]>
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 | ||
e467cde2 RR |
32 | mempool_t *pool; |
33 | ||
7a7c924c CH |
34 | /* Process context for config space updates */ |
35 | struct work_struct config_work; | |
36 | ||
4678d6f9 MT |
37 | /* Lock for config space updates */ |
38 | struct mutex config_lock; | |
39 | ||
40 | /* enable config space updates */ | |
41 | bool config_enable; | |
42 | ||
0864b79a RR |
43 | /* What host tells us, plus 2 for header & tailer. */ |
44 | unsigned int sg_elems; | |
45 | ||
5087a50e MT |
46 | /* Ida index - used to track minor number allocations. */ |
47 | int index; | |
48 | ||
e467cde2 | 49 | /* Scatterlist: can be too big for stack. */ |
0864b79a | 50 | struct scatterlist sg[/*sg_elems*/]; |
e467cde2 RR |
51 | }; |
52 | ||
53 | struct virtblk_req | |
54 | { | |
e467cde2 RR |
55 | struct request *req; |
56 | struct virtio_blk_outhdr out_hdr; | |
1cde26f9 | 57 | struct virtio_scsi_inhdr in_hdr; |
cb38fa23 | 58 | u8 status; |
e467cde2 RR |
59 | }; |
60 | ||
18445c4d | 61 | static void blk_done(struct virtqueue *vq) |
e467cde2 RR |
62 | { |
63 | struct virtio_blk *vblk = vq->vdev->priv; | |
64 | struct virtblk_req *vbr; | |
65 | unsigned int len; | |
66 | unsigned long flags; | |
67 | ||
68 | spin_lock_irqsave(&vblk->lock, flags); | |
09ec6b69 | 69 | while ((vbr = virtqueue_get_buf(vblk->vq, &len)) != NULL) { |
8316982a | 70 | int error; |
1cde26f9 | 71 | |
cb38fa23 | 72 | switch (vbr->status) { |
e467cde2 | 73 | case VIRTIO_BLK_S_OK: |
8316982a | 74 | error = 0; |
e467cde2 RR |
75 | break; |
76 | case VIRTIO_BLK_S_UNSUPP: | |
8316982a | 77 | error = -ENOTTY; |
e467cde2 RR |
78 | break; |
79 | default: | |
8316982a | 80 | error = -EIO; |
e467cde2 RR |
81 | break; |
82 | } | |
83 | ||
33659ebb CH |
84 | switch (vbr->req->cmd_type) { |
85 | case REQ_TYPE_BLOCK_PC: | |
1cde26f9 HR |
86 | vbr->req->resid_len = vbr->in_hdr.residual; |
87 | vbr->req->sense_len = vbr->in_hdr.sense_len; | |
88 | vbr->req->errors = vbr->in_hdr.errors; | |
33659ebb CH |
89 | break; |
90 | case REQ_TYPE_SPECIAL: | |
4cb2ea28 | 91 | vbr->req->errors = (error != 0); |
33659ebb | 92 | break; |
15fa6e81 JA |
93 | default: |
94 | break; | |
33659ebb | 95 | } |
1cde26f9 | 96 | |
40cbbb78 | 97 | __blk_end_request_all(vbr->req, error); |
e467cde2 RR |
98 | mempool_free(vbr, vblk->pool); |
99 | } | |
100 | /* In case queue is stopped waiting for more buffers. */ | |
101 | blk_start_queue(vblk->disk->queue); | |
102 | spin_unlock_irqrestore(&vblk->lock, flags); | |
e467cde2 RR |
103 | } |
104 | ||
105 | static bool do_req(struct request_queue *q, struct virtio_blk *vblk, | |
106 | struct request *req) | |
107 | { | |
1cde26f9 | 108 | unsigned long num, out = 0, in = 0; |
e467cde2 RR |
109 | struct virtblk_req *vbr; |
110 | ||
111 | vbr = mempool_alloc(vblk->pool, GFP_ATOMIC); | |
112 | if (!vbr) | |
113 | /* When another request finishes we'll try again. */ | |
