[linux-block.git] / drivers / md / dm-delay.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * Copyright (C) 2005-2007 Red Hat GmbH
 *
 * A target that delays reads and/or writes and can send
 * them to different devices.
 *
 * This file is released under the GPL.
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/blkdev.h>
#include <linux/bio.h>
#include <linux/slab.h>
#include <linux/kthread.h>

#include <linux/device-mapper.h>

#define DM_MSG_PREFIX "delay"

struct delay_class {
	struct dm_dev *dev;
	sector_t start;
	unsigned int delay;
	unsigned int ops;
};

struct delay_c {
	struct timer_list delay_timer;
	struct mutex timer_lock;
	struct workqueue_struct *kdelayd_wq;
	struct work_struct flush_expired_bios;
	struct list_head delayed_bios;
	struct task_struct *worker;
	bool may_delay;

	struct delay_class read;
	struct delay_class write;
	struct delay_class flush;

	int argc;
};

struct dm_delay_info {
	struct delay_c *context;
	struct delay_class *class;
	struct list_head list;
	unsigned long expires;
};

static DEFINE_MUTEX(delayed_bios_lock);

static void handle_delayed_timer(struct timer_list *t)
{
	struct delay_c *dc = from_timer(dc, t, delay_timer);

	queue_work(dc->kdelayd_wq, &dc->flush_expired_bios);
}

static void queue_timeout(struct delay_c *dc, unsigned long expires)
{
	mutex_lock(&dc->timer_lock);

	if (!timer_pending(&dc->delay_timer) || expires < dc->delay_timer.expires)
		mod_timer(&dc->delay_timer, expires);

	mutex_unlock(&dc->timer_lock);
}

static inline bool delay_is_fast(struct delay_c *dc)
{
	return !!dc->worker;
}

static void flush_bios(struct bio *bio)
{
	struct bio *n;

	while (bio) {
		n = bio->bi_next;
		bio->bi_next = NULL;
		dm_submit_bio_remap(bio, NULL);
		bio = n;
	}
}

static void flush_delayed_bios(struct delay_c *dc, bool flush_all)
{
	struct dm_delay_info *delayed, *next;
	struct bio_list flush_bio_list;
	unsigned long next_expires = 0;
	bool start_timer = false;
	bio_list_init(&flush_bio_list);

	mutex_lock(&delayed_bios_lock);
	list_for_each_entry_safe(delayed, next, &dc->delayed_bios, list) {
		cond_resched();
		if (flush_all || time_after_eq(jiffies, delayed->expires)) {
			struct bio *bio = dm_bio_from_per_bio_data(delayed,
						sizeof(struct dm_delay_info));
			list_del(&delayed->list);
			bio_list_add(&flush_bio_list, bio);
			delayed->class->ops--;
			continue;
		}

		if (!delay_is_fast(dc)) {
			if (!start_timer) {
				start_timer = true;
				next_expires = delayed->expires;
			} else {
				next_expires = min(next_expires, delayed->expires);
			}
		}
	}
	mutex_unlock(&delayed_bios_lock);

	if (start_timer)
		queue_timeout(dc, next_expires);

	flush_bios(bio_list_get(&flush_bio_list));
}

static int flush_worker_fn(void *data)
{
	struct delay_c *dc = data;

	while (!kthread_should_stop()) {
		flush_delayed_bios(dc, false);
		mutex_lock(&delayed_bios_lock);
		if (unlikely(list_empty(&dc->delayed_bios))) {
			set_current_state(TASK_INTERRUPTIBLE);
			mutex_unlock(&delayed_bios_lock);
			schedule();
		} else {
			mutex_unlock(&delayed_bios_lock);
			cond_resched();
		}
	}

	return 0;
}

static void flush_expired_bios(struct work_struct *work)
{
	struct delay_c *dc;

	dc = container_of(work, struct delay_c, flush_expired_bios);
	flush_delayed_bios(dc, false);
}

static void delay_dtr(struct dm_target *ti)
{
	struct delay_c *dc = ti->private;

