Commit | Line | Data |
---|---|---|
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 |
17 | static int major; |
18 | static DEFINE_IDA(vd_index_ida); | |
19 | ||
7a7c924c | 20 | struct workqueue_struct *virtblk_wq; |
4f3bf19c | 21 | |
e467cde2 RR |
22 | struct virtio_blk |
23 | { | |
24 | spinlock_t lock; | |
25 | ||
26 | struct virtio_device *vdev; | |
27 | struct virtqueue *vq; | |
28 | ||
29 | /* The disk structure for the kernel. */ | |
30 | struct gendisk *disk; | |
31 | ||
32 | /* Request tracking. */ | |
33 | struct list_head reqs; | |
34 | ||
35 | mempool_t *pool; | |
36 | ||
7a7c924c CH |
37 | /* Process context for config space updates */ |
38 | struct work_struct config_work; | |
39 | ||
4678d6f9 MT |
40 | /* Lock for config space updates */ |
41 | struct mutex config_lock; | |
42 | ||
43 | /* enable config space updates */ | |
44 | bool config_enable; | |
45 | ||
0864b79a RR |
46 | /* What host tells us, plus 2 for header & tailer. */ |
47 | unsigned int sg_elems; | |
48 | ||
5087a50e MT |
49 | /* Ida index - used to track minor number allocations. */ |
50 | int index; | |
51 | ||
e467cde2 | 52 | /* Scatterlist: can be too big for stack. */ |
0864b79a | 53 | struct scatterlist sg[/*sg_elems*/]; |
e467cde2 RR |
54 | }; |
55 | ||
56 | struct virtblk_req | |
57 | { | |
58 | struct list_head list; | |
59 | struct request *req; | |
60 | struct virtio_blk_outhdr out_hdr; | |
1cde26f9 | 61 | struct virtio_scsi_inhdr in_hdr; |
cb38fa23 | 62 | u8 status; |
e467cde2 RR |
63 | }; |
64 | ||
18445c4d | 65 | static void blk_done(struct virtqueue *vq) |
e467cde2 RR |
66 | { |
67 | struct virtio_blk *vblk = vq->vdev->priv; | |
68 | struct virtblk_req *vbr; | |
69 | unsigned int len; | |
70 | unsigned long flags; | |
71 | ||
72 | spin_lock_irqsave(&vblk->lock, flags); | |
09ec6b69 | 73 | while ((vbr = virtqueue_get_buf(vblk->vq, &len)) != NULL) { |
8316982a | 74 | int error; |
1cde26f9 | 75 | |
cb38fa23 | 76 | switch (vbr->status) { |
e467cde2 | 77 | case VIRTIO_BLK_S_OK: |
8316982a | 78 | error = 0; |
e467cde2 RR |
79 | break; |
80 | case VIRTIO_BLK_S_UNSUPP: | |
8316982a | 81 | error = -ENOTTY; |
e467cde2 RR |
82 | break; |
83 | default: | |
8316982a | 84 | error = -EIO; |
e467cde2 RR |
85 | break; |
86 | } | |
87 | ||
33659ebb CH |
88 | switch (vbr->req->cmd_type) { |
89 | case REQ_TYPE_BLOCK_PC: | |
1cde26f9 HR |
90 | vbr->req->resid_len = vbr->in_hdr.residual; |
91 | vbr->req->sense_len = vbr->in_hdr.sense_len; | |
92 | vbr->req->errors = vbr->in_hdr.errors; | |
33659ebb CH |
93 | break; |
94 | case REQ_TYPE_SPECIAL: | |
4cb2ea28 | 95 | vbr->req->errors = (error != 0); |
33659ebb | 96 | break; |
15fa6e81 JA |
97 | default: |
98 | break; | |
33659ebb | 99 | } |
1cde26f9 | 100 | |
40cbbb78 | 101 | __blk_end_request_all(vbr->req, error); |
e467cde2 RR |
102 | list_del(&vbr->list); |
103 | mempool_free(vbr, vblk->pool); | |
104 | } | |
105 | /* In case queue is stopped waiting for more buffers. */ | |
106 | blk_start_queue(vblk->disk->queue); | |
107 | spin_unlock_irqrestore(&vblk->lock, flags); | |
e467cde2 RR |
108 | } |
109 | ||
110 | static bool do_req(struct request_queue *q, struct virtio_blk *vblk, | |
111 | struct request *req) | |
