virtio_blk.c 23.4 KB
Newer Older
R
Rusty Russell 已提交
1 2
//#define DEBUG
#include <linux/spinlock.h>
3
#include <linux/slab.h>
R
Rusty Russell 已提交
4 5
#include <linux/blkdev.h>
#include <linux/hdreg.h>
6
#include <linux/module.h>
7
#include <linux/mutex.h>
8
#include <linux/interrupt.h>
R
Rusty Russell 已提交
9 10
#include <linux/virtio.h>
#include <linux/virtio_blk.h>
11
#include <linux/scatterlist.h>
12
#include <linux/string_helpers.h>
13
#include <scsi/scsi_cmnd.h>
14
#include <linux/idr.h>
J
Jens Axboe 已提交
15
#include <linux/blk-mq.h>
16
#include <linux/blk-mq-virtio.h>
J
Jens Axboe 已提交
17
#include <linux/numa.h>
18

19
#define PART_BITS 4
20
#define VQ_NAME_LEN 16
R
Rusty Russell 已提交
21

22 23 24
static int major;
static DEFINE_IDA(vd_index_ida);

25
static struct workqueue_struct *virtblk_wq;
26

27 28 29 30 31 32
struct virtio_blk_vq {
	struct virtqueue *vq;
	spinlock_t lock;
	char name[VQ_NAME_LEN];
} ____cacheline_aligned_in_smp;

33
struct virtio_blk {
R
Rusty Russell 已提交
34 35 36 37 38
	struct virtio_device *vdev;

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

39 40 41
	/* Block layer tags. */
	struct blk_mq_tag_set tag_set;

42 43 44
	/* Process context for config space updates */
	struct work_struct config_work;

45 46 47
	/* What host tells us, plus 2 for header & tailer. */
	unsigned int sg_elems;

48 49
	/* Ida index - used to track minor number allocations. */
	int index;
50 51 52 53

	/* num of vqs */
	int num_vqs;
	struct virtio_blk_vq *vqs;
R
Rusty Russell 已提交
54 55
};

56
struct virtblk_req {
R
Rusty Russell 已提交
57 58
	struct request *req;
	struct virtio_blk_outhdr out_hdr;
59
	struct virtio_scsi_inhdr in_hdr;
60
	u8 status;
61
	u8 sense[SCSI_SENSE_BUFFERSIZE];
62
	struct scatterlist sg[];
R
Rusty Russell 已提交
63 64
};

65 66 67 68 69 70 71 72 73 74 75 76
static inline int virtblk_result(struct virtblk_req *vbr)
{
	switch (vbr->status) {
	case VIRTIO_BLK_S_OK:
		return 0;
	case VIRTIO_BLK_S_UNSUPP:
		return -ENOTTY;
	default:
		return -EIO;
	}
}

77
static int __virtblk_add_req(struct virtqueue *vq,
78 79
			     struct virtblk_req *vbr,
			     struct scatterlist *data_sg,
R
Rusty Russell 已提交
80
			     bool have_data)
81
{
82
	struct scatterlist hdr, status, cmd, sense, inhdr, *sgs[6];
83
	unsigned int num_out = 0, num_in = 0;
M
Michael S. Tsirkin 已提交
84
	__virtio32 type = vbr->out_hdr.type & ~cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_OUT);
85 86 87 88

	sg_init_one(&hdr, &vbr->out_hdr, sizeof(vbr->out_hdr));
	sgs[num_out++] = &hdr;

89 90 91 92 93 94
	/*
	 * If this is a packet command we need a couple of additional headers.
	 * Behind the normal outhdr we put a segment with the scsi command
	 * block, and before the normal inhdr we put the sense data and the
	 * inhdr with additional status information.
	 */
M
Michael S. Tsirkin 已提交
95
	if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
96 97 98 99
		sg_init_one(&cmd, vbr->req->cmd, vbr->req->cmd_len);
		sgs[num_out++] = &cmd;
	}

R
Rusty Russell 已提交
100
	if (have_data) {
M
Michael S. Tsirkin 已提交
101
		if (vbr->out_hdr.type & cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_OUT))
102
			sgs[num_out++] = data_sg;
103
		else
104 105 106
			sgs[num_out + num_in++] = data_sg;
	}

