]>
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 | { | |
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 | ||
51 | struct 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 | 59 | static 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 | ||
103 | static 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 | ||
183 | static 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 | */ | |
208 | static 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 |
233 | static 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 */ |
250 | static 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 | 274 | static 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 |
280 | static int index_to_minor(int index) |
281 | { | |
282 | return index << PART_BITS; | |
283 | } | |
284 | ||
5087a50e MT |
285 | static int minor_to_index(int minor) |
286 | { | |
287 | return minor >> PART_BITS; | |
288 | } | |
289 | ||
a5eb9e4f RH |
290 | static 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 | } | |
309 | DEVICE_ATTR(serial, S_IRUGO, virtblk_serial_show, NULL); | |
310 | ||
7a7c924c CH |
311 | static 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 |
347 | done: |
348 | mutex_unlock(&vblk->config_lock); | |
7a7c924c CH |
349 | } |
350 | ||
351 | static 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 |
358 | static 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 | */ | |
374 | static 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 | ||
98e94444 | 398 | static int __devinit virtblk_probe(struct virtio_device *vdev) |
e467cde2 RR |
399 | { |
400 | struct virtio_blk *vblk; | |
69740c8b | 401 | struct request_queue *q; |
5087a50e | 402 | int err, index; |
e467cde2 | 403 | u64 cap; |
69740c8b CH |
404 | u32 v, blk_size, sg_elems, opt_io_size; |
405 | u16 min_io_size; | |
406 | u8 physical_block_exp, alignment_offset; | |
e467cde2 | 407 | |
5087a50e MT |
408 | err = ida_simple_get(&vd_index_ida, 0, minor_to_index(1 << MINORBITS), |
409 | GFP_KERNEL); | |
410 | if (err < 0) | |
411 | goto out; | |
412 | index = err; | |
4f3bf19c | 413 | |
0864b79a RR |
414 | /* We need to know how many segments before we allocate. */ |
415 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX, | |
416 | offsetof(struct virtio_blk_config, seg_max), | |
417 | &sg_elems); | |
a5b365a6 CH |
418 | |
419 | /* We need at least one SG element, whatever they say. */ | |
420 | if (err || !sg_elems) | |
0864b79a RR |
421 | sg_elems = 1; |
422 | ||
423 | /* We need an extra sg elements at head and tail. */ | |
424 | sg_elems += 2; | |
425 | vdev->priv = vblk = kmalloc(sizeof(*vblk) + | |
426 | sizeof(vblk->sg[0]) * sg_elems, GFP_KERNEL); | |
e467cde2 RR |
427 | if (!vblk) { |
428 | err = -ENOMEM; | |
5087a50e | 429 | goto out_free_index; |
e467cde2 RR |
430 | } |
431 | ||
e467cde2 | 432 | vblk->vdev = vdev; |
0864b79a RR |
433 | vblk->sg_elems = sg_elems; |
434 | sg_init_table(vblk->sg, vblk->sg_elems); | |
4678d6f9 | 435 | mutex_init(&vblk->config_lock); |
7a7c924c | 436 | INIT_WORK(&vblk->config_work, virtblk_config_changed_work); |
4678d6f9 | 437 | vblk->config_enable = true; |
e467cde2 | 438 | |
6abd6e5a AS |
439 | err = init_vq(vblk); |
440 | if (err) | |
e467cde2 | 441 | goto out_free_vblk; |
e467cde2 RR |
442 | |