	if (dc->kdelayd_wq)
		destroy_workqueue(dc->kdelayd_wq);

	if (dc->read.dev)
		dm_put_device(ti, dc->read.dev);
	if (dc->write.dev)
		dm_put_device(ti, dc->write.dev);
	if (dc->flush.dev)
		dm_put_device(ti, dc->flush.dev);
	if (dc->worker)
		kthread_stop(dc->worker);

	mutex_destroy(&dc->timer_lock);

	kfree(dc);
}

static int delay_class_ctr(struct dm_target *ti, struct delay_class *c, char **argv)
{
	int ret;
	unsigned long long tmpll;
	char dummy;

	if (sscanf(argv[1], "%llu%c", &tmpll, &dummy) != 1 || tmpll != (sector_t)tmpll) {
		ti->error = "Invalid device sector";
		return -EINVAL;
	}
	c->start = tmpll;

	if (sscanf(argv[2], "%u%c", &c->delay, &dummy) != 1) {
		ti->error = "Invalid delay";
		return -EINVAL;
	}

	ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &c->dev);
	if (ret) {
		ti->error = "Device lookup failed";
		return ret;
	}

	return 0;
}

/*
 * Mapping parameters:
 *    <device> <offset> <delay> [<write_device> <write_offset> <write_delay>]
 *
 * With separate write parameters, the first set is only used for reads.
 * Offsets are specified in sectors.
 * Delays are specified in milliseconds.
 */
static int delay_ctr(struct dm_target *ti, unsigned int argc, char **argv)
{
	struct delay_c *dc;
	int ret;
	unsigned int max_delay;

	if (argc != 3 && argc != 6 && argc != 9) {
		ti->error = "Requires exactly 3, 6 or 9 arguments";
		return -EINVAL;
	}

	dc = kzalloc(sizeof(*dc), GFP_KERNEL);
	if (!dc) {
		ti->error = "Cannot allocate context";
		return -ENOMEM;
	}

	ti->private = dc;
	INIT_LIST_HEAD(&dc->delayed_bios);
	mutex_init(&dc->timer_lock);
	dc->may_delay = true;
	dc->argc = argc;

	ret = delay_class_ctr(ti, &dc->read, argv);
	if (ret)
		goto bad;
	max_delay = dc->read.delay;

	if (argc == 3) {
		ret = delay_class_ctr(ti, &dc->write, argv);
		if (ret)
			goto bad;
		ret = delay_class_ctr(ti, &dc->flush, argv);
		if (ret)
			goto bad;
		max_delay = max(max_delay, dc->write.delay);
		max_delay = max(max_delay, dc->flush.delay);
		goto out;
	}

	ret = delay_class_ctr(ti, &dc->write, argv + 3);
	if (ret)
		goto bad;
	if (argc == 6) {
		ret = delay_class_ctr(ti, &dc->flush, argv + 3);
		if (ret)
			goto bad;
		max_delay = max(max_delay, dc->flush.delay);
		goto out;
	}

	ret = delay_class_ctr(ti, &dc->flush, argv + 6);
	if (ret)
		goto bad;
	max_delay = max(max_delay, dc->flush.delay);

out:
	if (max_delay < 50) {
		/*
		 * In case of small requested delays, use kthread instead of
		 * timers and workqueue to achieve better latency.
		 */
		dc->worker = kthread_create(&flush_worker_fn, dc,
					    "dm-delay-flush-worker");
		if (IS_ERR(dc->worker)) {
			ret = PTR_ERR(dc->worker);
			dc->worker = NULL;
			goto bad;
		}
	} else {
		timer_setup(&dc->delay_timer, handle_delayed_timer, 0);
		INIT_WORK(&dc->flush_expired_bios, flush_expired_bios);
		dc->kdelayd_wq = alloc_workqueue("kdelayd", WQ_MEM_RECLAIM, 0);
		if (!dc->kdelayd_wq) {
			ret = -EINVAL;
			DMERR("Couldn't start kdelayd");
			goto bad;
		}
	}