112 | { | |
1cde26f9 | 113 | unsigned long num, out = 0, in = 0; |
e467cde2 RR |
114 | struct virtblk_req *vbr; |
115 | ||
116 | vbr = mempool_alloc(vblk->pool, GFP_ATOMIC); | |
117 | if (!vbr) | |
118 | /* When another request finishes we'll try again. */ | |
119 | return false; | |
120 | ||
121 | vbr->req = req; | |
dd40e456 FT |
122 | |
123 | if (req->cmd_flags & REQ_FLUSH) { | |
124 | vbr->out_hdr.type = VIRTIO_BLK_T_FLUSH; | |
4cb2ea28 | 125 | vbr->out_hdr.sector = 0; |
126 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
dd40e456 FT |
127 | } else { |
128 | switch (req->cmd_type) { | |
129 | case REQ_TYPE_FS: | |
130 | vbr->out_hdr.type = 0; | |
131 | vbr->out_hdr.sector = blk_rq_pos(vbr->req); | |
132 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
133 | break; | |
134 | case REQ_TYPE_BLOCK_PC: | |
135 | vbr->out_hdr.type = VIRTIO_BLK_T_SCSI_CMD; | |
f1b0ef06 CH |
136 | vbr->out_hdr.sector = 0; |
137 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
138 | break; | |
dd40e456 FT |
139 | case REQ_TYPE_SPECIAL: |
140 | vbr->out_hdr.type = VIRTIO_BLK_T_GET_ID; | |
141 | vbr->out_hdr.sector = 0; | |
142 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
143 | break; | |
144 | default: | |
145 | /* We don't put anything else in the queue. */ | |
146 | BUG(); | |
f1b0ef06 | 147 | } |
e467cde2 RR |
148 | } |
149 | ||
1cde26f9 | 150 | sg_set_buf(&vblk->sg[out++], &vbr->out_hdr, sizeof(vbr->out_hdr)); |
e467cde2 | 151 | |
1cde26f9 HR |
152 | /* |
153 | * If this is a packet command we need a couple of additional headers. | |
154 | * Behind the normal outhdr we put a segment with the scsi command | |
155 | * block, and before the normal inhdr we put the sense data and the | |
156 | * inhdr with additional status information before the normal inhdr. | |
157 | */ | |
33659ebb | 158 | if (vbr->req->cmd_type == REQ_TYPE_BLOCK_PC) |
1cde26f9 HR |
159 | sg_set_buf(&vblk->sg[out++], vbr->req->cmd, vbr->req->cmd_len); |
160 | ||
161 | num = blk_rq_map_sg(q, vbr->req, vblk->sg + out); | |
162 | ||
33659ebb | 163 | if (vbr->req->cmd_type == REQ_TYPE_BLOCK_PC) { |
6917f83f | 164 | sg_set_buf(&vblk->sg[num + out + in++], vbr->req->sense, SCSI_SENSE_BUFFERSIZE); |
1cde26f9 HR |
165 | sg_set_buf(&vblk->sg[num + out + in++], &vbr->in_hdr, |
166 | sizeof(vbr->in_hdr)); | |
167 | } | |
168 | ||
169 | sg_set_buf(&vblk->sg[num + out + in++], &vbr->status, | |
170 | sizeof(vbr->status)); | |
171 | ||
172 | if (num) { | |
173 | if (rq_data_dir(vbr->req) == WRITE) { | |
174 | vbr->out_hdr.type |= VIRTIO_BLK_T_OUT; | |
175 | out += num; | |
176 | } else { | |
177 | vbr->out_hdr.type |= VIRTIO_BLK_T_IN; | |
178 | in += num; | |
179 | } | |
e467cde2 RR |
180 | } |
181 | ||
f96fde41 | 182 | if (virtqueue_add_buf(vblk->vq, vblk->sg, out, in, vbr, GFP_ATOMIC)<0) { |
e467cde2 RR |
183 | mempool_free(vbr, vblk->pool); |
184 | return false; | |
185 | } | |
186 | ||
187 | list_add_tail(&vbr->list, &vblk->reqs); | |
188 | return true; | |
189 | } | |
190 | ||
191 | static void do_virtblk_request(struct request_queue *q) | |
192 | { | |
6c3b46f7 | 193 | struct virtio_blk *vblk = q->queuedata; |
e467cde2 RR |
194 | struct request *req; |
195 | unsigned int issued = 0; | |