443 | vblk->pool = mempool_create_kmalloc_pool(1,sizeof(struct virtblk_req)); | |
444 | if (!vblk->pool) { | |
445 | err = -ENOMEM; | |
446 | goto out_free_vq; | |
447 | } | |
448 | ||
e467cde2 | 449 | /* FIXME: How many partitions? How long is a piece of string? */ |
4f3bf19c | 450 | vblk->disk = alloc_disk(1 << PART_BITS); |
e467cde2 RR |
451 | if (!vblk->disk) { |
452 | err = -ENOMEM; | |
4f3bf19c | 453 | goto out_mempool; |
e467cde2 RR |
454 | } |
455 | ||
2c95a329 | 456 | q = vblk->disk->queue = blk_init_queue(do_virtblk_request, NULL); |
69740c8b | 457 | if (!q) { |
e467cde2 RR |
458 | err = -ENOMEM; |
459 | goto out_put_disk; | |
460 | } | |
461 | ||
69740c8b | 462 | q->queuedata = vblk; |
7d116b62 | 463 | |
c0aa3e09 | 464 | virtblk_name_format("vd", index, vblk->disk->disk_name, DISK_NAME_LEN); |
d50ed907 | 465 | |
e467cde2 | 466 | vblk->disk->major = major; |
d50ed907 | 467 | vblk->disk->first_minor = index_to_minor(index); |
e467cde2 RR |
468 | vblk->disk->private_data = vblk; |
469 | vblk->disk->fops = &virtblk_fops; | |
c4839346 | 470 | vblk->disk->driverfs_dev = &vdev->dev; |
5087a50e | 471 | vblk->index = index; |
4f3bf19c | 472 | |
02c42b7a | 473 | /* configure queue flush support */ |
4913efe4 TH |
474 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_FLUSH)) |
475 | blk_queue_flush(q, REQ_FLUSH); | |
e467cde2 | 476 | |
3ef53609 CB |
477 | /* If disk is read-only in the host, the guest should obey */ |
478 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO)) | |
479 | set_disk_ro(vblk->disk, 1); | |
480 | ||
a586d4f6 | 481 | /* Host must always specify the capacity. */ |
72e61eb4 RR |
482 | vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity), |
483 | &cap, sizeof(cap)); | |
e467cde2 RR |
484 | |
485 | /* If capacity is too big, truncate with warning. */ | |
486 | if ((sector_t)cap != cap) { | |
487 | dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n", | |
488 | (unsigned long long)cap); | |
489 | cap = (sector_t)-1; | |
490 | } | |
491 | set_capacity(vblk->disk, cap); | |
492 | ||
0864b79a | 493 | /* We can handle whatever the host told us to handle. */ |
ee714f2d | 494 | blk_queue_max_segments(q, vblk->sg_elems-2); |
0864b79a | 495 | |
4eff3cae | 496 | /* No need to bounce any requests */ |
69740c8b | 497 | blk_queue_bounce_limit(q, BLK_BOUNCE_ANY); |
4eff3cae | 498 | |
4b7f7e20 | 499 | /* No real sector limit. */ |
ee714f2d | 500 | blk_queue_max_hw_sectors(q, -1U); |
4b7f7e20 | 501 | |
a586d4f6 RR |
502 | /* Host can optionally specify maximum segment size and number of |
503 | * segments. */ | |
504 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX, | |
505 | offsetof(struct virtio_blk_config, size_max), | |
506 | &v); | |
e467cde2 | 507 | if (!err) |
69740c8b | 508 | blk_queue_max_segment_size(q, v); |
4b7f7e20 | 509 | else |
69740c8b | 510 | blk_queue_max_segment_size(q, -1U); |
e467cde2 | 511 | |
066f4d82 CB |
512 | /* Host can optionally specify the block size of the device */ |
513 | err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE, | |
514 | offsetof(struct virtio_blk_config, blk_size), | |
515 | &blk_size); | |
516 | if (!err) | |
69740c8b CH |
517 | blk_queue_logical_block_size(q, blk_size); |
518 | else | |
519 | blk_size = queue_logical_block_size(q); | |
520 | ||
521 | /* Use topology information if available */ | |