	ti->num_flush_bios = 1;
	ti->num_discard_bios = 1;
	ti->accounts_remapped_io = true;
	ti->per_io_data_size = sizeof(struct dm_delay_info);
	return 0;

bad:
	delay_dtr(ti);
	return ret;
}

static int delay_bio(struct delay_c *dc, struct delay_class *c, struct bio *bio)
{
	struct dm_delay_info *delayed;
	unsigned long expires = 0;

	if (!c->delay)
		return DM_MAPIO_REMAPPED;

	delayed = dm_per_bio_data(bio, sizeof(struct dm_delay_info));

	delayed->context = dc;
	delayed->expires = expires = jiffies + msecs_to_jiffies(c->delay);

	mutex_lock(&delayed_bios_lock);
	if (unlikely(!dc->may_delay)) {
		mutex_unlock(&delayed_bios_lock);
		return DM_MAPIO_REMAPPED;
	}
	c->ops++;
	list_add_tail(&delayed->list, &dc->delayed_bios);
	mutex_unlock(&delayed_bios_lock);

	if (delay_is_fast(dc))
		wake_up_process(dc->worker);
	else
		queue_timeout(dc, expires);

	return DM_MAPIO_SUBMITTED;
}

static void delay_presuspend(struct dm_target *ti)
{
	struct delay_c *dc = ti->private;

	mutex_lock(&delayed_bios_lock);
	dc->may_delay = false;
	mutex_unlock(&delayed_bios_lock);

	if (!delay_is_fast(dc))
		del_timer_sync(&dc->delay_timer);
	flush_delayed_bios(dc, true);
}

static void delay_resume(struct dm_target *ti)
{
	struct delay_c *dc = ti->private;

	dc->may_delay = true;
}

static int delay_map(struct dm_target *ti, struct bio *bio)
{
	struct delay_c *dc = ti->private;
	struct delay_class *c;
	struct dm_delay_info *delayed = dm_per_bio_data(bio, sizeof(struct dm_delay_info));

	if (bio_data_dir(bio) == WRITE) {
		if (unlikely(bio->bi_opf & REQ_PREFLUSH))
			c = &dc->flush;
		else
			c = &dc->write;
	} else {
		c = &dc->read;
	}
	delayed->class = c;
	bio_set_dev(bio, c->dev->bdev);
	bio->bi_iter.bi_sector = c->start + dm_target_offset(ti, bio->bi_iter.bi_sector);

	return delay_bio(dc, c, bio);
}

#define DMEMIT_DELAY_CLASS(c) \
	DMEMIT("%s %llu %u", (c)->dev->name, (unsigned long long)(c)->start, (c)->delay)

static void delay_status(struct dm_target *ti, status_type_t type,
			 unsigned int status_flags, char *result, unsigned int maxlen)
{
	struct delay_c *dc = ti->private;
	int sz = 0;

	switch (type) {
	case STATUSTYPE_INFO:
		DMEMIT("%u %u %u", dc->read.ops, dc->write.ops, dc->flush.ops);
		break;

	case STATUSTYPE_TABLE:
		DMEMIT_DELAY_CLASS(&dc->read);
		if (dc->argc >= 6) {
			DMEMIT(" ");
			DMEMIT_DELAY_CLASS(&dc->write);
		}
		if (dc->argc >= 9) {
			DMEMIT(" ");
			DMEMIT_DELAY_CLASS(&dc->flush);
		}
		break;

	case STATUSTYPE_IMA:
		*result = '\0';
		break;
	}
}

static int delay_iterate_devices(struct dm_target *ti,
				 iterate_devices_callout_fn fn, void *data)
{
	struct delay_c *dc = ti->private;
	int ret = 0;

	ret = fn(ti, dc->read.dev, dc->read.start, ti->len, data);
	if (ret)
		goto out;
	ret = fn(ti, dc->write.dev, dc->write.start, ti->len, data);
	if (ret)
		goto out;
	ret = fn(ti, dc->flush.dev, dc->flush.start, ti->len, data);
	if (ret)
		goto out;

out:
	return ret;
}

static struct target_type delay_target = {
	.name	     = "delay",
	.version     = {1, 4, 0},
	.features    = DM_TARGET_PASSES_INTEGRITY,
	.module      = THIS_MODULE,
	.ctr	     = delay_ctr,
	.dtr	     = delay_dtr,
	.map	     = delay_map,
	.presuspend  = delay_presuspend,
	.resume	     = delay_resume,
	.status	     = delay_status,
	.iterate_devices = delay_iterate_devices,
};
module_dm(delay);

