[linux-2.6-block.git] / drivers / block / virtio_blk.c

//#define DEBUG
#include <linux/spinlock.h>
#include <linux/blkdev.h>
#include <linux/hdreg.h>
#include <linux/virtio.h>
#include <linux/virtio_blk.h>
#include <linux/scatterlist.h>

#define VIRTIO_MAX_SG	(3+MAX_PHYS_SEGMENTS)

static unsigned char virtblk_index = 'a';
struct virtio_blk
{
	spinlock_t lock;

	struct virtio_device *vdev;
	struct virtqueue *vq;

	/* The disk structure for the kernel. */
	struct gendisk *disk;

	/* Request tracking. */
	struct list_head reqs;

	mempool_t *pool;

	/* Scatterlist: can be too big for stack. */
	struct scatterlist sg[VIRTIO_MAX_SG];
};

struct virtblk_req
{
	struct list_head list;
	struct request *req;
	struct virtio_blk_outhdr out_hdr;
	struct virtio_blk_inhdr in_hdr;
};

static bool blk_done(struct virtqueue *vq)
{
	struct virtio_blk *vblk = vq->vdev->priv;
	struct virtblk_req *vbr;
	unsigned int len;
	unsigned long flags;

	spin_lock_irqsave(&vblk->lock, flags);
	while ((vbr = vblk->vq->vq_ops->get_buf(vblk->vq, &len)) != NULL) {
		int uptodate;
		switch (vbr->in_hdr.status) {
		case VIRTIO_BLK_S_OK:
			uptodate = 1;
			break;
		case VIRTIO_BLK_S_UNSUPP:
			uptodate = -ENOTTY;
			break;
		default:
			uptodate = 0;
			break;
		}

		end_dequeued_request(vbr->req, uptodate);
		list_del(&vbr->list);
		mempool_free(vbr, vblk->pool);
	}
	/* In case queue is stopped waiting for more buffers. */
	blk_start_queue(vblk->disk->queue);
	spin_unlock_irqrestore(&vblk->lock, flags);
	return true;
}

static bool do_req(struct request_queue *q, struct virtio_blk *vblk,
		   struct request *req)
{
	unsigned long num, out, in;
	struct virtblk_req *vbr;

	vbr = mempool_alloc(vblk->pool, GFP_ATOMIC);
	if (!vbr)
		/* When another request finishes we'll try again. */
		return false;

	vbr->req = req;
	if (blk_fs_request(vbr->req)) {
		vbr->out_hdr.type = 0;
		vbr->out_hdr.sector = vbr->req->sector;
		vbr->out_hdr.ioprio = vbr->req->ioprio;
	} else if (blk_pc_request(vbr->req)) {
		vbr->out_hdr.type = VIRTIO_BLK_T_SCSI_CMD;
		vbr->out_hdr.sector = 0;
		vbr->out_hdr.ioprio = vbr->req->ioprio;
	} else {
		/* We don't put anything else in the queue. */
		BUG();
	}

	if (blk_barrier_rq(vbr->req))
		vbr->out_hdr.type |= VIRTIO_BLK_T_BARRIER;

	/* This init could be done at vblk creation time */
	sg_init_table(vblk->sg, VIRTIO_MAX_SG);
	sg_set_buf(&vblk->sg[0], &vbr->out_hdr, sizeof(vbr->out_hdr));
	num = blk_rq_map_sg(q, vbr->req, vblk->sg+1);
	sg_set_buf(&vblk->sg[num+1], &vbr->in_hdr, sizeof(vbr->in_hdr));

	if (rq_data_dir(vbr->req) == WRITE) {
		vbr->out_hdr.type |= VIRTIO_BLK_T_OUT;
		out = 1 + num;
		in = 1;
	} else {
		vbr->out_hdr.type |= VIRTIO_BLK_T_IN;
		out = 1;
		in = 1 + num;
	}

	if (vblk->vq->vq_ops->add_buf(vblk->vq, vblk->sg, out, in, vbr)) {
		mempool_free(vbr, vblk->pool);
		return false;
	}

	list_add_tail(&vbr->list, &vblk->reqs);
	return true;
}

static void do_virtblk_request(struct request_queue *q)
{
	struct virtio_blk *vblk = NULL;
	struct request *req;
	unsigned int issued = 0;

	while ((req = elv_next_request(q)) != NULL) {
		vblk = req->rq_disk->private_data;
		BUG_ON(req->nr_phys_segments > ARRAY_SIZE(vblk->sg));

