[linux-2.6-block.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;

static int null_param_store_val(const char *str, int *val, int min, int max)
{
	int ret, new_val;

	ret = kstrtoint(str, 10, &new_val);
	if (ret)
		return -EINVAL;

	if (new_val < min || new_val > max)
		return -EINVAL;

	*val = new_val;
	return 0;
}

static int null_set_queue_mode(const char *str, const struct kernel_param *kp)
{
	return null_param_store_val(str, &queue_mode, NULL_Q_BIO, NULL_Q_MQ);
}

static struct kernel_param_ops null_queue_mode_param_ops = {
	.set	= null_set_queue_mode,
	.get	= param_get_int,
};

device_param_cb(queue_mode, &null_queue_mode_param_ops, &queue_mode, 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;

static int null_set_irqmode(const char *str, const struct kernel_param *kp)
{
	return null_param_store_val(str, &irqmode, NULL_IRQ_NONE,
					NULL_IRQ_TIMER);
}

static struct kernel_param_ops null_irqmode_param_ops = {
	.set	= null_set_irqmode,
	.get	= param_get_int,
};

device_param_cb(irqmode, &null_irqmode_param_ops, &irqmode, 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_request(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,
			 const struct blk_mq_queue_data *bd)
{
	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(bd->rq);

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

	blk_mq_start_request(bd->rq);

	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)
			return ret;
		nullb->nr_queues++;
	}
	return 0;
}

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 (IS_ERR(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);
		rv = init_driver_queues(nullb);
		if (rv)
			goto out_cleanup_blk_queue;
	} 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);
		rv = init_driver_queues(nullb);
		if (rv)
			goto out_cleanup_blk_queue;
	}