M
Michael S. Tsirkin 已提交
107
	if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
108 109
		memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
		sg_init_one(&sense, vbr->sense, SCSI_SENSE_BUFFERSIZE);
110 111 112
		sgs[num_out + num_in++] = &sense;
		sg_init_one(&inhdr, &vbr->in_hdr, sizeof(vbr->in_hdr));
		sgs[num_out + num_in++] = &inhdr;
113 114 115 116 117 118
	}

	sg_init_one(&status, &vbr->status, sizeof(vbr->status));
	sgs[num_out + num_in++] = &status;

	return virtqueue_add_sgs(vq, sgs, num_out, num_in, vbr, GFP_ATOMIC);
119 120
}

121
static inline void virtblk_request_done(struct request *req)
122
{
123
	struct virtblk_req *vbr = blk_mq_rq_to_pdu(req);
M
Michael S. Tsirkin 已提交
124
	struct virtio_blk *vblk = req->q->queuedata;
125 126 127
	int error = virtblk_result(vbr);

	if (req->cmd_type == REQ_TYPE_BLOCK_PC) {
M
Michael S. Tsirkin 已提交
128 129 130
		req->resid_len = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.residual);
		req->sense_len = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.sense_len);
		req->errors = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.errors);
131
	} else if (req->cmd_type == REQ_TYPE_DRV_PRIV) {
132 133 134
		req->errors = (error != 0);
	}

135
	blk_mq_end_request(req, error);
136 137 138
}

static void virtblk_done(struct virtqueue *vq)
R
Rusty Russell 已提交
139 140
{
	struct virtio_blk *vblk = vq->vdev->priv;
J
Jens Axboe 已提交
141
	bool req_done = false;
142
	int qid = vq->index;
R
Rusty Russell 已提交
143 144
	struct virtblk_req *vbr;
	unsigned long flags;
145
	unsigned int len;
R
Rusty Russell 已提交
146

147
	spin_lock_irqsave(&vblk->vqs[qid].lock, flags);
148 149
	do {
		virtqueue_disable_cb(vq);
150
		while ((vbr = virtqueue_get_buf(vblk->vqs[qid].vq, &len)) != NULL) {
151
			blk_mq_complete_request(vbr->req, vbr->req->errors);
J
Jens Axboe 已提交
152
			req_done = true;
153
		}
154 155
		if (unlikely(virtqueue_is_broken(vq)))
			break;
156
	} while (!virtqueue_enable_cb(vq));
J
Jens Axboe 已提交
157

R
Rusty Russell 已提交
158
	/* In case queue is stopped waiting for more buffers. */
159
	if (req_done)
160
		blk_mq_start_stopped_hw_queues(vblk->disk->queue, true);
161
	spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags);
162 163
}

164 165
static int virtio_queue_rq(struct blk_mq_hw_ctx *hctx,
			   const struct blk_mq_queue_data *bd)
R
Rusty Russell 已提交
166
{
J
Jens Axboe 已提交
167
	struct virtio_blk *vblk = hctx->queue->queuedata;
168
	struct request *req = bd->rq;
169
	struct virtblk_req *vbr = blk_mq_rq_to_pdu(req);
J
Jens Axboe 已提交
170
	unsigned long flags;
171
	unsigned int num;
172
	int qid = hctx->queue_num;
173
	int err;
174
	bool notify = false;
R
Rusty Russell 已提交
175

J
Jens Axboe 已提交
176
	BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems);
R
Rusty Russell 已提交
177 178

	vbr->req = req;
179
	if (req_op(req) == REQ_OP_FLUSH) {
M
Michael S. Tsirkin 已提交
180
		vbr->out_hdr.type = cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_FLUSH);
181
		vbr->out_hdr.sector = 0;
M
Michael S. Tsirkin 已提交
182
		vbr->out_hdr.ioprio = cpu_to_virtio32(vblk->vdev, req_get_ioprio(vbr->req));
183 184 185 186
	} else {
		switch (req->cmd_type) {
		case REQ_TYPE_FS:
			vbr->out_hdr.type = 0;
M
Michael S. Tsirkin 已提交
187 188
			vbr->out_hdr.sector = cpu_to_virtio64(vblk->vdev, blk_rq_pos(vbr->req));
			vbr->out_hdr.ioprio = cpu_to_virtio32(vblk->vdev, req_get_ioprio(vbr->req));
189 190
			break;
		case REQ_TYPE_BLOCK_PC:
M
Michael S. Tsirkin 已提交
191
			vbr->out_hdr.type = cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_SCSI_CMD);
192
			vbr->out_hdr.sector = 0;
M
Michael S. Tsirkin 已提交
193
			vbr->out_hdr.ioprio = cpu_to_virtio32(vblk->vdev, req_get_ioprio(vbr->req));
194
			break;
195
		case REQ_TYPE_DRV_PRIV:
M
Michael S. Tsirkin 已提交
196
			vbr->out_hdr.type = cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_GET_ID);
197
			vbr->out_hdr.sector = 0;
M
Michael S. Tsirkin 已提交
198
			vbr->out_hdr.ioprio = cpu_to_virtio32(vblk->vdev, req_get_ioprio(vbr->req));
199 200 201 202
			break;
		default:
			/* We don't put anything else in the queue. */
			BUG();
203
		}
R
Rusty Russell 已提交
204 205
	}

