[deliverable/linux.git] / drivers / block / null_blk.c

#include <linux/module.h>

#include <linux/moduleparam.h>
#include <linux/sched.h>
#include <linux/fs.h>
#include <linux/blkdev.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/blk-mq.h>
#include <linux/hrtimer.h>

struct nullb_cmd {
	struct list_head list;
	struct llist_node ll_list;
	struct call_single_data csd;
	struct request *rq;
	struct bio *bio;
	unsigned int tag;
	struct nullb_queue *nq;
};

struct nullb_queue {
	unsigned long *tag_map;
	wait_queue_head_t wait;
	unsigned int queue_depth;

	struct nullb_cmd *cmds;
};

struct nullb {
	struct list_head list;
	unsigned int index;
	struct request_queue *q;
	struct gendisk *disk;
	struct blk_mq_tag_set tag_set;
	struct hrtimer timer;
	unsigned int queue_depth;
	spinlock_t lock;

	struct nullb_queue *queues;
	unsigned int nr_queues;
};

static LIST_HEAD(nullb_list);
static struct mutex lock;
static int null_major;
static int nullb_indexes;

struct completion_queue {
	struct llist_head list;
	struct hrtimer timer;
};

/*
 * These are per-cpu for now, they will need to be configured by the
 * complete_queues parameter and appropriately mapped.
 */
static DEFINE_PER_CPU(struct completion_queue, completion_queues);

enum {
	NULL_IRQ_NONE		= 0,
	NULL_IRQ_SOFTIRQ	= 1,
	NULL_IRQ_TIMER		= 2,
};

enum {
	NULL_Q_BIO		= 0,
	NULL_Q_RQ		= 1,
	NULL_Q_MQ		= 2,
};

static int submit_queues;
module_param(submit_queues, int, S_IRUGO);
MODULE_PARM_DESC(submit_queues, "Number of submission queues");

static int home_node = NUMA_NO_NODE;
module_param(home_node, int, S_IRUGO);
MODULE_PARM_DESC(home_node, "Home node for the device");

static int queue_mode = NULL_Q_MQ;
module_param(queue_mode, int, S_IRUGO);
MODULE_PARM_DESC(queue_mode, "Block interface to use (0=bio,1=rq,2=multiqueue)");

static int gb = 250;
module_param(gb, int, S_IRUGO);
MODULE_PARM_DESC(gb, "Size in GB");

static int bs = 512;
module_param(bs, int, S_IRUGO);
MODULE_PARM_DESC(bs, "Block size (in bytes)");

static int nr_devices = 2;
module_param(nr_devices, int, S_IRUGO);
MODULE_PARM_DESC(nr_devices, "Number of devices to register");

static int irqmode = NULL_IRQ_SOFTIRQ;
module_param(irqmode, int, S_IRUGO);
MODULE_PARM_DESC(irqmode, "IRQ completion handler. 0-none, 1-softirq, 2-timer");

static int completion_nsec = 10000;
module_param(completion_nsec, int, S_IRUGO);
MODULE_PARM_DESC(completion_nsec, "Time in ns to complete a request in hardware. Default: 10,000ns");

static int hw_queue_depth = 64;
module_param(hw_queue_depth, int, S_IRUGO);
MODULE_PARM_DESC(hw_queue_depth, "Queue depth for each hardware queue. Default: 64");

static bool use_per_node_hctx = false;
module_param(use_per_node_hctx, bool, S_IRUGO);
MODULE_PARM_DESC(use_per_node_hctx, "Use per-node allocation for hardware context queues. Default: false");

static void put_tag(struct nullb_queue *nq, unsigned int tag)
{
	clear_bit_unlock(tag, nq->tag_map);

	if (waitqueue_active(&nq->wait))
		wake_up(&nq->wait);
}

static unsigned int get_tag(struct nullb_queue *nq)
{
	unsigned int tag;

	do {
		tag = find_first_zero_bit(nq->tag_map, nq->queue_depth);
		if (tag >= nq->queue_depth)
			return -1U;
	} while (test_and_set_bit_lock(tag, nq->tag_map));

	return tag;
}

static void free_cmd(struct nullb_cmd *cmd)
{
	put_tag(cmd->nq, cmd->tag);
}

static struct nullb_cmd *__alloc_cmd(struct nullb_queue *nq)
{
	struct nullb_cmd *cmd;
	unsigned int tag;

	tag = get_tag(nq);
	if (tag != -1U) {
		cmd = &nq->cmds[tag];
		cmd->tag = tag;
		cmd->nq = nq;
		return cmd;
	}

	return NULL;
}

static struct nullb_cmd *alloc_cmd(struct nullb_queue *nq, int can_wait)
{
	struct nullb_cmd *cmd;
	DEFINE_WAIT(wait);

	cmd = __alloc_cmd(nq);
	if (cmd || !can_wait)
		return cmd;

	do {
		prepare_to_wait(&nq->wait, &wait, TASK_UNINTERRUPTIBLE);
		cmd = __alloc_cmd(nq);
		if (cmd)
			break;

		io_schedule();
	} while (1);