	nullb->q->queuedata = nullb;
	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);
	queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, 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 | GENHD_FL_SUPPRESS_PARTITION_INFO;
	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;
709c8667 MB	81
	82	static int null_param_store_val(const char str, int val, int min, int max)
	83	{
	84	int ret, new_val;
	85
	86	ret = kstrtoint(str, 10, &new_val);
	87	if (ret)
	88	return -EINVAL;
	89
	90	if (new_val < min \|\| new_val > max)
	91	return -EINVAL;
	92
	93	*val = new_val;
	94	return 0;
	95	}
	96
	97	static int null_set_queue_mode(const char str, const struct kernel_param kp)
	98	{
	99	return null_param_store_val(str, &queue_mode, NULL_Q_BIO, NULL_Q_MQ);
	100	}
	101
	102	static struct kernel_param_ops null_queue_mode_param_ops = {
	103	.set = null_set_queue_mode,
	104	.get = param_get_int,
	105	};
	106
	107	device_param_cb(queue_mode, &null_queue_mode_param_ops, &queue_mode, S_IRUGO);
54ae81cd	108	MODULE_PARM_DESC(queue_mode, "Block interface to use (0=bio,1=rq,2=multiqueue)");
f2298c04 JA	109
	110	static int gb = 250;
	111	module_param(gb, int, S_IRUGO);
	112	MODULE_PARM_DESC(gb, "Size in GB");
	113
	114	static int bs = 512;
	115	module_param(bs, int, S_IRUGO);
	116	MODULE_PARM_DESC(bs, "Block size (in bytes)");
	117
	118	static int nr_devices = 2;
	119	module_param(nr_devices, int, S_IRUGO);
	120	MODULE_PARM_DESC(nr_devices, "Number of devices to register");
	121
	122	static int irqmode = NULL_IRQ_SOFTIRQ;
709c8667 MB	123
	124	static int null_set_irqmode(const char str, const struct kernel_param kp)
	125	{
	126	return null_param_store_val(str, &irqmode, NULL_IRQ_NONE,
	127	NULL_IRQ_TIMER);
	128	}
	129
	130	static struct kernel_param_ops null_irqmode_param_ops = {
	131	.set = null_set_irqmode,
	132	.get = param_get_int,
	133	};
	134
	135	device_param_cb(irqmode, &null_irqmode_param_ops, &irqmode, S_IRUGO);
f2298c04 JA	136	MODULE_PARM_DESC(irqmode, "IRQ completion handler. 0-none, 1-softirq, 2-timer");
	137
	138	static int completion_nsec = 10000;
	139	module_param(completion_nsec, int, S_IRUGO);
	140	MODULE_PARM_DESC(completion_nsec, "Time in ns to complete a request in hardware. Default: 10,000ns");
	141
	142	static int hw_queue_depth = 64;
	143	module_param(hw_queue_depth, int, S_IRUGO);
	144	MODULE_PARM_DESC(hw_queue_depth, "Queue depth for each hardware queue. Default: 64");
	145
20005244	146	static bool use_per_node_hctx = false;
f2298c04	147	module_param(use_per_node_hctx, bool, S_IRUGO);
20005244	148	MODULE_PARM_DESC(use_per_node_hctx, "Use per-node allocation for hardware context queues. Default: false");
f2298c04 JA	149
	150	static void put_tag(struct nullb_queue *nq, unsigned int tag)
	151	{
	152	clear_bit_unlock(tag, nq->tag_map);
	153
	154	if (waitqueue_active(&nq->wait))
	155	wake_up(&nq->wait);
	156	}
	157
	158	static unsigned int get_tag(struct nullb_queue *nq)
	159	{
	160	unsigned int tag;
	161
	162	do {
	163	tag = find_first_zero_bit(nq->tag_map, nq->queue_depth);
	164	if (tag >= nq->queue_depth)
	165	return -1U;
	166	} while (test_and_set_bit_lock(tag, nq->tag_map));
	167
	168	return tag;
	169	}
	170
	171	static void free_cmd(struct nullb_cmd *cmd)
	172	{
	173	put_tag(cmd->nq, cmd->tag);
	174	}
	175
	176	static struct nullb_cmd __alloc_cmd(struct nullb_queue nq)
	177	{
	178	struct nullb_cmd *cmd;