196 | ||
9934c8c0 | 197 | while ((req = blk_peek_request(q)) != NULL) { |
0864b79a | 198 | BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems); |
e467cde2 RR |
199 | |
200 | /* If this request fails, stop queue and wait for something to | |
201 | finish to restart it. */ | |
202 | if (!do_req(q, vblk, req)) { | |
203 | blk_stop_queue(q); | |
204 | break; | |
205 | } | |
9934c8c0 | 206 | blk_start_request(req); |
e467cde2 RR |
207 | issued++; |
208 | } | |
209 | ||
210 | if (issued) | |
09ec6b69 | 211 | virtqueue_kick(vblk->vq); |
e467cde2 RR |
212 | } |
213 | ||
4cb2ea28 | 214 | /* return id (s/n) string for *disk to *id_str |
215 | */ | |
216 | static int virtblk_get_id(struct gendisk *disk, char *id_str) | |
217 | { | |
218 | struct virtio_blk *vblk = disk->private_data; | |
219 | struct request *req; | |
220 | struct bio *bio; | |
e4c4776d | 221 | int err; |
4cb2ea28 | 222 | |
223 | bio = bio_map_kern(vblk->disk->queue, id_str, VIRTIO_BLK_ID_BYTES, | |
224 | GFP_KERNEL); | |
225 | if (IS_ERR(bio)) | |
226 | return PTR_ERR(bio); | |
227 | ||
228 | req = blk_make_request(vblk->disk->queue, bio, GFP_KERNEL); | |
229 | if (IS_ERR(req)) { | |
230 | bio_put(bio); | |
231 | return PTR_ERR(req); | |
232 | } | |
233 | ||
234 | req->cmd_type = REQ_TYPE_SPECIAL; | |
e4c4776d MS |
235 | err = blk_execute_rq(vblk->disk->queue, vblk->disk, req, false); |
236 | blk_put_request(req); | |
237 | ||
238 | return err; | |
4cb2ea28 | 239 | } |
240 | ||
fe5a50a1 CH |
241 | static int virtblk_ioctl(struct block_device *bdev, fmode_t mode, |
242 | unsigned int cmd, unsigned long data) | |
e467cde2 | 243 | { |
1cde26f9 HR |
244 | struct gendisk *disk = bdev->bd_disk; |
245 | struct virtio_blk *vblk = disk->private_data; | |
246 | ||
247 | /* | |
248 | * Only allow the generic SCSI ioctls if the host can support it. | |
249 | */ | |
250 | if (!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_SCSI)) | |
d9ecdea7 | 251 | return -ENOTTY; |
1cde26f9 | 252 | |
577ebb37 PB |
253 | return scsi_cmd_blk_ioctl(bdev, mode, cmd, |
254 | (void __user *)data); | |
e467cde2 RR |
255 | } |
256 | ||
135da0b0 CB |
257 | /* We provide getgeo only to please some old bootloader/partitioning tools */ |
258 | static int virtblk_getgeo(struct block_device *bd, struct hd_geometry *geo) | |
259 | { | |
48e4043d RH |
260 | struct virtio_blk *vblk = bd->bd_disk->private_data; |
261 | struct virtio_blk_geometry vgeo; | |
262 | int err; | |
263 | ||
264 | /* see if the host passed in geometry config */ | |
265 | err = virtio_config_val(vblk->vdev, VIRTIO_BLK_F_GEOMETRY, | |
266 | offsetof(struct virtio_blk_config, geometry), | |
267 | &vgeo); | |
268 | ||
269 | if (!err) { | |
270 | geo->heads = vgeo.heads; | |
271 | geo->sectors = vgeo.sectors; | |
272 | geo->cylinders = vgeo.cylinders; | |
273 | } else { | |
274 | /* some standard values, similar to sd */ | |
275 | geo->heads = 1 << 6; | |
276 | geo->sectors = 1 << 5; | |
277 | geo->cylinders = get_capacity(bd->bd_disk) >> 11; | |
278 | } | |
135da0b0 CB |
279 | return 0; |
280 | } | |
281 | ||
83d5cde4 | 282 | static const struct block_device_operations virtblk_fops = { |
8a6cfeb6 | 283 | .ioctl = virtblk_ioctl, |
135da0b0 CB |