MODULE_DESCRIPTION(DM_NAME " delay target");
MODULE_AUTHOR("Heinz Mauelshagen <mauelshagen@redhat.com>");
MODULE_LICENSE("GPL");
Commit	Line	Data
3bd94003	1	// SPDX-License-Identifier: GPL-2.0-only
26b9f228 HM	2	/*
	3	* Copyright (C) 2005-2007 Red Hat GmbH
	4	*
	5	* A target that delays reads and/or writes and can send
	6	* them to different devices.
	7	*
	8	* This file is released under the GPL.
	9	*/
	10
	11	#include <linux/module.h>
	12	#include <linux/init.h>
	13	#include <linux/blkdev.h>
	14	#include <linux/bio.h>
	15	#include <linux/slab.h>
70bbeb29	16	#include <linux/kthread.h>
26b9f228	17
586e80e6 MP	18	#include <linux/device-mapper.h>
586e80e6 MP	19
26b9f228 HM	20	#define DM_MSG_PREFIX "delay"
26b9f228 HM	21
3876ac76 MP	22	struct delay_class {
	23	struct dm_dev *dev;
	24	sector_t start;
86a3238c HM	25	unsigned int delay;
86a3238c HM	26	unsigned int ops;
3876ac76 MP	27	};
3876ac76 MP	28
26b9f228 HM	29	struct delay_c {
26b9f228 HM	30	struct timer_list delay_timer;
ac818646	31	struct mutex timer_lock;
718822c1	32	struct workqueue_struct *kdelayd_wq;
26b9f228 HM	33	struct work_struct flush_expired_bios;
26b9f228 HM	34	struct list_head delayed_bios;
70bbeb29	35	struct task_struct *worker;
6fc45b6e	36	bool may_delay;
26b9f228	37
3876ac76 MP	38	struct delay_class read;
3876ac76 MP	39	struct delay_class write;
cda6b5ab	40	struct delay_class flush;
26b9f228	41
3876ac76	42	int argc;
26b9f228 HM	43	};
26b9f228 HM	44
028867ac	45	struct dm_delay_info {
26b9f228	46	struct delay_c *context;
3876ac76	47	struct delay_class *class;
26b9f228	48	struct list_head list;
26b9f228 HM	49	unsigned long expires;
	50	};
	51
	52	static DEFINE_MUTEX(delayed_bios_lock);
	53
8376d3c1	54	static void handle_delayed_timer(struct timer_list *t)
26b9f228	55	{
8376d3c1	56	struct delay_c *dc = from_timer(dc, t, delay_timer);
26b9f228	57
718822c1	58	queue_work(dc->kdelayd_wq, &dc->flush_expired_bios);
26b9f228 HM	59	}
	60
	61	static void queue_timeout(struct delay_c *dc, unsigned long expires)
	62	{
ac818646	63	mutex_lock(&dc->timer_lock);
26b9f228 HM	64
	65	if (!timer_pending(&dc->delay_timer) \|\| expires < dc->delay_timer.expires)
	66	mod_timer(&dc->delay_timer, expires);
	67
ac818646	68	mutex_unlock(&dc->timer_lock);
26b9f228 HM	69	}
26b9f228 HM	70
70bbeb29 CL	71	static inline bool delay_is_fast(struct delay_c *dc)
	72	{
	73	return !!dc->worker;
	74	}
	75
38cfff56 MP	76	static void flush_bios(struct bio *bio)
	77	{
	78	struct bio *n;
	79
	80	while (bio) {
	81	n = bio->bi_next;
	82	bio->bi_next = NULL;
	83	dm_submit_bio_remap(bio, NULL);
	84	bio = n;
	85	}
	86	}
	87
ccadc8a2	88	static void flush_delayed_bios(struct delay_c *dc, bool flush_all)
70bbeb29 CL	89	{
70bbeb29 CL	90	struct dm_delay_info delayed, next;