		/* If this request fails, stop queue and wait for something to
		   finish to restart it. */
		if (!do_req(q, vblk, req)) {
			blk_stop_queue(q);
			break;
		}
		blkdev_dequeue_request(req);
		issued++;
	}

	if (issued)
		vblk->vq->vq_ops->kick(vblk->vq);
}

static int virtblk_ioctl(struct inode *inode, struct file *filp,
			 unsigned cmd, unsigned long data)
{
	return scsi_cmd_ioctl(filp, inode->i_bdev->bd_disk->queue,
			      inode->i_bdev->bd_disk, cmd,
			      (void __user *)data);
}

static struct block_device_operations virtblk_fops = {
	.ioctl = virtblk_ioctl,
	.owner = THIS_MODULE,
};

static int virtblk_probe(struct virtio_device *vdev)
{
	struct virtio_blk *vblk;
	int err, major;
	u64 cap;
	u32 v;

	vdev->priv = vblk = kmalloc(sizeof(*vblk), GFP_KERNEL);
	if (!vblk) {
		err = -ENOMEM;
		goto out;
	}

	INIT_LIST_HEAD(&vblk->reqs);
	spin_lock_init(&vblk->lock);
	vblk->vdev = vdev;

	/* We expect one virtqueue, for output. */
	vblk->vq = vdev->config->find_vq(vdev, 0, blk_done);
	if (IS_ERR(vblk->vq)) {
		err = PTR_ERR(vblk->vq);
		goto out_free_vblk;
	}

	vblk->pool = mempool_create_kmalloc_pool(1,sizeof(struct virtblk_req));
	if (!vblk->pool) {
		err = -ENOMEM;
		goto out_free_vq;
	}

	major = register_blkdev(0, "virtblk");
	if (major < 0) {
		err = major;
		goto out_mempool;
	}

	/* FIXME: How many partitions?  How long is a piece of string? */
	vblk->disk = alloc_disk(1 << 4);
	if (!vblk->disk) {
		err = -ENOMEM;
		goto out_unregister_blkdev;
	}

	vblk->disk->queue = blk_init_queue(do_virtblk_request, &vblk->lock);
	if (!vblk->disk->queue) {
		err = -ENOMEM;
		goto out_put_disk;
	}

	sprintf(vblk->disk->disk_name, "vd%c", virtblk_index++);
	vblk->disk->major = major;
	vblk->disk->first_minor = 0;
	vblk->disk->private_data = vblk;
	vblk->disk->fops = &virtblk_fops;

	/* If barriers are supported, tell block layer that queue is ordered */
	if (vdev->config->feature(vdev, VIRTIO_BLK_F_BARRIER))
		blk_queue_ordered(vblk->disk->queue, QUEUE_ORDERED_TAG, NULL);

	/* Host must always specify the capacity. */
	__virtio_config_val(vdev, offsetof(struct virtio_blk_config, capacity),
			    &cap);

	/* If capacity is too big, truncate with warning. */
	if ((sector_t)cap != cap) {
		dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n",
			 (unsigned long long)cap);
		cap = (sector_t)-1;
	}
	set_capacity(vblk->disk, cap);

	/* Host can optionally specify maximum segment size and number of
	 * segments. */
	err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX,
				offsetof(struct virtio_blk_config, size_max),
				&v);
	if (!err)
		blk_queue_max_segment_size(vblk->disk->queue, v);

	err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX,
				offsetof(struct virtio_blk_config, seg_max),
				&v);
	if (!err)
		blk_queue_max_hw_segments(vblk->disk->queue, v);

	add_disk(vblk->disk);
	return 0;

out_put_disk:
	put_disk(vblk->disk);
out_unregister_blkdev:
	unregister_blkdev(major, "virtblk");
out_mempool:
	mempool_destroy(vblk->pool);
out_free_vq:
	vdev->config->del_vq(vblk->vq);
out_free_vblk:
	kfree(vblk);
out:
	return err;
}

static void virtblk_remove(struct virtio_device *vdev)
{
	struct virtio_blk *vblk = vdev->priv;
	int major = vblk->disk->major;