284 | .owner = THIS_MODULE, |
285 | .getgeo = virtblk_getgeo, | |
e467cde2 RR |
286 | }; |
287 | ||
d50ed907 CB |
288 | static int index_to_minor(int index) |
289 | { | |
290 | return index << PART_BITS; | |
291 | } | |
292 | ||
5087a50e MT |
293 | static int minor_to_index(int minor) |
294 | { | |
295 | return minor >> PART_BITS; | |
296 | } | |
297 | ||
a5eb9e4f RH |
298 | static ssize_t virtblk_serial_show(struct device *dev, |
299 | struct device_attribute *attr, char *buf) | |
300 | { | |
301 | struct gendisk *disk = dev_to_disk(dev); | |
302 | int err; | |
303 | ||
304 | /* sysfs gives us a PAGE_SIZE buffer */ | |
305 | BUILD_BUG_ON(PAGE_SIZE < VIRTIO_BLK_ID_BYTES); | |
306 | ||
307 | buf[VIRTIO_BLK_ID_BYTES] = '\0'; | |
308 | err = virtblk_get_id(disk, buf); | |
309 | if (!err) | |
310 | return strlen(buf); | |
311 | ||
312 | if (err == -EIO) /* Unsupported? Make it empty. */ | |
313 | return 0; | |
314 | ||
315 | return err; | |
316 | } | |
317 | DEVICE_ATTR(serial, S_IRUGO, virtblk_serial_show, NULL); | |
318 | ||
7a7c924c CH |
319 | static void virtblk_config_changed_work(struct work_struct *work) |
320 | { | |
321 | struct virtio_blk *vblk = | |
322 | container_of(work, struct virtio_blk, config_work); | |
323 | struct virtio_device *vdev = vblk->vdev; | |
324 | struct request_queue *q = vblk->disk->queue; | |
325 | char cap_str_2[10], cap_str_10[10]; | |
326 | u64 capacity, size; | |
327 | ||
4678d6f9 MT |
328 | mutex_lock(&vblk->config_lock); |
329 | if (!vblk->config_enable) | |
330 | goto done; | |
331 | ||
7a7c924c CH |
332 | /* Host must always specify the capacity. */ |
333 | vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity), | |
334 | &capacity, sizeof(capacity)); | |
335 | ||
336 | /* If capacity is too big, truncate with warning. */ | |
337 | if ((sector_t)capacity != capacity) { | |
338 | dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n", | |
339 | (unsigned long long)capacity); | |
340 | capacity = (sector_t)-1; | |
341 | } | |
342 | ||
343 | size = capacity * queue_logical_block_size(q); | |
344 | string_get_size(size, STRING_UNITS_2, cap_str_2, sizeof(cap_str_2)); | |
345 | string_get_size(size, STRING_UNITS_10, cap_str_10, sizeof(cap_str_10)); | |
346 | ||
347 | dev_notice(&vdev->dev, | |
348 | "new size: %llu %d-byte logical blocks (%s/%s)\n", | |
349 | (unsigned long long)capacity, | |
350 | queue_logical_block_size(q), | |
351 | cap_str_10, cap_str_2); | |
352 | ||
353 | set_capacity(vblk->disk, capacity); | |
e9986f30 | 354 | revalidate_disk(vblk->disk); |
4678d6f9 MT |
355 | done: |
356 | mutex_unlock(&vblk->config_lock); | |
7a7c924c CH |
357 | } |
358 | ||
359 | static void virtblk_config_changed(struct virtio_device *vdev) | |
360 | { | |
361 | struct virtio_blk *vblk = vdev->priv; | |
362 | ||
363 | queue_work(virtblk_wq, &vblk->config_work); | |
364 | } | |
365 | ||
6abd6e5a AS |
366 | static int init_vq(struct virtio_blk *vblk) |
367 | { | |
368 | int err = 0; | |
369 | ||
370 | /* We expect one virtqueue, for output. */ | |
371 | vblk->vq = virtio_find_single_vq(vblk->vdev, blk_done, "requests"); | |
372 | if (IS_ERR(vblk->vq)) | |
373 | err = PTR_ERR(vblk->vq); | |
374 | ||
375 | return err; | |