38cfff56	91	struct bio_list flush_bio_list;
ccadc8a2 MP	92	unsigned long next_expires = 0;
ccadc8a2 MP	93	bool start_timer = false;
38cfff56	94	bio_list_init(&flush_bio_list);
70bbeb29 CL	95
	96	mutex_lock(&delayed_bios_lock);
	97	list_for_each_entry_safe(delayed, next, &dc->delayed_bios, list) {
ccadc8a2	98	cond_resched();
70bbeb29 CL	99	if (flush_all \|\| time_after_eq(jiffies, delayed->expires)) {
	100	struct bio *bio = dm_bio_from_per_bio_data(delayed,
	101	sizeof(struct dm_delay_info));
	102	list_del(&delayed->list);
38cfff56	103	bio_list_add(&flush_bio_list, bio);
70bbeb29	104	delayed->class->ops--;
ccadc8a2 MP	105	continue;
	106	}
	107
	108	if (!delay_is_fast(dc)) {
	109	if (!start_timer) {
	110	start_timer = true;
	111	next_expires = delayed->expires;
	112	} else {
	113	next_expires = min(next_expires, delayed->expires);
	114	}
70bbeb29 CL	115	}
	116	}
	117	mutex_unlock(&delayed_bios_lock);
38cfff56	118
ccadc8a2 MP	119	if (start_timer)
	120	queue_timeout(dc, next_expires);
	121
38cfff56	122	flush_bios(bio_list_get(&flush_bio_list));
70bbeb29 CL	123	}
	124
	125	static int flush_worker_fn(void *data)
	126	{
	127	struct delay_c *dc = data;
	128
38cfff56	129	while (!kthread_should_stop()) {
ccadc8a2	130	flush_delayed_bios(dc, false);
38cfff56	131	mutex_lock(&delayed_bios_lock);
70bbeb29 CL	132	if (unlikely(list_empty(&dc->delayed_bios))) {
70bbeb29 CL	133	set_current_state(TASK_INTERRUPTIBLE);
38cfff56	134	mutex_unlock(&delayed_bios_lock);
70bbeb29	135	schedule();
38cfff56 MP	136	} else {
38cfff56 MP	137	mutex_unlock(&delayed_bios_lock);
70bbeb29	138	cond_resched();
38cfff56	139	}
70bbeb29 CL	140	}
	141
	142	return 0;
	143	}
	144
26b9f228 HM	145	static void flush_expired_bios(struct work_struct *work)
	146	{
	147	struct delay_c *dc;
	148
	149	dc = container_of(work, struct delay_c, flush_expired_bios);
ccadc8a2	150	flush_delayed_bios(dc, false);
26b9f228 HM	151	}
26b9f228 HM	152
3876ac76 MP	153	static void delay_dtr(struct dm_target *ti)
	154	{
	155	struct delay_c *dc = ti->private;
	156
81bc6d15 MP	157	if (dc->kdelayd_wq)
81bc6d15 MP	158	destroy_workqueue(dc->kdelayd_wq);
3876ac76 MP	159
	160	if (dc->read.dev)
	161	dm_put_device(ti, dc->read.dev);
	162	if (dc->write.dev)
	163	dm_put_device(ti, dc->write.dev);
cda6b5ab MP	164	if (dc->flush.dev)
cda6b5ab MP	165	dm_put_device(ti, dc->flush.dev);
70bbeb29 CL	166	if (dc->worker)
70bbeb29 CL	167	kthread_stop(dc->worker);
3876ac76	168
38cfff56	169	mutex_destroy(&dc->timer_lock);
3876ac76 MP	170
	171	kfree(dc);
	172	}
	173
	174	static int delay_class_ctr(struct dm_target ti, struct delay_class c, char **argv)
	175	{
	176	int ret;
	177	unsigned long long tmpll;
	178	char dummy;
	179