	finish_wait(&nq->wait, &wait);
	return cmd;
}

static void end_cmd(struct nullb_cmd *cmd)
{
	switch (queue_mode)  {
	case NULL_Q_MQ:
		blk_mq_end_io(cmd->rq, 0);
		return;
	case NULL_Q_RQ:
		INIT_LIST_HEAD(&cmd->rq->queuelist);
		blk_end_request_all(cmd->rq, 0);
		break;
	case NULL_Q_BIO:
		bio_endio(cmd->bio, 0);
		break;
	}

	free_cmd(cmd);
}

static enum hrtimer_restart null_cmd_timer_expired(struct hrtimer *timer)
{
	struct completion_queue *cq;
	struct llist_node *entry;
	struct nullb_cmd *cmd;

	cq = &per_cpu(completion_queues, smp_processor_id());

	while ((entry = llist_del_all(&cq->list)) != NULL) {
		entry = llist_reverse_order(entry);
		do {
			cmd = container_of(entry, struct nullb_cmd, ll_list);
			entry = entry->next;
			end_cmd(cmd);
		} while (entry);
	}

	return HRTIMER_NORESTART;
}

static void null_cmd_end_timer(struct nullb_cmd *cmd)
{
	struct completion_queue *cq = &per_cpu(completion_queues, get_cpu());

	cmd->ll_list.next = NULL;
	if (llist_add(&cmd->ll_list, &cq->list)) {
		ktime_t kt = ktime_set(0, completion_nsec);

		hrtimer_start(&cq->timer, kt, HRTIMER_MODE_REL);
	}

	put_cpu();
}

static void null_softirq_done_fn(struct request *rq)
{
	if (queue_mode == NULL_Q_MQ)
		end_cmd(blk_mq_rq_to_pdu(rq));
	else
		end_cmd(rq->special);
}

static inline void null_handle_cmd(struct nullb_cmd *cmd)
{
	/* Complete IO by inline, softirq or timer */
	switch (irqmode) {
	case NULL_IRQ_SOFTIRQ:
		switch (queue_mode)  {
		case NULL_Q_MQ:
			blk_mq_complete_request(cmd->rq);
			break;
		case NULL_Q_RQ:
			blk_complete_request(cmd->rq);
			break;
		case NULL_Q_BIO:
			/*
			 * XXX: no proper submitting cpu information available.
			 */
			end_cmd(cmd);
			break;
		}
		break;
	case NULL_IRQ_NONE:
		end_cmd(cmd);
		break;
	case NULL_IRQ_TIMER:
		null_cmd_end_timer(cmd);
		break;
	}
}

static struct nullb_queue *nullb_to_queue(struct nullb *nullb)
{
	int index = 0;

	if (nullb->nr_queues != 1)
		index = raw_smp_processor_id() / ((nr_cpu_ids + nullb->nr_queues - 1) / nullb->nr_queues);

	return &nullb->queues[index];
}

static void null_queue_bio(struct request_queue *q, struct bio *bio)
{
	struct nullb *nullb = q->queuedata;
	struct nullb_queue *nq = nullb_to_queue(nullb);
	struct nullb_cmd *cmd;

	cmd = alloc_cmd(nq, 1);
	cmd->bio = bio;

	null_handle_cmd(cmd);
}

static int null_rq_prep_fn(struct request_queue *q, struct request *req)
{
	struct nullb *nullb = q->queuedata;
	struct nullb_queue *nq = nullb_to_queue(nullb);
	struct nullb_cmd *cmd;

	cmd = alloc_cmd(nq, 0);
	if (cmd) {
		cmd->rq = req;
		req->special = cmd;
		return BLKPREP_OK;
	}

	return BLKPREP_DEFER;
}

static void null_request_fn(struct request_queue *q)
{
	struct request *rq;

	while ((rq = blk_fetch_request(q)) != NULL) {
		struct nullb_cmd *cmd = rq->special;

		spin_unlock_irq(q->queue_lock);
		null_handle_cmd(cmd);
		spin_lock_irq(q->queue_lock);
	}
}

static int null_queue_rq(struct blk_mq_hw_ctx *hctx, struct request *rq,
		bool last)
{
	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(rq);

	cmd->rq = rq;
	cmd->nq = hctx->driver_data;

	null_handle_cmd(cmd);
	return BLK_MQ_RQ_QUEUE_OK;
}

static void null_init_queue(struct nullb *nullb, struct nullb_queue *nq)
{
	BUG_ON(!nullb);
	BUG_ON(!nq);

	init_waitqueue_head(&nq->wait);
	nq->queue_depth = nullb->queue_depth;
}

static int null_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
			  unsigned int index)
{
	struct nullb *nullb = data;
	struct nullb_queue *nq = &nullb->queues[index];

	hctx->driver_data = nq;
	null_init_queue(nullb, nq);
	nullb->nr_queues++;

	return 0;
}

static struct blk_mq_ops null_mq_ops = {
	.queue_rq       = null_queue_rq,
	.map_queue      = blk_mq_map_queue,
	.init_hctx	= null_init_hctx,
	.complete	= null_softirq_done_fn,
};

static void null_del_dev(struct nullb *nullb)
{
	list_del_init(&nullb->list);

	del_gendisk(nullb->disk);
	blk_cleanup_queue(nullb->q);
	if (queue_mode == NULL_Q_MQ)
		blk_mq_free_tag_set(&nullb->tag_set);
	put_disk(nullb->disk);
	kfree(nullb);
}

