[linux.git] / block / blk-barrier.c

/*
 * Functions related to barrier IO handling
 */
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/bio.h>
#include <linux/blkdev.h>
#include <linux/gfp.h>

#include "blk.h"

/**
 * blk_queue_ordered - does this queue support ordered writes
 * @q:        the request queue
 * @ordered:  one of QUEUE_ORDERED_*
 * @prepare_flush_fn: rq setup helper for cache flush ordered writes
 *
 * Description:
 *   For journalled file systems, doing ordered writes on a commit
 *   block instead of explicitly doing wait_on_buffer (which is bad
 *   for performance) can be a big win. Block drivers supporting this
 *   feature should call this function and indicate so.
 *
 **/
int blk_queue_ordered(struct request_queue *q, unsigned ordered,
		      prepare_flush_fn *prepare_flush_fn)
{
	if (!prepare_flush_fn && (ordered & (QUEUE_ORDERED_DO_PREFLUSH |
					     QUEUE_ORDERED_DO_POSTFLUSH))) {
		printk(KERN_ERR "%s: prepare_flush_fn required\n", __func__);
		return -EINVAL;
	}

	if (ordered != QUEUE_ORDERED_NONE &&
	    ordered != QUEUE_ORDERED_DRAIN &&
	    ordered != QUEUE_ORDERED_DRAIN_FLUSH &&
	    ordered != QUEUE_ORDERED_DRAIN_FUA &&
	    ordered != QUEUE_ORDERED_TAG &&
	    ordered != QUEUE_ORDERED_TAG_FLUSH &&
	    ordered != QUEUE_ORDERED_TAG_FUA) {
		printk(KERN_ERR "blk_queue_ordered: bad value %d\n", ordered);
		return -EINVAL;
	}

	q->ordered = ordered;
	q->next_ordered = ordered;
	q->prepare_flush_fn = prepare_flush_fn;

	return 0;
}
EXPORT_SYMBOL(blk_queue_ordered);

/*
 * Cache flushing for ordered writes handling
 */
unsigned blk_ordered_cur_seq(struct request_queue *q)
{
	if (!q->ordseq)
		return 0;
	return 1 << ffz(q->ordseq);
}

unsigned blk_ordered_req_seq(struct request *rq)
{
	struct request_queue *q = rq->q;

	BUG_ON(q->ordseq == 0);

	if (rq == &q->pre_flush_rq)
		return QUEUE_ORDSEQ_PREFLUSH;
	if (rq == &q->bar_rq)
		return QUEUE_ORDSEQ_BAR;
	if (rq == &q->post_flush_rq)
		return QUEUE_ORDSEQ_POSTFLUSH;

	/*
	 * !fs requests don't need to follow barrier ordering.  Always
	 * put them at the front.  This fixes the following deadlock.
	 *
	 * http://thread.gmane.org/gmane.linux.kernel/537473
	 */
	if (!blk_fs_request(rq))
		return QUEUE_ORDSEQ_DRAIN;

	if ((rq->cmd_flags & REQ_ORDERED_COLOR) ==
	    (q->orig_bar_rq->cmd_flags & REQ_ORDERED_COLOR))
		return QUEUE_ORDSEQ_DRAIN;
	else
		return QUEUE_ORDSEQ_DONE;
}

bool blk_ordered_complete_seq(struct request_queue *q, unsigned seq, int error)
{
	struct request *rq;

	if (error && !q->orderr)
		q->orderr = error;

	BUG_ON(q->ordseq & seq);
	q->ordseq |= seq;

	if (blk_ordered_cur_seq(q) != QUEUE_ORDSEQ_DONE)
		return false;

	/*
	 * Okay, sequence complete.
	 */
	q->ordseq = 0;
	rq = q->orig_bar_rq;
	__blk_end_request_all(rq, q->orderr);
	return true;
}

static void pre_flush_end_io(struct request *rq, int error)
{
	elv_completed_request(rq->q, rq);
	blk_ordered_complete_seq(rq->q, QUEUE_ORDSEQ_PREFLUSH, error);
}

static void bar_end_io(struct request *rq, int error)
{
	elv_completed_request(rq->q, rq);
	blk_ordered_complete_seq(rq->q, QUEUE_ORDSEQ_BAR, error);
}

