[linux-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 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_reg *reg, unsigned int hctx_index)
{
	int b_size = DIV_ROUND_UP(reg->nr_hw_queues, nr_online_nodes);
	int tip = (reg->nr_hw_queues % nr_online_nodes);
	int node = 0, i, n;

	/*
	 * Split submit queues evenly wrt to the number of nodes. If uneven,
	 * fill the first buckets with one extra, until the rest is filled with
	 * no extra.
	 */
	for (i = 0, n = 1; i < hctx_index; i++, n++) {
		if (n % b_size == 0) {
			n = 0;
			node++;

			tip--;
			if (!tip)
				b_size = reg->nr_hw_queues / nr_online_nodes;
		}
	}

	/*
	 * A node might not be online, therefore map the relative node id to the
	 * real node id.
	 */
	for_each_online_node(n) {
		if (!node)
			break;
		node--;
	}

	return kzalloc_node(sizeof(struct blk_mq_hw_ctx), GFP_KERNEL, n);
}

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,
};

static struct blk_mq_reg null_mq_reg = {
	.ops		= &null_mq_ops,
	.queue_depth	= 64,
	.cmd_size	= sizeof(struct nullb_cmd),
	.flags		= BLK_MQ_F_SHOULD_MERGE,
};

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

	del_gendisk(nullb->disk);
	blk_cleanup_queue(nullb->q);
	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)
		return -ENOMEM;

	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 err;

	if (queue_mode == NULL_Q_MQ) {
		null_mq_reg.numa_node = home_node;
		null_mq_reg.queue_depth = hw_queue_depth;
		null_mq_reg.nr_hw_queues = submit_queues;

		if (use_per_node_hctx) {
			null_mq_reg.ops->alloc_hctx = null_alloc_hctx;
			null_mq_reg.ops->free_hctx = null_free_hctx;
		} else {
			null_mq_reg.ops->alloc_hctx = blk_mq_alloc_single_hw_queue;
			null_mq_reg.ops->free_hctx = blk_mq_free_single_hw_queue;
		}

		nullb->q = blk_mq_init_queue(&null_mq_reg, nullb);
	} else if (queue_mode == NULL_Q_BIO) {
		nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
		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);
		blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
		if (nullb->q)
			blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
		init_driver_queues(nullb);
	}

	if (!nullb->q)
		goto queue_fail;

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

	disk = nullb->disk = alloc_disk_node(1, home_node);
	if (!disk) {
queue_fail:
		blk_cleanup_queue(nullb->q);
		cleanup_queues(nullb);
err:
		kfree(nullb);
		return -ENOMEM;
	}

