]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/block/virtio_blk.c
virtio: console: fix error handling in init() function
[mirror_ubuntu-bionic-kernel.git] / drivers / block / virtio_blk.c
CommitLineData
e467cde2
RR
1//#define DEBUG
2#include <linux/spinlock.h>
5a0e3ad6 3#include <linux/slab.h>
e467cde2
RR
4#include <linux/blkdev.h>
5#include <linux/hdreg.h>
0c8d44f2 6#include <linux/module.h>
4678d6f9 7#include <linux/mutex.h>
e467cde2
RR
8#include <linux/virtio.h>
9#include <linux/virtio_blk.h>
3d1266c7 10#include <linux/scatterlist.h>
7a7c924c 11#include <linux/string_helpers.h>
6917f83f 12#include <scsi/scsi_cmnd.h>
5087a50e 13#include <linux/idr.h>
3d1266c7 14
4f3bf19c 15#define PART_BITS 4
e467cde2 16
5087a50e
MT
17static int major;
18static DEFINE_IDA(vd_index_ida);
19
7a7c924c 20struct workqueue_struct *virtblk_wq;
4f3bf19c 21
e467cde2
RR
22struct virtio_blk
23{
e467cde2
RR
24 struct virtio_device *vdev;
25 struct virtqueue *vq;
26
27 /* The disk structure for the kernel. */
28 struct gendisk *disk;
29
e467cde2
RR
30 mempool_t *pool;
31
7a7c924c
CH
32 /* Process context for config space updates */
33 struct work_struct config_work;
34
4678d6f9
MT
35 /* Lock for config space updates */
36 struct mutex config_lock;
37
38 /* enable config space updates */
39 bool config_enable;
40
0864b79a
RR
41 /* What host tells us, plus 2 for header & tailer. */
42 unsigned int sg_elems;
43
5087a50e
MT
44 /* Ida index - used to track minor number allocations. */
45 int index;
46
e467cde2 47 /* Scatterlist: can be too big for stack. */
0864b79a 48 struct scatterlist sg[/*sg_elems*/];
e467cde2
RR
49};
50
51struct virtblk_req
52{
e467cde2
RR
53 struct request *req;
54 struct virtio_blk_outhdr out_hdr;
1cde26f9 55 struct virtio_scsi_inhdr in_hdr;
cb38fa23 56 u8 status;
e467cde2
RR
57};
58
18445c4d 59static void blk_done(struct virtqueue *vq)
e467cde2
RR
60{
61 struct virtio_blk *vblk = vq->vdev->priv;
62 struct virtblk_req *vbr;
63 unsigned int len;
64 unsigned long flags;
65
2c95a329 66 spin_lock_irqsave(vblk->disk->queue->queue_lock, flags);
09ec6b69 67 while ((vbr = virtqueue_get_buf(vblk->vq, &len)) != NULL) {
8316982a 68 int error;
1cde26f9 69
cb38fa23 70 switch (vbr->status) {
e467cde2 71 case VIRTIO_BLK_S_OK:
8316982a 72 error = 0;
e467cde2
RR
73 break;
74 case VIRTIO_BLK_S_UNSUPP:
8316982a 75 error = -ENOTTY;
e467cde2
RR
76 break;
77 default:
8316982a 78 error = -EIO;
e467cde2
RR
79 break;
80 }
81
33659ebb
CH
82 switch (vbr->req->cmd_type) {
83 case REQ_TYPE_BLOCK_PC:
1cde26f9
HR
84 vbr->req->resid_len = vbr->in_hdr.residual;
85 vbr->req->sense_len = vbr->in_hdr.sense_len;
86 vbr->req->errors = vbr->in_hdr.errors;
33659ebb
CH
87 break;
88 case REQ_TYPE_SPECIAL:
4cb2ea28 89 vbr->req->errors = (error != 0);
33659ebb 90 break;
15fa6e81
JA
91 default:
92 break;
33659ebb 93 }
1cde26f9 94
