[GitHub/mt8127/android_kernel_alcatel_ttab.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 <linux/slab.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;

	if (mddev_congested(mddev, bits))
		return 1;

	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, 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;

		disk_stack_limits(mddev->gendisk, rdev1->bdev,
				  rdev1->data_offset << 9);
		/* as we don't honour merge_bvec_fn, we must never risk
		 * violating it, so limit ->max_segments to 1, lying within
		 * a single page.
		 */

		if (rdev1->bdev->bd_disk->queue->merge_bvec_fn) {
			blk_queue_max_segments(mddev->queue, 1);
			blk_queue_segment_boundary(mddev->queue,
						   PAGE_CACHE_SIZE - 1);
		}
		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++)
	{
		int j;

		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++) {
			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;
	}

	blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
	blk_queue_io_opt(mddev->queue,
			 (mddev->chunk_sectors << 9) * mddev->raid_disks);

	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_hw_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);
	md_integrity_register(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_rw_flagged(bio, BIO_RW_BARRIER))) {
		md_barrier_request(mddev, bio);
		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_DESCRIPTION("RAID0 (striping) personality for MD");
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>
5a0e3ad6	23	#include <linux/slab.h>
43b2e5d8	24	#include "md.h"
ef740c37	25	#include "raid0.h"
1da177e4	26
165125e1	27	static void raid0_unplug(struct request_queue *q)
1da177e4 LT	28	{
1da177e4 LT	29	mddev_t *mddev = q->queuedata;
070ec55d	30	raid0_conf_t *conf = mddev->private;
b414579f	31	mdk_rdev_t **devlist = conf->devlist;
1da177e4 LT	32	int i;
	33
	34	for (i=0; i<mddev->raid_disks; i++) {
165125e1	35	struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev);
1da177e4	36
2ad8b1ef	37	blk_unplug(r_queue);
1da177e4 LT	38	}
	39	}
	40
26be34dc N	41	static int raid0_congested(void *data, int bits)
	42	{
	43	mddev_t *mddev = data;
070ec55d	44	raid0_conf_t *conf = mddev->private;
b414579f	45	mdk_rdev_t **devlist = conf->devlist;
26be34dc N	46	int i, ret = 0;
26be34dc N	47
3fa841d7 N	48	if (mddev_congested(mddev, bits))
	49	return 1;
	50
26be34dc	51	for (i = 0; i < mddev->raid_disks && !ret ; i++) {
165125e1	52	struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
26be34dc N	53
	54	ret \|= bdi_congested(&q->backing_dev_info, bits);
	55	}
	56	return ret;
	57	}
	58
46994191	59	/*
	60	* inform the user of the raid configuration
	61	*/
	62	static void dump_zones(mddev_t *mddev)
	63	{
	64	int j, k, h;
	65	sector_t zone_size = 0;
	66	sector_t zone_start = 0;
	67	char b[BDEVNAME_SIZE];
	68	raid0_conf_t *conf = mddev->private;
	69	printk(KERN_INFO "***** %s configuration *******\n",
	70	mdname(mddev));
	71	h = 0;
	72	for (j = 0; j < conf->nr_strip_zones; j++) {
	73	printk(KERN_INFO "zone%d=[", j);
	74	for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
	75	printk("%s/",
	76	bdevname(conf->devlist[j*mddev->raid_disks
	77	+ k]->bdev, b));
	78	printk("]\n");
	79
	80	zone_size = conf->strip_zone[j].zone_end - zone_start;
	81	printk(KERN_INFO " zone offset=%llukb "
	82	"device offset=%llukb size=%llukb\n",
	83	(unsigned long long)zone_start>>1,
	84	(unsigned long long)conf->strip_zone[j].dev_start>>1,
	85	(unsigned long long)zone_size>>1);
