[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(use_mq, "Use blk-mq interface (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);
			end_cmd(cmd);
			entry = entry->next;
		} 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)
{
	end_cmd(blk_mq_rq_to_pdu(rq));
}

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)
{
	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 struct blk_mq_hw_ctx *null_alloc_hctx(struct blk_mq_tag_set *set,
					     unsigned int hctx_index,
					     int node)
{
	return kzalloc_node(sizeof(struct blk_mq_hw_ctx), GFP_KERNEL, node);
}

static void null_free_hctx(struct blk_mq_hw_ctx *hctx, unsigned int hctx_index)
{
	kfree(hctx);
}

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,
	.alloc_hctx	= blk_mq_alloc_single_hw_queue,
	.free_hctx	= blk_mq_free_single_hw_queue,
};

static struct blk_mq_ops null_mq_ops_pernode = {
	.queue_rq       = null_queue_rq,
	.map_queue      = blk_mq_map_queue,
	.init_hctx	= null_init_hctx,
	.complete	= null_softirq_done_fn,
	.alloc_hctx	= null_alloc_hctx,
	.free_hctx	= null_free_hctx,
};

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;

	nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
	if (!nullb)
		goto out;

	spin_lock_init(&nullb->lock);

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

	if (setup_queues(nullb))
		goto out_free_nullb;

	if (queue_mode == NULL_Q_MQ) {
		if (use_per_node_hctx)
			nullb->tag_set.ops = &null_mq_ops_pernode;
		else
			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;

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

		nullb->q = blk_mq_init_queue(&nullb->tag_set);
		if (!nullb->q)
			goto out_cleanup_tags;
	} else if (queue_mode == NULL_Q_BIO) {
		nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
		if (!nullb->q)
			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)
			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)
		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 -ENOMEM;
}