40cbbb78 95 __blk_end_request_all(vbr->req, error);
e467cde2
RR
96 mempool_free(vbr, vblk->pool);
97 }
98 /* In case queue is stopped waiting for more buffers. */
99 blk_start_queue(vblk->disk->queue);
2c95a329 100 spin_unlock_irqrestore(vblk->disk->queue->queue_lock, flags);
e467cde2
RR
101}
102
103static bool do_req(struct request_queue *q, struct virtio_blk *vblk,
104 struct request *req)
105{
1cde26f9 106 unsigned long num, out = 0, in = 0;
e467cde2
RR
107 struct virtblk_req *vbr;
108
109 vbr = mempool_alloc(vblk->pool, GFP_ATOMIC);
110 if (!vbr)
111 /* When another request finishes we'll try again. */
112 return false;
113
114 vbr->req = req;
dd40e456
FT
115
116 if (req->cmd_flags & REQ_FLUSH) {
117 vbr->out_hdr.type = VIRTIO_BLK_T_FLUSH;
4cb2ea28 118 vbr->out_hdr.sector = 0;
119 vbr->out_hdr.ioprio = req_get_ioprio(vbr->req);
dd40e456
FT
120 } else {
121 switch (req->cmd_type) {
122 case REQ_TYPE_FS:
123 vbr->out_hdr.type = 0;
124 vbr->out_hdr.sector = blk_rq_pos(vbr->req);
125 vbr->out_hdr.ioprio = req_get_ioprio(vbr->req);
126 break;
127 case REQ_TYPE_BLOCK_PC:
128 vbr->out_hdr.type = VIRTIO_BLK_T_SCSI_CMD;
f1b0ef06
CH
129 vbr->out_hdr.sector = 0;
130 vbr->out_hdr.ioprio = req_get_ioprio(vbr->req);
131 break;
dd40e456
FT
132 case REQ_TYPE_SPECIAL:
133 vbr->out_hdr.type = VIRTIO_BLK_T_GET_ID;
134 vbr->out_hdr.sector = 0;
135 vbr->out_hdr.ioprio = req_get_ioprio(vbr->req);
136 break;
137 default:
138 /* We don't put anything else in the queue. */
139 BUG();
f1b0ef06 140 }
e467cde2
RR
141 }
142
1cde26f9 143 sg_set_buf(&vblk->sg[out++], &vbr->out_hdr, sizeof(vbr->out_hdr));
e467cde2 144
1cde26f9
HR
145 /*
146 * If this is a packet command we need a couple of additional headers.
147 * Behind the normal outhdr we put a segment with the scsi command
148 * block, and before the normal inhdr we put the sense data and the
149 * inhdr with additional status information before the normal inhdr.
150 */
33659ebb 151 if (vbr->req->cmd_type == REQ_TYPE_BLOCK_PC)
1cde26f9
HR
152 sg_set_buf(&vblk->sg[out++], vbr->req->cmd, vbr->req->cmd_len);
153
154 num = blk_rq_map_sg(q, vbr->req, vblk->sg + out);
155
33659ebb 156 if (vbr->req->cmd_type == REQ_TYPE_BLOCK_PC) {
6917f83f 157 sg_set_buf(&vblk->sg[num + out + in++], vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
1cde26f9
HR
158 sg_set_buf(&vblk->sg[num + out + in++], &vbr->in_hdr,
159 sizeof(vbr->in_hdr));
160 }
161
162 sg_set_buf(&vblk->sg[num + out + in++], &vbr->status,
163 sizeof(vbr->status));
164
165 if (num) {
166 if (rq_data_dir(vbr->req) == WRITE) {
167 vbr->out_hdr.type |= VIRTIO_BLK_T_OUT;
168 out += num;
169 } else {
170 vbr->out_hdr.type |= VIRTIO_BLK_T_IN;
171 in += num;
172 }
e467cde2
RR
173 }
174
f96fde41 175 if (virtqueue_add_buf(vblk->vq, vblk->sg, out, in, vbr, GFP_ATOMIC)<0) {
e467cde2
RR
176 mempool_free(vbr, vblk->pool);
177 return false;
178 }
179
e467cde2
