: a larger
	338	* chunksize should be used in that case.
	339	*/
	340	{
9d8f0363 AN	341	int stripe = mddev->raid_disks *
9d8f0363 AN	342	(mddev->chunk_sectors << 9) / PAGE_SIZE;
1da177e4 LT	343	if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
	344	mddev->queue->backing_dev_info.ra_pages = 2* stripe;
	345	}
	346
1da177e4	347	blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec);
46994191	348	dump_zones(mddev);
1da177e4	349	return 0;
1da177e4 LT	350	}
1da177e4 LT	351
fb5ab4b5	352	static int raid0_stop(mddev_t *mddev)
1da177e4	353	{
070ec55d	354	raid0_conf_t *conf = mddev->private;
1da177e4 LT	355
1da177e4 LT	356	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
990a8baf	357	kfree(conf->strip_zone);
fb5ab4b5	358	kfree(conf->devlist);
990a8baf	359	kfree(conf);
1da177e4	360	mddev->private = NULL;
1da177e4 LT	361	return 0;
	362	}
	363
49f357a2 N	364	/* Find the zone which holds a particular offset
	365	* Update *sectorp to be an offset in that zone
	366	*/
dc582663	367	static struct strip_zone find_zone(struct raid0_private_data conf,
49f357a2	368	sector_t *sectorp)
dc582663 AN	369	{
	370	int i;
	371	struct strip_zone *z = conf->strip_zone;
49f357a2	372	sector_t sector = *sectorp;
dc582663 AN	373
dc582663 AN	374	for (i = 0; i < conf->nr_strip_zones; i++)
49f357a2 N	375	if (sector < z[i].zone_end) {
	376	if (i)
	377	*sectorp = sector - z[i-1].zone_end;
dc582663	378	return z + i;
49f357a2	379	}
dc582663 AN	380	BUG();
	381	}
	382
fbb704ef	383	/*
	384	* remaps the bio to the target device. we separate two flows.
	385	* power 2 flow and a general flow for the sake of perfromance
	386	*/
	387	static mdk_rdev_t map_sector(mddev_t mddev, struct strip_zone *zone,
	388	sector_t sector, sector_t *sector_offset)
1da177e4	389	{
fbb704ef	390	unsigned int sect_in_chunk;
fbb704ef	391	sector_t chunk;
070ec55d	392	raid0_conf_t *conf = mddev->private;
9d8f0363	393	unsigned int chunk_sects = mddev->chunk_sectors;
fbb704ef	394
d6e412ea	395	if (is_power_of_2(chunk_sects)) {
fbb704ef	396	int chunksect_bits = ffz(~chunk_sects);
	397	/* find the sector offset inside the chunk */
	398	sect_in_chunk = sector & (chunk_sects - 1);
	399	sector >>= chunksect_bits;
	400	/* chunk in zone */
	401	chunk = *sector_offset;
	402	/* quotient is the chunk in real device*/
	403	sector_div(chunk, zone->nb_dev << chunksect_bits);
	404	} else{
	405	sect_in_chunk = sector_div(sector, chunk_sects);
	406	chunk = *sector_offset;
	407	sector_div(chunk, chunk_sects * zone->nb_dev);
	408	}
	409	/*
	410	* position the bio over the real device
	411	* real sector = chunk in device + starting of zone
	412	* + the position in the chunk
	413	*/
	414	sector_offset = (chunk chunk_sects) + sect_in_chunk;
	415	return conf->devlist[(zone - conf->strip_zone)*mddev->raid_disks
	416	+ sector_div(sector, zone->nb_dev)];
	417	}
	418
	419	/*
	420	* Is io distribute over 1 or more chunks ?
	421	*/
	422	static inline int is_io_in_chunk_boundary(mddev_t *mddev,
	423	unsigned int chunk_sects, struct bio *bio)
	424	{
d6e412ea	425	if (likely(is_power_of_2(chunk_sects))) {
fbb704ef	426	return chunk_sects >= ((bio->bi_sector & (chunk_sects-1))
	427	+ (bio->bi_size >> 9));
	428	} else{
	429	sector_t sector = bio->bi_sector;
	430	return chunk_sects >= (sector_div(sector, chunk_sects)
	431	+ (bio->bi_size >> 9));
	432	}
	433	}
	434
	435	static int raid0_make_request(struct request_queue q, struct bio bio)
	436	{
	437	mddev_t *mddev = q->queuedata;
	438	unsigned int chunk_sects;
	439	sector_t sector_offset;
1da177e4 LT	440	struct strip_zone *zone;
1da177e4 LT	441	mdk_rdev_t *tmp_dev;
a362357b	442	const int rw = bio_data_dir(bio);
c9959059	443	int cpu;
1da177e4	444
e5dcdd80	445	if (unlikely(bio_barrier(bio))) {
6712ecf8	446	bio_endio(bio, -EOPNOTSUPP);
e5dcdd80 N	447	return 0;
	448	}
	449
074a7aca TH	450	cpu = part_stat_lock();
	451	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	452	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
	453	bio_sectors(bio));
	454	part_stat_unlock();
1da177e4	455
9d8f0363	456	chunk_sects = mddev->chunk_sectors;
fbb704ef	457	if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) {
fbb704ef	458	sector_t sector = bio->bi_sector;
1da177e4 LT	459	struct bio_pair *bp;
	460	/* Sanity check -- queue functions should prevent this happening */
	461	if (bio->bi_vcnt != 1 \|\|
	462	bio->bi_idx != 0)
	463	goto bad_map;
	464	/* This is a one page bio that upper layers
	465	* refuse to split for us, so we need to split it.