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);
	82	MODULE_PARM_DESC(use_mq, "Use blk-mq interface (0=bio,1=rq,2=multiqueue)");
	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 {
	205	cmd = container_of(entry, struct nullb_cmd, ll_list);
	206	end_cmd(cmd);
	207	entry = entry->next;
	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	{
9d74e257	230	end_cmd(blk_mq_rq_to_pdu(rq));
f2298c04 JA	231	}
f2298c04 JA	232
f2298c04 JA	233	static inline void null_handle_cmd(struct nullb_cmd *cmd)
	234	{
	235	/* Complete IO by inline, softirq or timer */
	236	switch (irqmode) {
f2298c04	237	case NULL_IRQ_SOFTIRQ:
ce2c350b CH	238	switch (queue_mode) {
	239	case NULL_Q_MQ:
	240	blk_mq_complete_request(cmd->rq);
	241	break;
	242	case NULL_Q_RQ:
	243	blk_complete_request(cmd->rq);
	244	break;
	245	case NULL_Q_BIO:
	246	/*
	247	* XXX: no proper submitting cpu information available.
	248	*/
	249	end_cmd(cmd);
	250	break;
	251	}
	252	break;
	253	case NULL_IRQ_NONE:
f2298c04	254	end_cmd(cmd);
f2298c04 JA	255	break;
	256	case NULL_IRQ_TIMER:
	257	null_cmd_end_timer(cmd);
	258	break;
	259	}
	260	}
	261
	262	static struct nullb_queue nullb_to_queue(struct nullb nullb)
	263	{
	264	int index = 0;
	265
	266	if (nullb->nr_queues != 1)
	267	index = raw_smp_processor_id() / ((nr_cpu_ids + nullb->nr_queues - 1) / nullb->nr_queues);
	268
	269	return &nullb->queues[index];
	270	}
	271
	272	static void null_queue_bio(struct request_queue q, struct bio bio)
	273	{
	274	struct nullb *nullb = q->queuedata;
	275	struct nullb_queue *nq = nullb_to_queue(nullb);
	276	struct nullb_cmd *cmd;
	277
	278	cmd = alloc_cmd(nq, 1);
	279	cmd->bio = bio;
	280
	281	null_handle_cmd(cmd);
	282	}
	283
	284	static int null_rq_prep_fn(struct request_queue q, struct request req)
	285	{
	286	struct nullb *nullb = q->queuedata;
	287	struct nullb_queue *nq = nullb_to_queue(nullb);
	288	struct nullb_cmd *cmd;
	289
	290	cmd = alloc_cmd(nq, 0);
	291	if (cmd) {
	292	cmd->rq = req;
	293	req->special = cmd;
	294	return BLKPREP_OK;
	295	}
	296
	297	return BLKPREP_DEFER;
	298	}
	299
	300	static void null_request_fn(struct request_queue *q)
	301	{
	302	struct request *rq;
	303
	304	while ((rq = blk_fetch_request(q)) != NULL) {
	305	struct nullb_cmd *cmd = rq->special;
	306
	307	spin_unlock_irq(q->queue_lock);
	308	null_handle_cmd(cmd);
	309	spin_lock_irq(q->queue_lock);
	310	}
	311	}
	312
	313	static int null_queue_rq(struct blk_mq_hw_ctx hctx, struct request rq)
	314	{
9d74e257	315	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(rq);
f2298c04 JA	316
	317	cmd->rq = rq;
	318	cmd->nq = hctx->driver_data;
	319
	320	null_handle_cmd(cmd);
	321	return BLK_MQ_RQ_QUEUE_OK;
	322	}
	323
24d2f903	324	static struct blk_mq_hw_ctx null_alloc_hctx(struct blk_mq_tag_set set,
f14bbe77 JA	325	unsigned int hctx_index,
f14bbe77 JA	326	int node)
f2298c04	327	{
f14bbe77	328	return kzalloc_node(sizeof(struct blk_mq_hw_ctx), GFP_KERNEL, node);
f2298c04 JA	329	}
	330
	331	static void null_free_hctx(struct blk_mq_hw_ctx *hctx, unsigned int hctx_index)
	332	{
	333	kfree(hctx);
	334	}
	335
2d263a78 MB	336	static void null_init_queue(struct nullb nullb, struct nullb_queue nq)
	337	{
	338	BUG_ON(!nullb);
	339	BUG_ON(!nq);
	340
	341	init_waitqueue_head(&nq->wait);
	342	nq->queue_depth = nullb->queue_depth;
	343	}
	344
f2298c04 JA	345	static int null_init_hctx(struct blk_mq_hw_ctx hctx, void data,
	346	unsigned int index)
	347	{
	348	struct nullb *nullb = data;
	349	struct nullb_queue *nq = &nullb->queues[index];
	350
f2298c04	351	hctx->driver_data = nq;
2d263a78 MB	352	null_init_queue(nullb, nq);
2d263a78 MB	353	nullb->nr_queues++;
f2298c04 JA	354
	355	return 0;
	356	}
	357
	358	static struct blk_mq_ops null_mq_ops = {
	359	.queue_rq = null_queue_rq,
	360	.map_queue = blk_mq_map_queue,
	361	.init_hctx = null_init_hctx,
ce2c350b	362	.complete = null_softirq_done_fn,
24d2f903 CH	363	.alloc_hctx = blk_mq_alloc_single_hw_queue,
24d2f903 CH	364	.free_hctx = blk_mq_free_single_hw_queue,
