[linux-2.6-block.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_to_conf(mddev);
	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
	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_to_conf(mddev);
	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
	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;
}

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;
	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;
		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;
	zone->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 (zone->dev[j]) {
			printk(KERN_ERR "raid0: multiple devices for %d - "
				"aborting!\n", j);
			goto abort;
		}
		zone->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;
		zone->dev = conf->strip_zone[i-1].dev + 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->strip_zone[0].dev[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);
			zone->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;

	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_size >> 9;
	unsigned int bio_sectors = bvm->bi_size >> 9;

	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + 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_size == 0) {
		printk(KERN_ERR "md/raid0: non-zero chunk size required.\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);
	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_size / 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);
	return 0;
}

static int raid0_stop(mddev_t *mddev)
{
	raid0_conf_t *conf = mddev_to_conf(mddev);

	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();
}

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);
	struct strip_zone *zone;
	mdk_rdev_t *tmp_dev;
	sector_t chunk;
	sector_t sector, rsect, sector_offset;
	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_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))) {
		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.
		 */
		bp = bio_split(bio, chunk_sects - (bio->bi_sector & (chunk_sects - 1)));
		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 = sector;
	zone = find_zone(conf, &sector_offset);
	sect_in_chunk = bio->bi_sector & (chunk_sects - 1);
	{
		sector_t x = sector_offset >> 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;
 
	bio->bi_bdev = tmp_dev->bdev;
	bio->bi_sector = rsect + 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_to_conf(mddev);

	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->strip_zone[j].dev[k]->bdev,b));