static void post_flush_end_io(struct request *rq, int error)
{
	elv_completed_request(rq->q, rq);
	blk_ordered_complete_seq(rq->q, QUEUE_ORDSEQ_POSTFLUSH, error);
}

static void queue_flush(struct request_queue *q, unsigned which)
{
	struct request *rq;
	rq_end_io_fn *end_io;

	if (which == QUEUE_ORDERED_DO_PREFLUSH) {
		rq = &q->pre_flush_rq;
		end_io = pre_flush_end_io;
	} else {
		rq = &q->post_flush_rq;
		end_io = post_flush_end_io;
	}

	blk_rq_init(q, rq);
	rq->cmd_flags = REQ_HARDBARRIER;
	rq->rq_disk = q->bar_rq.rq_disk;
	rq->end_io = end_io;
	q->prepare_flush_fn(q, rq);

	elv_insert(q, rq, ELEVATOR_INSERT_FRONT);
}

static inline bool start_ordered(struct request_queue *q, struct request **rqp)
{
	struct request *rq = *rqp;
	unsigned skip = 0;

	q->orderr = 0;
	q->ordered = q->next_ordered;
	q->ordseq |= QUEUE_ORDSEQ_STARTED;

	/*
	 * For an empty barrier, there's no actual BAR request, which
	 * in turn makes POSTFLUSH unnecessary.  Mask them off.
	 */
	if (!blk_rq_sectors(rq)) {
		q->ordered &= ~(QUEUE_ORDERED_DO_BAR |
				QUEUE_ORDERED_DO_POSTFLUSH);
		/*
		 * Empty barrier on a write-through device w/ ordered
		 * tag has no command to issue and without any command
		 * to issue, ordering by tag can't be used.  Drain
		 * instead.
		 */
		if ((q->ordered & QUEUE_ORDERED_BY_TAG) &&
		    !(q->ordered & QUEUE_ORDERED_DO_PREFLUSH)) {
			q->ordered &= ~QUEUE_ORDERED_BY_TAG;
			q->ordered |= QUEUE_ORDERED_BY_DRAIN;
		}
	}

	/* stash away the original request */
	blk_dequeue_request(rq);
	q->orig_bar_rq = rq;
	rq = NULL;

	/*
	 * Queue ordered sequence.  As we stack them at the head, we
	 * need to queue in reverse order.  Note that we rely on that
	 * no fs request uses ELEVATOR_INSERT_FRONT and thus no fs
	 * request gets inbetween ordered sequence.
	 */
	if (q->ordered & QUEUE_ORDERED_DO_POSTFLUSH) {
		queue_flush(q, QUEUE_ORDERED_DO_POSTFLUSH);
		rq = &q->post_flush_rq;
	} else
		skip |= QUEUE_ORDSEQ_POSTFLUSH;

	if (q->ordered & QUEUE_ORDERED_DO_BAR) {
		rq = &q->bar_rq;

		/* initialize proxy request and queue it */
		blk_rq_init(q, rq);
		if (bio_data_dir(q->orig_bar_rq->bio) == WRITE)
			rq->cmd_flags |= REQ_RW;
		if (q->ordered & QUEUE_ORDERED_DO_FUA)
			rq->cmd_flags |= REQ_FUA;
		init_request_from_bio(rq, q->orig_bar_rq->bio);
		rq->end_io = bar_end_io;

		elv_insert(q, rq, ELEVATOR_INSERT_FRONT);
	} else
		skip |= QUEUE_ORDSEQ_BAR;

	if (q->ordered & QUEUE_ORDERED_DO_PREFLUSH) {
		queue_flush(q, QUEUE_ORDERED_DO_PREFLUSH);
		rq = &q->pre_flush_rq;
	} else
		skip |= QUEUE_ORDSEQ_PREFLUSH;

	if ((q->ordered & QUEUE_ORDERED_BY_DRAIN) && queue_in_flight(q))
		rq = NULL;
	else
		skip |= QUEUE_ORDSEQ_DRAIN;

	*rqp = rq;