376 | } | |
377 | ||
c0aa3e09 RM |
378 | /* |
379 | * Legacy naming scheme used for virtio devices. We are stuck with it for | |
380 | * virtio blk but don't ever use it for any new driver. | |
381 | */ | |
382 | static int virtblk_name_format(char *prefix, int index, char *buf, int buflen) | |
383 | { | |
384 | const int base = 'z' - 'a' + 1; | |
385 | char *begin = buf + strlen(prefix); | |
386 | char *end = buf + buflen; | |
387 | char *p; | |
388 | int unit; | |
389 | ||
390 | p = end - 1; | |
391 | *p = '\0'; | |
392 | unit = base; | |
393 | do { | |
394 | if (p == begin) | |
395 | return -EINVAL; | |
396 | *--p = 'a' + (index % unit); | |
397 | index = (index / unit) - 1; | |
398 | } while (index >= 0); | |
399 | ||
400 | memmove(begin, p, end - p); | |
401 | memcpy(buf, prefix, strlen(prefix)); | |
402 | ||
403 | return 0; | |
404 | } | |
405 | ||
98e94444 | 406 | static int __devinit virtblk_probe(struct virtio_device *vdev) |
e467cde2 RR |
407 | { |
408 | struct virtio_blk *vblk; | |
69740c8b | 409 | struct request_queue *q; |
5087a50e | 410 | int err, index; |
e467cde2 | 411 | u64 cap; |
69740c8b CH |
412 | u32 v, blk_size, sg_elems, opt_io_size; |
413 | u16 min_io_size; | |
414 | u8 physical_block_exp, alignment_offset; | |
e467cde2 | 415 | |
5087a50e MT |
416 | err = ida_simple_get(&vd_index_ida, 0, minor_to_index(1 << MINORBITS), |
417 | GFP_KERNEL); | |
418 | if (err < 0) | |
419 | goto out; | |
420 | index = err; | |
4f3bf19c | 421 | |
0864b79a RR |
422 | /* We need to know how many segments before we allocate. */ |
423 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX, | |
424 | offsetof(struct virtio_blk_config, seg_max), | |
425 | &sg_elems); | |
a5b365a6 CH |
426 | |
427 | /* We need at least one SG element, whatever they say. */ | |
428 | if (err || !sg_elems) | |
0864b79a RR |
429 | sg_elems = 1; |
430 | ||
431 | /* We need an extra sg elements at head and tail. */ | |
432 | sg_elems += 2; | |
433 | vdev->priv = vblk = kmalloc(sizeof(*vblk) + | |
434 | sizeof(vblk->sg[0]) * sg_elems, GFP_KERNEL); | |
e467cde2 RR |
435 | if (!vblk) { |
436 | err = -ENOMEM; | |
5087a50e | 437 | goto out_free_index; |
e467cde2 RR |
438 | } |
439 | ||
440 | INIT_LIST_HEAD(&vblk->reqs); | |
441 | spin_lock_init(&vblk->lock); | |
442 | vblk->vdev = vdev; | |
0864b79a RR |
443 | vblk->sg_elems = sg_elems; |
444 | sg_init_table(vblk->sg, vblk->sg_elems); | |
4678d6f9 | 445 | mutex_init(&vblk->config_lock); |
7a7c924c | 446 | INIT_WORK(&vblk->config_work, virtblk_config_changed_work); |
4678d6f9 | 447 | vblk->config_enable = true; |
e467cde2 | 448 | |
6abd6e5a AS |
449 | err = init_vq(vblk); |
450 | if (err) | |
e467cde2 | 451 | goto out_free_vblk; |
e467cde2 RR |
452 | |
453 | vblk->pool = mempool_create_kmalloc_pool(1,sizeof(struct virtblk_req)); | |
454 | if (!vblk->pool) { | |
455 | err = -ENOMEM; | |
456 | goto out_free_vq; | |
457 | } | |
458 | ||
e467cde2 | 459 | /* FIXME: How many partitions? How long is a piece of string? */ |
4f3bf19c | 460 | vblk->disk = alloc_disk(1 << PART_BITS); |
e467cde2 RR |
461 | if (!vblk->disk) { |
462 | err = -ENOMEM; | |
4f3bf19c | 463 | goto out_mempool; |
e467cde2 RR |
464 | } |
465 | ||
69740c8b CH |