206 207
	blk_mq_start_request(req);

J
Jens Axboe 已提交
208
	num = blk_rq_map_sg(hctx->queue, vbr->req, vbr->sg);
209
	if (num) {
210
		if (rq_data_dir(vbr->req) == WRITE)
M
Michael S. Tsirkin 已提交
211
			vbr->out_hdr.type |= cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_OUT);
212
		else
M
Michael S. Tsirkin 已提交
213
			vbr->out_hdr.type |= cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_IN);
R
Rusty Russell 已提交
214 215
	}

216 217
	spin_lock_irqsave(&vblk->vqs[qid].lock, flags);
	err = __virtblk_add_req(vblk->vqs[qid].vq, vbr, vbr->sg, num);
218
	if (err) {
219
		virtqueue_kick(vblk->vqs[qid].vq);
J
Jens Axboe 已提交
220
		blk_mq_stop_hw_queue(hctx);
221
		spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags);
222 223 224 225 226
		/* Out of mem doesn't actually happen, since we fall back
		 * to direct descriptors */
		if (err == -ENOMEM || err == -ENOSPC)
			return BLK_MQ_RQ_QUEUE_BUSY;
		return BLK_MQ_RQ_QUEUE_ERROR;
227 228
	}

229
	if (bd->last && virtqueue_kick_prepare(vblk->vqs[qid].vq))
230
		notify = true;
231
	spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags);
232 233

	if (notify)
234
		virtqueue_notify(vblk->vqs[qid].vq);
J
Jens Axboe 已提交
235
	return BLK_MQ_RQ_QUEUE_OK;
236 237
}

238 239 240 241 242
/* return id (s/n) string for *disk to *id_str
 */
static int virtblk_get_id(struct gendisk *disk, char *id_str)
{
	struct virtio_blk *vblk = disk->private_data;
243
	struct request_queue *q = vblk->disk->queue;
244
	struct request *req;
M
Mike Snitzer 已提交
245
	int err;
246

247 248
	req = blk_get_request(q, READ, GFP_KERNEL);
	if (IS_ERR(req))
249
		return PTR_ERR(req);
250
	req->cmd_type = REQ_TYPE_DRV_PRIV;
251 252 253 254 255

	err = blk_rq_map_kern(q, req, id_str, VIRTIO_BLK_ID_BYTES, GFP_KERNEL);
	if (err)
		goto out;

M
Mike Snitzer 已提交
256
	err = blk_execute_rq(vblk->disk->queue, vblk->disk, req, false);
257
out:
M
Mike Snitzer 已提交
258 259
	blk_put_request(req);
	return err;
260 261
}

262 263
static int virtblk_ioctl(struct block_device *bdev, fmode_t mode,
			     unsigned int cmd, unsigned long data)
R
Rusty Russell 已提交
264
{
265 266 267 268 269 270 271
	struct gendisk *disk = bdev->bd_disk;
	struct virtio_blk *vblk = disk->private_data;

	/*
	 * Only allow the generic SCSI ioctls if the host can support it.
	 */
	if (!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_SCSI))
272
		return -ENOTTY;
273

274 275
	return scsi_cmd_blk_ioctl(bdev, mode, cmd,
				  (void __user *)data);
R
Rusty Russell 已提交
276 277
}

278 279 280
/* We provide getgeo only to please some old bootloader/partitioning tools */
static int virtblk_getgeo(struct block_device *bd, struct hd_geometry *geo)
{
281 282 283
	struct virtio_blk *vblk = bd->bd_disk->private_data;

	/* see if the host passed in geometry config */
284 285 286 287 288 289 290
	if (virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_GEOMETRY)) {
		virtio_cread(vblk->vdev, struct virtio_blk_config,
			     geometry.cylinders, &geo->cylinders);
		virtio_cread(vblk->vdev, struct virtio_blk_config,
			     geometry.heads, &geo->heads);
		virtio_cread(vblk->vdev, struct virtio_blk_config,
			     geometry.sectors, &geo->sectors);
291 292 293 294 295 296
	} else {
		/* some standard values, similar to sd */
		geo->heads = 1 << 6;
		geo->sectors = 1 << 5;
		geo->cylinders = get_capacity(bd->bd_disk) >> 11;
	}
297 298 299
	return 0;
}