	/*
	 * Complete skipped sequences.  If whole sequence is complete,
	 * return false to tell elevator that this request is gone.
	 */
	return !blk_ordered_complete_seq(q, skip, 0);
}

bool blk_do_ordered(struct request_queue *q, struct request **rqp)
{
	struct request *rq = *rqp;
	const int is_barrier = blk_fs_request(rq) && blk_barrier_rq(rq);

	if (!q->ordseq) {
		if (!is_barrier)
			return true;

		if (q->next_ordered != QUEUE_ORDERED_NONE)
			return start_ordered(q, rqp);
		else {
			/*
			 * Queue ordering not supported.  Terminate
			 * with prejudice.
			 */
			blk_dequeue_request(rq);
			__blk_end_request_all(rq, -EOPNOTSUPP);
			*rqp = NULL;
			return false;
		}
	}

	/*
	 * Ordered sequence in progress
	 */

	/* Special requests are not subject to ordering rules. */
	if (!blk_fs_request(rq) &&
	    rq != &q->pre_flush_rq && rq != &q->post_flush_rq)
		return true;

	if (q->ordered & QUEUE_ORDERED_BY_TAG) {
		/* Ordered by tag.  Blocking the next barrier is enough. */
		if (is_barrier && rq != &q->bar_rq)
			*rqp = NULL;
	} else {
		/* Ordered by draining.  Wait for turn. */
		WARN_ON(blk_ordered_req_seq(rq) < blk_ordered_cur_seq(q));
		if (blk_ordered_req_seq(rq) > blk_ordered_cur_seq(q))
			*rqp = NULL;
	}

	return true;
}

static void bio_end_empty_barrier(struct bio *bio, int err)
{
	if (err) {
		if (err == -EOPNOTSUPP)
			set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
	}

	complete(bio->bi_private);
}

/**
 * blkdev_issue_flush - queue a flush
 * @bdev:	blockdev to issue flush for
 * @error_sector:	error sector
 *
 * Description:
 *    Issue a flush for the block device in question. Caller can supply
 *    room for storing the error offset in case of a flush error, if they
 *    wish to.
 */
int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector)
{
	DECLARE_COMPLETION_ONSTACK(wait);
	struct request_queue *q;
	struct bio *bio;
	int ret;

	if (bdev->bd_disk == NULL)
		return -ENXIO;

	q = bdev_get_queue(bdev);
	if (!q)
		return -ENXIO;

	bio = bio_alloc(GFP_KERNEL, 0);
	bio->bi_end_io = bio_end_empty_barrier;
	bio->bi_private = &wait;
	bio->bi_bdev = bdev;
	submit_bio(WRITE_BARRIER, bio);

	wait_for_completion(&wait);

	/*
	 * The driver must store the error location in ->bi_sector, if
	 * it supports it. For non-stacked drivers, this should be copied
	 * from blk_rq_pos(rq).
	 */
	if (error_sector)
		*error_sector = bio->bi_sector;

	ret = 0;
	if (bio_flagged(bio, BIO_EOPNOTSUPP))
		ret = -EOPNOTSUPP;
	else if (!bio_flagged(bio, BIO_UPTODATE))
		ret = -EIO;

	bio_put(bio);
	return ret;
}
EXPORT_SYMBOL(blkdev_issue_flush);

static void blkdev_discard_end_io(struct bio *bio, int err)
{
	if (err) {
		if (err == -EOPNOTSUPP)
			set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
	}

	if (bio->bi_private)
		complete(bio->bi_private);
	__free_page(bio_page(bio));

	bio_put(bio);
}

/**
 * blkdev_issue_discard - queue a discard
 * @bdev:	blockdev to issue discard for
 * @sector:	start sector
 * @nr_sects:	number of sectors to discard
 * @gfp_mask:	memory allocation flags (for bio_alloc)
 * @flags:	DISCARD_FL_* flags to control behaviour
 *
 * Description:
 *    Issue a discard request for the sectors in question.
 */
int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
		sector_t nr_sects, gfp_t gfp_mask, int flags)
{
	DECLARE_COMPLETION_ONSTACK(wait);
	struct request_queue *q = bdev_get_queue(bdev);
	int type = flags & DISCARD_FL_BARRIER ?
		DISCARD_BARRIER : DISCARD_NOBARRIER;
	struct bio *bio;
	struct page *page;
	int ret = 0;

	if (!q)
		return -ENXIO;

	if (!blk_queue_discard(q))
		return -EOPNOTSUPP;

	while (nr_sects && !ret) {
		unsigned int sector_size = q->limits.logical_block_size;
		unsigned int max_discard_sectors =
			min(q->limits.max_discard_sectors, UINT_MAX >> 9);

		bio = bio_alloc(gfp_mask, 1);
		if (!bio)
			goto out;
		bio->bi_sector = sector;
		bio->bi_end_io = blkdev_discard_end_io;
		bio->bi_bdev = bdev;
		if (flags & DISCARD_FL_WAIT)
			bio->bi_private = &wait;