static int null_open(struct block_device *bdev, fmode_t mode)
{
	return 0;
}

static void null_release(struct gendisk *disk, fmode_t mode)
{
}

static const struct block_device_operations null_fops = {
	.owner =	THIS_MODULE,
	.open =		null_open,
	.release =	null_release,
};

static int setup_commands(struct nullb_queue *nq)
{
	struct nullb_cmd *cmd;
	int i, tag_size;

	nq->cmds = kzalloc(nq->queue_depth * sizeof(*cmd), GFP_KERNEL);
	if (!nq->cmds)
		return -ENOMEM;

	tag_size = ALIGN(nq->queue_depth, BITS_PER_LONG) / BITS_PER_LONG;
	nq->tag_map = kzalloc(tag_size * sizeof(unsigned long), GFP_KERNEL);
	if (!nq->tag_map) {
		kfree(nq->cmds);
		return -ENOMEM;
	}

	for (i = 0; i < nq->queue_depth; i++) {
		cmd = &nq->cmds[i];
		INIT_LIST_HEAD(&cmd->list);
		cmd->ll_list.next = NULL;
		cmd->tag = -1U;
	}

	return 0;
}

static void cleanup_queue(struct nullb_queue *nq)
{
	kfree(nq->tag_map);
	kfree(nq->cmds);
}

static void cleanup_queues(struct nullb *nullb)
{
	int i;

	for (i = 0; i < nullb->nr_queues; i++)
		cleanup_queue(&nullb->queues[i]);

	kfree(nullb->queues);
}

static int setup_queues(struct nullb *nullb)
{
	nullb->queues = kzalloc(submit_queues * sizeof(struct nullb_queue),
								GFP_KERNEL);
	if (!nullb->queues)
		return -ENOMEM;

	nullb->nr_queues = 0;
	nullb->queue_depth = hw_queue_depth;

	return 0;
}

static int init_driver_queues(struct nullb *nullb)
{
	struct nullb_queue *nq;
	int i, ret = 0;

	for (i = 0; i < submit_queues; i++) {
		nq = &nullb->queues[i];

		null_init_queue(nullb, nq);

		ret = setup_commands(nq);
		if (ret)
			goto err_queue;
		nullb->nr_queues++;
	}

	return 0;
err_queue:
	cleanup_queues(nullb);
	return ret;
}

static int null_add_dev(void)
{
	struct gendisk *disk;
	struct nullb *nullb;
	sector_t size;
	int rv;

	nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
	if (!nullb) {
		rv = -ENOMEM;
		goto out;
	}

	spin_lock_init(&nullb->lock);

	if (queue_mode == NULL_Q_MQ && use_per_node_hctx)
		submit_queues = nr_online_nodes;

	rv = setup_queues(nullb);
	if (rv)
		goto out_free_nullb;

	if (queue_mode == NULL_Q_MQ) {
		nullb->tag_set.ops = &null_mq_ops;
		nullb->tag_set.nr_hw_queues = submit_queues;
		nullb->tag_set.queue_depth = hw_queue_depth;
		nullb->tag_set.numa_node = home_node;
		nullb->tag_set.cmd_size	= sizeof(struct nullb_cmd);
		nullb->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
		nullb->tag_set.driver_data = nullb;

		rv = blk_mq_alloc_tag_set(&nullb->tag_set);
		if (rv)
			goto out_cleanup_queues;

		nullb->q = blk_mq_init_queue(&nullb->tag_set);
		if (!nullb->q) {
			rv = -ENOMEM;
			goto out_cleanup_tags;
		}
	} else if (queue_mode == NULL_Q_BIO) {
		nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
		if (!nullb->q) {
			rv = -ENOMEM;
			goto out_cleanup_queues;
		}
		blk_queue_make_request(nullb->q, null_queue_bio);
		init_driver_queues(nullb);
	} else {
		nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
		if (!nullb->q) {
			rv = -ENOMEM;
			goto out_cleanup_queues;
		}
		blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
		blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
		init_driver_queues(nullb);
	}

	nullb->q->queuedata = nullb;
	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);

	disk = nullb->disk = alloc_disk_node(1, home_node);
	if (!disk) {
		rv = -ENOMEM;
		goto out_cleanup_blk_queue;
	}

	mutex_lock(&lock);
	list_add_tail(&nullb->list, &nullb_list);
	nullb->index = nullb_indexes++;
	mutex_unlock(&lock);

	blk_queue_logical_block_size(nullb->q, bs);
	blk_queue_physical_block_size(nullb->q, bs);

	size = gb * 1024 * 1024 * 1024ULL;
	sector_div(size, bs);
	set_capacity(disk, size);

	disk->flags |= GENHD_FL_EXT_DEVT;
	disk->major		= null_major;
	disk->first_minor	= nullb->index;
	disk->fops		= &null_fops;
	disk->private_data	= nullb;
	disk->queue		= nullb->q;
	sprintf(disk->disk_name, "nullb%d", nullb->index);
	add_disk(disk);
	return 0;