ef87bfc2	180	if (sscanf(argv[1], "%llu%c", &tmpll, &dummy) != 1 \|\| tmpll != (sector_t)tmpll) {
3876ac76 MP	181	ti->error = "Invalid device sector";
	182	return -EINVAL;
	183	}
	184	c->start = tmpll;
	185
	186	if (sscanf(argv[2], "%u%c", &c->delay, &dummy) != 1) {
	187	ti->error = "Invalid delay";
	188	return -EINVAL;
	189	}
	190
	191	ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &c->dev);
	192	if (ret) {
	193	ti->error = "Device lookup failed";
	194	return ret;
	195	}
	196
	197	return 0;
	198	}
	199
26b9f228 HM	200	/*
	201	* Mapping parameters:
	202	* <device> <offset> <delay> [<write_device> <write_offset> <write_delay>]
	203	*
	204	* With separate write parameters, the first set is only used for reads.
f49e869a	205	* Offsets are specified in sectors.
26b9f228 HM	206	* Delays are specified in milliseconds.
	207	*/
	208	static int delay_ctr(struct dm_target ti, unsigned int argc, char *argv)
	209	{
	210	struct delay_c *dc;
e80d1c80	211	int ret;
70bbeb29	212	unsigned int max_delay;
26b9f228	213
cda6b5ab MP	214	if (argc != 3 && argc != 6 && argc != 9) {
cda6b5ab MP	215	ti->error = "Requires exactly 3, 6 or 9 arguments";
26b9f228 HM	216	return -EINVAL;
	217	}
	218
3876ac76	219	dc = kzalloc(sizeof(*dc), GFP_KERNEL);
26b9f228 HM	220	if (!dc) {
	221	ti->error = "Cannot allocate context";
	222	return -ENOMEM;
	223	}
	224
3876ac76	225	ti->private = dc;
3876ac76	226	INIT_LIST_HEAD(&dc->delayed_bios);
38cfff56	227	mutex_init(&dc->timer_lock);
6fc45b6e	228	dc->may_delay = true;
3876ac76	229	dc->argc = argc;
26b9f228	230
3876ac76 MP	231	ret = delay_class_ctr(ti, &dc->read, argv);
3876ac76 MP	232	if (ret)
26b9f228	233	goto bad;
70bbeb29	234	max_delay = dc->read.delay;
26b9f228	235
3876ac76 MP	236	if (argc == 3) {
3876ac76 MP	237	ret = delay_class_ctr(ti, &dc->write, argv);
cda6b5ab MP	238	if (ret)
	239	goto bad;
	240	ret = delay_class_ctr(ti, &dc->flush, argv);
3876ac76 MP	241	if (ret)
3876ac76 MP	242	goto bad;
70bbeb29 CL	243	max_delay = max(max_delay, dc->write.delay);
70bbeb29 CL	244	max_delay = max(max_delay, dc->flush.delay);
26b9f228	245	goto out;
26b9f228	246	}
26b9f228	247
3876ac76 MP	248	ret = delay_class_ctr(ti, &dc->write, argv + 3);
	249	if (ret)
	250	goto bad;
cda6b5ab MP	251	if (argc == 6) {
	252	ret = delay_class_ctr(ti, &dc->flush, argv + 3);
	253	if (ret)
	254	goto bad;
70bbeb29	255	max_delay = max(max_delay, dc->flush.delay);
cda6b5ab MP	256	goto out;
	257	}
	258
	259	ret = delay_class_ctr(ti, &dc->flush, argv + 6);
	260	if (ret)
	261	goto bad;
70bbeb29	262	max_delay = max(max_delay, dc->flush.delay);
26b9f228 HM	263
26b9f228 HM	264	out:
70bbeb29 CL	265	if (max_delay < 50) {
	266	/*
	267	* In case of small requested delays, use kthread instead of
	268	* timers and workqueue to achieve better latency.