		/*
		 * Add a zeroed one-sector payload as that's what
		 * our current implementations need.  If we'll ever need
		 * more the interface will need revisiting.
		 */
		page = alloc_page(gfp_mask | __GFP_ZERO);
		if (!page)
			goto out_free_bio;
		if (bio_add_pc_page(q, bio, page, sector_size, 0) < sector_size)
			goto out_free_page;

		/*
		 * And override the bio size - the way discard works we
		 * touch many more blocks on disk than the actual payload
		 * length.
		 */
		if (nr_sects > max_discard_sectors) {
			bio->bi_size = max_discard_sectors << 9;
			nr_sects -= max_discard_sectors;
			sector += max_discard_sectors;
		} else {
			bio->bi_size = nr_sects << 9;
			nr_sects = 0;
		}

		bio_get(bio);
		submit_bio(type, bio);

		if (flags & DISCARD_FL_WAIT)
			wait_for_completion(&wait);

		if (bio_flagged(bio, BIO_EOPNOTSUPP))
			ret = -EOPNOTSUPP;
		else if (!bio_flagged(bio, BIO_UPTODATE))
			ret = -EIO;
		bio_put(bio);
	}
	return ret;
out_free_page:
	__free_page(page);
out_free_bio:
	bio_put(bio);
out:
	return -ENOMEM;
}
EXPORT_SYMBOL(blkdev_issue_discard);
Commit	Line	Data
86db1e29 JA	1	/*
	2	* Functions related to barrier IO handling
	3	*/
	4	#include <linux/kernel.h>
	5	#include <linux/module.h>
	6	#include <linux/bio.h>
	7	#include <linux/blkdev.h>
5a0e3ad6	8	#include <linux/gfp.h>
86db1e29 JA	9
	10	#include "blk.h"
	11
	12	/**
	13	* blk_queue_ordered - does this queue support ordered writes
	14	* @q: the request queue
	15	* @ordered: one of QUEUE_ORDERED_*
	16	* @prepare_flush_fn: rq setup helper for cache flush ordered writes
	17	*
	18	* Description:
	19	* For journalled file systems, doing ordered writes on a commit
	20	* block instead of explicitly doing wait_on_buffer (which is bad
	21	* for performance) can be a big win. Block drivers supporting this
	22	* feature should call this function and indicate so.
	23	*
	24	**/
	25	int blk_queue_ordered(struct request_queue *q, unsigned ordered,
	26	prepare_flush_fn *prepare_flush_fn)
	27	{
313e4299 TH	28	if (!prepare_flush_fn && (ordered & (QUEUE_ORDERED_DO_PREFLUSH \|
313e4299 TH	29	QUEUE_ORDERED_DO_POSTFLUSH))) {
24c03d47	30	printk(KERN_ERR "%s: prepare_flush_fn required\n", __func__);
86db1e29 JA	31	return -EINVAL;
	32	}
	33
	34	if (ordered != QUEUE_ORDERED_NONE &&
	35	ordered != QUEUE_ORDERED_DRAIN &&
	36	ordered != QUEUE_ORDERED_DRAIN_FLUSH &&
	37	ordered != QUEUE_ORDERED_DRAIN_FUA &&
	38	ordered != QUEUE_ORDERED_TAG &&
	39	ordered != QUEUE_ORDERED_TAG_FLUSH &&
	40	ordered != QUEUE_ORDERED_TAG_FUA) {
	41	printk(KERN_ERR "blk_queue_ordered: bad value %d\n", ordered);
	42	return -EINVAL;
	43	}
	44
	45	q->ordered = ordered;
	46	q->next_ordered = ordered;
	47	q->prepare_flush_fn = prepare_flush_fn;
	48
	49	return 0;
	50	}
86db1e29 JA	51	EXPORT_SYMBOL(blk_queue_ordered);
	52
	53	/*
	54	* Cache flushing for ordered writes handling
	55	*/
6f6a036e	56	unsigned blk_ordered_cur_seq(struct request_queue *q)
86db1e29 JA	57	{
	58	if (!q->ordseq)
	59	return 0;