RR
180 return true;
181}
182
183static void do_virtblk_request(struct request_queue *q)
184{
6c3b46f7 185 struct virtio_blk *vblk = q->queuedata;
e467cde2
RR
186 struct request *req;
187 unsigned int issued = 0;
188
9934c8c0 189 while ((req = blk_peek_request(q)) != NULL) {
0864b79a 190 BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems);
e467cde2
RR
191
192 /* If this request fails, stop queue and wait for something to
193 finish to restart it. */
194 if (!do_req(q, vblk, req)) {
195 blk_stop_queue(q);
196 break;
197 }
9934c8c0 198 blk_start_request(req);
e467cde2
RR
199 issued++;
200 }
201
202 if (issued)
09ec6b69 203 virtqueue_kick(vblk->vq);
e467cde2
RR
204}
205
4cb2ea28 206/* return id (s/n) string for *disk to *id_str
207 */
208static int virtblk_get_id(struct gendisk *disk, char *id_str)
209{
210 struct virtio_blk *vblk = disk->private_data;
211 struct request *req;
212 struct bio *bio;
e4c4776d 213 int err;
4cb2ea28 214
215 bio = bio_map_kern(vblk->disk->queue, id_str, VIRTIO_BLK_ID_BYTES,
216 GFP_KERNEL);
217 if (IS_ERR(bio))
218 return PTR_ERR(bio);
219
220 req = blk_make_request(vblk->disk->queue, bio, GFP_KERNEL);
221 if (IS_ERR(req)) {
222 bio_put(bio);
223 return PTR_ERR(req);
224 }
225
226 req->cmd_type = REQ_TYPE_SPECIAL;
e4c4776d
MS
227 err = blk_execute_rq(vblk->disk->queue, vblk->disk, req, false);
228 blk_put_request(req);
229
230 return err;
4cb2ea28 231}
232
fe5a50a1
CH
233static int virtblk_ioctl(struct block_device *bdev, fmode_t mode,
234 unsigned int cmd, unsigned long data)
e467cde2 235{
1cde26f9
HR
236 struct gendisk *disk = bdev->bd_disk;
237 struct virtio_blk *vblk = disk->private_data;
238
239 /*
240 * Only allow the generic SCSI ioctls if the host can support it.
241 */
242 if (!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_SCSI))
d9ecdea7 243 return -ENOTTY;
1cde26f9 244
577ebb37
PB
245 return scsi_cmd_blk_ioctl(bdev, mode, cmd,
246 (void __user *)data);
e467cde2
RR
247}
248
135da0b0
CB
249/* We provide getgeo only to please some old bootloader/partitioning tools */
250static int virtblk_getgeo(struct block_device *bd, struct hd_geometry *geo)
251{
48e4043d
RH
252 struct virtio_blk *vblk = bd->bd_disk->private_data;
253 struct virtio_blk_geometry vgeo;
254 int err;
255
256 /* see if the host passed in geometry config */
257 err = virtio_config_val(vblk->vdev, VIRTIO_BLK_F_GEOMETRY,
258 offsetof(struct virtio_blk_config, geometry),
259 &vgeo);
260
261 if (!err) {
262 geo->heads = vgeo.heads;
263 geo->sectors = vgeo.sectors;
264 geo->cylinders = vgeo.cylinders;
265 } else {
266 /* some standard values, similar to sd */
267 geo->heads = 1 << 6;
268 geo->sectors = 1 << 5;
269 geo->cylinders = get_capacity(bd->bd_disk) >> 11;
270 }
135da0b0
CB
271 return 0;
272}
273
83d5cde4 274static const struct block_device_operations virtblk_fops = {
8a6cfeb6 275 .ioctl = virtblk_ioctl,