	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;
}

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;
	35	struct hrtimer timer;
	36	unsigned int queue_depth;
	37	spinlock_t lock;
	38
	39	struct nullb_queue *queues;
	40	unsigned int nr_queues;
	41	};
	42
	43	static LIST_HEAD(nullb_list);
	44	static struct mutex lock;
	45	static int null_major;
	46	static int nullb_indexes;
	47
	48	struct completion_queue {
	49	struct llist_head list;
	50	struct hrtimer timer;
	51	};
	52
	53	/*
	54	* These are per-cpu for now, they will need to be configured by the
	55	* complete_queues parameter and appropriately mapped.
	56	*/
	57	static DEFINE_PER_CPU(struct completion_queue, completion_queues);
	58
	59	enum {
	60	NULL_IRQ_NONE = 0,
	61	NULL_IRQ_SOFTIRQ = 1,
	62	NULL_IRQ_TIMER = 2,
ce2c350b	63	};
f2298c04	64
ce2c350b	65	enum {
f2298c04 JA	66	NULL_Q_BIO = 0,
	67	NULL_Q_RQ = 1,
	68	NULL_Q_MQ = 2,
	69	};
	70
2d263a78	71	static int submit_queues;
f2298c04 JA	72	module_param(submit_queues, int, S_IRUGO);
	73	MODULE_PARM_DESC(submit_queues, "Number of submission queues");
	74
	75	static int home_node = NUMA_NO_NODE;
	76	module_param(home_node, int, S_IRUGO);
	77	MODULE_PARM_DESC(home_node, "Home node for the device");
	78
	79	static int queue_mode = NULL_Q_MQ;
	80	module_param(queue_mode, int, S_IRUGO);
	81	MODULE_PARM_DESC(use_mq, "Use blk-mq interface (0=bio,1=rq,2=multiqueue)");
	82
	83	static int gb = 250;
	84	module_param(gb, int, S_IRUGO);
	85	MODULE_PARM_DESC(gb, "Size in GB");
	86
	87	static int bs = 512;
	88	module_param(bs, int, S_IRUGO);
	89	MODULE_PARM_DESC(bs, "Block size (in bytes)");
	90
	91	static int nr_devices = 2;
	92	module_param(nr_devices, int, S_IRUGO);
	93	MODULE_PARM_DESC(nr_devices, "Number of devices to register");
	94
	95	static int irqmode = NULL_IRQ_SOFTIRQ;
	96	module_param(irqmode, int, S_IRUGO);
	97	MODULE_PARM_DESC(irqmode, "IRQ completion handler. 0-none, 1-softirq, 2-timer");
	98
	99	static int completion_nsec = 10000;
	100	module_param(completion_nsec, int, S_IRUGO);
	101	MODULE_PARM_DESC(completion_nsec, "Time in ns to complete a request in hardware. Default: 10,000ns");
	102
	103	static int hw_queue_depth = 64;
	104	module_param(hw_queue_depth, int, S_IRUGO);
	105	MODULE_PARM_DESC(hw_queue_depth, "Queue depth for each hardware queue. Default: 64");
	106
20005244	107	static bool use_per_node_hctx = false;
f2298c04	108	module_param(use_per_node_hctx, bool, S_IRUGO);
20005244	109	MODULE_PARM_DESC(use_per_node_hctx, "Use per-node allocation for hardware context queues. Default: false");
f2298c04 JA	110
	111	static void put_tag(struct nullb_queue *nq, unsigned int tag)
	112	{
	113	clear_bit_unlock(tag, nq->tag_map);
	114
	115	if (waitqueue_active(&nq->wait))
	116	wake_up(&nq->wait);
	117	}
	118
	119	static unsigned int get_tag(struct nullb_queue *nq)
	120	{
	121	unsigned int tag;
	122
	123	do {
	124	tag = find_first_zero_bit(nq->tag_map, nq->queue_depth);
	125	if (tag >= nq->queue_depth)
	126	return -1U;
	127	} while (test_and_set_bit_lock(tag, nq->tag_map));
	128
	129	return tag;
	130	}
	131
	132	static void free_cmd(struct nullb_cmd *cmd)
	133	{
	134	put_tag(cmd->nq, cmd->tag);
	135	}
	136
	137	static struct nullb_cmd __alloc_cmd(struct nullb_queue nq)
	138	{
	139	struct nullb_cmd *cmd;
	140	unsigned int tag;
	141
	142	tag = get_tag(nq);
	143	if (tag != -1U) {
	144	cmd = &nq->cmds[tag];
	145	cmd->tag = tag;
	146	cmd->nq = nq;
	147	return cmd;
	148	}
	149
	150	return NULL;
	151	}
	152
	153	static struct nullb_cmd alloc_cmd(struct nullb_queue nq, int can_wait)
	154	{
	155	struct nullb_cmd *cmd;
	156	DEFINE_WAIT(wait);
	157
	158	cmd = __alloc_cmd(nq);
	159	if (cmd \|\| !can_wait)