	60	return 1 << ffz(q->ordseq);
	61	}
	62
	63	unsigned blk_ordered_req_seq(struct request *rq)
	64	{
	65	struct request_queue *q = rq->q;
	66
	67	BUG_ON(q->ordseq == 0);
	68
	69	if (rq == &q->pre_flush_rq)
	70	return QUEUE_ORDSEQ_PREFLUSH;
	71	if (rq == &q->bar_rq)
	72	return QUEUE_ORDSEQ_BAR;
	73	if (rq == &q->post_flush_rq)
	74	return QUEUE_ORDSEQ_POSTFLUSH;
	75
	76	/*
	77	* !fs requests don't need to follow barrier ordering. Always
	78	* put them at the front. This fixes the following deadlock.
	79	*
	80	* http://thread.gmane.org/gmane.linux.kernel/537473
	81	*/
	82	if (!blk_fs_request(rq))
	83	return QUEUE_ORDSEQ_DRAIN;
	84
	85	if ((rq->cmd_flags & REQ_ORDERED_COLOR) ==
	86	(q->orig_bar_rq->cmd_flags & REQ_ORDERED_COLOR))
	87	return QUEUE_ORDSEQ_DRAIN;
	88	else
	89	return QUEUE_ORDSEQ_DONE;
	90	}
	91
8f11b3e9	92	bool blk_ordered_complete_seq(struct request_queue *q, unsigned seq, int error)
86db1e29 JA	93	{
	94	struct request *rq;
	95
	96	if (error && !q->orderr)
	97	q->orderr = error;
	98
	99	BUG_ON(q->ordseq & seq);
	100	q->ordseq \|= seq;
	101
	102	if (blk_ordered_cur_seq(q) != QUEUE_ORDSEQ_DONE)
8f11b3e9	103	return false;
86db1e29 JA	104
	105	/*
	106	* Okay, sequence complete.
	107	*/
	108	q->ordseq = 0;
	109	rq = q->orig_bar_rq;
40cbbb78	110	__blk_end_request_all(rq, q->orderr);
8f11b3e9	111	return true;
86db1e29 JA	112	}
	113
	114	static void pre_flush_end_io(struct request *rq, int error)
	115	{
	116	elv_completed_request(rq->q, rq);
	117	blk_ordered_complete_seq(rq->q, QUEUE_ORDSEQ_PREFLUSH, error);
	118	}
	119
	120	static void bar_end_io(struct request *rq, int error)
	121	{
	122	elv_completed_request(rq->q, rq);
	123	blk_ordered_complete_seq(rq->q, QUEUE_ORDSEQ_BAR, error);
	124	}
	125
	126	static void post_flush_end_io(struct request *rq, int error)
	127	{
	128	elv_completed_request(rq->q, rq);
	129	blk_ordered_complete_seq(rq->q, QUEUE_ORDSEQ_POSTFLUSH, error);
	130	}
	131
	132	static void queue_flush(struct request_queue *q, unsigned which)
	133	{
	134	struct request *rq;
	135	rq_end_io_fn *end_io;
	136
313e4299	137	if (which == QUEUE_ORDERED_DO_PREFLUSH) {
86db1e29 JA	138	rq = &q->pre_flush_rq;
	139	end_io = pre_flush_end_io;
	140	} else {
	141	rq = &q->post_flush_rq;
	142	end_io = post_flush_end_io;
	143	}
	144
2a4aa30c	145	blk_rq_init(q, rq);
1afb20f3	146	rq->cmd_flags = REQ_HARDBARRIER;
86db1e29 JA	147	rq->rq_disk = q->bar_rq.rq_disk;
	148	rq->end_io = end_io;
	149	q->prepare_flush_fn(q, rq);
	150
	151	elv_insert(q, rq, ELEVATOR_INSERT_FRONT);
	152	}
	153
8f11b3e9	154	static inline bool start_ordered(struct request_queue q, struct request *rqp)
86db1e29	155	{
8f11b3e9 TH	156	struct request rq = rqp;
	157	unsigned skip = 0;
	158
86db1e29 JA	159	q->orderr = 0;
	160	q->ordered = q->next_ordered;
	161	q->ordseq \|= QUEUE_ORDSEQ_STARTED;
	162
58eea927 TH	163	/*
	164	* For an empty barrier, there's no actual BAR request, which
	165	* in turn makes POSTFLUSH unnecessary. Mask them off.
	166	*/