135da0b0
CB
276 .owner = THIS_MODULE,
277 .getgeo = virtblk_getgeo,
e467cde2
RR
278};
279
d50ed907
CB
280static int index_to_minor(int index)
281{
282 return index << PART_BITS;
283}
284
5087a50e
MT
285static int minor_to_index(int minor)
286{
287 return minor >> PART_BITS;
288}
289
a5eb9e4f
RH
290static ssize_t virtblk_serial_show(struct device *dev,
291 struct device_attribute *attr, char *buf)
292{
293 struct gendisk *disk = dev_to_disk(dev);
294 int err;
295
296 /* sysfs gives us a PAGE_SIZE buffer */
297 BUILD_BUG_ON(PAGE_SIZE < VIRTIO_BLK_ID_BYTES);
298
299 buf[VIRTIO_BLK_ID_BYTES] = '\0';
300 err = virtblk_get_id(disk, buf);
301 if (!err)
302 return strlen(buf);
303
304 if (err == -EIO) /* Unsupported? Make it empty. */
305 return 0;
306
307 return err;
308}
309DEVICE_ATTR(serial, S_IRUGO, virtblk_serial_show, NULL);
310
7a7c924c
CH
311static void virtblk_config_changed_work(struct work_struct *work)
312{
313 struct virtio_blk *vblk =
314 container_of(work, struct virtio_blk, config_work);
315 struct virtio_device *vdev = vblk->vdev;
316 struct request_queue *q = vblk->disk->queue;
317 char cap_str_2[10], cap_str_10[10];
318 u64 capacity, size;
319
4678d6f9
MT
320 mutex_lock(&vblk->config_lock);
321 if (!vblk->config_enable)
322 goto done;
323
7a7c924c
CH
324 /* Host must always specify the capacity. */
325 vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity),
326 &capacity, sizeof(capacity));
327
328 /* If capacity is too big, truncate with warning. */
329 if ((sector_t)capacity != capacity) {
330 dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n",
331 (unsigned long long)capacity);
332 capacity = (sector_t)-1;
333 }
334
335 size = capacity * queue_logical_block_size(q);
336 string_get_size(size, STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
337 string_get_size(size, STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
338
339 dev_notice(&vdev->dev,
340 "new size: %llu %d-byte logical blocks (%s/%s)\n",
341 (unsigned long long)capacity,
342 queue_logical_block_size(q),
343 cap_str_10, cap_str_2);
344
345 set_capacity(vblk->disk, capacity);
e9986f30 346 revalidate_disk(vblk->disk);
4678d6f9
MT
347done:
348 mutex_unlock(&vblk->config_lock);
7a7c924c
CH
349}
350
351static void virtblk_config_changed(struct virtio_device *vdev)
352{
353 struct virtio_blk *vblk = vdev->priv;
354
355 queue_work(virtblk_wq, &vblk->config_work);
356}
357
6abd6e5a
AS
358static int init_vq(struct virtio_blk *vblk)
359{
360 int err = 0;
361
362 /* We expect one virtqueue, for output. */
363 vblk->vq = virtio_find_single_vq(vblk->vdev, blk_done, "requests");
364 if (IS_ERR(vblk->vq))
365 err = PTR_ERR(vblk->vq);
366
367 return err;
368}
369
c0aa3e09
RM
370/*
371 * Legacy naming scheme used for virtio devices. We are stuck with it for
372 * virtio blk but don't ever use it for any new driver.