	160	return cmd;
	161
	162	do {
	163	prepare_to_wait(&nq->wait, &wait, TASK_UNINTERRUPTIBLE);
	164	cmd = __alloc_cmd(nq);
	165	if (cmd)
	166	break;
	167
	168	io_schedule();
	169	} while (1);
	170
	171	finish_wait(&nq->wait, &wait);
	172	return cmd;
	173	}
174
175	static void end_cmd(struct nullb_cmd *cmd)
176	{
ce2c350b CH	177	switch (queue_mode) {
	178	case NULL_Q_MQ:
	179	blk_mq_end_io(cmd->rq, 0);
	180	return;
	181	case NULL_Q_RQ:
	182	INIT_LIST_HEAD(&cmd->rq->queuelist);
	183	blk_end_request_all(cmd->rq, 0);
	184	break;
	185	case NULL_Q_BIO:
f2298c04	186	bio_endio(cmd->bio, 0);
ce2c350b CH	187	break;
ce2c350b CH	188	}
f2298c04	189
ce2c350b	190	free_cmd(cmd);
f2298c04 JA	191	}
	192
	193	static enum hrtimer_restart null_cmd_timer_expired(struct hrtimer *timer)
	194	{
	195	struct completion_queue *cq;
	196	struct llist_node *entry;
	197	struct nullb_cmd *cmd;
	198
	199	cq = &per_cpu(completion_queues, smp_processor_id());
	200
	201	while ((entry = llist_del_all(&cq->list)) != NULL) {
d7790b92	202	entry = llist_reverse_order(entry);
f2298c04 JA	203	do {
	204	cmd = container_of(entry, struct nullb_cmd, ll_list);
	205	end_cmd(cmd);
	206	entry = entry->next;
	207	} while (entry);
	208	}
	209
	210	return HRTIMER_NORESTART;
	211	}
	212
	213	static void null_cmd_end_timer(struct nullb_cmd *cmd)
	214	{
	215	struct completion_queue *cq = &per_cpu(completion_queues, get_cpu());
	216
	217	cmd->ll_list.next = NULL;
	218	if (llist_add(&cmd->ll_list, &cq->list)) {
	219	ktime_t kt = ktime_set(0, completion_nsec);
	220
	221	hrtimer_start(&cq->timer, kt, HRTIMER_MODE_REL);
	222	}
	223
	224	put_cpu();
	225	}
	226
	227	static void null_softirq_done_fn(struct request *rq)
	228	{
9d74e257	229	end_cmd(blk_mq_rq_to_pdu(rq));
f2298c04 JA	230	}
f2298c04 JA	231
f2298c04 JA	232	static inline void null_handle_cmd(struct nullb_cmd *cmd)
	233	{
	234	/* Complete IO by inline, softirq or timer */
	235	switch (irqmode) {
f2298c04	236	case NULL_IRQ_SOFTIRQ:
ce2c350b CH	237	switch (queue_mode) {
	238	case NULL_Q_MQ:
	239	blk_mq_complete_request(cmd->rq);
	240	break;
	241	case NULL_Q_RQ:
	242	blk_complete_request(cmd->rq);
	243	break;
	244	case NULL_Q_BIO:
	245	/*
	246	* XXX: no proper submitting cpu information available.
	247	*/
	248	end_cmd(cmd);
	249	break;
	250	}
	251	break;
	252	case NULL_IRQ_NONE:
f2298c04	253	end_cmd(cmd);
f2298c04 JA	254	break;
	255	case NULL_IRQ_TIMER:
	256	null_cmd_end_timer(cmd);
	257	break;
	258	}
	259	}
	260
	261	static struct nullb_queue nullb_to_queue(struct nullb nullb)
	262	{
	263	int index = 0;
	264
	265	if (nullb->nr_queues != 1)
	266	index = raw_smp_processor_id() / ((nr_cpu_ids + nullb->nr_queues - 1) / nullb->nr_queues);
	267
	268	return &nullb->queues[index];
	269	}
	270
	271	static void null_queue_bio(struct request_queue q, struct bio bio)
	272	{
	273	struct nullb *nullb = q->queuedata;
	274	struct nullb_queue *nq = nullb_to_queue(nullb);
	275	struct nullb_cmd *cmd;
	276
	277	cmd = alloc_cmd(nq, 1);
	278	cmd->bio = bio;
	279
	280	null_handle_cmd(cmd);
	281	}
	282
	283	static int null_rq_prep_fn(struct request_queue q, struct request req)
	284	{
	285	struct nullb *nullb = q->queuedata;
	286	struct nullb_queue *nq = nullb_to_queue(nullb);
	287	struct nullb_cmd *cmd;
	288
	289	cmd = alloc_cmd(nq, 0);
	290	if (cmd) {
	291	cmd->rq = req;
	292	req->special = cmd;
	293	return BLKPREP_OK;
	294	}
	295
	296	return BLKPREP_DEFER;
	297	}
	298
	299	static void null_request_fn(struct request_queue *q)
	300	{
	301	struct request *rq;
	302
	303	while ((rq = blk_fetch_request(q)) != NULL) {
	304	struct nullb_cmd *cmd = rq->special;
	305
	306	spin_unlock_irq(q->queue_lock);
	307	null_handle_cmd(cmd);