f2298c04 JA	365	};
f2298c04 JA	366
24d2f903 CH	367	static struct blk_mq_ops null_mq_ops_pernode = {
	368	.queue_rq = null_queue_rq,
	369	.map_queue = blk_mq_map_queue,
	370	.init_hctx = null_init_hctx,
	371	.complete = null_softirq_done_fn,
	372	.alloc_hctx = null_alloc_hctx,
	373	.free_hctx = null_free_hctx,
f2298c04 JA	374	};
	375
	376	static void null_del_dev(struct nullb *nullb)
	377	{
	378	list_del_init(&nullb->list);
	379
	380	del_gendisk(nullb->disk);
518d00b7	381	blk_cleanup_queue(nullb->q);
24d2f903 CH	382	if (queue_mode == NULL_Q_MQ)
24d2f903 CH	383	blk_mq_free_tag_set(&nullb->tag_set);
f2298c04 JA	384	put_disk(nullb->disk);
	385	kfree(nullb);
	386	}
	387
	388	static int null_open(struct block_device *bdev, fmode_t mode)
	389	{
	390	return 0;
	391	}
	392
	393	static void null_release(struct gendisk *disk, fmode_t mode)
	394	{
	395	}
	396
	397	static const struct block_device_operations null_fops = {
	398	.owner = THIS_MODULE,
	399	.open = null_open,
	400	.release = null_release,
	401	};
	402
	403	static int setup_commands(struct nullb_queue *nq)
	404	{
	405	struct nullb_cmd *cmd;
	406	int i, tag_size;
	407
	408	nq->cmds = kzalloc(nq->queue_depth * sizeof(*cmd), GFP_KERNEL);
	409	if (!nq->cmds)
2d263a78	410	return -ENOMEM;
f2298c04 JA	411
	412	tag_size = ALIGN(nq->queue_depth, BITS_PER_LONG) / BITS_PER_LONG;
	413	nq->tag_map = kzalloc(tag_size * sizeof(unsigned long), GFP_KERNEL);
	414	if (!nq->tag_map) {
	415	kfree(nq->cmds);
2d263a78	416	return -ENOMEM;
f2298c04 JA	417	}
	418
	419	for (i = 0; i < nq->queue_depth; i++) {
	420	cmd = &nq->cmds[i];
	421	INIT_LIST_HEAD(&cmd->list);
	422	cmd->ll_list.next = NULL;
	423	cmd->tag = -1U;
	424	}
	425
	426	return 0;
	427	}
	428
	429	static void cleanup_queue(struct nullb_queue *nq)
	430	{
	431	kfree(nq->tag_map);
	432	kfree(nq->cmds);
	433	}
	434
	435	static void cleanup_queues(struct nullb *nullb)
	436	{
	437	int i;
	438
	439	for (i = 0; i < nullb->nr_queues; i++)
	440	cleanup_queue(&nullb->queues[i]);
	441
	442	kfree(nullb->queues);
	443	}
	444
	445	static int setup_queues(struct nullb *nullb)
	446	{
2d263a78 MB	447	nullb->queues = kzalloc(submit_queues * sizeof(struct nullb_queue),
2d263a78 MB	448	GFP_KERNEL);
f2298c04	449	if (!nullb->queues)
2d263a78	450	return -ENOMEM;
f2298c04 JA	451
	452	nullb->nr_queues = 0;
	453	nullb->queue_depth = hw_queue_depth;
	454
2d263a78 MB	455	return 0;
	456	}
	457
	458	static int init_driver_queues(struct nullb *nullb)
	459	{
	460	struct nullb_queue *nq;
	461	int i, ret = 0;
f2298c04 JA	462
	463	for (i = 0; i < submit_queues; i++) {
	464	nq = &nullb->queues[i];
2d263a78 MB	465
	466	null_init_queue(nullb, nq);
	467
	468	ret = setup_commands(nq);
	469	if (ret)
	470	goto err_queue;
f2298c04 JA	471	nullb->nr_queues++;
	472	}
	473
2d263a78 MB	474	return 0;
2d263a78 MB	475	err_queue:
f2298c04	476	cleanup_queues(nullb);
2d263a78	477	return ret;
f2298c04 JA	478	}
	479
	480	static int null_add_dev(void)
	481	{
	482	struct gendisk *disk;
	483	struct nullb *nullb;
	484	sector_t size;
	485
	486	nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
	487	if (!nullb)
24d2f903	488	goto out;
f2298c04 JA	489
	490	spin_lock_init(&nullb->lock);
	491
57053d8c MB	492	if (queue_mode == NULL_Q_MQ && use_per_node_hctx)
	493	submit_queues = nr_online_nodes;
	494
f2298c04	495	if (setup_queues(nullb))
24d2f903	496	goto out_free_nullb;
f2298c04 JA	497
f2298c04 JA	498	if (queue_mode == NULL_Q_MQ) {
24d2f903 CH	499	if (use_per_node_hctx)
	500	nullb->tag_set.ops = &null_mq_ops_pernode;
	501	else
	502	nullb->tag_set.ops = &null_mq_ops;
	503	nullb->tag_set.nr_hw_queues = submit_queues;
	504	nullb->tag_set.queue_depth = hw_queue_depth;
	505	nullb->tag_set.numa_node = home_node;
	506	nullb->tag_set.cmd_size = sizeof(struct nullb_cmd);
	507	nullb->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
	508	nullb->tag_set.driver_data = nullb;
	509
	510	if (blk_mq_alloc_tag_set(&nullb->tag_set))