300
static const struct block_device_operations virtblk_fops = {
301
	.ioctl  = virtblk_ioctl,
302 303
	.owner  = THIS_MODULE,
	.getgeo = virtblk_getgeo,
R
Rusty Russell 已提交
304 305
};

306 307 308 309 310
static int index_to_minor(int index)
{
	return index << PART_BITS;
}

311 312 313 314 315
static int minor_to_index(int minor)
{
	return minor >> PART_BITS;
}

316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
static ssize_t virtblk_serial_show(struct device *dev,
				struct device_attribute *attr, char *buf)
{
	struct gendisk *disk = dev_to_disk(dev);
	int err;

	/* sysfs gives us a PAGE_SIZE buffer */
	BUILD_BUG_ON(PAGE_SIZE < VIRTIO_BLK_ID_BYTES);

	buf[VIRTIO_BLK_ID_BYTES] = '\0';
	err = virtblk_get_id(disk, buf);
	if (!err)
		return strlen(buf);

	if (err == -EIO) /* Unsupported? Make it empty. */
		return 0;

	return err;
}
335 336

static DEVICE_ATTR(serial, S_IRUGO, virtblk_serial_show, NULL);
337

338 339 340 341 342 343 344
static void virtblk_config_changed_work(struct work_struct *work)
{
	struct virtio_blk *vblk =
		container_of(work, struct virtio_blk, config_work);
	struct virtio_device *vdev = vblk->vdev;
	struct request_queue *q = vblk->disk->queue;
	char cap_str_2[10], cap_str_10[10];
345
	char *envp[] = { "RESIZE=1", NULL };
346
	u64 capacity;
347 348

	/* Host must always specify the capacity. */
349
	virtio_cread(vdev, struct virtio_blk_config, capacity, &capacity);
350 351 352 353 354 355 356 357

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

358 359 360 361
	string_get_size(capacity, queue_logical_block_size(q),
			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
	string_get_size(capacity, queue_logical_block_size(q),
			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
362 363 364 365 366 367 368 369

	dev_notice(&vdev->dev,
		  "new size: %llu %d-byte logical blocks (%s/%s)\n",
		  (unsigned long long)capacity,
		  queue_logical_block_size(q),
		  cap_str_10, cap_str_2);

	set_capacity(vblk->disk, capacity);
370
	revalidate_disk(vblk->disk);
371
	kobject_uevent_env(&disk_to_dev(vblk->disk)->kobj, KOBJ_CHANGE, envp);
372 373 374 375 376 377 378 379 380
}

static void virtblk_config_changed(struct virtio_device *vdev)
{
	struct virtio_blk *vblk = vdev->priv;

	queue_work(virtblk_wq, &vblk->config_work);
}

381 382
static int init_vq(struct virtio_blk *vblk)
{
383
	int err;
384 385 386 387 388 389
	int i;
	vq_callback_t **callbacks;
	const char **names;
	struct virtqueue **vqs;
	unsigned short num_vqs;
	struct virtio_device *vdev = vblk->vdev;
390
	struct irq_affinity desc = { 0, };
391 392 393 394 395 396 397

	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_MQ,
				   struct virtio_blk_config, num_queues,
				   &num_vqs);
	if (err)
		num_vqs = 1;

398
	vblk->vqs = kmalloc_array(num_vqs, sizeof(*vblk->vqs), GFP_KERNEL);
399 400
	if (!vblk->vqs)
		return -ENOMEM;
401

402 403 404
	names = kmalloc_array(num_vqs, sizeof(*names), GFP_KERNEL);
	callbacks = kmalloc_array(num_vqs, sizeof(*callbacks), GFP_KERNEL);
	vqs = kmalloc_array(num_vqs, sizeof(*vqs), GFP_KERNEL);
405 406 407 408
	if (!names || !callbacks || !vqs) {
		err = -ENOMEM;
		goto out;
	}
409

410 411 412 413 414 415 416
	for (i = 0; i < num_vqs; i++) {
		callbacks[i] = virtblk_done;
		snprintf(vblk->vqs[i].name, VQ_NAME_LEN, "req.%d", i);
		names[i] = vblk->vqs[i].name;
	}

	/* Discover virtqueues and write information to configuration.  */
417
	err = vdev->config->find_vqs(vdev, num_vqs, vqs, callbacks, names,
418
			&desc);
419
	if (err)
420
		goto out;
421

422 423 424 425 426 427
	for (i = 0; i < num_vqs; i++) {
		spin_lock_init(&vblk->vqs[i].lock);
		vblk->vqs[i].vq = vqs[i];
	}
	vblk->num_vqs = num_vqs;