114 | return false; | |
115 | ||
116 | vbr->req = req; | |
dd40e456 FT |
117 | |
118 | if (req->cmd_flags & REQ_FLUSH) { | |
119 | vbr->out_hdr.type = VIRTIO_BLK_T_FLUSH; | |
4cb2ea28 | 120 | vbr->out_hdr.sector = 0; |
121 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
dd40e456 FT |
122 | } else { |
123 | switch (req->cmd_type) { | |
124 | case REQ_TYPE_FS: | |
125 | vbr->out_hdr.type = 0; | |
126 | vbr->out_hdr.sector = blk_rq_pos(vbr->req); | |
127 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
128 | break; | |
129 | case REQ_TYPE_BLOCK_PC: | |
130 | vbr->out_hdr.type = VIRTIO_BLK_T_SCSI_CMD; | |
f1b0ef06 CH |
131 | vbr->out_hdr.sector = 0; |
132 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
133 | break; | |
dd40e456 FT |
134 | case REQ_TYPE_SPECIAL: |
135 | vbr->out_hdr.type = VIRTIO_BLK_T_GET_ID; | |
136 | vbr->out_hdr.sector = 0; | |
137 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); | |
138 | break; | |
139 | default: | |
140 | /* We don't put anything else in the queue. */ | |
141 | BUG(); | |
f1b0ef06 | 142 | } |
e467cde2 RR |
143 | } |
144 | ||
1cde26f9 | 145 | sg_set_buf(&vblk->sg[out++], &vbr->out_hdr, sizeof(vbr->out_hdr)); |
e467cde2 | 146 | |
1cde26f9 HR |
147 | /* |
148 | * If this is a packet command we need a couple of additional headers. | |
149 | * Behind the normal outhdr we put a segment with the scsi command | |
150 | * block, and before the normal inhdr we put the sense data and the | |
151 | * inhdr with additional status information before the normal inhdr. | |
152 | */ | |
33659ebb | 153 | if (vbr->req->cmd_type == REQ_TYPE_BLOCK_PC) |
1cde26f9 HR |
154 | sg_set_buf(&vblk->sg[out++], vbr->req->cmd, vbr->req->cmd_len); |
155 | ||
156 | num = blk_rq_map_sg(q, vbr->req, vblk->sg + out); | |
157 | ||
33659ebb | 158 | if (vbr->req->cmd_type == REQ_TYPE_BLOCK_PC) { |
6917f83f | 159 | sg_set_buf(&vblk->sg[num + out + in++], vbr->req->sense, SCSI_SENSE_BUFFERSIZE); |
1cde26f9 HR |
160 | sg_set_buf(&vblk->sg[num + out + in++], &vbr->in_hdr, |
161 | sizeof(vbr->in_hdr)); | |
162 | } | |
163 | ||
164 | sg_set_buf(&vblk->sg[num + out + in++], &vbr->status, | |
165 | sizeof(vbr->status)); | |
166 | ||
167 | if (num) { | |
168 | if (rq_data_dir(vbr->req) == WRITE) { | |
169 | vbr->out_hdr.type |= VIRTIO_BLK_T_OUT; | |
170 | out += num; | |
171 | } else { | |
172 | vbr->out_hdr.type |= VIRTIO_BLK_T_IN; | |
173 | in += num; | |
174 | } | |
e467cde2 RR |
175 | } |
176 | ||
f96fde41 | 177 | if (virtqueue_add_buf(vblk->vq, vblk->sg, out, in, vbr, GFP_ATOMIC)<0) { |
e467cde2 RR |
178 | mempool_free(vbr, vblk->pool); |
179 | return false; | |
180 | } | |
181 | ||
e467cde2 RR |
182 | return true; |
183 | } | |
184 | ||
185 | static void do_virtblk_request(struct request_queue *q) | |
186 | { | |
6c3b46f7 | 187 | struct virtio_blk *vblk = q->queuedata; |
e467cde2 RR |
188 | struct request *req; |