		seq_printf(seq, "] ze=%d ds=%d s=%d\n",
				conf->strip_zone[j].zone_end,
				conf->strip_zone[j].dev_start,
				conf->strip_zone[j].sectors);
	}
#endif
	seq_printf(seq, " %dk chunks", mddev->chunk_size/1024);
	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	{
	28	mddev_t *mddev = q->queuedata;
	29	raid0_conf_t *conf = mddev_to_conf(mddev);
	30	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
	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;
	43	raid0_conf_t *conf = mddev_to_conf(mddev);
	44	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
	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
ed7b0038	55	static int create_strip_zones(mddev_t *mddev)
1da177e4	56	{
ed7b0038	57	int i, c, j, err;
49f357a2	58	sector_t curr_zone_end, sectors;
1da177e4	59	mdk_rdev_t smallest, rdev1, rdev2, rdev;
1da177e4 LT	60	struct strip_zone *zone;
	61	int cnt;
	62	char b[BDEVNAME_SIZE];
ed7b0038 AN	63	raid0_conf_t conf = kzalloc(sizeof(conf), GFP_KERNEL);
	64
	65	if (!conf)
	66	return -ENOMEM;
159ec1fc	67	list_for_each_entry(rdev1, &mddev->disks, same_set) {
0825b87a	68	printk(KERN_INFO "raid0: looking at %s\n",
1da177e4 LT	69	bdevname(rdev1->bdev,b));
1da177e4 LT	70	c = 0;
159ec1fc	71	list_for_each_entry(rdev2, &mddev->disks, same_set) {
0825b87a	72	printk(KERN_INFO "raid0: comparing %s(%llu)",
1da177e4	73	bdevname(rdev1->bdev,b),
dd8ac336	74	(unsigned long long)rdev1->sectors);
0825b87a	75	printk(KERN_INFO " with %s(%llu)\n",
1da177e4	76	bdevname(rdev2->bdev,b),
dd8ac336	77	(unsigned long long)rdev2->sectors);
1da177e4	78	if (rdev2 == rdev1) {
0825b87a	79	printk(KERN_INFO "raid0: END\n");
1da177e4 LT	80	break;
1da177e4 LT	81	}
dd8ac336	82	if (rdev2->sectors == rdev1->sectors) {
1da177e4 LT	83	/*
	84	* Not unique, don't count it as a new
	85	* group
	86	*/
0825b87a	87	printk(KERN_INFO "raid0: EQUAL\n");
1da177e4 LT	88	c = 1;
	89	break;
	90	}
0825b87a	91	printk(KERN_INFO "raid0: NOT EQUAL\n");
1da177e4 LT	92	}
1da177e4 LT	93	if (!c) {
0825b87a	94	printk(KERN_INFO "raid0: ==> UNIQUE\n");
1da177e4	95	conf->nr_strip_zones++;
0825b87a AN	96	printk(KERN_INFO "raid0: %d zones\n",
0825b87a AN	97	conf->nr_strip_zones);
1da177e4 LT	98	}
1da177e4 LT	99	}
0825b87a	100	printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones);
ed7b0038	101	err = -ENOMEM;
9ffae0cf	102	conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
1da177e4 LT	103	conf->nr_strip_zones, GFP_KERNEL);
1da177e4 LT	104	if (!conf->strip_zone)
ed7b0038	105	goto abort;
9ffae0cf	106	conf->devlist = kzalloc(sizeof(mdk_rdev_t)
1da177e4 LT	107	conf->nr_strip_zones*mddev->raid_disks,
	108	GFP_KERNEL);
	109	if (!conf->devlist)
ed7b0038	110	goto abort;
1da177e4	111
1da177e4 LT	112	/* The first zone must contain all devices, so here we check that
	113	* there is a proper alignment of slots to devices and find them all
	114	*/
	115	zone = &conf->strip_zone[0];
	116	cnt = 0;
	117	smallest = NULL;
	118	zone->dev = conf->devlist;
ed7b0038	119	err = -EINVAL;
159ec1fc	120	list_for_each_entry(rdev1, &mddev->disks, same_set) {
1da177e4 LT	121	int j = rdev1->raid_disk;
	122
	123	if (j < 0 \|\| j >= mddev->raid_disks) {
0825b87a AN	124	printk(KERN_ERR "raid0: bad disk number %d - "
0825b87a AN	125	"aborting!\n", j);
1da177e4 LT	126	goto abort;
	127	}
	128	if (zone->dev[j]) {
0825b87a AN	129	printk(KERN_ERR "raid0: multiple devices for %d - "
0825b87a AN	130	"aborting!\n", j);
1da177e4 LT	131	goto abort;
	132	}
	133	zone->dev[j] = rdev1;
	134
	135	blk_queue_stack_limits(mddev->queue,
	136	rdev1->bdev->bd_disk->queue);
	137	/* as we don't honour merge_bvec_fn, we must never risk
	138	* violating it, so limit ->max_sector to one PAGE, as
	139	* a one page request is never in violation.
	140	*/
	141