	179	unsigned int tag;
	180
	181	tag = get_tag(nq);
	182	if (tag != -1U) {
	183	cmd = &nq->cmds[tag];
	184	cmd->tag = tag;
	185	cmd->nq = nq;
	186	return cmd;
	187	}
	188
	189	return NULL;
	190	}
	191
	192	static struct nullb_cmd alloc_cmd(struct nullb_queue nq, int can_wait)
	193	{
	194	struct nullb_cmd *cmd;
	195	DEFINE_WAIT(wait);
	196
	197	cmd = __alloc_cmd(nq);
	198	if (cmd \|\| !can_wait)
	199	return cmd;
	200
	201	do {
	202	prepare_to_wait(&nq->wait, &wait, TASK_UNINTERRUPTIBLE);
	203	cmd = __alloc_cmd(nq);
	204	if (cmd)
	205	break;
	206
	207	io_schedule();
	208	} while (1);
	209
	210	finish_wait(&nq->wait, &wait);
	211	return cmd;
	212	}
213
214	static void end_cmd(struct nullb_cmd *cmd)
215	{
ce2c350b CH	216	switch (queue_mode) {
ce2c350b CH	217	case NULL_Q_MQ:
c8a446ad	218	blk_mq_end_request(cmd->rq, 0);
ce2c350b CH	219	return;
	220	case NULL_Q_RQ:
	221	INIT_LIST_HEAD(&cmd->rq->queuelist);
	222	blk_end_request_all(cmd->rq, 0);
	223	break;
	224	case NULL_Q_BIO:
f2298c04	225	bio_endio(cmd->bio, 0);
ce2c350b CH	226	break;
ce2c350b CH	227	}
f2298c04	228
ce2c350b	229	free_cmd(cmd);
f2298c04 JA	230	}
	231
	232	static enum hrtimer_restart null_cmd_timer_expired(struct hrtimer *timer)
	233	{
	234	struct completion_queue *cq;
	235	struct llist_node *entry;
	236	struct nullb_cmd *cmd;
	237
	238	cq = &per_cpu(completion_queues, smp_processor_id());
	239
	240	while ((entry = llist_del_all(&cq->list)) != NULL) {
d7790b92	241	entry = llist_reverse_order(entry);
f2298c04 JA	242	do {
f2298c04 JA	243	cmd = container_of(entry, struct nullb_cmd, ll_list);
f2298c04	244	entry = entry->next;
fc27691f	245	end_cmd(cmd);
f2298c04 JA	246	} while (entry);
	247	}
	248
	249	return HRTIMER_NORESTART;
	250	}
	251
	252	static void null_cmd_end_timer(struct nullb_cmd *cmd)
	253	{
	254	struct completion_queue *cq = &per_cpu(completion_queues, get_cpu());
	255
	256	cmd->ll_list.next = NULL;
	257	if (llist_add(&cmd->ll_list, &cq->list)) {
	258	ktime_t kt = ktime_set(0, completion_nsec);
	259
	260	hrtimer_start(&cq->timer, kt, HRTIMER_MODE_REL);
	261	}
	262
	263	put_cpu();
	264	}
	265
	266	static void null_softirq_done_fn(struct request *rq)
	267	{
d891fa70 JA	268	if (queue_mode == NULL_Q_MQ)
	269	end_cmd(blk_mq_rq_to_pdu(rq));
	270	else
	271	end_cmd(rq->special);
f2298c04 JA	272	}
f2298c04 JA	273
f2298c04 JA	274	static inline void null_handle_cmd(struct nullb_cmd *cmd)
	275	{
	276	/* Complete IO by inline, softirq or timer */
	277	switch (irqmode) {
f2298c04	278	case NULL_IRQ_SOFTIRQ:
ce2c350b CH	279	switch (queue_mode) {
	280	case NULL_Q_MQ:
	281	blk_mq_complete_request(cmd->rq);
	282	break;
	283	case NULL_Q_RQ:
	284	blk_complete_request(cmd->rq);
	285	break;
	286	case NULL_Q_BIO:
	287	/*
	288	* XXX: no proper submitting cpu information available.
	289	*/
	290	end_cmd(cmd);
	291	break;
	292	}
	293	break;
	294	case NULL_IRQ_NONE:
f2298c04	295	end_cmd(cmd);
f2298c04 JA	296	break;
	297	case NULL_IRQ_TIMER:
	298	null_cmd_end_timer(cmd);
	299	break;
	300	}
	301	}
	302
	303	static struct nullb_queue nullb_to_queue(struct nullb nullb)
	304	{
	305	int index = 0;
	306
	307	if (nullb->nr_queues != 1)
	308	index = raw_smp_processor_id() / ((nr_cpu_ids + nullb->nr_queues - 1) / nullb->nr_queues);
	309
	310	return &nullb->queues[index];
	311	}
	312
	313	static void null_queue_bio(struct request_queue q, struct bio bio)
	314	{
	315	struct nullb *nullb = q->queuedata;