428
out:
429 430 431 432 433
	kfree(vqs);
	kfree(callbacks);
	kfree(names);
	if (err)
		kfree(vblk->vqs);
434 435 436
	return err;
}

437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464
/*
 * Legacy naming scheme used for virtio devices.  We are stuck with it for
 * virtio blk but don't ever use it for any new driver.
 */
static int virtblk_name_format(char *prefix, int index, char *buf, int buflen)
{
	const int base = 'z' - 'a' + 1;
	char *begin = buf + strlen(prefix);
	char *end = buf + buflen;
	char *p;
	int unit;

	p = end - 1;
	*p = '\0';
	unit = base;
	do {
		if (p == begin)
			return -EINVAL;
		*--p = 'a' + (index % unit);
		index = (index / unit) - 1;
	} while (index >= 0);

	memmove(begin, p, end - p);
	memcpy(buf, prefix, strlen(prefix));

	return 0;
}

465 466 467 468 469
static int virtblk_get_cache_mode(struct virtio_device *vdev)
{
	u8 writeback;
	int err;

470 471 472
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_CONFIG_WCE,
				   struct virtio_blk_config, wce,
				   &writeback);
473 474 475 476 477

	/*
	 * If WCE is not configurable and flush is not available,
	 * assume no writeback cache is in use.
	 */
478
	if (err)
479
		writeback = virtio_has_feature(vdev, VIRTIO_BLK_F_FLUSH);
480 481 482 483 484 485 486 487 488

	return writeback;
}

static void virtblk_update_cache_mode(struct virtio_device *vdev)
{
	u8 writeback = virtblk_get_cache_mode(vdev);
	struct virtio_blk *vblk = vdev->priv;

489
	blk_queue_write_cache(vblk->disk->queue, writeback, false);
490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513
	revalidate_disk(vblk->disk);
}

static const char *const virtblk_cache_types[] = {
	"write through", "write back"
};

static ssize_t
virtblk_cache_type_store(struct device *dev, struct device_attribute *attr,
			 const char *buf, size_t count)
{
	struct gendisk *disk = dev_to_disk(dev);
	struct virtio_blk *vblk = disk->private_data;
	struct virtio_device *vdev = vblk->vdev;
	int i;

	BUG_ON(!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_CONFIG_WCE));
	for (i = ARRAY_SIZE(virtblk_cache_types); --i >= 0; )
		if (sysfs_streq(buf, virtblk_cache_types[i]))
			break;

	if (i < 0)
		return -EINVAL;

514
	virtio_cwrite8(vdev, offsetof(struct virtio_blk_config, wce), i);
515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
	virtblk_update_cache_mode(vdev);
	return count;
}

static ssize_t
virtblk_cache_type_show(struct device *dev, struct device_attribute *attr,
			 char *buf)
{
	struct gendisk *disk = dev_to_disk(dev);
	struct virtio_blk *vblk = disk->private_data;
	u8 writeback = virtblk_get_cache_mode(vblk->vdev);

	BUG_ON(writeback >= ARRAY_SIZE(virtblk_cache_types));
	return snprintf(buf, 40, "%s\n", virtblk_cache_types[writeback]);
}

static const struct device_attribute dev_attr_cache_type_ro =
	__ATTR(cache_type, S_IRUGO,
	       virtblk_cache_type_show, NULL);
static const struct device_attribute dev_attr_cache_type_rw =
	__ATTR(cache_type, S_IRUGO|S_IWUSR,
	       virtblk_cache_type_show, virtblk_cache_type_store);

538 539 540
static int virtblk_init_request(void *data, struct request *rq,
		unsigned int hctx_idx, unsigned int request_idx,
		unsigned int numa_node)
541 542 543 544 545 546 547 548
{
	struct virtio_blk *vblk = data;
	struct virtblk_req *vbr = blk_mq_rq_to_pdu(rq);

	sg_init_table(vbr->sg, vblk->sg_elems);
	return 0;
}

549 550 551 552 553 554 555
static int virtblk_map_queues(struct blk_mq_tag_set *set)
{
	struct virtio_blk *vblk = set->driver_data;

	return blk_mq_virtio_map_queues(set, vblk->vdev, 0);
}

J
Jens Axboe 已提交
556 557
static struct blk_mq_ops virtio_mq_ops = {
	.queue_rq	= virtio_queue_rq,
558
	.complete	= virtblk_request_done,
559
	.init_request	= virtblk_init_request,
560
	.map_queues	= virtblk_map_queues,
J
Jens Axboe 已提交
561 562
};

563 564
static unsigned int virtblk_queue_depth;
module_param_named(queue_depth, virtblk_queue_depth, uint, 0444);
J
Jens Axboe 已提交
565