	86	zone_start = conf->strip_zone[j].zone_end;
	87	}
	88	printk(KERN_INFO "**********************************\n\n");
	89	}
	90
ed7b0038	91	static int create_strip_zones(mddev_t *mddev)
1da177e4	92	{
a9f326eb	93	int i, c, err;
49f357a2	94	sector_t curr_zone_end, sectors;
b414579f	95	mdk_rdev_t smallest, rdev1, rdev2, rdev, **dev;
1da177e4 LT	96	struct strip_zone *zone;
	97	int cnt;
	98	char b[BDEVNAME_SIZE];
ed7b0038 AN	99	raid0_conf_t conf = kzalloc(sizeof(conf), GFP_KERNEL);
	100
	101	if (!conf)
	102	return -ENOMEM;
159ec1fc	103	list_for_each_entry(rdev1, &mddev->disks, same_set) {
0825b87a	104	printk(KERN_INFO "raid0: looking at %s\n",
1da177e4 LT	105	bdevname(rdev1->bdev,b));
1da177e4 LT	106	c = 0;
13f2682b N	107
	108	/* round size to chunk_size */
	109	sectors = rdev1->sectors;
	110	sector_div(sectors, mddev->chunk_sectors);
	111	rdev1->sectors = sectors * mddev->chunk_sectors;
	112
159ec1fc	113	list_for_each_entry(rdev2, &mddev->disks, same_set) {
0825b87a	114	printk(KERN_INFO "raid0: comparing %s(%llu)",
1da177e4	115	bdevname(rdev1->bdev,b),
dd8ac336	116	(unsigned long long)rdev1->sectors);
0825b87a	117	printk(KERN_INFO " with %s(%llu)\n",
1da177e4	118	bdevname(rdev2->bdev,b),
dd8ac336	119	(unsigned long long)rdev2->sectors);
1da177e4	120	if (rdev2 == rdev1) {
0825b87a	121	printk(KERN_INFO "raid0: END\n");
1da177e4 LT	122	break;
1da177e4 LT	123	}
dd8ac336	124	if (rdev2->sectors == rdev1->sectors) {
1da177e4 LT	125	/*
	126	* Not unique, don't count it as a new
	127	* group
	128	*/
0825b87a	129	printk(KERN_INFO "raid0: EQUAL\n");
1da177e4 LT	130	c = 1;
	131	break;
	132	}
0825b87a	133	printk(KERN_INFO "raid0: NOT EQUAL\n");
1da177e4 LT	134	}
1da177e4 LT	135	if (!c) {
0825b87a	136	printk(KERN_INFO "raid0: ==> UNIQUE\n");
1da177e4	137	conf->nr_strip_zones++;
0825b87a AN	138	printk(KERN_INFO "raid0: %d zones\n",
0825b87a AN	139	conf->nr_strip_zones);
1da177e4 LT	140	}
1da177e4 LT	141	}
0825b87a	142	printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones);
ed7b0038	143	err = -ENOMEM;
9ffae0cf	144	conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
1da177e4 LT	145	conf->nr_strip_zones, GFP_KERNEL);
1da177e4 LT	146	if (!conf->strip_zone)
ed7b0038	147	goto abort;
9ffae0cf	148	conf->devlist = kzalloc(sizeof(mdk_rdev_t)
1da177e4 LT	149	conf->nr_strip_zones*mddev->raid_disks,
	150	GFP_KERNEL);
	151	if (!conf->devlist)
ed7b0038	152	goto abort;
1da177e4	153
1da177e4 LT	154	/* The first zone must contain all devices, so here we check that
	155	* there is a proper alignment of slots to devices and find them all
	156	*/
	157	zone = &conf->strip_zone[0];
	158	cnt = 0;
	159	smallest = NULL;
b414579f	160	dev = conf->devlist;
ed7b0038	161	err = -EINVAL;
159ec1fc	162	list_for_each_entry(rdev1, &mddev->disks, same_set) {
1da177e4 LT	163	int j = rdev1->raid_disk;
	164
	165	if (j < 0 \|\| j >= mddev->raid_disks) {
0825b87a AN	166	printk(KERN_ERR "raid0: bad disk number %d - "
0825b87a AN	167	"aborting!\n", j);
1da177e4 LT	168	goto abort;
1da177e4 LT	169	}
b414579f	170	if (dev[j]) {
0825b87a AN	171	printk(KERN_ERR "raid0: multiple devices for %d - "
0825b87a AN	172	"aborting!\n", j);
1da177e4 LT	173	goto abort;
1da177e4 LT	174	}
b414579f	175	dev[j] = rdev1;
1da177e4	176
8f6c2e4b MP	177	disk_stack_limits(mddev->gendisk, rdev1->bdev,
8f6c2e4b MP	178	rdev1->data_offset << 9);
1da177e4	179	/* as we don't honour merge_bvec_fn, we must never risk
627a2d3c N	180	* violating it, so limit ->max_segments to 1, lying within
627a2d3c N	181	* a single page.