373 */
374static int virtblk_name_format(char *prefix, int index, char *buf, int buflen)
375{
376 const int base = 'z' - 'a' + 1;
377 char *begin = buf + strlen(prefix);
378 char *end = buf + buflen;
379 char *p;
380 int unit;
381
382 p = end - 1;
383 *p = '\0';
384 unit = base;
385 do {
386 if (p == begin)
387 return -EINVAL;
388 *--p = 'a' + (index % unit);
389 index = (index / unit) - 1;
390 } while (index >= 0);
391
392 memmove(begin, p, end - p);
393 memcpy(buf, prefix, strlen(prefix));
394
395 return 0;
396}
397
cd5d5038
PB
398static int virtblk_get_cache_mode(struct virtio_device *vdev)
399{
400 u8 writeback;
401 int err;
402
403 err = virtio_config_val(vdev, VIRTIO_BLK_F_CONFIG_WCE,
404 offsetof(struct virtio_blk_config, wce),
405 &writeback);
406 if (err)
407 writeback = virtio_has_feature(vdev, VIRTIO_BLK_F_WCE);
408
409 return writeback;
410}
411
412static void virtblk_update_cache_mode(struct virtio_device *vdev)
413{
414 u8 writeback = virtblk_get_cache_mode(vdev);
415 struct virtio_blk *vblk = vdev->priv;
416
417 if (writeback)
418 blk_queue_flush(vblk->disk->queue, REQ_FLUSH);
419 else
420 blk_queue_flush(vblk->disk->queue, 0);
421
422 revalidate_disk(vblk->disk);
423}
424
425static const char *const virtblk_cache_types[] = {
426 "write through", "write back"
427};
428
429static ssize_t
430virtblk_cache_type_store(struct device *dev, struct device_attribute *attr,
431 const char *buf, size_t count)
432{
433 struct gendisk *disk = dev_to_disk(dev);
434 struct virtio_blk *vblk = disk->private_data;
435 struct virtio_device *vdev = vblk->vdev;
436 int i;
437 u8 writeback;
438
439 BUG_ON(!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_CONFIG_WCE));
440 for (i = ARRAY_SIZE(virtblk_cache_types); --i >= 0; )
441 if (sysfs_streq(buf, virtblk_cache_types[i]))
442 break;
443
444 if (i < 0)
445 return -EINVAL;
446
447 writeback = i;
448 vdev->config->set(vdev,
449 offsetof(struct virtio_blk_config, wce),
450 &writeback, sizeof(writeback));
451
452 virtblk_update_cache_mode(vdev);
453 return count;
454}
455
456static ssize_t
457virtblk_cache_type_show(struct device *dev, struct device_attribute *attr,
458 char *buf)
459{
460 struct gendisk *disk = dev_to_disk(dev);
461 struct virtio_blk *vblk = disk->private_data;
462 u8 writeback = virtblk_get_cache_mode(vblk->vdev);
463
464 BUG_ON(writeback >= ARRAY_SIZE(virtblk_cache_types));
465 return snprintf(buf, 40, "%s\n", virtblk_cache_types[writeback]);
466}
467
468static const struct device_attribute dev_attr_cache_type_ro =
469 __ATTR(cache_type, S_IRUGO,
470 virtblk_cache_type_show, NULL);
471static const struct device_attribute dev_attr_cache_type_rw =
472 __ATTR(cache_type, S_IRUGO|S_IWUSR,
473 virtblk_cache_type_show, virtblk_cache_type_store);
474
98e94444 475static int __devinit virtblk_probe(struct virtio_device *vdev)
e467cde2
RR
476{
477 struct virtio_blk *vblk;
69740c8b 478 struct request_queue *q;
5087a50e 479 int err, index;
e467cde2 480 u64 cap;
69740c8b
CH
481 u32 v, blk_size, sg_elems, opt_io_size;
482 u16 min_io_size;
483 u8 physical_block_exp, alignment_offset;
e467cde2 484
5087a50e
MT
485 err = ida_simple_get(&vd_index_ida, 0, minor_to_index(1 << MINORBITS),
486 GFP_KERNEL);
487 if (err < 0)
488 goto out;