	316	struct nullb_queue *nq = nullb_to_queue(nullb);
	317	struct nullb_cmd *cmd;
	318
	319	cmd = alloc_cmd(nq, 1);
	320	cmd->bio = bio;
	321
	322	null_handle_cmd(cmd);
	323	}
	324
	325	static int null_rq_prep_fn(struct request_queue q, struct request req)
	326	{
	327	struct nullb *nullb = q->queuedata;
	328	struct nullb_queue *nq = nullb_to_queue(nullb);
	329	struct nullb_cmd *cmd;
	330
	331	cmd = alloc_cmd(nq, 0);
	332	if (cmd) {
	333	cmd->rq = req;
	334	req->special = cmd;
	335	return BLKPREP_OK;
	336	}
	337
	338	return BLKPREP_DEFER;
	339	}
	340
	341	static void null_request_fn(struct request_queue *q)
	342	{
	343	struct request *rq;
	344
	345	while ((rq = blk_fetch_request(q)) != NULL) {
	346	struct nullb_cmd *cmd = rq->special;
	347
	348	spin_unlock_irq(q->queue_lock);
	349	null_handle_cmd(cmd);
	350	spin_lock_irq(q->queue_lock);
	351	}
	352	}
	353
74c45052 JA	354	static int null_queue_rq(struct blk_mq_hw_ctx *hctx,
74c45052 JA	355	const struct blk_mq_queue_data *bd)
f2298c04	356	{
74c45052	357	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(bd->rq);
f2298c04	358
74c45052	359	cmd->rq = bd->rq;
f2298c04 JA	360	cmd->nq = hctx->driver_data;
f2298c04 JA	361
74c45052	362	blk_mq_start_request(bd->rq);
e2490073	363
f2298c04 JA	364	null_handle_cmd(cmd);
	365	return BLK_MQ_RQ_QUEUE_OK;
	366	}
	367
2d263a78 MB	368	static void null_init_queue(struct nullb nullb, struct nullb_queue nq)
	369	{
	370	BUG_ON(!nullb);
	371	BUG_ON(!nq);
	372
	373	init_waitqueue_head(&nq->wait);
	374	nq->queue_depth = nullb->queue_depth;
	375	}
	376
f2298c04 JA	377	static int null_init_hctx(struct blk_mq_hw_ctx hctx, void data,
	378	unsigned int index)
	379	{
	380	struct nullb *nullb = data;
	381	struct nullb_queue *nq = &nullb->queues[index];
	382
f2298c04	383	hctx->driver_data = nq;
2d263a78 MB	384	null_init_queue(nullb, nq);
2d263a78 MB	385	nullb->nr_queues++;
f2298c04 JA	386
	387	return 0;
	388	}
	389
	390	static struct blk_mq_ops null_mq_ops = {
	391	.queue_rq = null_queue_rq,
	392	.map_queue = blk_mq_map_queue,
	393	.init_hctx = null_init_hctx,
ce2c350b	394	.complete = null_softirq_done_fn,
f2298c04 JA	395	};
	396
	397	static void null_del_dev(struct nullb *nullb)
	398	{
	399	list_del_init(&nullb->list);
	400
	401	del_gendisk(nullb->disk);
518d00b7	402	blk_cleanup_queue(nullb->q);
24d2f903 CH	403	if (queue_mode == NULL_Q_MQ)
24d2f903 CH	404	blk_mq_free_tag_set(&nullb->tag_set);
f2298c04 JA	405	put_disk(nullb->disk);
	406	kfree(nullb);
	407	}
	408
	409	static int null_open(struct block_device *bdev, fmode_t mode)
	410	{
	411	return 0;
	412	}
	413
	414	static void null_release(struct gendisk *disk, fmode_t mode)
	415	{
	416	}
	417
	418	static const struct block_device_operations null_fops = {
	419	.owner = THIS_MODULE,
	420	.open = null_open,
	421	.release = null_release,
	422	};
	423
	424	static int setup_commands(struct nullb_queue *nq)
	425	{
	426	struct nullb_cmd *cmd;
	427	int i, tag_size;
	428
	429	nq->cmds = kzalloc(nq->queue_depth * sizeof(*cmd), GFP_KERNEL);
	430	if (!nq->cmds)
2d263a78	431	return -ENOMEM;
f2298c04 JA	432
	433	tag_size = ALIGN(nq->queue_depth, BITS_PER_LONG) / BITS_PER_LONG;
	434	nq->tag_map = kzalloc(tag_size * sizeof(unsigned long), GFP_KERNEL);
	435	if (!nq->tag_map) {
	436	kfree(nq->cmds);
2d263a78	437	return -ENOMEM;
f2298c04 JA	438	}
	439
	440	for (i = 0; i < nq->queue_depth; i++) {
	441	cmd = &nq->cmds[i];
	442	INIT_LIST_HEAD(&cmd->list);