1da177e4 LT	182	*/
1da177e4 LT	183
627a2d3c N	184	if (rdev1->bdev->bd_disk->queue->merge_bvec_fn) {
	185	blk_queue_max_segments(mddev->queue, 1);
	186	blk_queue_segment_boundary(mddev->queue,
	187	PAGE_CACHE_SIZE - 1);
	188	}
dd8ac336	189	if (!smallest \|\| (rdev1->sectors < smallest->sectors))
1da177e4 LT	190	smallest = rdev1;
	191	cnt++;
	192	}
	193	if (cnt != mddev->raid_disks) {
0825b87a AN	194	printk(KERN_ERR "raid0: too few disks (%d of %d) - "
0825b87a AN	195	"aborting!\n", cnt, mddev->raid_disks);
1da177e4 LT	196	goto abort;
	197	}
	198	zone->nb_dev = cnt;
49f357a2	199	zone->zone_end = smallest->sectors * cnt;
1da177e4	200
49f357a2	201	curr_zone_end = zone->zone_end;
1da177e4 LT	202
	203	/* now do the other zones */
	204	for (i = 1; i < conf->nr_strip_zones; i++)
	205	{
a9f326eb N	206	int j;
a9f326eb N	207
1da177e4	208	zone = conf->strip_zone + i;
b414579f	209	dev = conf->devlist + i * mddev->raid_disks;
1da177e4	210
0825b87a	211	printk(KERN_INFO "raid0: zone %d\n", i);
d27a43ab	212	zone->dev_start = smallest->sectors;
1da177e4 LT	213	smallest = NULL;
	214	c = 0;
	215
	216	for (j=0; j<cnt; j++) {
b414579f	217	rdev = conf->devlist[j];
0825b87a AN	218	printk(KERN_INFO "raid0: checking %s ...",
0825b87a AN	219	bdevname(rdev->bdev, b));
d27a43ab	220	if (rdev->sectors <= zone->dev_start) {
0825b87a	221	printk(KERN_INFO " nope.\n");
dd8ac336 AN	222	continue;
	223	}
	224	printk(KERN_INFO " contained as device %d\n", c);
b414579f	225	dev[c] = rdev;
dd8ac336 AN	226	c++;
	227	if (!smallest \|\| rdev->sectors < smallest->sectors) {
	228	smallest = rdev;
	229	printk(KERN_INFO " (%llu) is smallest!.\n",
	230	(unsigned long long)rdev->sectors);
	231	}
1da177e4 LT	232	}
	233
	234	zone->nb_dev = c;
49f357a2	235	sectors = (smallest->sectors - zone->dev_start) * c;
83838ed8	236	printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n",
49f357a2	237	zone->nb_dev, (unsigned long long)sectors);
1da177e4	238
49f357a2	239	curr_zone_end += sectors;
d27a43ab	240	zone->zone_end = curr_zone_end;
1da177e4	241
6b8796cc	242	printk(KERN_INFO "raid0: current zone start: %llu\n",
d27a43ab	243	(unsigned long long)smallest->sectors);
1da177e4	244	}
1da177e4	245	mddev->queue->unplug_fn = raid0_unplug;
26be34dc N	246	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
26be34dc N	247	mddev->queue->backing_dev_info.congested_data = mddev;
1da177e4	248
92e59b6b	249	/*
	250	* now since we have the hard sector sizes, we can make sure
	251	* chunk size is a multiple of that sector size
	252	*/
9d8f0363	253	if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) {
92e59b6b	254	printk(KERN_ERR "%s chunk_size of %d not valid\n",
92e59b6b	255	mdname(mddev),
9d8f0363	256	mddev->chunk_sectors << 9);
92e59b6b	257	goto abort;
92e59b6b	258	}
8f6c2e4b MP	259
	260	blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
	261	blk_queue_io_opt(mddev->queue,
	262	(mddev->chunk_sectors << 9) * mddev->raid_disks);
	263
0825b87a	264	printk(KERN_INFO "raid0: done.\n");
ed7b0038	265	mddev->private = conf;
1da177e4	266	return 0;
5568a603	267	abort:
ed7b0038 AN	268	kfree(conf->strip_zone);
	269	kfree(conf->devlist);
	270	kfree(conf);
	271	mddev->private = NULL;
	272	return err;
1da177e4 LT	273	}
	274
	275	/**
	276	* raid0_mergeable_bvec -- tell bio layer if a two requests can be merged
	277	* @q: request queue
cc371e66	278	* @bvm: properties of new bio
1da177e4 LT	279	* @biovec: the request that could be merged to it.