566
static int virtblk_probe(struct virtio_device *vdev)
R
Rusty Russell 已提交
567 568
{
	struct virtio_blk *vblk;
569
	struct request_queue *q;
570
	int err, index;
571

R
Rusty Russell 已提交
572
	u64 cap;
573 574 575
	u32 v, blk_size, sg_elems, opt_io_size;
	u16 min_io_size;
	u8 physical_block_exp, alignment_offset;
R
Rusty Russell 已提交
576

577 578 579 580 581 582
	if (!vdev->config->get) {
		dev_err(&vdev->dev, "%s failure: config access disabled\n",
			__func__);
		return -EINVAL;
	}

583 584 585 586 587
	err = ida_simple_get(&vd_index_ida, 0, minor_to_index(1 << MINORBITS),
			     GFP_KERNEL);
	if (err < 0)
		goto out;
	index = err;
588

589
	/* We need to know how many segments before we allocate. */
590 591 592
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_SEG_MAX,
				   struct virtio_blk_config, seg_max,
				   &sg_elems);
593 594 595

	/* We need at least one SG element, whatever they say. */
	if (err || !sg_elems)
596 597 598 599
		sg_elems = 1;

	/* We need an extra sg elements at head and tail. */
	sg_elems += 2;
J
Jens Axboe 已提交
600
	vdev->priv = vblk = kmalloc(sizeof(*vblk), GFP_KERNEL);
R
Rusty Russell 已提交
601 602
	if (!vblk) {
		err = -ENOMEM;
603
		goto out_free_index;
R
Rusty Russell 已提交
604 605 606
	}

	vblk->vdev = vdev;
607
	vblk->sg_elems = sg_elems;
608

609
	INIT_WORK(&vblk->config_work, virtblk_config_changed_work);
R
Rusty Russell 已提交
610

611 612
	err = init_vq(vblk);
	if (err)
R
Rusty Russell 已提交
613 614 615
		goto out_free_vblk;

	/* FIXME: How many partitions?  How long is a piece of string? */
616
	vblk->disk = alloc_disk(1 << PART_BITS);
R
Rusty Russell 已提交
617 618
	if (!vblk->disk) {
		err = -ENOMEM;
J
Jens Axboe 已提交
619
		goto out_free_vq;
R
Rusty Russell 已提交
620 621
	}

622
	/* Default queue sizing is to fill the ring. */
623
	if (!virtblk_queue_depth) {
624
		virtblk_queue_depth = vblk->vqs[0].vq->num_free;
625 626
		/* ... but without indirect descs, we use 2 descs per req */
		if (!virtio_has_feature(vdev, VIRTIO_RING_F_INDIRECT_DESC))
627
			virtblk_queue_depth /= 2;
628
	}
629 630 631 632 633 634 635

	memset(&vblk->tag_set, 0, sizeof(vblk->tag_set));
	vblk->tag_set.ops = &virtio_mq_ops;
	vblk->tag_set.queue_depth = virtblk_queue_depth;
	vblk->tag_set.numa_node = NUMA_NO_NODE;
	vblk->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
	vblk->tag_set.cmd_size =
J
Jens Axboe 已提交
636 637
		sizeof(struct virtblk_req) +
		sizeof(struct scatterlist) * sg_elems;
638
	vblk->tag_set.driver_data = vblk;
639
	vblk->tag_set.nr_hw_queues = vblk->num_vqs;
J
Jens Axboe 已提交
640

641 642 643 644
	err = blk_mq_alloc_tag_set(&vblk->tag_set);
	if (err)
		goto out_put_disk;

645
	q = blk_mq_init_queue(&vblk->tag_set);
646
	if (IS_ERR(q)) {
R
Rusty Russell 已提交
647
		err = -ENOMEM;
648
		goto out_free_tags;
R
Rusty Russell 已提交
649
	}
650
	vblk->disk->queue = q;
R
Rusty Russell 已提交
651

652
	q->queuedata = vblk;
653

654
	virtblk_name_format("vd", index, vblk->disk->disk_name, DISK_NAME_LEN);
655

R
Rusty Russell 已提交
656
	vblk->disk->major = major;
657
	vblk->disk->first_minor = index_to_minor(index);
R
Rusty Russell 已提交
658 659
	vblk->disk->private_data = vblk;
	vblk->disk->fops = &virtblk_fops;
660
	vblk->disk->flags |= GENHD_FL_EXT_DEVT;
661
	vblk->index = index;
662

663
	/* configure queue flush support */
664
	virtblk_update_cache_mode(vdev);
R
Rusty Russell 已提交
665