	308	spin_lock_irq(q->queue_lock);
	309	}
	310	}
	311
	312	static int null_queue_rq(struct blk_mq_hw_ctx hctx, struct request rq)
	313	{
9d74e257	314	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(rq);
f2298c04 JA	315
	316	cmd->rq = rq;
	317	cmd->nq = hctx->driver_data;
	318
	319	null_handle_cmd(cmd);
	320	return BLK_MQ_RQ_QUEUE_OK;
	321	}
	322
	323	static struct blk_mq_hw_ctx null_alloc_hctx(struct blk_mq_reg reg, unsigned int hctx_index)
	324	{
fc1bc354 MB	325	int b_size = DIV_ROUND_UP(reg->nr_hw_queues, nr_online_nodes);
	326	int tip = (reg->nr_hw_queues % nr_online_nodes);
	327	int node = 0, i, n;
	328
	329	/*
	330	* Split submit queues evenly wrt to the number of nodes. If uneven,
	331	* fill the first buckets with one extra, until the rest is filled with
	332	* no extra.
	333	*/
	334	for (i = 0, n = 1; i < hctx_index; i++, n++) {
	335	if (n % b_size == 0) {
	336	n = 0;
	337	node++;
	338
	339	tip--;
	340	if (!tip)
	341	b_size = reg->nr_hw_queues / nr_online_nodes;
	342	}
	343	}
	344
	345	/*
	346	* A node might not be online, therefore map the relative node id to the
	347	* real node id.
	348	*/
	349	for_each_online_node(n) {
	350	if (!node)
	351	break;
	352	node--;
	353	}
	354
	355	return kzalloc_node(sizeof(struct blk_mq_hw_ctx), GFP_KERNEL, n);
f2298c04 JA	356	}
	357
	358	static void null_free_hctx(struct blk_mq_hw_ctx *hctx, unsigned int hctx_index)
	359	{
	360	kfree(hctx);
	361	}
	362
2d263a78 MB	363	static void null_init_queue(struct nullb nullb, struct nullb_queue nq)
	364	{
	365	BUG_ON(!nullb);
	366	BUG_ON(!nq);
	367
	368	init_waitqueue_head(&nq->wait);
	369	nq->queue_depth = nullb->queue_depth;
	370	}
	371
f2298c04 JA	372	static int null_init_hctx(struct blk_mq_hw_ctx hctx, void data,
	373	unsigned int index)
	374	{
	375	struct nullb *nullb = data;
	376	struct nullb_queue *nq = &nullb->queues[index];
	377
f2298c04	378	hctx->driver_data = nq;
2d263a78 MB	379	null_init_queue(nullb, nq);
2d263a78 MB	380	nullb->nr_queues++;
f2298c04 JA	381
	382	return 0;
	383	}
	384
	385	static struct blk_mq_ops null_mq_ops = {
	386	.queue_rq = null_queue_rq,
	387	.map_queue = blk_mq_map_queue,
	388	.init_hctx = null_init_hctx,
ce2c350b	389	.complete = null_softirq_done_fn,
f2298c04 JA	390	};
	391
	392	static struct blk_mq_reg null_mq_reg = {
	393	.ops = &null_mq_ops,
	394	.queue_depth = 64,
	395	.cmd_size = sizeof(struct nullb_cmd),
	396	.flags = BLK_MQ_F_SHOULD_MERGE,
	397	};
	398
	399	static void null_del_dev(struct nullb *nullb)
	400	{
	401	list_del_init(&nullb->list);
	402
	403	del_gendisk(nullb->disk);
518d00b7	404	blk_cleanup_queue(nullb->q);
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)
	491	goto err_queue;
f2298c04 JA	492	nullb->nr_queues++;
	493	}
	494
2d263a78 MB	495	return 0;
2d263a78 MB	496	err_queue:
f2298c04	497	cleanup_queues(nullb);
2d263a78	498	return ret;
f2298c04 JA	499	}
	500
	501	static int null_add_dev(void)
	502	{
	503	struct gendisk *disk;
	504	struct nullb *nullb;
	505	sector_t size;
	506
	507	nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
	508	if (!nullb)
	509	return -ENOMEM;
	510
	511	spin_lock_init(&nullb->lock);
	512
57053d8c MB	513	if (queue_mode == NULL_Q_MQ && use_per_node_hctx)
	514	submit_queues = nr_online_nodes;
	515
f2298c04 JA	516	if (setup_queues(nullb))
	517	goto err;
	518
	519	if (queue_mode == NULL_Q_MQ) {
	520	null_mq_reg.numa_node = home_node;
	521	null_mq_reg.queue_depth = hw_queue_depth;
57053d8c	522	null_mq_reg.nr_hw_queues = submit_queues;
f2298c04 JA	523
	524	if (use_per_node_hctx) {
	525	null_mq_reg.ops->alloc_hctx = null_alloc_hctx;
	526	null_mq_reg.ops->free_hctx = null_free_hctx;
f2298c04 JA	527	} else {