	280	*
	281	* Return amount of bytes we can accept at this offset
	282	*/
cc371e66 AK	283	static int raid0_mergeable_bvec(struct request_queue *q,
	284	struct bvec_merge_data *bvm,
	285	struct bio_vec *biovec)
1da177e4 LT	286	{
1da177e4 LT	287	mddev_t *mddev = q->queuedata;
cc371e66	288	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
1da177e4	289	int max;
9d8f0363	290	unsigned int chunk_sectors = mddev->chunk_sectors;
cc371e66	291	unsigned int bio_sectors = bvm->bi_size >> 9;
1da177e4	292
d6e412ea	293	if (is_power_of_2(chunk_sectors))
fbb704ef	294	max = (chunk_sectors - ((sector & (chunk_sectors-1))
	295	+ bio_sectors)) << 9;
	296	else
	297	max = (chunk_sectors - (sector_div(sector, chunk_sectors)
	298	+ bio_sectors)) << 9;
1da177e4 LT	299	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
	300	if (max <= biovec->bv_len && bio_sectors == 0)
	301	return biovec->bv_len;
	302	else
	303	return max;
	304	}
	305
80c3a6ce DW	306	static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks)
	307	{
	308	sector_t array_sectors = 0;
	309	mdk_rdev_t *rdev;
	310
	311	WARN_ONCE(sectors \|\| raid_disks,
	312	"%s does not support generic reshape\n", __func__);
	313
	314	list_for_each_entry(rdev, &mddev->disks, same_set)
	315	array_sectors += rdev->sectors;
	316
	317	return array_sectors;
	318	}
	319
8f79cfcd	320	static int raid0_run(mddev_t *mddev)
1da177e4	321	{
5568a603	322	int ret;
1da177e4	323
9d8f0363	324	if (mddev->chunk_sectors == 0) {
fbb704ef	325	printk(KERN_ERR "md/raid0: chunk size must be set.\n");
2604b703 N	326	return -EINVAL;
2604b703 N	327	}
0894cc30 AN	328	if (md_check_no_bitmap(mddev))
0894cc30 AN	329	return -EINVAL;
086fa5ff	330	blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors);
e7e72bf6	331	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
1da177e4	332
5568a603 AN	333	ret = create_strip_zones(mddev);
5568a603 AN	334	if (ret < 0)
ed7b0038	335	return ret;
1da177e4 LT	336
1da177e4 LT	337	/* calculate array device size */
1f403624	338	md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
1da177e4	339
ccacc7d2 AN	340	printk(KERN_INFO "raid0 : md_size is %llu sectors.\n",
ccacc7d2 AN	341	(unsigned long long)mddev->array_sectors);
1da177e4 LT	342	/* calculate the max read-ahead size.
	343	* For read-ahead of large files to be effective, we need to
	344	* readahead at least twice a whole stripe. i.e. number of devices
	345	* multiplied by chunk size times 2.
	346	* If an individual device has an ra_pages greater than the
	347	* chunk size, then we will not drive that device as hard as it
	348	* wants. We consider this a configuration error: a larger
	349	* chunksize should be used in that case.
	350	*/
	351	{
9d8f0363 AN	352	int stripe = mddev->raid_disks *
9d8f0363 AN	353	(mddev->chunk_sectors << 9) / PAGE_SIZE;
1da177e4 LT	354	if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
	355	mddev->queue->backing_dev_info.ra_pages = 2* stripe;
	356	}
	357
1da177e4	358	blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec);
46994191	359	dump_zones(mddev);
ac5e7113	360	md_integrity_register(mddev);
1da177e4	361	return 0;
1da177e4 LT	362	}
1da177e4 LT	363
fb5ab4b5	364	static int raid0_stop(mddev_t *mddev)
1da177e4	365	{
070ec55d	366	raid0_conf_t *conf = mddev->private;
1da177e4 LT	367
1da177e4 LT	368	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
990a8baf	369	kfree(conf->strip_zone);
fb5ab4b5	370	kfree(conf->devlist);
990a8baf	371	kfree(conf);
1da177e4	372	mddev->private = NULL;
1da177e4 LT	373	return 0;
	374	}
	375
49f357a2 N	376	/* Find the zone which holds a particular offset
	377	* Update *sectorp to be an offset in that zone
	378	*/
dc582663	379	static struct strip_zone find_zone(struct raid0_private_data conf,
49f357a2	380	sector_t *sectorp)
dc582663 AN	381	{
	382	int i;
	383	struct strip_zone *z = conf->strip_zone;
49f357a2	384	sector_t sector = *sectorp;
dc582663 AN	385
dc582663 AN	386	for (i = 0; i < conf->nr_strip_zones; i++)
49f357a2 N	387	if (sector < z[i].zone_end) {
	388	if (i)
	389	*sectorp = sector - z[i-1].zone_end;
dc582663	390	return z + i;
49f357a2	391	}
dc582663 AN	392	BUG();
	393	}
	394
fbb704ef	395	/*
	396	* remaps the bio to the target device. we separate two flows.