	511	goto out_cleanup_queues;
	512
	513	nullb->q = blk_mq_init_queue(&nullb->tag_set);
	514	if (!nullb->q)
	515	goto out_cleanup_tags;
f2298c04 JA	516	} else if (queue_mode == NULL_Q_BIO) {
f2298c04 JA	517	nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
24d2f903 CH	518	if (!nullb->q)
24d2f903 CH	519	goto out_cleanup_queues;
f2298c04	520	blk_queue_make_request(nullb->q, null_queue_bio);
2d263a78	521	init_driver_queues(nullb);
f2298c04 JA	522	} else {
f2298c04 JA	523	nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
24d2f903 CH	524	if (!nullb->q)
24d2f903 CH	525	goto out_cleanup_queues;
f2298c04	526	blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
24d2f903	527	blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
2d263a78	528	init_driver_queues(nullb);
f2298c04 JA	529	}
f2298c04 JA	530
f2298c04 JA	531	nullb->q->queuedata = nullb;
	532	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);
	533
	534	disk = nullb->disk = alloc_disk_node(1, home_node);
24d2f903 CH	535	if (!disk)
24d2f903 CH	536	goto out_cleanup_blk_queue;
f2298c04 JA	537
	538	mutex_lock(&lock);
	539	list_add_tail(&nullb->list, &nullb_list);
	540	nullb->index = nullb_indexes++;
	541	mutex_unlock(&lock);
	542
	543	blk_queue_logical_block_size(nullb->q, bs);
	544	blk_queue_physical_block_size(nullb->q, bs);
	545
	546	size = gb * 1024 * 1024 * 1024ULL;
	547	sector_div(size, bs);
	548	set_capacity(disk, size);
	549
	550	disk->flags \|= GENHD_FL_EXT_DEVT;
	551	disk->major = null_major;
	552	disk->first_minor = nullb->index;
	553	disk->fops = &null_fops;
	554	disk->private_data = nullb;
	555	disk->queue = nullb->q;
	556	sprintf(disk->disk_name, "nullb%d", nullb->index);
	557	add_disk(disk);
	558	return 0;
24d2f903 CH	559
	560	out_cleanup_blk_queue:
	561	blk_cleanup_queue(nullb->q);
	562	out_cleanup_tags:
	563	if (queue_mode == NULL_Q_MQ)
	564	blk_mq_free_tag_set(&nullb->tag_set);
	565	out_cleanup_queues:
	566	cleanup_queues(nullb);
	567	out_free_nullb:
	568	kfree(nullb);
	569	out:
	570	return -ENOMEM;
f2298c04 JA	571	}
	572
	573	static int __init null_init(void)
	574	{
	575	unsigned int i;
	576
9967d8ac R	577	if (bs > PAGE_SIZE) {
	578	pr_warn("null_blk: invalid block size\n");
	579	pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);
	580	bs = PAGE_SIZE;
	581	}
f2298c04	582
d15ee6b1	583	if (queue_mode == NULL_Q_MQ && use_per_node_hctx) {
fc1bc354	584	if (submit_queues < nr_online_nodes) {
d15ee6b1 MB	585	pr_warn("null_blk: submit_queues param is set to %u.",
d15ee6b1 MB	586	nr_online_nodes);
fc1bc354 MB	587	submit_queues = nr_online_nodes;
fc1bc354 MB	588	}
d15ee6b1	589	} else if (submit_queues > nr_cpu_ids)
f2298c04 JA	590	submit_queues = nr_cpu_ids;
	591	else if (!submit_queues)
	592	submit_queues = 1;
	593
	594	mutex_init(&lock);
	595
	596	/* Initialize a separate list for each CPU for issuing softirqs */
	597	for_each_possible_cpu(i) {
	598	struct completion_queue *cq = &per_cpu(completion_queues, i);
	599
	600	init_llist_head(&cq->list);
	601
	602	if (irqmode != NULL_IRQ_TIMER)
	603	continue;
	604
	605	hrtimer_init(&cq->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	606	cq->timer.function = null_cmd_timer_expired;
	607	}
	608
	609	null_major = register_blkdev(0, "nullb");
	610	if (null_major < 0)
	611	return null_major;
	612
	613	for (i = 0; i < nr_devices; i++) {
	614	if (null_add_dev()) {
	615	unregister_blkdev(null_major, "nullb");
	616	return -EINVAL;
	617	}
	618	}
	619
	620	pr_info("null: module loaded\n");
	621	return 0;
	622	}
	623
	624	static void __exit null_exit(void)
	625	{
	626	struct nullb *nullb;
	627
	628	unregister_blkdev(null_major, "nullb");
	629
	630	mutex_lock(&lock);
	631	while (!list_empty(&nullb_list)) {
	632	nullb = list_entry(nullb_list.next, struct nullb, list);
	633	null_del_dev(nullb);
	634	}
	635	mutex_unlock(&lock);
	636	}
	637
	638	module_init(null_init);
	639	module_exit(null_exit);
	640
	641	MODULE_AUTHOR("Jens Axboe <jaxboe@fusionio.com>");
	642	MODULE_LICENSE("GPL");