	142	if (rdev1->bdev->bd_disk->queue->merge_bvec_fn &&
ae03bf63	143	queue_max_sectors(mddev->queue) > (PAGE_SIZE>>9))
1da177e4 LT	144	blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
1da177e4 LT	145
dd8ac336	146	if (!smallest \|\| (rdev1->sectors < smallest->sectors))
1da177e4 LT	147	smallest = rdev1;
	148	cnt++;
	149	}
	150	if (cnt != mddev->raid_disks) {
0825b87a AN	151	printk(KERN_ERR "raid0: too few disks (%d of %d) - "
0825b87a AN	152	"aborting!\n", cnt, mddev->raid_disks);
1da177e4 LT	153	goto abort;
	154	}
	155	zone->nb_dev = cnt;
49f357a2	156	zone->zone_end = smallest->sectors * cnt;
1da177e4	157
49f357a2	158	curr_zone_end = zone->zone_end;
1da177e4 LT	159
	160	/* now do the other zones */
	161	for (i = 1; i < conf->nr_strip_zones; i++)
	162	{
	163	zone = conf->strip_zone + i;
	164	zone->dev = conf->strip_zone[i-1].dev + mddev->raid_disks;
	165
0825b87a	166	printk(KERN_INFO "raid0: zone %d\n", i);
d27a43ab	167	zone->dev_start = smallest->sectors;
1da177e4 LT	168	smallest = NULL;
	169	c = 0;
	170
	171	for (j=0; j<cnt; j++) {
	172	char b[BDEVNAME_SIZE];
	173	rdev = conf->strip_zone[0].dev[j];
0825b87a AN	174	printk(KERN_INFO "raid0: checking %s ...",
0825b87a AN	175	bdevname(rdev->bdev, b));
d27a43ab	176	if (rdev->sectors <= zone->dev_start) {
0825b87a	177	printk(KERN_INFO " nope.\n");
dd8ac336 AN	178	continue;
	179	}
	180	printk(KERN_INFO " contained as device %d\n", c);
	181	zone->dev[c] = rdev;
	182	c++;
	183	if (!smallest \|\| rdev->sectors < smallest->sectors) {
	184	smallest = rdev;
	185	printk(KERN_INFO " (%llu) is smallest!.\n",
	186	(unsigned long long)rdev->sectors);
	187	}
1da177e4 LT	188	}
	189
	190	zone->nb_dev = c;
49f357a2	191	sectors = (smallest->sectors - zone->dev_start) * c;
83838ed8	192	printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n",
49f357a2	193	zone->nb_dev, (unsigned long long)sectors);
1da177e4	194
49f357a2	195	curr_zone_end += sectors;
d27a43ab	196	zone->zone_end = curr_zone_end;
1da177e4	197
6b8796cc	198	printk(KERN_INFO "raid0: current zone start: %llu\n",
d27a43ab	199	(unsigned long long)smallest->sectors);
1da177e4	200	}
1da177e4	201	mddev->queue->unplug_fn = raid0_unplug;
26be34dc N	202	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
26be34dc N	203	mddev->queue->backing_dev_info.congested_data = mddev;
1da177e4	204
0825b87a	205	printk(KERN_INFO "raid0: done.\n");
ed7b0038	206	mddev->private = conf;
1da177e4	207	return 0;
5568a603	208	abort:
ed7b0038 AN	209	kfree(conf->strip_zone);
	210	kfree(conf->devlist);
	211	kfree(conf);
	212	mddev->private = NULL;
	213	return err;
1da177e4 LT	214	}
	215
	216	/**
	217	* raid0_mergeable_bvec -- tell bio layer if a two requests can be merged
	218	* @q: request queue
cc371e66	219	* @bvm: properties of new bio
1da177e4 LT	220	* @biovec: the request that could be merged to it.
	221	*
	222	* Return amount of bytes we can accept at this offset
	223	*/
cc371e66 AK	224	static int raid0_mergeable_bvec(struct request_queue *q,
	225	struct bvec_merge_data *bvm,
	226	struct bio_vec *biovec)
1da177e4 LT	227	{
1da177e4 LT	228	mddev_t *mddev = q->queuedata;
cc371e66	229	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
1da177e4 LT	230	int max;
1da177e4 LT	231	unsigned int chunk_sectors = mddev->chunk_size >> 9;
cc371e66	232	unsigned int bio_sectors = bvm->bi_size >> 9;
1da177e4 LT	233
	234	max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
	235	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
	236	if (max <= biovec->bv_len && bio_sectors == 0)
	237	return biovec->bv_len;
	238	else
	239	return max;
	240	}
	241
80c3a6ce DW	242	static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks)
	243	{
	244	sector_t array_sectors = 0;
	245	mdk_rdev_t *rdev;
	246
	247	WARN_ONCE(sectors \|\| raid_disks,