	397	* power 2 flow and a general flow for the sake of perfromance
	398	*/
	399	static mdk_rdev_t map_sector(mddev_t mddev, struct strip_zone *zone,
	400	sector_t sector, sector_t *sector_offset)
1da177e4	401	{
fbb704ef	402	unsigned int sect_in_chunk;
fbb704ef	403	sector_t chunk;
070ec55d	404	raid0_conf_t *conf = mddev->private;
9d8f0363	405	unsigned int chunk_sects = mddev->chunk_sectors;
fbb704ef	406
d6e412ea	407	if (is_power_of_2(chunk_sects)) {
fbb704ef	408	int chunksect_bits = ffz(~chunk_sects);
	409	/* find the sector offset inside the chunk */
	410	sect_in_chunk = sector & (chunk_sects - 1);
	411	sector >>= chunksect_bits;
	412	/* chunk in zone */
	413	chunk = *sector_offset;
	414	/* quotient is the chunk in real device*/
	415	sector_div(chunk, zone->nb_dev << chunksect_bits);
	416	} else{
	417	sect_in_chunk = sector_div(sector, chunk_sects);
	418	chunk = *sector_offset;
	419	sector_div(chunk, chunk_sects * zone->nb_dev);
	420	}
	421	/*
	422	* position the bio over the real device
	423	* real sector = chunk in device + starting of zone
	424	* + the position in the chunk
	425	*/
	426	sector_offset = (chunk chunk_sects) + sect_in_chunk;
	427	return conf->devlist[(zone - conf->strip_zone)*mddev->raid_disks
	428	+ sector_div(sector, zone->nb_dev)];
	429	}
	430
	431	/*
	432	* Is io distribute over 1 or more chunks ?
	433	*/
	434	static inline int is_io_in_chunk_boundary(mddev_t *mddev,
	435	unsigned int chunk_sects, struct bio *bio)
	436	{
d6e412ea	437	if (likely(is_power_of_2(chunk_sects))) {
fbb704ef	438	return chunk_sects >= ((bio->bi_sector & (chunk_sects-1))
	439	+ (bio->bi_size >> 9));
	440	} else{
	441	sector_t sector = bio->bi_sector;
	442	return chunk_sects >= (sector_div(sector, chunk_sects)
	443	+ (bio->bi_size >> 9));
	444	}
	445	}
	446
	447	static int raid0_make_request(struct request_queue q, struct bio bio)
	448	{
	449	mddev_t *mddev = q->queuedata;
	450	unsigned int chunk_sects;
	451	sector_t sector_offset;
1da177e4 LT	452	struct strip_zone *zone;
1da177e4 LT	453	mdk_rdev_t *tmp_dev;
a362357b	454	const int rw = bio_data_dir(bio);
c9959059	455	int cpu;
1da177e4	456
1f98a13f	457	if (unlikely(bio_rw_flagged(bio, BIO_RW_BARRIER))) {
a2826aa9	458	md_barrier_request(mddev, bio);
e5dcdd80 N	459	return 0;
	460	}
	461
074a7aca TH	462	cpu = part_stat_lock();
	463	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	464	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
	465	bio_sectors(bio));
	466	part_stat_unlock();
1da177e4	467
9d8f0363	468	chunk_sects = mddev->chunk_sectors;
fbb704ef	469	if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) {
fbb704ef	470	sector_t sector = bio->bi_sector;
1da177e4 LT	471	struct bio_pair *bp;
	472	/* Sanity check -- queue functions should prevent this happening */
	473	if (bio->bi_vcnt != 1 \|\|
	474	bio->bi_idx != 0)
	475	goto bad_map;
	476	/* This is a one page bio that upper layers
	477	* refuse to split for us, so we need to split it.