666 667 668 669
	/* If disk is read-only in the host, the guest should obey */
	if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO))
		set_disk_ro(vblk->disk, 1);

670
	/* Host must always specify the capacity. */
671
	virtio_cread(vdev, struct virtio_blk_config, capacity, &cap);
R
Rusty Russell 已提交
672 673 674 675 676 677 678 679 680

	/* 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);

681
	/* We can handle whatever the host told us to handle. */
682
	blk_queue_max_segments(q, vblk->sg_elems-2);
683

684
	/* No need to bounce any requests */
685
	blk_queue_bounce_limit(q, BLK_BOUNCE_ANY);
686

687
	/* No real sector limit. */
688
	blk_queue_max_hw_sectors(q, -1U);
689

690 691
	/* Host can optionally specify maximum segment size and number of
	 * segments. */
692 693
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_SIZE_MAX,
				   struct virtio_blk_config, size_max, &v);
R
Rusty Russell 已提交
694
	if (!err)
695
		blk_queue_max_segment_size(q, v);
696
	else
697
		blk_queue_max_segment_size(q, -1U);
R
Rusty Russell 已提交
698

699
	/* Host can optionally specify the block size of the device */
700 701 702
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_BLK_SIZE,
				   struct virtio_blk_config, blk_size,
				   &blk_size);
703
	if (!err)
704 705 706 707 708
		blk_queue_logical_block_size(q, blk_size);
	else
		blk_size = queue_logical_block_size(q);

	/* Use topology information if available */
709 710 711
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY,
				   struct virtio_blk_config, physical_block_exp,
				   &physical_block_exp);
712 713 714 715
	if (!err && physical_block_exp)
		blk_queue_physical_block_size(q,
				blk_size * (1 << physical_block_exp));

716 717 718
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY,
				   struct virtio_blk_config, alignment_offset,
				   &alignment_offset);
719 720 721
	if (!err && alignment_offset)
		blk_queue_alignment_offset(q, blk_size * alignment_offset);

722 723 724
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY,
				   struct virtio_blk_config, min_io_size,
				   &min_io_size);
725 726 727
	if (!err && min_io_size)
		blk_queue_io_min(q, blk_size * min_io_size);

728 729 730
	err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY,
				   struct virtio_blk_config, opt_io_size,
				   &opt_io_size);
731 732 733
	if (!err && opt_io_size)
		blk_queue_io_opt(q, blk_size * opt_io_size);

M
Michael S. Tsirkin 已提交
734 735
	virtio_device_ready(vdev);

736
	device_add_disk(&vdev->dev, vblk->disk);
737 738 739 740
	err = device_create_file(disk_to_dev(vblk->disk), &dev_attr_serial);
	if (err)
		goto out_del_disk;

741 742 743 744 745 746 747 748
	if (virtio_has_feature(vdev, VIRTIO_BLK_F_CONFIG_WCE))
		err = device_create_file(disk_to_dev(vblk->disk),
					 &dev_attr_cache_type_rw);
	else
		err = device_create_file(disk_to_dev(vblk->disk),
					 &dev_attr_cache_type_ro);
	if (err)
		goto out_del_disk;
R
Rusty Russell 已提交
749 750
	return 0;

751 752 753
out_del_disk:
	del_gendisk(vblk->disk);
	blk_cleanup_queue(vblk->disk->queue);
754 755
out_free_tags:
	blk_mq_free_tag_set(&vblk->tag_set);
R
Rusty Russell 已提交
756 757 758
out_put_disk:
	put_disk(vblk->disk);
out_free_vq:
759
	vdev->config->del_vqs(vdev);
R
Rusty Russell 已提交
760 761
out_free_vblk:
	kfree(vblk);
762 763
out_free_index:
	ida_simple_remove(&vd_index_ida, index);
R
Rusty Russell 已提交
764 765 766 767
out:
	return err;
}

768
static void virtblk_remove(struct virtio_device *vdev)
R
Rusty Russell 已提交
769 770
{
	struct virtio_blk *vblk = vdev->priv;
771
	int index = vblk->index;
772
	int refc;
R
Rusty Russell 已提交
773

774 775
	/* Make sure no work handler is accessing the device. */
	flush_work(&vblk->config_work);
776

777
	del_gendisk(vblk->disk);
778
	blk_cleanup_queue(vblk->disk->queue);
779

780 781
	blk_mq_free_tag_set(&vblk->tag_set);

R
Rusty Russell 已提交
782 783 784
	/* Stop all the virtqueues. */
	vdev->config->reset(vdev);