5b93629b	167	if (!blk_rq_sectors(rq)) {
58eea927 TH	168	q->ordered &= ~(QUEUE_ORDERED_DO_BAR \|
58eea927 TH	169	QUEUE_ORDERED_DO_POSTFLUSH);
a185eb4b TH	170	/*
	171	* Empty barrier on a write-through device w/ ordered
	172	* tag has no command to issue and without any command
	173	* to issue, ordering by tag can't be used. Drain
	174	* instead.
	175	*/
	176	if ((q->ordered & QUEUE_ORDERED_BY_TAG) &&
	177	!(q->ordered & QUEUE_ORDERED_DO_PREFLUSH)) {
	178	q->ordered &= ~QUEUE_ORDERED_BY_TAG;
	179	q->ordered \|= QUEUE_ORDERED_BY_DRAIN;
	180	}
	181	}
58eea927	182
f671620e	183	/* stash away the original request */
9934c8c0	184	blk_dequeue_request(rq);
86db1e29	185	q->orig_bar_rq = rq;
f671620e	186	rq = NULL;
86db1e29 JA	187
	188	/*
	189	* Queue ordered sequence. As we stack them at the head, we
	190	* need to queue in reverse order. Note that we rely on that
	191	* no fs request uses ELEVATOR_INSERT_FRONT and thus no fs
58eea927	192	* request gets inbetween ordered sequence.
86db1e29	193	*/
58eea927	194	if (q->ordered & QUEUE_ORDERED_DO_POSTFLUSH) {
313e4299	195	queue_flush(q, QUEUE_ORDERED_DO_POSTFLUSH);
f671620e TH	196	rq = &q->post_flush_rq;
f671620e TH	197	} else
8f11b3e9	198	skip \|= QUEUE_ORDSEQ_POSTFLUSH;
86db1e29	199
f671620e TH	200	if (q->ordered & QUEUE_ORDERED_DO_BAR) {
	201	rq = &q->bar_rq;
	202
	203	/* initialize proxy request and queue it */
	204	blk_rq_init(q, rq);
	205	if (bio_data_dir(q->orig_bar_rq->bio) == WRITE)
	206	rq->cmd_flags \|= REQ_RW;
	207	if (q->ordered & QUEUE_ORDERED_DO_FUA)
	208	rq->cmd_flags \|= REQ_FUA;
	209	init_request_from_bio(rq, q->orig_bar_rq->bio);
	210	rq->end_io = bar_end_io;
	211
	212	elv_insert(q, rq, ELEVATOR_INSERT_FRONT);
	213	} else
8f11b3e9	214	skip \|= QUEUE_ORDSEQ_BAR;
86db1e29	215
313e4299 TH	216	if (q->ordered & QUEUE_ORDERED_DO_PREFLUSH) {
313e4299 TH	217	queue_flush(q, QUEUE_ORDERED_DO_PREFLUSH);
86db1e29 JA	218	rq = &q->pre_flush_rq;
86db1e29 JA	219	} else
8f11b3e9	220	skip \|= QUEUE_ORDSEQ_PREFLUSH;
86db1e29	221
0a7ae2ff	222	if ((q->ordered & QUEUE_ORDERED_BY_DRAIN) && queue_in_flight(q))
86db1e29	223	rq = NULL;
f671620e	224	else
8f11b3e9	225	skip \|= QUEUE_ORDSEQ_DRAIN;
86db1e29	226
8f11b3e9 TH	227	*rqp = rq;
	228
	229	/*
	230	* Complete skipped sequences. If whole sequence is complete,
	231	* return false to tell elevator that this request is gone.
	232	*/
	233	return !blk_ordered_complete_seq(q, skip, 0);
86db1e29 JA	234	}
86db1e29 JA	235
8f11b3e9	236	bool blk_do_ordered(struct request_queue q, struct request *rqp)
86db1e29 JA	237	{
	238	struct request rq = rqp;
	239	const int is_barrier = blk_fs_request(rq) && blk_barrier_rq(rq);
	240
	241	if (!q->ordseq) {
	242	if (!is_barrier)
8f11b3e9	243	return true;
86db1e29	244
8f11b3e9 TH	245	if (q->next_ordered != QUEUE_ORDERED_NONE)
	246	return start_ordered(q, rqp);
	247	else {
86db1e29	248	/*
a7384677 TH	249	* Queue ordering not supported. Terminate
a7384677 TH	250	* with prejudice.
86db1e29	251	*/
9934c8c0	252	blk_dequeue_request(rq);