	478	*/
d6e412ea	479	if (likely(is_power_of_2(chunk_sects)))
fbb704ef	480	bp = bio_split(bio, chunk_sects - (sector &
	481	(chunk_sects-1)));
	482	else
	483	bp = bio_split(bio, chunk_sects -
	484	sector_div(sector, chunk_sects));
1da177e4 LT	485	if (raid0_make_request(q, &bp->bio1))
	486	generic_make_request(&bp->bio1);
	487	if (raid0_make_request(q, &bp->bio2))
	488	generic_make_request(&bp->bio2);
	489
	490	bio_pair_release(bp);
	491	return 0;
	492	}
1da177e4	493
fbb704ef	494	sector_offset = bio->bi_sector;
	495	zone = find_zone(mddev->private, &sector_offset);
	496	tmp_dev = map_sector(mddev, zone, bio->bi_sector,
	497	&sector_offset);
1da177e4	498	bio->bi_bdev = tmp_dev->bdev;
fbb704ef	499	bio->bi_sector = sector_offset + zone->dev_start +
fbb704ef	500	tmp_dev->data_offset;
1da177e4 LT	501	/*
	502	* Let the main block layer submit the IO and resolve recursion:
	503	*/
	504	return 1;
	505
	506	bad_map:
	507	printk("raid0_make_request bug: can't convert block across chunks"
a4712005	508	" or bigger than %dk %llu %d\n", chunk_sects / 2,
1da177e4 LT	509	(unsigned long long)bio->bi_sector, bio->bi_size >> 10);
1da177e4 LT	510
6712ecf8	511	bio_io_error(bio);
1da177e4 LT	512	return 0;
1da177e4 LT	513	}
8299d7f7	514
1b961429	515	static void raid0_status(struct seq_file seq, mddev_t mddev)
1da177e4 LT	516	{
	517	#undef MD_DEBUG
	518	#ifdef MD_DEBUG
	519	int j, k, h;
	520	char b[BDEVNAME_SIZE];
070ec55d	521	raid0_conf_t *conf = mddev->private;
8299d7f7	522
1b961429	523	sector_t zone_size;
1b961429	524	sector_t zone_start = 0;
1da177e4	525	h = 0;
1b961429	526
1da177e4 LT	527	for (j = 0; j < conf->nr_strip_zones; j++) {
1da177e4 LT	528	seq_printf(seq, " z%d", j);
1da177e4 LT	529	seq_printf(seq, "=[");
1da177e4 LT	530	for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
8299d7f7	531	seq_printf(seq, "%s/", bdevname(
1b961429	532	conf->devlist[j*mddev->raid_disks + k]
	533	->bdev, b));
	534
	535	zone_size = conf->strip_zone[j].zone_end - zone_start;
	536	seq_printf(seq, "] ze=%lld ds=%lld s=%lld\n",
	537	(unsigned long long)zone_start>>1,
	538	(unsigned long long)conf->strip_zone[j].dev_start>>1,
	539	(unsigned long long)zone_size>>1);
	540	zone_start = conf->strip_zone[j].zone_end;
1da177e4 LT	541	}
1da177e4 LT	542	#endif
9d8f0363	543	seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2);
1da177e4 LT	544	return;
	545	}
	546
2604b703	547	static struct mdk_personality raid0_personality=
1da177e4 LT	548	{
1da177e4 LT	549	.name = "raid0",
2604b703	550	.level = 0,
1da177e4 LT	551	.owner = THIS_MODULE,
	552	.make_request = raid0_make_request,
	553	.run = raid0_run,
	554	.stop = raid0_stop,
	555	.status = raid0_status,
80c3a6ce	556	.size = raid0_size,
1da177e4 LT	557	};
	558
	559	static int __init raid0_init (void)
	560	{
2604b703	561	return register_md_personality (&raid0_personality);
1da177e4 LT	562	}
	563
	564	static void raid0_exit (void)
	565	{
2604b703	566	unregister_md_personality (&raid0_personality);
1da177e4 LT	567	}
	568
	569	module_init(raid0_init);
	570	module_exit(raid0_exit);
	571	MODULE_LICENSE("GPL");
0efb9e61	572	MODULE_DESCRIPTION("RAID0 (striping) personality for MD");
1da177e4	573	MODULE_ALIAS("md-personality-2"); /* RAID0 */
d9d166c2	574	MODULE_ALIAS("md-raid0");
2604b703	575	MODULE_ALIAS("md-level-0");