	466	*/
d6e412ea	467	if (likely(is_power_of_2(chunk_sects)))
fbb704ef	468	bp = bio_split(bio, chunk_sects - (sector &
	469	(chunk_sects-1)));
	470	else
	471	bp = bio_split(bio, chunk_sects -
	472	sector_div(sector, chunk_sects));
1da177e4 LT	473	if (raid0_make_request(q, &bp->bio1))
	474	generic_make_request(&bp->bio1);
	475	if (raid0_make_request(q, &bp->bio2))
	476	generic_make_request(&bp->bio2);
	477
	478	bio_pair_release(bp);
	479	return 0;
	480	}
1da177e4	481
fbb704ef	482	sector_offset = bio->bi_sector;
	483	zone = find_zone(mddev->private, &sector_offset);
	484	tmp_dev = map_sector(mddev, zone, bio->bi_sector,
	485	&sector_offset);
1da177e4	486	bio->bi_bdev = tmp_dev->bdev;
fbb704ef	487	bio->bi_sector = sector_offset + zone->dev_start +
fbb704ef	488	tmp_dev->data_offset;
1da177e4 LT	489	/*
	490	* Let the main block layer submit the IO and resolve recursion:
	491	*/
	492	return 1;
	493
	494	bad_map:
	495	printk("raid0_make_request bug: can't convert block across chunks"
a4712005	496	" or bigger than %dk %llu %d\n", chunk_sects / 2,
1da177e4 LT	497	(unsigned long long)bio->bi_sector, bio->bi_size >> 10);
1da177e4 LT	498
6712ecf8	499	bio_io_error(bio);
1da177e4 LT	500	return 0;
1da177e4 LT	501	}
8299d7f7	502
1b961429	503	static void raid0_status(struct seq_file seq, mddev_t mddev)
1da177e4 LT	504	{
	505	#undef MD_DEBUG
	506	#ifdef MD_DEBUG
	507	int j, k, h;
	508	char b[BDEVNAME_SIZE];
070ec55d	509	raid0_conf_t *conf = mddev->private;
8299d7f7	510
1b961429	511	sector_t zone_size;
1b961429	512	sector_t zone_start = 0;
1da177e4	513	h = 0;
1b961429	514
1da177e4 LT	515	for (j = 0; j < conf->nr_strip_zones; j++) {
1da177e4 LT	516	seq_printf(seq, " z%d", j);
1da177e4 LT	517	seq_printf(seq, "=[");
1da177e4 LT	518	for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
8299d7f7	519	seq_printf(seq, "%s/", bdevname(
1b961429	520	conf->devlist[j*mddev->raid_disks + k]
	521	->bdev, b));
	522
	523	zone_size = conf->strip_zone[j].zone_end - zone_start;
	524	seq_printf(seq, "] ze=%lld ds=%lld s=%lld\n",
	525	(unsigned long long)zone_start>>1,
	526	(unsigned long long)conf->strip_zone[j].dev_start>>1,
	527	(unsigned long long)zone_size>>1);
	528	zone_start = conf->strip_zone[j].zone_end;
1da177e4 LT	529	}
1da177e4 LT	530	#endif
9d8f0363	531	seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2);
1da177e4 LT	532	return;
	533	}
	534
2604b703	535	static struct mdk_personality raid0_personality=
1da177e4 LT	536	{
1da177e4 LT	537	.name = "raid0",
2604b703	538	.level = 0,
1da177e4 LT	539	.owner = THIS_MODULE,
	540	.make_request = raid0_make_request,
	541	.run = raid0_run,
	542	.stop = raid0_stop,
	543	.status = raid0_status,
80c3a6ce	544	.size = raid0_size,
1da177e4 LT	545	};
	546
	547	static int __init raid0_init (void)
	548	{
2604b703	549	return register_md_personality (&raid0_personality);
1da177e4 LT	550	}
	551
	552	static void raid0_exit (void)
	553	{
2604b703	554	unregister_md_personality (&raid0_personality);
1da177e4 LT	555	}
	556
	557	module_init(raid0_init);
	558	module_exit(raid0_exit);
	559	MODULE_LICENSE("GPL");
	560	MODULE_ALIAS("md-personality-2"); /* RAID0 */
d9d166c2	561	MODULE_ALIAS("md-raid0");
2604b703	562	MODULE_ALIAS("md-level-0");