40cbbb78	253	__blk_end_request_all(rq, -EOPNOTSUPP);
86db1e29	254	*rqp = NULL;
8f11b3e9	255	return false;
86db1e29 JA	256	}
	257	}
	258
	259	/*
	260	* Ordered sequence in progress
	261	*/
	262
	263	/* Special requests are not subject to ordering rules. */
	264	if (!blk_fs_request(rq) &&
	265	rq != &q->pre_flush_rq && rq != &q->post_flush_rq)
8f11b3e9	266	return true;
86db1e29	267
313e4299	268	if (q->ordered & QUEUE_ORDERED_BY_TAG) {
86db1e29 JA	269	/* Ordered by tag. Blocking the next barrier is enough. */
	270	if (is_barrier && rq != &q->bar_rq)
	271	*rqp = NULL;
	272	} else {
	273	/* Ordered by draining. Wait for turn. */
	274	WARN_ON(blk_ordered_req_seq(rq) < blk_ordered_cur_seq(q));
	275	if (blk_ordered_req_seq(rq) > blk_ordered_cur_seq(q))
	276	*rqp = NULL;
	277	}
	278
8f11b3e9	279	return true;
86db1e29 JA	280	}
	281
	282	static void bio_end_empty_barrier(struct bio *bio, int err)
	283	{
cc66b451 JA	284	if (err) {
	285	if (err == -EOPNOTSUPP)
	286	set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
86db1e29	287	clear_bit(BIO_UPTODATE, &bio->bi_flags);
cc66b451	288	}
86db1e29 JA	289
	290	complete(bio->bi_private);
	291	}
	292
	293	/**
	294	* blkdev_issue_flush - queue a flush
	295	* @bdev: blockdev to issue flush for
	296	* @error_sector: error sector
	297	*
	298	* Description:
	299	* Issue a flush for the block device in question. Caller can supply
	300	* room for storing the error offset in case of a flush error, if they
dbdac9b7	301	* wish to.
86db1e29 JA	302	*/
	303	int blkdev_issue_flush(struct block_device bdev, sector_t error_sector)
	304	{
	305	DECLARE_COMPLETION_ONSTACK(wait);
	306	struct request_queue *q;
	307	struct bio *bio;
	308	int ret;
	309
	310	if (bdev->bd_disk == NULL)
	311	return -ENXIO;
	312
	313	q = bdev_get_queue(bdev);
	314	if (!q)
	315	return -ENXIO;
	316
	317	bio = bio_alloc(GFP_KERNEL, 0);
86db1e29 JA	318	bio->bi_end_io = bio_end_empty_barrier;
	319	bio->bi_private = &wait;
	320	bio->bi_bdev = bdev;
2ebca85a	321	submit_bio(WRITE_BARRIER, bio);
86db1e29 JA	322
	323	wait_for_completion(&wait);
	324
	325	/*
	326	* The driver must store the error location in ->bi_sector, if
	327	* it supports it. For non-stacked drivers, this should be copied
83096ebf	328	* from blk_rq_pos(rq).
86db1e29 JA	329	*/
	330	if (error_sector)
	331	*error_sector = bio->bi_sector;
	332
	333	ret = 0;
cc66b451 JA	334	if (bio_flagged(bio, BIO_EOPNOTSUPP))
	335	ret = -EOPNOTSUPP;
	336	else if (!bio_flagged(bio, BIO_UPTODATE))
86db1e29 JA	337	ret = -EIO;
	338
	339	bio_put(bio);
	340	return ret;
	341	}
86db1e29	342	EXPORT_SYMBOL(blkdev_issue_flush);
fb2dce86 DW	343
	344	static void blkdev_discard_end_io(struct bio *bio, int err)
	345	{
	346	if (err) {
	347	if (err == -EOPNOTSUPP)
	348	set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
	349	clear_bit(BIO_UPTODATE, &bio->bi_flags);
	350	}
	351
746cd1e7 CH	352	if (bio->bi_private)
746cd1e7 CH	353	complete(bio->bi_private);
c15227de	354	__free_page(bio_page(bio));