466 | q = vblk->disk->queue = blk_init_queue(do_virtblk_request, &vblk->lock); |
467 | if (!q) { | |
e467cde2 RR |
468 | err = -ENOMEM; |
469 | goto out_put_disk; | |
470 | } | |
471 | ||
69740c8b | 472 | q->queuedata = vblk; |
7d116b62 | 473 | |
c0aa3e09 | 474 | virtblk_name_format("vd", index, vblk->disk->disk_name, DISK_NAME_LEN); |
d50ed907 | 475 | |
e467cde2 | 476 | vblk->disk->major = major; |
d50ed907 | 477 | vblk->disk->first_minor = index_to_minor(index); |
e467cde2 RR |
478 | vblk->disk->private_data = vblk; |
479 | vblk->disk->fops = &virtblk_fops; | |
c4839346 | 480 | vblk->disk->driverfs_dev = &vdev->dev; |
5087a50e | 481 | vblk->index = index; |
4f3bf19c | 482 | |
02c42b7a | 483 | /* configure queue flush support */ |
4913efe4 TH |
484 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_FLUSH)) |
485 | blk_queue_flush(q, REQ_FLUSH); | |
e467cde2 | 486 | |
3ef53609 CB |
487 | /* If disk is read-only in the host, the guest should obey */ |
488 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO)) | |
489 | set_disk_ro(vblk->disk, 1); | |
490 | ||
a586d4f6 | 491 | /* Host must always specify the capacity. */ |
72e61eb4 RR |
492 | vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity), |
493 | &cap, sizeof(cap)); | |
e467cde2 RR |
494 | |
495 | /* If capacity is too big, truncate with warning. */ | |
496 | if ((sector_t)cap != cap) { | |
497 | dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n", | |
498 | (unsigned long long)cap); | |
499 | cap = (sector_t)-1; | |
500 | } | |
501 | set_capacity(vblk->disk, cap); | |
502 | ||
0864b79a | 503 | /* We can handle whatever the host told us to handle. */ |
ee714f2d | 504 | blk_queue_max_segments(q, vblk->sg_elems-2); |
0864b79a | 505 | |
4eff3cae | 506 | /* No need to bounce any requests */ |
69740c8b | 507 | blk_queue_bounce_limit(q, BLK_BOUNCE_ANY); |
4eff3cae | 508 | |
4b7f7e20 | 509 | /* No real sector limit. */ |
ee714f2d | 510 | blk_queue_max_hw_sectors(q, -1U); |
4b7f7e20 | 511 | |
a586d4f6 RR |
512 | /* Host can optionally specify maximum segment size and number of |
513 | * segments. */ | |
514 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX, | |
515 | offsetof(struct virtio_blk_config, size_max), | |
516 | &v); | |
e467cde2 | 517 | if (!err) |
69740c8b | 518 | blk_queue_max_segment_size(q, v); |
4b7f7e20 | 519 | else |
69740c8b | 520 | blk_queue_max_segment_size(q, -1U); |
e467cde2 | 521 | |
066f4d82 CB |
522 | /* Host can optionally specify the block size of the device */ |
523 | err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE, | |
524 | offsetof(struct virtio_blk_config, blk_size), | |
525 | &blk_size); | |
526 | if (!err) | |
69740c8b CH |
527 | blk_queue_logical_block_size(q, blk_size); |
528 | else | |
529 | blk_size = queue_logical_block_size(q); | |
530 | ||
531 | /* Use topology information if available */ | |
532 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
533 | offsetof(struct virtio_blk_config, physical_block_exp), | |
534 | &physical_block_exp); | |
535 | if (!err && physical_block_exp) | |
536 | blk_queue_physical_block_size(q, | |
537 | blk_size * (1 << physical_block_exp)); | |
538 | ||
539 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