	528	null_mq_reg.ops->alloc_hctx = blk_mq_alloc_single_hw_queue;
	529	null_mq_reg.ops->free_hctx = blk_mq_free_single_hw_queue;
f2298c04 JA	530	}
	531
	532	nullb->q = blk_mq_init_queue(&null_mq_reg, nullb);
	533	} else if (queue_mode == NULL_Q_BIO) {
	534	nullb->q = blk_alloc_queue_node(GFP_KERNEL, home_node);
	535	blk_queue_make_request(nullb->q, null_queue_bio);
2d263a78	536	init_driver_queues(nullb);
f2298c04 JA	537	} else {
	538	nullb->q = blk_init_queue_node(null_request_fn, &nullb->lock, home_node);
	539	blk_queue_prep_rq(nullb->q, null_rq_prep_fn);
	540	if (nullb->q)
	541	blk_queue_softirq_done(nullb->q, null_softirq_done_fn);
2d263a78	542	init_driver_queues(nullb);
f2298c04 JA	543	}
	544
	545	if (!nullb->q)
	546	goto queue_fail;
	547
	548	nullb->q->queuedata = nullb;
	549	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, nullb->q);
	550
	551	disk = nullb->disk = alloc_disk_node(1, home_node);
	552	if (!disk) {
	553	queue_fail:
518d00b7	554	blk_cleanup_queue(nullb->q);
f2298c04 JA	555	cleanup_queues(nullb);
	556	err:
	557	kfree(nullb);
	558	return -ENOMEM;
	559	}
	560
	561	mutex_lock(&lock);
	562	list_add_tail(&nullb->list, &nullb_list);
	563	nullb->index = nullb_indexes++;
	564	mutex_unlock(&lock);
	565
	566	blk_queue_logical_block_size(nullb->q, bs);
	567	blk_queue_physical_block_size(nullb->q, bs);
	568
	569	size = gb * 1024 * 1024 * 1024ULL;
	570	sector_div(size, bs);
	571	set_capacity(disk, size);
	572
	573	disk->flags \|= GENHD_FL_EXT_DEVT;
	574	disk->major = null_major;
	575	disk->first_minor = nullb->index;
	576	disk->fops = &null_fops;
	577	disk->private_data = nullb;
	578	disk->queue = nullb->q;
	579	sprintf(disk->disk_name, "nullb%d", nullb->index);
	580	add_disk(disk);
	581	return 0;
	582	}
	583
	584	static int __init null_init(void)
	585	{
	586	unsigned int i;
	587
9967d8ac R	588	if (bs > PAGE_SIZE) {
	589	pr_warn("null_blk: invalid block size\n");
	590	pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);
	591	bs = PAGE_SIZE;
	592	}
f2298c04	593
d15ee6b1	594	if (queue_mode == NULL_Q_MQ && use_per_node_hctx) {
fc1bc354	595	if (submit_queues < nr_online_nodes) {
d15ee6b1 MB	596	pr_warn("null_blk: submit_queues param is set to %u.",
d15ee6b1 MB	597	nr_online_nodes);
fc1bc354 MB	598	submit_queues = nr_online_nodes;
fc1bc354 MB	599	}
d15ee6b1	600	} else if (submit_queues > nr_cpu_ids)
f2298c04 JA	601	submit_queues = nr_cpu_ids;
	602	else if (!submit_queues)
	603	submit_queues = 1;
	604
	605	mutex_init(&lock);
	606
	607	/* Initialize a separate list for each CPU for issuing softirqs */
	608	for_each_possible_cpu(i) {
	609	struct completion_queue *cq = &per_cpu(completion_queues, i);
	610
	611	init_llist_head(&cq->list);
	612
	613	if (irqmode != NULL_IRQ_TIMER)
	614	continue;
	615
	616	hrtimer_init(&cq->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	617	cq->timer.function = null_cmd_timer_expired;
	618	}
	619
	620	null_major = register_blkdev(0, "nullb");
	621	if (null_major < 0)
	622	return null_major;
	623
	624	for (i = 0; i < nr_devices; i++) {
	625	if (null_add_dev()) {
	626	unregister_blkdev(null_major, "nullb");
	627	return -EINVAL;
	628	}
	629	}
	630
	631	pr_info("null: module loaded\n");
	632	return 0;
	633	}
	634
	635	static void __exit null_exit(void)
	636	{
	637	struct nullb *nullb;
	638
	639	unregister_blkdev(null_major, "nullb");
	640
	641	mutex_lock(&lock);
	642	while (!list_empty(&nullb_list)) {
	643	nullb = list_entry(nullb_list.next, struct nullb, list);
	644	null_del_dev(nullb);
	645	}
	646	mutex_unlock(&lock);
	647	}
	648
	649	module_init(null_init);
	650	module_exit(null_exit);
	651
	652	MODULE_AUTHOR("Jens Axboe <jaxboe@fusionio.com>");
	653	MODULE_LICENSE("GPL");