189 | unsigned int issued = 0; | |
190 | ||
9934c8c0 | 191 | while ((req = blk_peek_request(q)) != NULL) { |
0864b79a | 192 | BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems); |
e467cde2 RR |
193 | |
194 | /* If this request fails, stop queue and wait for something to | |
195 | finish to restart it. */ | |
196 | if (!do_req(q, vblk, req)) { | |
197 | blk_stop_queue(q); | |
198 | break; | |
199 | } | |
9934c8c0 | 200 | blk_start_request(req); |
e467cde2 RR |
201 | issued++; |
202 | } | |
203 | ||
204 | if (issued) | |
09ec6b69 | 205 | virtqueue_kick(vblk->vq); |
e467cde2 RR |
206 | } |
207 | ||
4cb2ea28 | 208 | /* return id (s/n) string for *disk to *id_str |
209 | */ | |
210 | static int virtblk_get_id(struct gendisk *disk, char *id_str) | |
211 | { | |
212 | struct virtio_blk *vblk = disk->private_data; | |
213 | struct request *req; | |
214 | struct bio *bio; | |
e4c4776d | 215 | int err; |
4cb2ea28 | 216 | |
217 | bio = bio_map_kern(vblk->disk->queue, id_str, VIRTIO_BLK_ID_BYTES, | |
218 | GFP_KERNEL); | |
219 | if (IS_ERR(bio)) | |
220 | return PTR_ERR(bio); | |
221 | ||
222 | req = blk_make_request(vblk->disk->queue, bio, GFP_KERNEL); | |
223 | if (IS_ERR(req)) { | |
224 | bio_put(bio); | |
225 | return PTR_ERR(req); | |
226 | } | |
227 | ||
228 | req->cmd_type = REQ_TYPE_SPECIAL; | |
e4c4776d MS |
229 | err = blk_execute_rq(vblk->disk->queue, vblk->disk, req, false); |
230 | blk_put_request(req); | |
231 | ||
232 | return err; | |
4cb2ea28 | 233 | } |
234 | ||
fe5a50a1 CH |
235 | static int virtblk_ioctl(struct block_device *bdev, fmode_t mode, |
236 | unsigned int cmd, unsigned long data) | |
e467cde2 | 237 | { |
1cde26f9 HR |
238 | struct gendisk *disk = bdev->bd_disk; |
239 | struct virtio_blk *vblk = disk->private_data; | |
240 | ||
241 | /* | |
242 | * Only allow the generic SCSI ioctls if the host can support it. | |
243 | */ | |
244 | if (!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_SCSI)) | |
d9ecdea7 | 245 | return -ENOTTY; |
1cde26f9 | 246 | |
577ebb37 PB |
247 | return scsi_cmd_blk_ioctl(bdev, mode, cmd, |
248 | (void __user *)data); | |
e467cde2 RR |
249 | } |
250 | ||
135da0b0 CB |
251 | /* We provide getgeo only to please some old bootloader/partitioning tools */ |
252 | static int virtblk_getgeo(struct block_device *bd, struct hd_geometry *geo) | |
253 | { | |
48e4043d RH |
254 | struct virtio_blk *vblk = bd->bd_disk->private_data; |
255 | struct virtio_blk_geometry vgeo; | |
256 | int err; | |
257 | ||
258 | /* see if the host passed in geometry config */ | |
259 | err = virtio_config_val(vblk->vdev, VIRTIO_BLK_F_GEOMETRY, | |
260 | offsetof(struct virtio_blk_config, geometry), | |
261 | &vgeo); | |
262 | ||
263 | if (!err) { | |
264 | geo->heads = vgeo.heads; | |
265 | geo->sectors = vgeo.sectors; | |
266 | geo->cylinders = vgeo.cylinders; | |
267 | } else { | |
268 | /* some standard values, similar to sd */ | |