	248	"%s does not support generic reshape\n", __func__);
	249
	250	list_for_each_entry(rdev, &mddev->disks, same_set)
	251	array_sectors += rdev->sectors;
	252
	253	return array_sectors;
	254	}
	255
8f79cfcd	256	static int raid0_run(mddev_t *mddev)
1da177e4	257	{
5568a603	258	int ret;
1da177e4	259
2604b703 N	260	if (mddev->chunk_size == 0) {
	261	printk(KERN_ERR "md/raid0: non-zero chunk size required.\n");
	262	return -EINVAL;
	263	}
	264	printk(KERN_INFO "%s: setting max_sectors to %d, segment boundary to %d\n",
1da177e4 LT	265	mdname(mddev),
	266	mddev->chunk_size >> 9,
	267	(mddev->chunk_size>>1)-1);
	268	blk_queue_max_sectors(mddev->queue, mddev->chunk_size >> 9);
	269	blk_queue_segment_boundary(mddev->queue, (mddev->chunk_size>>1) - 1);
e7e72bf6	270	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
1da177e4	271
5568a603 AN	272	ret = create_strip_zones(mddev);
5568a603 AN	273	if (ret < 0)
ed7b0038	274	return ret;
1da177e4 LT	275
1da177e4 LT	276	/* calculate array device size */
1f403624	277	md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
1da177e4	278
ccacc7d2 AN	279	printk(KERN_INFO "raid0 : md_size is %llu sectors.\n",
ccacc7d2 AN	280	(unsigned long long)mddev->array_sectors);
1da177e4 LT	281	/* calculate the max read-ahead size.
	282	* For read-ahead of large files to be effective, we need to
	283	* readahead at least twice a whole stripe. i.e. number of devices
	284	* multiplied by chunk size times 2.
	285	* If an individual device has an ra_pages greater than the
	286	* chunk size, then we will not drive that device as hard as it
	287	* wants. We consider this a configuration error: a larger
	288	* chunksize should be used in that case.
	289	*/
	290	{
2d1f3b5d	291	int stripe = mddev->raid_disks * mddev->chunk_size / PAGE_SIZE;
1da177e4 LT	292	if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
	293	mddev->queue->backing_dev_info.ra_pages = 2* stripe;
	294	}
	295
1da177e4 LT	296	blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec);
1da177e4 LT	297	return 0;
1da177e4 LT	298	}
1da177e4 LT	299
fb5ab4b5	300	static int raid0_stop(mddev_t *mddev)
1da177e4 LT	301	{
	302	raid0_conf_t *conf = mddev_to_conf(mddev);
	303
	304	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
990a8baf	305	kfree(conf->strip_zone);
fb5ab4b5	306	kfree(conf->devlist);
990a8baf	307	kfree(conf);
1da177e4	308	mddev->private = NULL;
1da177e4 LT	309	return 0;
	310	}
	311
49f357a2 N	312	/* Find the zone which holds a particular offset
	313	* Update *sectorp to be an offset in that zone
	314	*/
dc582663	315	static struct strip_zone find_zone(struct raid0_private_data conf,
49f357a2	316	sector_t *sectorp)
dc582663 AN	317	{
	318	int i;
	319	struct strip_zone *z = conf->strip_zone;
49f357a2	320	sector_t sector = *sectorp;
dc582663 AN	321
dc582663 AN	322	for (i = 0; i < conf->nr_strip_zones; i++)
49f357a2 N	323	if (sector < z[i].zone_end) {
	324	if (i)
	325	*sectorp = sector - z[i-1].zone_end;
dc582663	326	return z + i;
49f357a2	327	}
dc582663 AN	328	BUG();
	329	}
	330
165125e1	331	static int raid0_make_request (struct request_queue q, struct bio bio)
1da177e4 LT	332	{
1da177e4 LT	333	mddev_t *mddev = q->queuedata;
a4712005	334	unsigned int sect_in_chunk, chunksect_bits, chunk_sects;
1da177e4 LT	335	raid0_conf_t *conf = mddev_to_conf(mddev);
	336	struct strip_zone *zone;
	337	mdk_rdev_t *tmp_dev;
787f17fe	338	sector_t chunk;
49f357a2	339	sector_t sector, rsect, sector_offset;
a362357b	340	const int rw = bio_data_dir(bio);
c9959059	341	int cpu;
1da177e4	342
e5dcdd80	343	if (unlikely(bio_barrier(bio))) {
6712ecf8	344	bio_endio(bio, -EOPNOTSUPP);
e5dcdd80 N	345	return 0;
	346	}
	347
074a7aca TH	348	cpu = part_stat_lock();