540 | offsetof(struct virtio_blk_config, alignment_offset), | |
541 | &alignment_offset); | |
542 | if (!err && alignment_offset) | |
543 | blk_queue_alignment_offset(q, blk_size * alignment_offset); | |
544 | ||
545 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
546 | offsetof(struct virtio_blk_config, min_io_size), | |
547 | &min_io_size); | |
548 | if (!err && min_io_size) | |
549 | blk_queue_io_min(q, blk_size * min_io_size); | |
550 | ||
551 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
552 | offsetof(struct virtio_blk_config, opt_io_size), | |
553 | &opt_io_size); | |
554 | if (!err && opt_io_size) | |
555 | blk_queue_io_opt(q, blk_size * opt_io_size); | |
556 | ||
066f4d82 | 557 | |
e467cde2 | 558 | add_disk(vblk->disk); |
a5eb9e4f RH |
559 | err = device_create_file(disk_to_dev(vblk->disk), &dev_attr_serial); |
560 | if (err) | |
561 | goto out_del_disk; | |
562 | ||
e467cde2 RR |
563 | return 0; |
564 | ||
a5eb9e4f RH |
565 | out_del_disk: |
566 | del_gendisk(vblk->disk); | |
567 | blk_cleanup_queue(vblk->disk->queue); | |
e467cde2 RR |
568 | out_put_disk: |
569 | put_disk(vblk->disk); | |
e467cde2 RR |
570 | out_mempool: |
571 | mempool_destroy(vblk->pool); | |
572 | out_free_vq: | |
d2a7ddda | 573 | vdev->config->del_vqs(vdev); |
e467cde2 RR |
574 | out_free_vblk: |
575 | kfree(vblk); | |
5087a50e MT |
576 | out_free_index: |
577 | ida_simple_remove(&vd_index_ida, index); | |
e467cde2 RR |
578 | out: |
579 | return err; | |
580 | } | |
581 | ||
98e94444 | 582 | static void __devexit virtblk_remove(struct virtio_device *vdev) |
e467cde2 RR |
583 | { |
584 | struct virtio_blk *vblk = vdev->priv; | |
5087a50e | 585 | int index = vblk->index; |
e467cde2 | 586 | |
4678d6f9 MT |
587 | /* Prevent config work handler from accessing the device. */ |
588 | mutex_lock(&vblk->config_lock); | |
589 | vblk->config_enable = false; | |
590 | mutex_unlock(&vblk->config_lock); | |
7a7c924c | 591 | |
6e5aa7ef | 592 | /* Nothing should be pending. */ |
e467cde2 | 593 | BUG_ON(!list_empty(&vblk->reqs)); |
6e5aa7ef RR |
594 | |
595 | /* Stop all the virtqueues. */ | |
596 | vdev->config->reset(vdev); | |
597 | ||
4678d6f9 MT |
598 | flush_work(&vblk->config_work); |
599 | ||
ac9d463a | 600 | del_gendisk(vblk->disk); |
e467cde2 RR |
601 | blk_cleanup_queue(vblk->disk->queue); |
602 | put_disk(vblk->disk); | |
e467cde2 | 603 | mempool_destroy(vblk->pool); |
d2a7ddda | 604 | vdev->config->del_vqs(vdev); |
e467cde2 | 605 | kfree(vblk); |
5087a50e | 606 | ida_simple_remove(&vd_index_ida, index); |
e467cde2 RR |
607 | } |
608 | ||
f8fb5bc2 AS |
609 | #ifdef CONFIG_PM |
610 | static int virtblk_freeze(struct virtio_device *vdev) | |
611 | { | |
612 | struct virtio_blk *vblk = vdev->priv; | |
613 | ||
614 | /* Ensure we don't receive any more interrupts */ | |
615 | vdev->config->reset(vdev); | |
616 | ||
617 | /* Prevent config work handler from accessing the device. */ | |
618 | mutex_lock(&vblk->config_lock); | |
619 | vblk->config_enable = false; | |
620 | mutex_unlock(&vblk->config_lock); | |
621 | ||
622 | flush_work(&vblk->config_work); | |
623 | ||
624 | spin_lock_irq(vblk->disk->queue->queue_lock); | |