	443	cmd->ll_list.next = NULL;
	444	cmd->tag = -1U;
	445	}
	446
	447	return 0;
	448	}
	449
	450	static void cleanup_queue(struct nullb_queue *nq)
	451	{
	452	kfree(nq->tag_map);
	453	kfree(nq->cmds);
	454	}
	455
	456	static void cleanup_queues(struct nullb *nullb)
	457	{
	458	int i;
	459
	460	for (i = 0; i < nullb->nr_queues; i++)
	461	cleanup_queue(&nullb->queues[i]);
	462
	463	kfree(nullb->queues);
	464	}
	465
	466	static int setup_queues(struct nullb *nullb)
	467	{
2d263a78 MB	468	nullb->queues = kzalloc(submit_queues * sizeof(struct nullb_queue),
2d263a78 MB	469	GFP_KERNEL);
f2298c04	470	if (!nullb->queues)
2d263a78	471	return -ENOMEM;
f2298c04 JA	472
	473	nullb->nr_queues = 0;
	474	nullb->queue_depth = hw_queue_depth;
	475
2d263a78 MB	476	return 0;
	477	}
	478
	479	static int init_driver_queues(struct nullb *nullb)
	480	{
	481	struct nullb_queue *nq;
	482	int i, ret = 0;
f2298c04 JA	483
	484	for (i = 0; i < submit_queues; i++) {
	485	nq = &nullb->queues[i];
2d263a78 MB	486
	487	null_init_queue(nullb, nq);
	488
	489	ret = setup_commands(nq);
	490	if (ret)
31f9690e	491	return ret;
f2298c04 JA	492	nullb->nr_queues++;
f2298c04 JA	493	}
2d263a78	494	return 0;
f2298c04 JA	495	}
	496
	497	static int null_add_dev(void)
	498	{
	499	struct gendisk *disk;
	500	struct nullb *nullb;
	501	sector_t size;
dc501dc0	502	int rv;
f2298c04 JA	503
f2298c04 JA	504	nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
dc501dc0 RE	505	if (!nullb) {
dc501dc0 RE	506	rv = -ENOMEM;
24d2f903	507	goto out;
dc501dc0	508	}
f2298c04 JA	509
	510	spin_lock_init(&nullb->lock);
	511
57053d8c MB	512	if (queue_mode == NULL_Q_MQ && use_per_node_hctx)
	513	submit_queues = nr_online_nodes;
	514
dc501dc0 RE	515	rv = setup_queues(nullb);
dc501dc0 RE	516	if (rv)
24d2f903	517	goto out_free_nullb;
f2298c04 JA	518
f2298c04 JA	519	if (queue_mode == NULL_Q_MQ) {
cdef54dd	520	nullb->tag_set.ops = &null_mq_ops;
24d2f903 CH	521	nullb->tag_set.nr_hw_queues = submit_queues;
	522	nullb->tag_set.queue_depth = hw_queue_depth;
	523	nullb->tag_set.numa_node = home_node;
	524	nullb->tag_set.cmd_size = sizeof(struct nullb_cmd);
	525	nullb->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
	526	nullb->tag_set.driver_data = nullb;
	527
dc501dc0 RE	528	rv = blk_mq_alloc_tag_set(&nullb->tag_set);
dc501dc0 RE	529	if (rv)
24d2f903 CH	530	goto out_cleanup_queues;
	531
	532	nullb->q = blk_mq_init_queue(&nullb->tag_set);
35b489d3	533	if (IS_ERR(nullb->q)) {
dc501dc0	534	rv = -ENOMEM;
24d2f903	535	goto out_cleanup_tags;
dc501dc0	536	}
f2298c04 JA	537	} else if (queue_mode == NULL_Q_BIO) {
f2298c04 JA	538	nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
dc501dc0 RE	539	if (!nullb->q) {
dc501dc0 RE	540	rv = -ENOMEM;
24d2f903	541	goto out_cleanup_queues;
dc501dc0	542	}
f2298c04	543	blk_queue_make_request(nullb->q, null_queue_bio);
31f9690e JK	544	rv = init_driver_queues(nullb);
	545	if (rv)
	546	goto out_cleanup_blk_queue;
f2298c04 JA	547	} else {
f2298c04 JA	548	nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
dc501dc0 RE	549	if (!nullb->q) {
dc501dc0 RE	550	rv = -ENOMEM;
24d2f903	551	goto out_cleanup_queues;
dc501dc0	552	}
f2298c04	553	blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
24d2f903	554	blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
31f9690e JK	555	rv = init_driver_queues(nullb);
	556	if (rv)
	557	goto out_cleanup_blk_queue;
f2298c04 JA	558	}