	349	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	350	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
	351	bio_sectors(bio));
	352	part_stat_unlock();
1da177e4	353
1da177e4	354	chunk_sects = mddev->chunk_size >> 9;
1b7fdf8f	355	chunksect_bits = ffz(~chunk_sects);
e0f06868	356	sector = bio->bi_sector;
1da177e4 LT	357
	358	if (unlikely(chunk_sects < (bio->bi_sector & (chunk_sects - 1)) + (bio->bi_size >> 9))) {
	359	struct bio_pair *bp;
	360	/* Sanity check -- queue functions should prevent this happening */
	361	if (bio->bi_vcnt != 1 \|\|
	362	bio->bi_idx != 0)
	363	goto bad_map;
	364	/* This is a one page bio that upper layers
	365	* refuse to split for us, so we need to split it.
	366	*/
6feef531	367	bp = bio_split(bio, chunk_sects - (bio->bi_sector & (chunk_sects - 1)));
1da177e4 LT	368	if (raid0_make_request(q, &bp->bio1))
	369	generic_make_request(&bp->bio1);
	370	if (raid0_make_request(q, &bp->bio2))
	371	generic_make_request(&bp->bio2);
	372
	373	bio_pair_release(bp);
	374	return 0;
	375	}
49f357a2 N	376	sector_offset = sector;
49f357a2 N	377	zone = find_zone(conf, &sector_offset);
a4712005	378	sect_in_chunk = bio->bi_sector & (chunk_sects - 1);
1da177e4	379	{
49f357a2	380	sector_t x = sector_offset >> chunksect_bits;
1da177e4 LT	381
	382	sector_div(x, zone->nb_dev);
	383	chunk = x;
1da177e4	384
e0f06868	385	x = sector >> chunksect_bits;
1da177e4 LT	386	tmp_dev = zone->dev[sector_div(x, zone->nb_dev)];
1da177e4 LT	387	}
019c4e2f	388	rsect = (chunk << chunksect_bits) + zone->dev_start + sect_in_chunk;
1da177e4 LT	389
	390	bio->bi_bdev = tmp_dev->bdev;
	391	bio->bi_sector = rsect + tmp_dev->data_offset;
	392
	393	/*
	394	* Let the main block layer submit the IO and resolve recursion:
	395	*/
	396	return 1;
	397
	398	bad_map:
	399	printk("raid0_make_request bug: can't convert block across chunks"
a4712005	400	" or bigger than %dk %llu %d\n", chunk_sects / 2,
1da177e4 LT	401	(unsigned long long)bio->bi_sector, bio->bi_size >> 10);
1da177e4 LT	402
6712ecf8	403	bio_io_error(bio);
1da177e4 LT	404	return 0;
1da177e4 LT	405	}
8299d7f7	406
1da177e4 LT	407	static void raid0_status (struct seq_file seq, mddev_t mddev)
	408	{
	409	#undef MD_DEBUG
	410	#ifdef MD_DEBUG
	411	int j, k, h;
	412	char b[BDEVNAME_SIZE];
	413	raid0_conf_t *conf = mddev_to_conf(mddev);
8299d7f7	414
1da177e4 LT	415	h = 0;
	416	for (j = 0; j < conf->nr_strip_zones; j++) {
	417	seq_printf(seq, " z%d", j);
1da177e4 LT	418	seq_printf(seq, "=[");
1da177e4 LT	419	for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
8299d7f7	420	seq_printf(seq, "%s/", bdevname(
1da177e4 LT	421	conf->strip_zone[j].dev[k]->bdev,b));
1da177e4 LT	422
dc582663 AN	423	seq_printf(seq, "] ze=%d ds=%d s=%d\n",
dc582663 AN	424	conf->strip_zone[j].zone_end,
019c4e2f	425	conf->strip_zone[j].dev_start,
83838ed8	426	conf->strip_zone[j].sectors);
1da177e4 LT	427	}
	428	#endif
	429	seq_printf(seq, " %dk chunks", mddev->chunk_size/1024);
	430	return;
	431	}
	432
2604b703	433	static struct mdk_personality raid0_personality=
1da177e4 LT	434	{
1da177e4 LT	435	.name = "raid0",
2604b703	436	.level = 0,
1da177e4 LT	437	.owner = THIS_MODULE,
	438	.make_request = raid0_make_request,
	439	.run = raid0_run,
	440	.stop = raid0_stop,
	441	.status = raid0_status,
80c3a6ce	442	.size = raid0_size,
1da177e4 LT	443	};
	444
	445	static int __init raid0_init (void)
	446	{
2604b703	447	return register_md_personality (&raid0_personality);
1da177e4 LT	448	}
	449
	450	static void raid0_exit (void)
	451	{
2604b703	452	unregister_md_personality (&raid0_personality);
1da177e4 LT	453	}
	454
	455	module_init(raid0_init);
	456	module_exit(raid0_exit);
	457	MODULE_LICENSE("GPL");
	458	MODULE_ALIAS("md-personality-2"); /* RAID0 */
d9d166c2	459	MODULE_ALIAS("md-raid0");
2604b703	460	MODULE_ALIAS("md-level-0");