	269	*/
	270	dc->worker = kthread_create(&flush_worker_fn, dc,
	271	"dm-delay-flush-worker");
	272	if (IS_ERR(dc->worker)) {
	273	ret = PTR_ERR(dc->worker);
38cfff56	274	dc->worker = NULL;
70bbeb29 CL	275	goto bad;
	276	}
	277	} else {
	278	timer_setup(&dc->delay_timer, handle_delayed_timer, 0);
	279	INIT_WORK(&dc->flush_expired_bios, flush_expired_bios);
70bbeb29 CL	280	dc->kdelayd_wq = alloc_workqueue("kdelayd", WQ_MEM_RECLAIM, 0);
	281	if (!dc->kdelayd_wq) {
	282	ret = -EINVAL;
	283	DMERR("Couldn't start kdelayd");
	284	goto bad;
	285	}
718822c1 MP	286	}
718822c1 MP	287
55a62eef AK	288	ti->num_flush_bios = 1;
55a62eef AK	289	ti->num_discard_bios = 1;
c3573421	290	ti->accounts_remapped_io = true;
30187e1d	291	ti->per_io_data_size = sizeof(struct dm_delay_info);
26b9f228 HM	292	return 0;
	293
	294	bad:
3876ac76	295	delay_dtr(ti);
e80d1c80	296	return ret;
26b9f228 HM	297	}
26b9f228 HM	298
3876ac76	299	static int delay_bio(struct delay_c dc, struct delay_class c, struct bio *bio)
26b9f228	300	{
028867ac	301	struct dm_delay_info *delayed;
26b9f228 HM	302	unsigned long expires = 0;
26b9f228 HM	303
6fc45b6e	304	if (!c->delay)
340c9ec0	305	return DM_MAPIO_REMAPPED;
26b9f228	306
42065460	307	delayed = dm_per_bio_data(bio, sizeof(struct dm_delay_info));
26b9f228 HM	308
26b9f228 HM	309	delayed->context = dc;
3876ac76	310	delayed->expires = expires = jiffies + msecs_to_jiffies(c->delay);
26b9f228 HM	311
26b9f228 HM	312	mutex_lock(&delayed_bios_lock);
6fc45b6e MP	313	if (unlikely(!dc->may_delay)) {
	314	mutex_unlock(&delayed_bios_lock);
	315	return DM_MAPIO_REMAPPED;
	316	}
3876ac76	317	c->ops++;
26b9f228	318	list_add_tail(&delayed->list, &dc->delayed_bios);
26b9f228 HM	319	mutex_unlock(&delayed_bios_lock);
26b9f228 HM	320
70bbeb29 CL	321	if (delay_is_fast(dc))
	322	wake_up_process(dc->worker);
	323	else
	324	queue_timeout(dc, expires);
26b9f228	325
340c9ec0	326	return DM_MAPIO_SUBMITTED;
26b9f228 HM	327	}
	328
	329	static void delay_presuspend(struct dm_target *ti)
	330	{
	331	struct delay_c *dc = ti->private;
	332
6fc45b6e MP	333	mutex_lock(&delayed_bios_lock);
	334	dc->may_delay = false;
	335	mutex_unlock(&delayed_bios_lock);
70bbeb29	336
ccadc8a2	337	if (!delay_is_fast(dc))
70bbeb29	338	del_timer_sync(&dc->delay_timer);
ccadc8a2	339	flush_delayed_bios(dc, true);
26b9f228 HM	340	}
	341
	342	static void delay_resume(struct dm_target *ti)
	343	{
	344	struct delay_c *dc = ti->private;
	345
6fc45b6e	346	dc->may_delay = true;
26b9f228 HM	347	}
26b9f228 HM	348
7de3ee57	349	static int delay_map(struct dm_target ti, struct bio bio)
26b9f228 HM	350	{
26b9f228 HM	351	struct delay_c *dc = ti->private;
3876ac76 MP	352	struct delay_class *c;