489 index = err;
4f3bf19c 490
0864b79a
RR
491 /* We need to know how many segments before we allocate. */
492 err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX,
493 offsetof(struct virtio_blk_config, seg_max),
494 &sg_elems);
a5b365a6
CH
495
496 /* We need at least one SG element, whatever they say. */
497 if (err || !sg_elems)
0864b79a
RR
498 sg_elems = 1;
499
500 /* We need an extra sg elements at head and tail. */
501 sg_elems += 2;
502 vdev->priv = vblk = kmalloc(sizeof(*vblk) +
503 sizeof(vblk->sg[0]) * sg_elems, GFP_KERNEL);
e467cde2
RR
504 if (!vblk) {
505 err = -ENOMEM;
5087a50e 506 goto out_free_index;
e467cde2
RR
507 }
508
e467cde2 509 vblk->vdev = vdev;
0864b79a
RR
510 vblk->sg_elems = sg_elems;
511 sg_init_table(vblk->sg, vblk->sg_elems);
4678d6f9 512 mutex_init(&vblk->config_lock);
7a7c924c 513 INIT_WORK(&vblk->config_work, virtblk_config_changed_work);
4678d6f9 514 vblk->config_enable = true;
e467cde2 515
6abd6e5a
AS
516 err = init_vq(vblk);
517 if (err)
e467cde2 518 goto out_free_vblk;
e467cde2
RR
519
520 vblk->pool = mempool_create_kmalloc_pool(1,sizeof(struct virtblk_req));
521 if (!vblk->pool) {
522 err = -ENOMEM;
523 goto out_free_vq;
524 }
525
e467cde2 526 /* FIXME: How many partitions? How long is a piece of string? */
4f3bf19c 527 vblk->disk = alloc_disk(1 << PART_BITS);
e467cde2
RR
528 if (!vblk->disk) {
529 err = -ENOMEM;
4f3bf19c 530 goto out_mempool;
e467cde2
RR
531 }
532
2c95a329 533 q = vblk->disk->queue = blk_init_queue(do_virtblk_request, NULL);
69740c8b 534 if (!q) {
e467cde2
RR
535 err = -ENOMEM;
536 goto out_put_disk;
537 }
538
69740c8b 539 q->queuedata = vblk;
7d116b62 540
c0aa3e09 541 virtblk_name_format("vd", index, vblk->disk->disk_name, DISK_NAME_LEN);
d50ed907 542
e467cde2 543 vblk->disk->major = major;
d50ed907 544 vblk->disk->first_minor = index_to_minor(index);
e467cde2
RR
545 vblk->disk->private_data = vblk;
546 vblk->disk->fops = &virtblk_fops;
c4839346 547 vblk->disk->driverfs_dev = &vdev->dev;
5087a50e 548 vblk->index = index;
4f3bf19c 549
02c42b7a 550 /* configure queue flush support */
cd5d5038 551 virtblk_update_cache_mode(vdev);
e467cde2 552
3ef53609
CB
553 /* If disk is read-only in the host, the guest should obey */
554 if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO))
555 set_disk_ro(vblk->disk, 1);
556
a586d4f6 557 /* Host must always specify the capacity. */
72e61eb4
RR
558 vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity),
559 &cap, sizeof(cap));
e467cde2
RR
560
561 /* If capacity is too big, truncate with warning. */
562 if ((sector_t)cap != cap) {
563 dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n",
564 (unsigned long long)cap);
565 cap = (sector_t)-1;
566 }
567 set_capacity(vblk->disk, cap);
568
0864b79a 569 /* We can handle whatever the host told us to handle. */
ee714f2d 570 blk_queue_max_segments(q, vblk->sg_elems-2);
0864b79a 571
4eff3cae 572 /* No need to bounce any requests */
69740c8b 573 blk_queue_bounce_limit(q, BLK_BOUNCE_ANY);
4eff3cae 574
4b7f7e20 575 /* No real sector limit. */
ee714f2d 576 blk_queue_max_hw_sectors(q, -1U);
4b7f7e20 577
a586d4f6
RR
578 /* Host can optionally specify maximum segment size and number of
579 * segments. */
580 err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX,