746cd1e7	355
fb2dce86 DW	356	bio_put(bio);
	357	}
	358
	359	/**
	360	* blkdev_issue_discard - queue a discard
	361	* @bdev: blockdev to issue discard for
	362	* @sector: start sector
	363	* @nr_sects: number of sectors to discard
3e6053d7	364	* @gfp_mask: memory allocation flags (for bio_alloc)
746cd1e7	365	* @flags: DISCARD_FL_* flags to control behaviour
fb2dce86 DW	366	*
fb2dce86 DW	367	* Description:
746cd1e7	368	* Issue a discard request for the sectors in question.
fb2dce86	369	*/
746cd1e7 CH	370	int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
746cd1e7 CH	371	sector_t nr_sects, gfp_t gfp_mask, int flags)
fb2dce86	372	{
746cd1e7 CH	373	DECLARE_COMPLETION_ONSTACK(wait);
	374	struct request_queue *q = bdev_get_queue(bdev);
	375	int type = flags & DISCARD_FL_BARRIER ?
	376	DISCARD_BARRIER : DISCARD_NOBARRIER;
c15227de CH	377	struct bio *bio;
c15227de CH	378	struct page *page;
fb2dce86 DW	379	int ret = 0;
fb2dce86 DW	380
fb2dce86 DW	381	if (!q)
	382	return -ENXIO;
	383
c15227de	384	if (!blk_queue_discard(q))
fb2dce86 DW	385	return -EOPNOTSUPP;
	386
	387	while (nr_sects && !ret) {
c15227de	388	unsigned int sector_size = q->limits.logical_block_size;
67efc925 CH	389	unsigned int max_discard_sectors =
67efc925 CH	390	min(q->limits.max_discard_sectors, UINT_MAX >> 9);
fb2dce86	391
c15227de CH	392	bio = bio_alloc(gfp_mask, 1);
	393	if (!bio)
	394	goto out;
	395	bio->bi_sector = sector;
fb2dce86 DW	396	bio->bi_end_io = blkdev_discard_end_io;
fb2dce86 DW	397	bio->bi_bdev = bdev;
746cd1e7 CH	398	if (flags & DISCARD_FL_WAIT)
746cd1e7 CH	399	bio->bi_private = &wait;
fb2dce86	400
c15227de CH	401	/*
	402	* Add a zeroed one-sector payload as that's what
	403	* our current implementations need. If we'll ever need
	404	* more the interface will need revisiting.
	405	*/
e79e95db	406	page = alloc_page(gfp_mask \| __GFP_ZERO);
c15227de CH	407	if (!page)
	408	goto out_free_bio;
	409	if (bio_add_pc_page(q, bio, page, sector_size, 0) < sector_size)
	410	goto out_free_page;
fb2dce86	411
c15227de CH	412	/*
	413	* And override the bio size - the way discard works we
	414	* touch many more blocks on disk than the actual payload
	415	* length.
	416	*/
67efc925 CH	417	if (nr_sects > max_discard_sectors) {
	418	bio->bi_size = max_discard_sectors << 9;
	419	nr_sects -= max_discard_sectors;
	420	sector += max_discard_sectors;
fb2dce86 DW	421	} else {
	422	bio->bi_size = nr_sects << 9;
	423	nr_sects = 0;
	424	}
746cd1e7	425
fb2dce86	426	bio_get(bio);
746cd1e7 CH	427	submit_bio(type, bio);
	428
	429	if (flags & DISCARD_FL_WAIT)
	430	wait_for_completion(&wait);
fb2dce86	431
fb2dce86 DW	432	if (bio_flagged(bio, BIO_EOPNOTSUPP))
	433	ret = -EOPNOTSUPP;
	434	else if (!bio_flagged(bio, BIO_UPTODATE))
	435	ret = -EIO;
	436	bio_put(bio);
	437	}
	438	return ret;
c15227de CH	439	out_free_page:
	440	__free_page(page);
	441	out_free_bio:
	442	bio_put(bio);
	443	out:
	444	return -ENOMEM;
fb2dce86 DW	445	}
fb2dce86 DW	446	EXPORT_SYMBOL(blkdev_issue_discard);