out_cleanup_blk_queue:
	blk_cleanup_queue(nullb->q);
out_cleanup_tags:
	if (queue_mode == NULL_Q_MQ)
		blk_mq_free_tag_set(&nullb->tag_set);
out_cleanup_queues:
	cleanup_queues(nullb);
out_free_nullb:
	kfree(nullb);
out:
	return rv;
}

static int __init null_init(void)
{
	unsigned int i;

	if (bs > PAGE_SIZE) {
		pr_warn("null_blk: invalid block size\n");
		pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);
		bs = PAGE_SIZE;
	}

	if (queue_mode == NULL_Q_MQ && use_per_node_hctx) {
		if (submit_queues < nr_online_nodes) {
			pr_warn("null_blk: submit_queues param is set to %u.",
							nr_online_nodes);
			submit_queues = nr_online_nodes;
		}
	} else if (submit_queues > nr_cpu_ids)
		submit_queues = nr_cpu_ids;
	else if (!submit_queues)
		submit_queues = 1;

	mutex_init(&lock);

	/* Initialize a separate list for each CPU for issuing softirqs */
	for_each_possible_cpu(i) {
		struct completion_queue *cq = &per_cpu(completion_queues, i);

		init_llist_head(&cq->list);

		if (irqmode != NULL_IRQ_TIMER)
			continue;

		hrtimer_init(&cq->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
		cq->timer.function = null_cmd_timer_expired;
	}

	null_major = register_blkdev(0, "nullb");
	if (null_major < 0)
		return null_major;

	for (i = 0; i < nr_devices; i++) {
		if (null_add_dev()) {
			unregister_blkdev(null_major, "nullb");
			return -EINVAL;
		}
	}

	pr_info("null: module loaded\n");
	return 0;
}

static void __exit null_exit(void)
{
	struct nullb *nullb;

	unregister_blkdev(null_major, "nullb");

	mutex_lock(&lock);
	while (!list_empty(&nullb_list)) {
		nullb = list_entry(nullb_list.next, struct nullb, list);
		null_del_dev(nullb);
	}
	mutex_unlock(&lock);
}

module_init(null_init);
module_exit(null_exit);