581 offsetof(struct virtio_blk_config, size_max),
582 &v);
e467cde2 583 if (!err)
69740c8b 584 blk_queue_max_segment_size(q, v);
4b7f7e20 585 else
69740c8b 586 blk_queue_max_segment_size(q, -1U);
e467cde2 587
066f4d82
CB
588 /* Host can optionally specify the block size of the device */
589 err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE,
590 offsetof(struct virtio_blk_config, blk_size),
591 &blk_size);
592 if (!err)
69740c8b
CH
593 blk_queue_logical_block_size(q, blk_size);
594 else
595 blk_size = queue_logical_block_size(q);
596
597 /* Use topology information if available */
598 err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY,
599 offsetof(struct virtio_blk_config, physical_block_exp),
600 &physical_block_exp);
601 if (!err && physical_block_exp)
602 blk_queue_physical_block_size(q,
603 blk_size * (1 << physical_block_exp));
604
605 err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY,
606 offsetof(struct virtio_blk_config, alignment_offset),
607 &alignment_offset);
608 if (!err && alignment_offset)
609 blk_queue_alignment_offset(q, blk_size * alignment_offset);
610
611 err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY,
612 offsetof(struct virtio_blk_config, min_io_size),
613 &min_io_size);
614 if (!err && min_io_size)
615 blk_queue_io_min(q, blk_size * min_io_size);
616
617 err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY,
618 offsetof(struct virtio_blk_config, opt_io_size),
619 &opt_io_size);
620 if (!err && opt_io_size)
621 blk_queue_io_opt(q, blk_size * opt_io_size);
622
066f4d82 623
e467cde2 624 add_disk(vblk->disk);
a5eb9e4f
RH
625 err = device_create_file(disk_to_dev(vblk->disk), &dev_attr_serial);
626 if (err)
627 goto out_del_disk;
628
cd5d5038
PB
629 if (virtio_has_feature(vdev, VIRTIO_BLK_F_CONFIG_WCE))
630 err = device_create_file(disk_to_dev(vblk->disk),
631 &dev_attr_cache_type_rw);
632 else
633 err = device_create_file(disk_to_dev(vblk->disk),
634 &dev_attr_cache_type_ro);
635 if (err)
636 goto out_del_disk;
e467cde2
RR
637 return 0;
638
a5eb9e4f
RH
639out_del_disk:
640 del_gendisk(vblk->disk);
641 blk_cleanup_queue(vblk->disk->queue);
e467cde2
RR
642out_put_disk:
643 put_disk(vblk->disk);
e467cde2
RR
644out_mempool:
645 mempool_destroy(vblk->pool);
646out_free_vq:
d2a7ddda 647 vdev->config->del_vqs(vdev);
e467cde2
RR
648out_free_vblk:
649 kfree(vblk);
5087a50e
MT
650out_free_index:
651 ida_simple_remove(&vd_index_ida, index);
e467cde2
RR
652out:
653 return err;
654}
655
98e94444 656static void __devexit virtblk_remove(struct virtio_device *vdev)
e467cde2
RR
657{
658 struct virtio_blk *vblk = vdev->priv;
5087a50e 659 int index = vblk->index;
e467cde2 660
4678d6f9
MT
661 /* Prevent config work handler from accessing the device. */
662 mutex_lock(&vblk->config_lock);
663 vblk->config_enable = false;
664 mutex_unlock(&vblk->config_lock);
7a7c924c 665
02e2b124 666 del_gendisk(vblk->disk);
483001c7 667 blk_cleanup_queue(vblk->disk->queue);
02e2b124 668
6e5aa7ef
RR
669 /* Stop all the virtqueues. */
670 vdev->config->reset(vdev);
671
4678d6f9
MT
672 flush_work(&vblk->config_work);
673
e467cde2 674 put_disk(vblk->disk);
e467cde2 675 mempool_destroy(vblk->pool);
d2a7ddda 676 vdev->config->del_vqs(vdev);
e467cde2 677 kfree(vblk);