3876ac76 MP	353	struct dm_delay_info *delayed = dm_per_bio_data(bio, sizeof(struct dm_delay_info));
26b9f228	354
3876ac76	355	if (bio_data_dir(bio) == WRITE) {
cda6b5ab MP	356	if (unlikely(bio->bi_opf & REQ_PREFLUSH))
	357	c = &dc->flush;
	358	else
	359	c = &dc->write;
3876ac76 MP	360	} else {
3876ac76 MP	361	c = &dc->read;
26b9f228	362	}
3876ac76 MP	363	delayed->class = c;
3876ac76 MP	364	bio_set_dev(bio, c->dev->bdev);
e86f2b00	365	bio->bi_iter.bi_sector = c->start + dm_target_offset(ti, bio->bi_iter.bi_sector);
26b9f228	366
3876ac76	367	return delay_bio(dc, c, bio);
26b9f228 HM	368	}
26b9f228 HM	369
3876ac76 MP	370	#define DMEMIT_DELAY_CLASS(c) \
	371	DMEMIT("%s %llu %u", (c)->dev->name, (unsigned long long)(c)->start, (c)->delay)
	372
fd7c092e	373	static void delay_status(struct dm_target *ti, status_type_t type,
86a3238c	374	unsigned int status_flags, char *result, unsigned int maxlen)
26b9f228 HM	375	{
	376	struct delay_c *dc = ti->private;
	377	int sz = 0;
	378
	379	switch (type) {
	380	case STATUSTYPE_INFO:
cda6b5ab	381	DMEMIT("%u %u %u", dc->read.ops, dc->write.ops, dc->flush.ops);
26b9f228 HM	382	break;
	383
	384	case STATUSTYPE_TABLE:
3876ac76 MP	385	DMEMIT_DELAY_CLASS(&dc->read);
	386	if (dc->argc >= 6) {
	387	DMEMIT(" ");
	388	DMEMIT_DELAY_CLASS(&dc->write);
	389	}
cda6b5ab MP	390	if (dc->argc >= 9) {
	391	DMEMIT(" ");
	392	DMEMIT_DELAY_CLASS(&dc->flush);
	393	}
26b9f228	394	break;
8ec45662 TS	395
	396	case STATUSTYPE_IMA:
	397	*result = '\0';
	398	break;
26b9f228	399	}
26b9f228 HM	400	}
26b9f228 HM	401
af4874e0 MS	402	static int delay_iterate_devices(struct dm_target *ti,
	403	iterate_devices_callout_fn fn, void *data)
	404	{
	405	struct delay_c *dc = ti->private;
	406	int ret = 0;
	407
3876ac76 MP	408	ret = fn(ti, dc->read.dev, dc->read.start, ti->len, data);
	409	if (ret)
	410	goto out;
	411	ret = fn(ti, dc->write.dev, dc->write.start, ti->len, data);
af4874e0 MS	412	if (ret)
af4874e0 MS	413	goto out;
cda6b5ab MP	414	ret = fn(ti, dc->flush.dev, dc->flush.start, ti->len, data);
	415	if (ret)
	416	goto out;
af4874e0 MS	417
	418	out:
	419	return ret;
	420	}
	421
26b9f228 HM	422	static struct target_type delay_target = {
26b9f228 HM	423	.name = "delay",
70bbeb29	424	.version = {1, 4, 0},
e2460f2a	425	.features = DM_TARGET_PASSES_INTEGRITY,
26b9f228 HM	426	.module = THIS_MODULE,
	427	.ctr = delay_ctr,
	428	.dtr = delay_dtr,
	429	.map = delay_map,
	430	.presuspend = delay_presuspend,
	431	.resume = delay_resume,
	432	.status = delay_status,
af4874e0	433	.iterate_devices = delay_iterate_devices,
26b9f228	434	};
3664ff82	435	module_dm(delay);
26b9f228 HM	436
	437	MODULE_DESCRIPTION(DM_NAME " delay target");
	438	MODULE_AUTHOR("Heinz Mauelshagen <mauelshagen@redhat.com>");
	439	MODULE_LICENSE("GPL");