MODULE_AUTHOR("Jens Axboe <jaxboe@fusionio.com>");
MODULE_LICENSE("GPL");
Commit	Line	Data
f2298c04	1	#include <linux/module.h>
fc1bc354	2
f2298c04 JA	3	#include <linux/moduleparam.h>
	4	#include <linux/sched.h>
	5	#include <linux/fs.h>
	6	#include <linux/blkdev.h>
	7	#include <linux/init.h>
	8	#include <linux/slab.h>
	9	#include <linux/blk-mq.h>
	10	#include <linux/hrtimer.h>
	11
	12	struct nullb_cmd {
	13	struct list_head list;
	14	struct llist_node ll_list;
	15	struct call_single_data csd;
	16	struct request *rq;
	17	struct bio *bio;
	18	unsigned int tag;
	19	struct nullb_queue *nq;
	20	};
	21
	22	struct nullb_queue {
	23	unsigned long *tag_map;
	24	wait_queue_head_t wait;
	25	unsigned int queue_depth;
	26
	27	struct nullb_cmd *cmds;
	28	};
	29
	30	struct nullb {
	31	struct list_head list;
	32	unsigned int index;
	33	struct request_queue *q;
	34	struct gendisk *disk;
24d2f903	35	struct blk_mq_tag_set tag_set;
f2298c04 JA	36	struct hrtimer timer;
	37	unsigned int queue_depth;
	38	spinlock_t lock;
	39
	40	struct nullb_queue *queues;
	41	unsigned int nr_queues;
	42	};
	43
	44	static LIST_HEAD(nullb_list);
	45	static struct mutex lock;
	46	static int null_major;
	47	static int nullb_indexes;
	48
	49	struct completion_queue {
	50	struct llist_head list;
	51	struct hrtimer timer;
	52	};
	53
	54	/*
	55	* These are per-cpu for now, they will need to be configured by the
	56	* complete_queues parameter and appropriately mapped.
	57	*/
	58	static DEFINE_PER_CPU(struct completion_queue, completion_queues);
	59
	60	enum {
	61	NULL_IRQ_NONE = 0,
	62	NULL_IRQ_SOFTIRQ = 1,
	63	NULL_IRQ_TIMER = 2,
ce2c350b	64	};
f2298c04	65
ce2c350b	66	enum {
f2298c04 JA	67	NULL_Q_BIO = 0,
	68	NULL_Q_RQ = 1,
	69	NULL_Q_MQ = 2,
	70	};
	71
2d263a78	72	static int submit_queues;
f2298c04 JA	73	module_param(submit_queues, int, S_IRUGO);
	74	MODULE_PARM_DESC(submit_queues, "Number of submission queues");
	75
	76	static int home_node = NUMA_NO_NODE;
	77	module_param(home_node, int, S_IRUGO);
	78	MODULE_PARM_DESC(home_node, "Home node for the device");
	79
	80	static int queue_mode = NULL_Q_MQ;
	81	module_param(queue_mode, int, S_IRUGO);
54ae81cd	82	MODULE_PARM_DESC(queue_mode, "Block interface to use (0=bio,1=rq,2=multiqueue)");
f2298c04 JA	83
	84	static int gb = 250;
	85	module_param(gb, int, S_IRUGO);
	86	MODULE_PARM_DESC(gb, "Size in GB");
	87
	88	static int bs = 512;
	89	module_param(bs, int, S_IRUGO);
	90	MODULE_PARM_DESC(bs, "Block size (in bytes)");
	91
	92	static int nr_devices = 2;
	93	module_param(nr_devices, int, S_IRUGO);
	94	MODULE_PARM_DESC(nr_devices, "Number of devices to register");
	95
	96	static int irqmode = NULL_IRQ_SOFTIRQ;
	97	module_param(irqmode, int, S_IRUGO);
	98	MODULE_PARM_DESC(irqmode, "IRQ completion handler. 0-none, 1-softirq, 2-timer");
	99
	100	static int completion_nsec = 10000;
	101	module_param(completion_nsec, int, S_IRUGO);
	102	MODULE_PARM_DESC(completion_nsec, "Time in ns to complete a request in hardware. Default: 10,000ns");
	103
	104	static int hw_queue_depth = 64;
	105	module_param(hw_queue_depth, int, S_IRUGO);
	106	MODULE_PARM_DESC(hw_queue_depth, "Queue depth for each hardware queue. Default: 64");
	107
20005244	108	static bool use_per_node_hctx = false;
f2298c04	109	module_param(use_per_node_hctx, bool, S_IRUGO);
20005244	110	MODULE_PARM_DESC(use_per_node_hctx, "Use per-node allocation for hardware context queues. Default: false");
f2298c04 JA	111
	112	static void put_tag(struct nullb_queue *nq, unsigned int tag)
	113	{
	114	clear_bit_unlock(tag, nq->tag_map);
	115
	116	if (waitqueue_active(&nq->wait))
	117	wake_up(&nq->wait);
	118	}
	119
	120	static unsigned int get_tag(struct nullb_queue *nq)
	121	{
	122	unsigned int tag;
	123
	124	do {
	125	tag = find_first_zero_bit(nq->tag_map, nq->queue_depth);
	126	if (tag >= nq->queue_depth)
	127	return -1U;
	128	} while (test_and_set_bit_lock(tag, nq->tag_map));
	129
	130	return tag;
	131	}
	132
	133	static void free_cmd(struct nullb_cmd *cmd)
	134	{
	135	put_tag(cmd->nq, cmd->tag);
	136	}
	137
	138	static struct nullb_cmd __alloc_cmd(struct nullb_queue nq)
	139	{
	140	struct nullb_cmd *cmd;
	141	unsigned int tag;
	142
	143	tag = get_tag(nq);
	144	if (tag != -1U) {
	145	cmd = &nq->cmds[tag];
	146	cmd->tag = tag;
	147	cmd->nq = nq;
	148	return cmd;
	149	}
	150
	151	return NULL;
	152	}
	153