5087a50e 678 ida_simple_remove(&vd_index_ida, index);
e467cde2
RR
679}
680
f8fb5bc2
AS
681#ifdef CONFIG_PM
682static int virtblk_freeze(struct virtio_device *vdev)
683{
684 struct virtio_blk *vblk = vdev->priv;
685
686 /* Ensure we don't receive any more interrupts */
687 vdev->config->reset(vdev);
688
689 /* Prevent config work handler from accessing the device. */
690 mutex_lock(&vblk->config_lock);
691 vblk->config_enable = false;
692 mutex_unlock(&vblk->config_lock);
693
694 flush_work(&vblk->config_work);
695
696 spin_lock_irq(vblk->disk->queue->queue_lock);
697 blk_stop_queue(vblk->disk->queue);
698 spin_unlock_irq(vblk->disk->queue->queue_lock);
699 blk_sync_queue(vblk->disk->queue);
700
701 vdev->config->del_vqs(vdev);
702 return 0;
703}
704
705static int virtblk_restore(struct virtio_device *vdev)
706{
707 struct virtio_blk *vblk = vdev->priv;
708 int ret;
709
710 vblk->config_enable = true;
711 ret = init_vq(vdev->priv);
712 if (!ret) {
713 spin_lock_irq(vblk->disk->queue->queue_lock);
714 blk_start_queue(vblk->disk->queue);
715 spin_unlock_irq(vblk->disk->queue->queue_lock);
716 }
717 return ret;
718}
719#endif
720
47483e25 721static const struct virtio_device_id id_table[] = {
e467cde2
RR
722 { VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID },
723 { 0 },
724};
725
c45a6816 726static unsigned int features[] = {
02c42b7a
TH
727 VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY,
728 VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, VIRTIO_BLK_F_SCSI,
cd5d5038 729 VIRTIO_BLK_F_WCE, VIRTIO_BLK_F_TOPOLOGY, VIRTIO_BLK_F_CONFIG_WCE
c45a6816
RR
730};
731
4fbfff76
RM
732/*
733 * virtio_blk causes spurious section mismatch warning by
734 * simultaneously referring to a __devinit and a __devexit function.
735 * Use __refdata to avoid this warning.
736 */
737static struct virtio_driver __refdata virtio_blk = {
7a7c924c
CH
738 .feature_table = features,
739 .feature_table_size = ARRAY_SIZE(features),
740 .driver.name = KBUILD_MODNAME,
741 .driver.owner = THIS_MODULE,
742 .id_table = id_table,
743 .probe = virtblk_probe,
744 .remove = __devexit_p(virtblk_remove),
745 .config_changed = virtblk_config_changed,
f8fb5bc2
AS
746#ifdef CONFIG_PM
747 .freeze = virtblk_freeze,
748 .restore = virtblk_restore,
749#endif
e467cde2
RR
750};
751
752static int __init init(void)
753{
7a7c924c
CH
754 int error;
755
756 virtblk_wq = alloc_workqueue("virtio-blk", 0, 0);
757 if (!virtblk_wq)
758 return -ENOMEM;
759
4f3bf19c 760 major = register_blkdev(0, "virtblk");
7a7c924c
CH
761 if (major < 0) {
762 error = major;
763 goto out_destroy_workqueue;
764 }
765
766 error = register_virtio_driver(&virtio_blk);
767 if (error)
768 goto out_unregister_blkdev;
769 return 0;
770
771out_unregister_blkdev:
772 unregister_blkdev(major, "virtblk");
773out_destroy_workqueue:
774 destroy_workqueue(virtblk_wq);
775 return error;
e467cde2
RR
776}
777
778static void __exit fini(void)
779{
4f3bf19c 780 unregister_blkdev(major, "virtblk");
e467cde2 781 unregister_virtio_driver(&virtio_blk);
7a7c924c 782 destroy_workqueue(virtblk_wq);
e467cde2
RR
783}
784module_init(init);
785module_exit(fini);
786
787MODULE_DEVICE_TABLE(virtio, id_table);
788MODULE_DESCRIPTION("Virtio block driver");
789MODULE_LICENSE("GPL");