625 | blk_stop_queue(vblk->disk->queue); | |
626 | spin_unlock_irq(vblk->disk->queue->queue_lock); | |
627 | blk_sync_queue(vblk->disk->queue); | |
628 | ||
629 | vdev->config->del_vqs(vdev); | |
630 | return 0; | |
631 | } | |
632 | ||
633 | static int virtblk_restore(struct virtio_device *vdev) | |
634 | { | |
635 | struct virtio_blk *vblk = vdev->priv; | |
636 | int ret; | |
637 | ||
638 | vblk->config_enable = true; | |
639 | ret = init_vq(vdev->priv); | |
640 | if (!ret) { | |
641 | spin_lock_irq(vblk->disk->queue->queue_lock); | |
642 | blk_start_queue(vblk->disk->queue); | |
643 | spin_unlock_irq(vblk->disk->queue->queue_lock); | |
644 | } | |
645 | return ret; | |
646 | } | |
647 | #endif | |
648 | ||
47483e25 | 649 | static const struct virtio_device_id id_table[] = { |
e467cde2 RR |
650 | { VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID }, |
651 | { 0 }, | |
652 | }; | |
653 | ||
c45a6816 | 654 | static unsigned int features[] = { |
02c42b7a TH |
655 | VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY, |
656 | VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, VIRTIO_BLK_F_SCSI, | |
657 | VIRTIO_BLK_F_FLUSH, VIRTIO_BLK_F_TOPOLOGY | |
c45a6816 RR |
658 | }; |
659 | ||
4fbfff76 RM |
660 | /* |
661 | * virtio_blk causes spurious section mismatch warning by | |
662 | * simultaneously referring to a __devinit and a __devexit function. | |
663 | * Use __refdata to avoid this warning. | |
664 | */ | |
665 | static struct virtio_driver __refdata virtio_blk = { | |
7a7c924c CH |
666 | .feature_table = features, |
667 | .feature_table_size = ARRAY_SIZE(features), | |
668 | .driver.name = KBUILD_MODNAME, | |
669 | .driver.owner = THIS_MODULE, | |
670 | .id_table = id_table, | |
671 | .probe = virtblk_probe, | |
672 | .remove = __devexit_p(virtblk_remove), | |
673 | .config_changed = virtblk_config_changed, | |
f8fb5bc2 AS |
674 | #ifdef CONFIG_PM |
675 | .freeze = virtblk_freeze, | |
676 | .restore = virtblk_restore, | |
677 | #endif | |
e467cde2 RR |
678 | }; |
679 | ||
680 | static int __init init(void) | |
681 | { | |
7a7c924c CH |
682 | int error; |
683 | ||
684 | virtblk_wq = alloc_workqueue("virtio-blk", 0, 0); | |
685 | if (!virtblk_wq) | |
686 | return -ENOMEM; | |
687 | ||
4f3bf19c | 688 | major = register_blkdev(0, "virtblk"); |
7a7c924c CH |
689 | if (major < 0) { |
690 | error = major; | |
691 | goto out_destroy_workqueue; | |
692 | } | |
693 | ||
694 | error = register_virtio_driver(&virtio_blk); | |
695 | if (error) | |
696 | goto out_unregister_blkdev; | |
697 | return 0; | |
698 | ||
699 | out_unregister_blkdev: | |
700 | unregister_blkdev(major, "virtblk"); | |
701 | out_destroy_workqueue: | |
702 | destroy_workqueue(virtblk_wq); | |
703 | return error; | |
e467cde2 RR |
704 | } |
705 | ||
706 | static void __exit fini(void) | |
707 | { | |
4f3bf19c | 708 | unregister_blkdev(major, "virtblk"); |
e467cde2 | 709 | unregister_virtio_driver(&virtio_blk); |
7a7c924c | 710 | destroy_workqueue(virtblk_wq); |
e467cde2 RR |
711 | } |
712 | module_init(init); | |
713 | module_exit(fini); | |
714 | ||
715 | MODULE_DEVICE_TABLE(virtio, id_table); | |
716 | MODULE_DESCRIPTION("Virtio block driver"); | |
717 | MODULE_LICENSE("GPL"); |