	154	static struct nullb_cmd alloc_cmd(struct nullb_queue nq, int can_wait)
	155	{
	156	struct nullb_cmd *cmd;
	157	DEFINE_WAIT(wait);
	158
	159	cmd = __alloc_cmd(nq);
	160	if (cmd \|\| !can_wait)
	161	return cmd;
	162
	163	do {
	164	prepare_to_wait(&nq->wait, &wait, TASK_UNINTERRUPTIBLE);
	165	cmd = __alloc_cmd(nq);
	166	if (cmd)
	167	break;
	168
	169	io_schedule();
	170	} while (1);
	171
	172	finish_wait(&nq->wait, &wait);
	173	return cmd;
	174	}
175
176	static void end_cmd(struct nullb_cmd *cmd)
177	{
ce2c350b CH	178	switch (queue_mode) {
	179	case NULL_Q_MQ:
	180	blk_mq_end_io(cmd->rq, 0);
	181	return;
	182	case NULL_Q_RQ:
	183	INIT_LIST_HEAD(&cmd->rq->queuelist);
	184	blk_end_request_all(cmd->rq, 0);
	185	break;
	186	case NULL_Q_BIO:
f2298c04	187	bio_endio(cmd->bio, 0);
ce2c350b CH	188	break;
ce2c350b CH	189	}
f2298c04	190
ce2c350b	191	free_cmd(cmd);
f2298c04 JA	192	}
	193
	194	static enum hrtimer_restart null_cmd_timer_expired(struct hrtimer *timer)
	195	{
	196	struct completion_queue *cq;
	197	struct llist_node *entry;
	198	struct nullb_cmd *cmd;
	199
	200	cq = &per_cpu(completion_queues, smp_processor_id());
	201
	202	while ((entry = llist_del_all(&cq->list)) != NULL) {
d7790b92	203	entry = llist_reverse_order(entry);
f2298c04 JA	204	do {
f2298c04 JA	205	cmd = container_of(entry, struct nullb_cmd, ll_list);
f2298c04	206	entry = entry->next;
fc27691f	207	end_cmd(cmd);
f2298c04 JA	208	} while (entry);
	209	}
	210
	211	return HRTIMER_NORESTART;
	212	}
	213
	214	static void null_cmd_end_timer(struct nullb_cmd *cmd)
	215	{
	216	struct completion_queue *cq = &per_cpu(completion_queues, get_cpu());
	217
	218	cmd->ll_list.next = NULL;
	219	if (llist_add(&cmd->ll_list, &cq->list)) {
	220	ktime_t kt = ktime_set(0, completion_nsec);
	221
	222	hrtimer_start(&cq->timer, kt, HRTIMER_MODE_REL);
	223	}
	224
	225	put_cpu();
	226	}
	227
	228	static void null_softirq_done_fn(struct request *rq)
	229	{
d891fa70 JA	230	if (queue_mode == NULL_Q_MQ)
	231	end_cmd(blk_mq_rq_to_pdu(rq));
	232	else
	233	end_cmd(rq->special);
f2298c04 JA	234	}
f2298c04 JA	235
f2298c04 JA	236	static inline void null_handle_cmd(struct nullb_cmd *cmd)
	237	{
	238	/* Complete IO by inline, softirq or timer */
	239	switch (irqmode) {
f2298c04	240	case NULL_IRQ_SOFTIRQ:
ce2c350b CH	241	switch (queue_mode) {
	242	case NULL_Q_MQ:
	243	blk_mq_complete_request(cmd->rq);
	244	break;
	245	case NULL_Q_RQ:
	246	blk_complete_request(cmd->rq);
	247	break;
	248	case NULL_Q_BIO:
	249	/*
	250	* XXX: no proper submitting cpu information available.
	251	*/
	252	end_cmd(cmd);
	253	break;
	254	}
	255	break;
	256	case NULL_IRQ_NONE:
f2298c04	257	end_cmd(cmd);
f2298c04 JA	258	break;
	259	case NULL_IRQ_TIMER:
	260	null_cmd_end_timer(cmd);
	261	break;
	262	}
	263	}
	264
	265	static struct nullb_queue nullb_to_queue(struct nullb nullb)
	266	{
	267	int index = 0;
	268
	269	if (nullb->nr_queues != 1)
	270	index = raw_smp_processor_id() / ((nr_cpu_ids + nullb->nr_queues - 1) / nullb->nr_queues);
	271
	272	return &nullb->queues[index];
	273	}
	274
	275	static void null_queue_bio(struct request_queue q, struct bio bio)
	276	{
	277	struct nullb *nullb = q->queuedata;
	278	struct nullb_queue *nq = nullb_to_queue(nullb);
	279	struct nullb_cmd *cmd;
	280
	281	cmd = alloc_cmd(nq, 1);
	282	cmd->bio = bio;
	283
	284	null_handle_cmd(cmd);
	285	}
	286
	287	static int null_rq_prep_fn(struct request_queue q, struct request req)
	288	{
	289	struct nullb *nullb = q->queuedata;
	290	struct nullb_queue *nq = nullb_to_queue(nullb);
	291	struct nullb_cmd *cmd;
	292
	293	cmd = alloc_cmd(nq, 0);
	294	if (cmd) {
	295	cmd->rq = req;
	296	req->special = cmd;
	297	return BLKPREP_OK;
	298	}
	299
	300	return BLKPREP_DEFER;
	301	}
	302
	303	static void null_request_fn(struct request_queue *q)
	304	{
	305	struct request *rq;
	306
	307	while ((rq = blk_fetch_request(q)) != NULL) {
	308	struct nullb_cmd *cmd = rq->special;
	309
	310	spin_unlock_irq(q->queue_lock);
	311	null_handle_cmd(cmd);
	312	spin_lock_irq(q->queue_lock);
	313	}
	314	}
	315
bf572297 CH	316	static int null_queue_rq(struct blk_mq_hw_ctx hctx, struct request rq,
bf572297 CH	317	bool last)
f2298c04	318	{
9d74e257	319	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(rq);
f2298c04 JA	320
	321	cmd->rq = rq;
	322	cmd->nq = hctx->driver_data;
	323
	324	null_handle_cmd(cmd);
	325	return BLK_MQ_RQ_QUEUE_OK;