	BUG_ON(!list_empty(&vblk->reqs));
	blk_cleanup_queue(vblk->disk->queue);
	put_disk(vblk->disk);
	unregister_blkdev(major, "virtblk");
	mempool_destroy(vblk->pool);
	/* There should be nothing in the queue now, so no need to shutdown */
	vdev->config->del_vq(vblk->vq);
	kfree(vblk);
}

static struct virtio_device_id id_table[] = {
	{ VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID },
	{ 0 },
};

static struct virtio_driver virtio_blk = {
	.driver.name =	KBUILD_MODNAME,
	.driver.owner =	THIS_MODULE,
	.id_table =	id_table,
	.probe =	virtblk_probe,
	.remove =	__devexit_p(virtblk_remove),
};

static int __init init(void)
{
	return register_virtio_driver(&virtio_blk);
}

static void __exit fini(void)
{
	unregister_virtio_driver(&virtio_blk);
}
module_init(init);
module_exit(fini);

MODULE_DEVICE_TABLE(virtio, id_table);
MODULE_DESCRIPTION("Virtio block driver");
MODULE_LICENSE("GPL");
Commit	Line	Data
e467cde2 RR	1	//#define DEBUG
	2	#include <linux/spinlock.h>
	3	#include <linux/blkdev.h>
	4	#include <linux/hdreg.h>
	5	#include <linux/virtio.h>
	6	#include <linux/virtio_blk.h>
3d1266c7 JA	7	#include <linux/scatterlist.h>
	8
	9	#define VIRTIO_MAX_SG (3+MAX_PHYS_SEGMENTS)
e467cde2 RR	10
	11	static unsigned char virtblk_index = 'a';
	12	struct virtio_blk
	13	{
	14	spinlock_t lock;
	15
	16	struct virtio_device *vdev;
	17	struct virtqueue *vq;
	18
	19	/* The disk structure for the kernel. */
	20	struct gendisk *disk;
	21
	22	/* Request tracking. */
	23	struct list_head reqs;
	24
	25	mempool_t *pool;
	26
	27	/* Scatterlist: can be too big for stack. */
3d1266c7	28	struct scatterlist sg[VIRTIO_MAX_SG];
e467cde2 RR	29	};
	30
	31	struct virtblk_req
	32	{
	33	struct list_head list;
	34	struct request *req;
	35	struct virtio_blk_outhdr out_hdr;
	36	struct virtio_blk_inhdr in_hdr;
	37	};
	38
	39	static bool blk_done(struct virtqueue *vq)
	40	{
	41	struct virtio_blk *vblk = vq->vdev->priv;
	42	struct virtblk_req *vbr;
	43	unsigned int len;
	44	unsigned long flags;
	45
	46	spin_lock_irqsave(&vblk->lock, flags);
	47	while ((vbr = vblk->vq->vq_ops->get_buf(vblk->vq, &len)) != NULL) {
	48	int uptodate;
	49	switch (vbr->in_hdr.status) {
	50	case VIRTIO_BLK_S_OK:
	51	uptodate = 1;
	52	break;
	53	case VIRTIO_BLK_S_UNSUPP:
	54	uptodate = -ENOTTY;
	55	break;
	56	default:
	57	uptodate = 0;
	58	break;
	59	}
	60
	61	end_dequeued_request(vbr->req, uptodate);
	62	list_del(&vbr->list);
	63	mempool_free(vbr, vblk->pool);
	64	}
	65	/* In case queue is stopped waiting for more buffers. */
	66	blk_start_queue(vblk->disk->queue);
	67	spin_unlock_irqrestore(&vblk->lock, flags);
	68	return true;
	69	}
	70
	71	static bool do_req(struct request_queue q, struct virtio_blk vblk,
	72	struct request *req)
	73	{
	74	unsigned long num, out, in;
	75	struct virtblk_req *vbr;
	76
	77	vbr = mempool_alloc(vblk->pool, GFP_ATOMIC);
	78	if (!vbr)
	79	/* When another request finishes we'll try again. */
	80	return false;
	81
	82	vbr->req = req;
	83	if (blk_fs_request(vbr->req)) {
	84	vbr->out_hdr.type = 0;
	85	vbr->out_hdr.sector = vbr->req->sector;
	86	vbr->out_hdr.ioprio = vbr->req->ioprio;
	87	} else if (blk_pc_request(vbr->req)) {