785
	refc = atomic_read(&disk_to_dev(vblk->disk)->kobj.kref.refcount);
R
Rusty Russell 已提交
786
	put_disk(vblk->disk);
787
	vdev->config->del_vqs(vdev);
788
	kfree(vblk->vqs);
R
Rusty Russell 已提交
789
	kfree(vblk);
790 791 792 793

	/* Only free device id if we don't have any users */
	if (refc == 1)
		ida_simple_remove(&vd_index_ida, index);
R
Rusty Russell 已提交
794 795
}

796
#ifdef CONFIG_PM_SLEEP
797 798 799 800 801 802 803
static int virtblk_freeze(struct virtio_device *vdev)
{
	struct virtio_blk *vblk = vdev->priv;

	/* Ensure we don't receive any more interrupts */
	vdev->config->reset(vdev);

804
	/* Make sure no work handler is accessing the device. */
805 806
	flush_work(&vblk->config_work);

J
Jens Axboe 已提交
807
	blk_mq_stop_hw_queues(vblk->disk->queue);
808 809 810 811 812 813 814 815 816 817 818

	vdev->config->del_vqs(vdev);
	return 0;
}

static int virtblk_restore(struct virtio_device *vdev)
{
	struct virtio_blk *vblk = vdev->priv;
	int ret;

	ret = init_vq(vdev->priv);
819 820 821 822
	if (ret)
		return ret;

	virtio_device_ready(vdev);
J
Jens Axboe 已提交
823

824 825
	blk_mq_start_stopped_hw_queues(vblk->disk->queue, true);
	return 0;
826 827 828
}
#endif

829
static const struct virtio_device_id id_table[] = {
R
Rusty Russell 已提交
830 831 832 833
	{ VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID },
	{ 0 },
};

M
Michael S. Tsirkin 已提交
834
static unsigned int features_legacy[] = {
835 836
	VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY,
	VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, VIRTIO_BLK_F_SCSI,
837
	VIRTIO_BLK_F_FLUSH, VIRTIO_BLK_F_TOPOLOGY, VIRTIO_BLK_F_CONFIG_WCE,
838
	VIRTIO_BLK_F_MQ,
M
Michael S. Tsirkin 已提交
839 840 841 842 843
}
;
static unsigned int features[] = {
	VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY,
	VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE,
844
	VIRTIO_BLK_F_FLUSH, VIRTIO_BLK_F_TOPOLOGY, VIRTIO_BLK_F_CONFIG_WCE,
M
Michael S. Tsirkin 已提交
845
	VIRTIO_BLK_F_MQ,
846 847
};

848
static struct virtio_driver virtio_blk = {
M
Michael S. Tsirkin 已提交
849 850 851 852 853 854 855 856 857 858
	.feature_table			= features,
	.feature_table_size		= ARRAY_SIZE(features),
	.feature_table_legacy		= features_legacy,
	.feature_table_size_legacy	= ARRAY_SIZE(features_legacy),
	.driver.name			= KBUILD_MODNAME,
	.driver.owner			= THIS_MODULE,
	.id_table			= id_table,
	.probe				= virtblk_probe,
	.remove				= virtblk_remove,
	.config_changed			= virtblk_config_changed,
859
#ifdef CONFIG_PM_SLEEP
M
Michael S. Tsirkin 已提交
860 861
	.freeze				= virtblk_freeze,
	.restore			= virtblk_restore,
862
#endif
R
Rusty Russell 已提交
863 864 865 866
};

static int __init init(void)
{
867 868 869 870 871 872
	int error;

	virtblk_wq = alloc_workqueue("virtio-blk", 0, 0);
	if (!virtblk_wq)
		return -ENOMEM;

873
	major = register_blkdev(0, "virtblk");
874 875 876 877 878 879 880 881 882 883 884 885 886 887 888
	if (major < 0) {
		error = major;
		goto out_destroy_workqueue;
	}

	error = register_virtio_driver(&virtio_blk);
	if (error)
		goto out_unregister_blkdev;
	return 0;

out_unregister_blkdev:
	unregister_blkdev(major, "virtblk");
out_destroy_workqueue:
	destroy_workqueue(virtblk_wq);
	return error;
R
Rusty Russell 已提交
889 890 891 892 893
}

static void __exit fini(void)
{
	unregister_virtio_driver(&virtio_blk);
894
	unregister_blkdev(major, "virtblk");
895
	destroy_workqueue(virtblk_wq);
R
Rusty Russell 已提交
896 897 898 899 900 901 902
}
module_init(init);
module_exit(fini);

MODULE_DEVICE_TABLE(virtio, id_table);
MODULE_DESCRIPTION("Virtio block driver");
MODULE_LICENSE("GPL");