	326	}
	327
2d263a78 MB	328	static void null_init_queue(struct nullb nullb, struct nullb_queue nq)
	329	{
	330	BUG_ON(!nullb);
	331	BUG_ON(!nq);
	332
	333	init_waitqueue_head(&nq->wait);
	334	nq->queue_depth = nullb->queue_depth;
	335	}
	336
f2298c04 JA	337	static int null_init_hctx(struct blk_mq_hw_ctx hctx, void data,
	338	unsigned int index)
	339	{
	340	struct nullb *nullb = data;
	341	struct nullb_queue *nq = &nullb->queues[index];
	342
f2298c04	343	hctx->driver_data = nq;
2d263a78 MB	344	null_init_queue(nullb, nq);
2d263a78 MB	345	nullb->nr_queues++;
f2298c04 JA	346
	347	return 0;
	348	}
	349
	350	static struct blk_mq_ops null_mq_ops = {
	351	.queue_rq = null_queue_rq,
	352	.map_queue = blk_mq_map_queue,
	353	.init_hctx = null_init_hctx,
ce2c350b	354	.complete = null_softirq_done_fn,
f2298c04 JA	355	};
	356
	357	static void null_del_dev(struct nullb *nullb)
	358	{
	359	list_del_init(&nullb->list);
	360
	361	del_gendisk(nullb->disk);
518d00b7	362	blk_cleanup_queue(nullb->q);
24d2f903 CH	363	if (queue_mode == NULL_Q_MQ)
24d2f903 CH	364	blk_mq_free_tag_set(&nullb->tag_set);
f2298c04 JA	365	put_disk(nullb->disk);
	366	kfree(nullb);
	367	}
	368
	369	static int null_open(struct block_device *bdev, fmode_t mode)
	370	{
	371	return 0;
	372	}
	373
	374	static void null_release(struct gendisk *disk, fmode_t mode)
	375	{
	376	}
	377
	378	static const struct block_device_operations null_fops = {
	379	.owner = THIS_MODULE,
	380	.open = null_open,
	381	.release = null_release,
	382	};
	383
	384	static int setup_commands(struct nullb_queue *nq)
	385	{
	386	struct nullb_cmd *cmd;
	387	int i, tag_size;
	388
	389	nq->cmds = kzalloc(nq->queue_depth * sizeof(*cmd), GFP_KERNEL);
	390	if (!nq->cmds)
2d263a78	391	return -ENOMEM;
f2298c04 JA	392
	393	tag_size = ALIGN(nq->queue_depth, BITS_PER_LONG) / BITS_PER_LONG;
	394	nq->tag_map = kzalloc(tag_size * sizeof(unsigned long), GFP_KERNEL);
	395	if (!nq->tag_map) {
	396	kfree(nq->cmds);
2d263a78	397	return -ENOMEM;
f2298c04 JA	398	}
	399
	400	for (i = 0; i < nq->queue_depth; i++) {
	401	cmd = &nq->cmds[i];
	402	INIT_LIST_HEAD(&cmd->list);
	403	cmd->ll_list.next = NULL;
	404	cmd->tag = -1U;
	405	}
	406
	407	return 0;
	408	}
	409
	410	static void cleanup_queue(struct nullb_queue *nq)
	411	{
	412	kfree(nq->tag_map);
	413	kfree(nq->cmds);
	414	}
	415
	416	static void cleanup_queues(struct nullb *nullb)
	417	{
	418	int i;
	419
	420	for (i = 0; i < nullb->nr_queues; i++)
	421	cleanup_queue(&nullb->queues[i]);
	422
	423	kfree(nullb->queues);
	424	}
	425
	426	static int setup_queues(struct nullb *nullb)
	427	{
2d263a78 MB	428	nullb->queues = kzalloc(submit_queues * sizeof(struct nullb_queue),
2d263a78 MB	429	GFP_KERNEL);
f2298c04	430	if (!nullb->queues)
2d263a78	431	return -ENOMEM;
f2298c04 JA	432
	433	nullb->nr_queues = 0;
	434	nullb->queue_depth = hw_queue_depth;
	435
2d263a78 MB	436	return 0;
	437	}
	438
	439	static int init_driver_queues(struct nullb *nullb)
	440	{
	441	struct nullb_queue *nq;
	442	int i, ret = 0;
f2298c04 JA	443
	444	for (i = 0; i < submit_queues; i++) {
	445	nq = &nullb->queues[i];
2d263a78 MB	446
	447	null_init_queue(nullb, nq);
	448
	449	ret = setup_commands(nq);
	450	if (ret)
	451	goto err_queue;
f2298c04 JA	452	nullb->nr_queues++;
	453	}
	454
2d263a78 MB	455	return 0;
2d263a78 MB	456	err_queue:
f2298c04	457	cleanup_queues(nullb);
2d263a78	458	return ret;
f2298c04 JA	459	}
	460
	461	static int null_add_dev(void)
	462	{
	463	struct gendisk *disk;
	464	struct nullb *nullb;
	465	sector_t size;
dc501dc0	466	int rv;
f2298c04 JA	467
f2298c04 JA	468	nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
dc501dc0 RE	469	if (!nullb) {
dc501dc0 RE	470	rv = -ENOMEM;
24d2f903	471	goto out;
dc501dc0	472	}
f2298c04 JA	473
	474	spin_lock_init(&nullb->lock);
	475
57053d8c MB	476	if (queue_mode == NULL_Q_MQ && use_per_node_hctx)
	477	submit_queues = nr_online_nodes;
	478
dc501dc0 RE	479	rv = setup_queues(nullb);
dc501dc0 RE	480	if (rv)
24d2f903	481	goto out_free_nullb;
f2298c04 JA	482
f2298c04 JA	483	if (queue_mode == NULL_Q_MQ) {
cdef54dd	484	nullb->tag_set.ops = &null_mq_ops;
24d2f903 CH	485	nullb->tag_set.nr_hw_queues = submit_queues;
	486	nullb->tag_set.queue_depth = hw_queue_depth;
	487	nullb->tag_set.numa_node = home_node;
	488	nullb->tag_set.cmd_size = sizeof(struct nullb_cmd);
	489	nullb->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