	88	vbr->out_hdr.type = VIRTIO_BLK_T_SCSI_CMD;
	89	vbr->out_hdr.sector = 0;
	90	vbr->out_hdr.ioprio = vbr->req->ioprio;
	91	} else {
	92	/* We don't put anything else in the queue. */
93	BUG();
94	}
95
96	if (blk_barrier_rq(vbr->req))
97	vbr->out_hdr.type \|= VIRTIO_BLK_T_BARRIER;
98
3d1266c7 JA	99	/* This init could be done at vblk creation time */
3d1266c7 JA	100	sg_init_table(vblk->sg, VIRTIO_MAX_SG);
e467cde2 RR	101	sg_set_buf(&vblk->sg[0], &vbr->out_hdr, sizeof(vbr->out_hdr));
	102	num = blk_rq_map_sg(q, vbr->req, vblk->sg+1);
	103	sg_set_buf(&vblk->sg[num+1], &vbr->in_hdr, sizeof(vbr->in_hdr));
	104
	105	if (rq_data_dir(vbr->req) == WRITE) {
	106	vbr->out_hdr.type \|= VIRTIO_BLK_T_OUT;
	107	out = 1 + num;
	108	in = 1;
	109	} else {
	110	vbr->out_hdr.type \|= VIRTIO_BLK_T_IN;
	111	out = 1;
	112	in = 1 + num;
	113	}
	114
	115	if (vblk->vq->vq_ops->add_buf(vblk->vq, vblk->sg, out, in, vbr)) {
	116	mempool_free(vbr, vblk->pool);
	117	return false;
	118	}
	119
	120	list_add_tail(&vbr->list, &vblk->reqs);
	121	return true;
	122	}
	123
	124	static void do_virtblk_request(struct request_queue *q)
	125	{
	126	struct virtio_blk *vblk = NULL;
	127	struct request *req;
	128	unsigned int issued = 0;
	129
	130	while ((req = elv_next_request(q)) != NULL) {
	131	vblk = req->rq_disk->private_data;
	132	BUG_ON(req->nr_phys_segments > ARRAY_SIZE(vblk->sg));
	133
	134	/* If this request fails, stop queue and wait for something to
	135	finish to restart it. */
	136	if (!do_req(q, vblk, req)) {
	137	blk_stop_queue(q);
	138	break;
	139	}
	140	blkdev_dequeue_request(req);
	141	issued++;
	142	}
	143
	144	if (issued)
	145	vblk->vq->vq_ops->kick(vblk->vq);
	146	}
	147
	148	static int virtblk_ioctl(struct inode inode, struct file filp,
	149	unsigned cmd, unsigned long data)
	150	{
	151	return scsi_cmd_ioctl(filp, inode->i_bdev->bd_disk->queue,
	152	inode->i_bdev->bd_disk, cmd,
	153	(void __user *)data);
	154	}
	155
	156	static struct block_device_operations virtblk_fops = {
	157	.ioctl = virtblk_ioctl,
	158	.owner = THIS_MODULE,
	159	};
	160
	161	static int virtblk_probe(struct virtio_device *vdev)
	162	{
	163	struct virtio_blk *vblk;
	164	int err, major;
e467cde2 RR	165	u64 cap;
	166	u32 v;
	167
	168	vdev->priv = vblk = kmalloc(sizeof(*vblk), GFP_KERNEL);
	169	if (!vblk) {
	170	err = -ENOMEM;
	171	goto out;
	172	}
	173
	174	INIT_LIST_HEAD(&vblk->reqs);
	175	spin_lock_init(&vblk->lock);
	176	vblk->vdev = vdev;
	177
	178	/* We expect one virtqueue, for output. */
a586d4f6	179	vblk->vq = vdev->config->find_vq(vdev, 0, blk_done);
e467cde2 RR	180	if (IS_ERR(vblk->vq)) {
	181	err = PTR_ERR(vblk->vq);
	182	goto out_free_vblk;
	183	}
	184
	185	vblk->pool = mempool_create_kmalloc_pool(1,sizeof(struct virtblk_req));
	186	if (!vblk->pool) {
	187	err = -ENOMEM;
	188	goto out_free_vq;
	189	}
	190
	191	major = register_blkdev(0, "virtblk");
	192	if (major < 0) {
	193	err = major;
	194	goto out_mempool;
	195	}
	196
	197	/* FIXME: How many partitions? How long is a piece of string? */
	198	vblk->disk = alloc_disk(1 << 4);
	199	if (!vblk->disk) {
	200	err = -ENOMEM;
	201	goto out_unregister_blkdev;
	202	}
	203
	204	vblk->disk->queue = blk_init_queue(do_virtblk_request, &vblk->lock);