f2298c04 JA	559
f2298c04 JA	560	nullb->q->queuedata = nullb;
f2298c04 JA	561	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);
b277da0a	562	queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, nullb->q);
f2298c04 JA	563
f2298c04 JA	564	disk = nullb->disk = alloc_disk_node(1, home_node);
dc501dc0 RE	565	if (!disk) {
dc501dc0 RE	566	rv = -ENOMEM;
24d2f903	567	goto out_cleanup_blk_queue;
dc501dc0	568	}
f2298c04 JA	569
	570	mutex_lock(&lock);
	571	list_add_tail(&nullb->list, &nullb_list);
	572	nullb->index = nullb_indexes++;
	573	mutex_unlock(&lock);
	574
	575	blk_queue_logical_block_size(nullb->q, bs);
	576	blk_queue_physical_block_size(nullb->q, bs);
	577
	578	size = gb * 1024 * 1024 * 1024ULL;
	579	sector_div(size, bs);
	580	set_capacity(disk, size);
	581
227290b4	582	disk->flags \|= GENHD_FL_EXT_DEVT \| GENHD_FL_SUPPRESS_PARTITION_INFO;
f2298c04 JA	583	disk->major = null_major;
	584	disk->first_minor = nullb->index;
	585	disk->fops = &null_fops;
	586	disk->private_data = nullb;
	587	disk->queue = nullb->q;
	588	sprintf(disk->disk_name, "nullb%d", nullb->index);
	589	add_disk(disk);
	590	return 0;
24d2f903 CH	591
	592	out_cleanup_blk_queue:
	593	blk_cleanup_queue(nullb->q);
	594	out_cleanup_tags:
	595	if (queue_mode == NULL_Q_MQ)
	596	blk_mq_free_tag_set(&nullb->tag_set);
	597	out_cleanup_queues:
	598	cleanup_queues(nullb);
	599	out_free_nullb:
	600	kfree(nullb);
	601	out:
dc501dc0	602	return rv;
f2298c04 JA	603	}
	604
	605	static int __init null_init(void)
	606	{
	607	unsigned int i;
	608
9967d8ac R	609	if (bs > PAGE_SIZE) {
	610	pr_warn("null_blk: invalid block size\n");
	611	pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);
	612	bs = PAGE_SIZE;
	613	}
f2298c04	614
d15ee6b1	615	if (queue_mode == NULL_Q_MQ && use_per_node_hctx) {
fc1bc354	616	if (submit_queues < nr_online_nodes) {
d15ee6b1 MB	617	pr_warn("null_blk: submit_queues param is set to %u.",
d15ee6b1 MB	618	nr_online_nodes);
fc1bc354 MB	619	submit_queues = nr_online_nodes;
fc1bc354 MB	620	}
d15ee6b1	621	} else if (submit_queues > nr_cpu_ids)
f2298c04 JA	622	submit_queues = nr_cpu_ids;
	623	else if (!submit_queues)
	624	submit_queues = 1;
	625
	626	mutex_init(&lock);
	627
	628	/* Initialize a separate list for each CPU for issuing softirqs */
	629	for_each_possible_cpu(i) {
	630	struct completion_queue *cq = &per_cpu(completion_queues, i);
	631
	632	init_llist_head(&cq->list);
	633
	634	if (irqmode != NULL_IRQ_TIMER)
	635	continue;
	636
	637	hrtimer_init(&cq->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	638	cq->timer.function = null_cmd_timer_expired;
	639	}
	640
	641	null_major = register_blkdev(0, "nullb");
	642	if (null_major < 0)
	643	return null_major;
	644
	645	for (i = 0; i < nr_devices; i++) {
	646	if (null_add_dev()) {
	647	unregister_blkdev(null_major, "nullb");
	648	return -EINVAL;
	649	}
	650	}
	651
	652	pr_info("null: module loaded\n");
	653	return 0;
	654	}
	655
	656	static void __exit null_exit(void)
	657	{
	658	struct nullb *nullb;
	659
	660	unregister_blkdev(null_major, "nullb");
	661
	662	mutex_lock(&lock);
	663	while (!list_empty(&nullb_list)) {
	664	nullb = list_entry(nullb_list.next, struct nullb, list);
	665	null_del_dev(nullb);
	666	}
	667	mutex_unlock(&lock);
	668	}
	669
	670	module_init(null_init);
	671	module_exit(null_exit);
	672
	673	MODULE_AUTHOR("Jens Axboe <jaxboe@fusionio.com>");
	674	MODULE_LICENSE("GPL");