	490	nullb->tag_set.driver_data = nullb;
	491
dc501dc0 RE	492	rv = blk_mq_alloc_tag_set(&nullb->tag_set);
dc501dc0 RE	493	if (rv)
24d2f903 CH	494	goto out_cleanup_queues;
	495
	496	nullb->q = blk_mq_init_queue(&nullb->tag_set);
dc501dc0 RE	497	if (!nullb->q) {
dc501dc0 RE	498	rv = -ENOMEM;
24d2f903	499	goto out_cleanup_tags;
dc501dc0	500	}
f2298c04 JA	501	} else if (queue_mode == NULL_Q_BIO) {
f2298c04 JA	502	nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
dc501dc0 RE	503	if (!nullb->q) {
dc501dc0 RE	504	rv = -ENOMEM;
24d2f903	505	goto out_cleanup_queues;
dc501dc0	506	}
f2298c04	507	blk_queue_make_request(nullb->q, null_queue_bio);
2d263a78	508	init_driver_queues(nullb);
f2298c04 JA	509	} else {
f2298c04 JA	510	nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
dc501dc0 RE	511	if (!nullb->q) {
dc501dc0 RE	512	rv = -ENOMEM;
24d2f903	513	goto out_cleanup_queues;
dc501dc0	514	}
f2298c04	515	blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
24d2f903	516	blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
2d263a78	517	init_driver_queues(nullb);
f2298c04 JA	518	}
f2298c04 JA	519
f2298c04 JA	520	nullb->q->queuedata = nullb;
	521	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);
	522
	523	disk = nullb->disk = alloc_disk_node(1, home_node);
dc501dc0 RE	524	if (!disk) {
dc501dc0 RE	525	rv = -ENOMEM;
24d2f903	526	goto out_cleanup_blk_queue;
dc501dc0	527	}
f2298c04 JA	528
	529	mutex_lock(&lock);
	530	list_add_tail(&nullb->list, &nullb_list);
	531	nullb->index = nullb_indexes++;
	532	mutex_unlock(&lock);
	533
	534	blk_queue_logical_block_size(nullb->q, bs);
	535	blk_queue_physical_block_size(nullb->q, bs);
	536
	537	size = gb * 1024 * 1024 * 1024ULL;
	538	sector_div(size, bs);
	539	set_capacity(disk, size);
	540
	541	disk->flags \|= GENHD_FL_EXT_DEVT;
	542	disk->major = null_major;
	543	disk->first_minor = nullb->index;
	544	disk->fops = &null_fops;
	545	disk->private_data = nullb;
	546	disk->queue = nullb->q;
	547	sprintf(disk->disk_name, "nullb%d", nullb->index);
	548	add_disk(disk);
	549	return 0;
24d2f903 CH	550
	551	out_cleanup_blk_queue:
	552	blk_cleanup_queue(nullb->q);
	553	out_cleanup_tags:
	554	if (queue_mode == NULL_Q_MQ)
	555	blk_mq_free_tag_set(&nullb->tag_set);
	556	out_cleanup_queues:
	557	cleanup_queues(nullb);
	558	out_free_nullb:
	559	kfree(nullb);
	560	out:
dc501dc0	561	return rv;
f2298c04 JA	562	}
	563
	564	static int __init null_init(void)
	565	{
	566	unsigned int i;
	567
9967d8ac R	568	if (bs > PAGE_SIZE) {
	569	pr_warn("null_blk: invalid block size\n");
	570	pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);
	571	bs = PAGE_SIZE;
	572	}
f2298c04	573
d15ee6b1	574	if (queue_mode == NULL_Q_MQ && use_per_node_hctx) {
fc1bc354	575	if (submit_queues < nr_online_nodes) {
d15ee6b1 MB	576	pr_warn("null_blk: submit_queues param is set to %u.",
d15ee6b1 MB	577	nr_online_nodes);
fc1bc354 MB	578	submit_queues = nr_online_nodes;
fc1bc354 MB	579	}
d15ee6b1	580	} else if (submit_queues > nr_cpu_ids)
f2298c04 JA	581	submit_queues = nr_cpu_ids;
	582	else if (!submit_queues)
	583	submit_queues = 1;
	584
	585	mutex_init(&lock);
	586
	587	/* Initialize a separate list for each CPU for issuing softirqs */
	588	for_each_possible_cpu(i) {
	589	struct completion_queue *cq = &per_cpu(completion_queues, i);
	590
	591	init_llist_head(&cq->list);
	592
	593	if (irqmode != NULL_IRQ_TIMER)
	594	continue;
	595
	596	hrtimer_init(&cq->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	597	cq->timer.function = null_cmd_timer_expired;
	598	}
	599
	600	null_major = register_blkdev(0, "nullb");
	601	if (null_major < 0)
	602	return null_major;
	603
	604	for (i = 0; i < nr_devices; i++) {
	605	if (null_add_dev()) {
	606	unregister_blkdev(null_major, "nullb");
	607	return -EINVAL;
	608	}
	609	}
	610
	611	pr_info("null: module loaded\n");
	612	return 0;
	613	}
	614
	615	static void __exit null_exit(void)
	616	{
	617	struct nullb *nullb;
	618
	619	unregister_blkdev(null_major, "nullb");
	620
	621	mutex_lock(&lock);
	622	while (!list_empty(&nullb_list)) {
	623	nullb = list_entry(nullb_list.next, struct nullb, list);
	624	null_del_dev(nullb);
	625	}
	626	mutex_unlock(&lock);
	627	}
	628
	629	module_init(null_init);
	630	module_exit(null_exit);
	631
	632	MODULE_AUTHOR("Jens Axboe <jaxboe@fusionio.com>");
	633	MODULE_LICENSE("GPL");