522 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
523 | offsetof(struct virtio_blk_config, physical_block_exp), | |
524 | &physical_block_exp); | |
525 | if (!err && physical_block_exp) | |
526 | blk_queue_physical_block_size(q, | |
527 | blk_size * (1 << physical_block_exp)); | |
528 | ||
529 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
530 | offsetof(struct virtio_blk_config, alignment_offset), | |
531 | &alignment_offset); | |
532 | if (!err && alignment_offset) | |
533 | blk_queue_alignment_offset(q, blk_size * alignment_offset); | |
534 | ||
535 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
536 | offsetof(struct virtio_blk_config, min_io_size), | |
537 | &min_io_size); | |
538 | if (!err && min_io_size) | |
539 | blk_queue_io_min(q, blk_size * min_io_size); | |
540 | ||
541 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
542 | offsetof(struct virtio_blk_config, opt_io_size), | |
543 | &opt_io_size); | |
544 | if (!err && opt_io_size) | |
545 | blk_queue_io_opt(q, blk_size * opt_io_size); | |
546 | ||
066f4d82 | 547 | |
e467cde2 | 548 | add_disk(vblk->disk); |
a5eb9e4f RH |
549 | err = device_create_file(disk_to_dev(vblk->disk), &dev_attr_serial); |
550 | if (err) | |
551 | goto out_del_disk; | |
552 | ||
e467cde2 RR |
553 | return 0; |
554 | ||
a5eb9e4f RH |
555 | out_del_disk: |
556 | del_gendisk(vblk->disk); | |
557 | blk_cleanup_queue(vblk->disk->queue); | |
e467cde2 RR |
558 | out_put_disk: |
559 | put_disk(vblk->disk); | |
e467cde2 RR |
560 | out_mempool: |
561 | mempool_destroy(vblk->pool); | |
562 | out_free_vq: | |
d2a7ddda | 563 | vdev->config->del_vqs(vdev); |
e467cde2 RR |
564 | out_free_vblk: |
565 | kfree(vblk); | |
5087a50e MT |
566 | out_free_index: |
567 | ida_simple_remove(&vd_index_ida, index); | |
e467cde2 RR |
568 | out: |
569 | return err; | |
570 | } | |
571 | ||
98e94444 | 572 | static void __devexit virtblk_remove(struct virtio_device *vdev) |
e467cde2 RR |
573 | { |
574 | struct virtio_blk *vblk = vdev->priv; | |
5087a50e | 575 | int index = vblk->index; |
e467cde2 | 576 | |
4678d6f9 MT |
577 | /* Prevent config work handler from accessing the device. */ |
578 | mutex_lock(&vblk->config_lock); | |
579 | vblk->config_enable = false; | |
580 | mutex_unlock(&vblk->config_lock); | |
7a7c924c | 581 | |
02e2b124 | 582 | del_gendisk(vblk->disk); |
483001c7 | 583 | blk_cleanup_queue(vblk->disk->queue); |
02e2b124 | 584 | |
6e5aa7ef RR |
585 | /* Stop all the virtqueues. */ |
586 | vdev->config->reset(vdev); | |
587 | ||
4678d6f9 MT |
588 | flush_work(&vblk->config_work); |
589 | ||
e467cde2 | 590 | put_disk(vblk->disk); |
e467cde2 | 591 | mempool_destroy(vblk->pool); |
d2a7ddda | 592 | vdev->config->del_vqs(vdev); |
e467cde2 | 593 | kfree(vblk); |
5087a50e | 594 | ida_simple_remove(&vd_index_ida, index); |
e467cde2 RR |
595 | } |
596 | ||
f8fb5bc2 AS |
597 | #ifdef CONFIG_PM |
598 | static int virtblk_freeze(struct virtio_device *vdev) | |
599 | { | |
600 | struct virtio_blk *vblk = vdev->priv; | |
601 | ||
602 | /* Ensure we don't receive any more interrupts */ | |
603 | vdev->config->reset(vdev); | |
604 | ||
605 | /* Prevent config work handler from accessing the device. */ | |
606 | mutex_lock(&vblk->config_lock); | |
607 | vblk->config_enable = false; | |
608 | mutex_unlock(&vblk->config_lock); | |
609 | ||