	205	if (!vblk->disk->queue) {
	206	err = -ENOMEM;
	207	goto out_put_disk;
	208	}
	209
	210	sprintf(vblk->disk->disk_name, "vd%c", virtblk_index++);
	211	vblk->disk->major = major;
	212	vblk->disk->first_minor = 0;
	213	vblk->disk->private_data = vblk;
	214	vblk->disk->fops = &virtblk_fops;
	215
	216	/* If barriers are supported, tell block layer that queue is ordered */
a586d4f6	217	if (vdev->config->feature(vdev, VIRTIO_BLK_F_BARRIER))
e467cde2 RR	218	blk_queue_ordered(vblk->disk->queue, QUEUE_ORDERED_TAG, NULL);
e467cde2 RR	219
a586d4f6 RR	220	/* Host must always specify the capacity. */
	221	__virtio_config_val(vdev, offsetof(struct virtio_blk_config, capacity),
	222	&cap);
e467cde2 RR	223
	224	/* If capacity is too big, truncate with warning. */
	225	if ((sector_t)cap != cap) {
	226	dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n",
	227	(unsigned long long)cap);
	228	cap = (sector_t)-1;
	229	}
	230	set_capacity(vblk->disk, cap);
	231
a586d4f6 RR	232	/* Host can optionally specify maximum segment size and number of
	233	* segments. */
	234	err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX,
	235	offsetof(struct virtio_blk_config, size_max),
	236	&v);
e467cde2 RR	237	if (!err)
e467cde2 RR	238	blk_queue_max_segment_size(vblk->disk->queue, v);
e467cde2	239
a586d4f6 RR	240	err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX,
	241	offsetof(struct virtio_blk_config, seg_max),
	242	&v);
e467cde2 RR	243	if (!err)
e467cde2 RR	244	blk_queue_max_hw_segments(vblk->disk->queue, v);
e467cde2 RR	245
	246	add_disk(vblk->disk);
	247	return 0;
	248
	249	out_put_disk:
	250	put_disk(vblk->disk);
	251	out_unregister_blkdev:
	252	unregister_blkdev(major, "virtblk");
	253	out_mempool:
	254	mempool_destroy(vblk->pool);
	255	out_free_vq:
	256	vdev->config->del_vq(vblk->vq);
	257	out_free_vblk:
	258	kfree(vblk);
	259	out:
	260	return err;
	261	}
	262
	263	static void virtblk_remove(struct virtio_device *vdev)
	264	{
	265	struct virtio_blk *vblk = vdev->priv;
	266	int major = vblk->disk->major;
	267
	268	BUG_ON(!list_empty(&vblk->reqs));
	269	blk_cleanup_queue(vblk->disk->queue);
	270	put_disk(vblk->disk);
	271	unregister_blkdev(major, "virtblk");
	272	mempool_destroy(vblk->pool);
74b2553f RR	273	/* There should be nothing in the queue now, so no need to shutdown */
74b2553f RR	274	vdev->config->del_vq(vblk->vq);
e467cde2 RR	275	kfree(vblk);
	276	}
	277
	278	static struct virtio_device_id id_table[] = {
	279	{ VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID },
	280	{ 0 },
	281	};
	282
	283	static struct virtio_driver virtio_blk = {
	284	.driver.name = KBUILD_MODNAME,
	285	.driver.owner = THIS_MODULE,
	286	.id_table = id_table,
	287	.probe = virtblk_probe,
	288	.remove = __devexit_p(virtblk_remove),
	289	};
	290
	291	static int __init init(void)
	292	{
	293	return register_virtio_driver(&virtio_blk);
	294	}
	295
	296	static void __exit fini(void)
	297	{
	298	unregister_virtio_driver(&virtio_blk);
	299	}
	300	module_init(init);
	301	module_exit(fini);
	302
	303	MODULE_DEVICE_TABLE(virtio, id_table);
	304	MODULE_DESCRIPTION("Virtio block driver");
	305	MODULE_LICENSE("GPL");