610 | flush_work(&vblk->config_work); | |
611 | ||
612 | spin_lock_irq(vblk->disk->queue->queue_lock); | |
613 | blk_stop_queue(vblk->disk->queue); | |
614 | spin_unlock_irq(vblk->disk->queue->queue_lock); | |
615 | blk_sync_queue(vblk->disk->queue); | |
616 | ||
617 | vdev->config->del_vqs(vdev); | |
618 | return 0; | |
619 | } | |
620 | ||
621 | static int virtblk_restore(struct virtio_device *vdev) | |
622 | { | |
623 | struct virtio_blk *vblk = vdev->priv; | |
624 | int ret; | |
625 | ||
626 | vblk->config_enable = true; | |
627 | ret = init_vq(vdev->priv); | |
628 | if (!ret) { | |
629 | spin_lock_irq(vblk->disk->queue->queue_lock); | |
630 | blk_start_queue(vblk->disk->queue); | |
631 | spin_unlock_irq(vblk->disk->queue->queue_lock); | |
632 | } | |
633 | return ret; | |
634 | } | |
635 | #endif | |
636 | ||
47483e25 | 637 | static const struct virtio_device_id id_table[] = { |
e467cde2 RR |
638 | { VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID }, |
639 | { 0 }, | |
640 | }; | |
641 | ||
c45a6816 | 642 | static unsigned int features[] = { |
02c42b7a TH |
643 | VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY, |
644 | VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, VIRTIO_BLK_F_SCSI, | |
645 | VIRTIO_BLK_F_FLUSH, VIRTIO_BLK_F_TOPOLOGY | |
c45a6816 RR |
646 | }; |
647 | ||
4fbfff76 RM |
648 | /* |
649 | * virtio_blk causes spurious section mismatch warning by | |
650 | * simultaneously referring to a __devinit and a __devexit function. | |
651 | * Use __refdata to avoid this warning. | |
652 | */ | |
653 | static struct virtio_driver __refdata virtio_blk = { | |
7a7c924c CH |
654 | .feature_table = features, |
655 | .feature_table_size = ARRAY_SIZE(features), | |
656 | .driver.name = KBUILD_MODNAME, | |
657 | .driver.owner = THIS_MODULE, | |
658 | .id_table = id_table, | |
659 | .probe = virtblk_probe, | |
660 | .remove = __devexit_p(virtblk_remove), | |
661 | .config_changed = virtblk_config_changed, | |
f8fb5bc2 AS |
662 | #ifdef CONFIG_PM |
663 | .freeze = virtblk_freeze, | |
664 | .restore = virtblk_restore, | |
665 | #endif | |
e467cde2 RR |
666 | }; |
667 | ||
668 | static int __init init(void) | |
669 | { | |
7a7c924c CH |
670 | int error; |
671 | ||
672 | virtblk_wq = alloc_workqueue("virtio-blk", 0, 0); | |
673 | if (!virtblk_wq) | |
674 | return -ENOMEM; | |
675 | ||
4f3bf19c | 676 | major = register_blkdev(0, "virtblk"); |
7a7c924c CH |
677 | if (major < 0) { |
678 | error = major; | |
679 | goto out_destroy_workqueue; | |
680 | } | |
681 | ||
682 | error = register_virtio_driver(&virtio_blk); | |
683 | if (error) | |
684 | goto out_unregister_blkdev; | |
685 | return 0; | |
686 | ||
687 | out_unregister_blkdev: | |
688 | unregister_blkdev(major, "virtblk"); | |
689 | out_destroy_workqueue: | |
690 | destroy_workqueue(virtblk_wq); | |
691 | return error; | |
e467cde2 RR |
692 | } |
693 | ||
694 | static void __exit fini(void) | |
695 | { | |
4f3bf19c | 696 | unregister_blkdev(major, "virtblk"); |
e467cde2 | 697 | unregister_virtio_driver(&virtio_blk); |
7a7c924c | 698 | destroy_workqueue(virtblk_wq); |
e467cde2 RR |
699 | } |
700 | module_init(init); | |
701 | module_exit(fini); | |
702 | ||
703 | MODULE_DEVICE_TABLE(virtio, id_table); | |
704 | MODULE_DESCRIPTION("Virtio block driver"); | |
705 | MODULE_LICENSE("GPL"); |