269 | geo->heads = 1 << 6; | |
270 | geo->sectors = 1 << 5; | |
271 | geo->cylinders = get_capacity(bd->bd_disk) >> 11; | |
272 | } | |
135da0b0 CB |
273 | return 0; |
274 | } | |
275 | ||
83d5cde4 | 276 | static const struct block_device_operations virtblk_fops = { |
8a6cfeb6 | 277 | .ioctl = virtblk_ioctl, |
135da0b0 CB |
278 | .owner = THIS_MODULE, |
279 | .getgeo = virtblk_getgeo, | |
e467cde2 RR |
280 | }; |
281 | ||
d50ed907 CB |
282 | static int index_to_minor(int index) |
283 | { | |
284 | return index << PART_BITS; | |
285 | } | |
286 | ||
5087a50e MT |
287 | static int minor_to_index(int minor) |
288 | { | |
289 | return minor >> PART_BITS; | |
290 | } | |
291 | ||
a5eb9e4f RH |
292 | static ssize_t virtblk_serial_show(struct device *dev, |
293 | struct device_attribute *attr, char *buf) | |
294 | { | |
295 | struct gendisk *disk = dev_to_disk(dev); | |
296 | int err; | |
297 | ||
298 | /* sysfs gives us a PAGE_SIZE buffer */ | |
299 | BUILD_BUG_ON(PAGE_SIZE < VIRTIO_BLK_ID_BYTES); | |
300 | ||
301 | buf[VIRTIO_BLK_ID_BYTES] = '\0'; | |
302 | err = virtblk_get_id(disk, buf); | |
303 | if (!err) | |
304 | return strlen(buf); | |
305 | ||
306 | if (err == -EIO) /* Unsupported? Make it empty. */ | |
307 | return 0; | |
308 | ||
309 | return err; | |
310 | } | |
311 | DEVICE_ATTR(serial, S_IRUGO, virtblk_serial_show, NULL); | |
312 | ||
7a7c924c CH |
313 | static void virtblk_config_changed_work(struct work_struct *work) |
314 | { | |
315 | struct virtio_blk *vblk = | |
316 | container_of(work, struct virtio_blk, config_work); | |
317 | struct virtio_device *vdev = vblk->vdev; | |
318 | struct request_queue *q = vblk->disk->queue; | |
319 | char cap_str_2[10], cap_str_10[10]; | |
320 | u64 capacity, size; | |
321 | ||
4678d6f9 MT |
322 | mutex_lock(&vblk->config_lock); |
323 | if (!vblk->config_enable) | |
324 | goto done; | |
325 | ||
7a7c924c CH |
326 | /* Host must always specify the capacity. */ |
327 | vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity), | |
328 | &capacity, sizeof(capacity)); | |
329 | ||
330 | /* If capacity is too big, truncate with warning. */ | |
331 | if ((sector_t)capacity != capacity) { | |
332 | dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n", | |
333 | (unsigned long long)capacity); | |
334 | capacity = (sector_t)-1; | |
335 | } | |
336 | ||
337 | size = capacity * queue_logical_block_size(q); | |
338 | string_get_size(size, STRING_UNITS_2, cap_str_2, sizeof(cap_str_2)); | |
339 | string_get_size(size, STRING_UNITS_10, cap_str_10, sizeof(cap_str_10)); | |
340 | ||
341 | dev_notice(&vdev->dev, | |
342 | "new size: %llu %d-byte logical blocks (%s/%s)\n", | |
343 | (unsigned long long)capacity, | |
344 | queue_logical_block_size(q), | |
345 | cap_str_10, cap_str_2); | |
346 | ||
347 | set_capacity(vblk->disk, capacity); | |
e9986f30 | 348 | revalidate_disk(vblk->disk); |
4678d6f9 MT |
349 | done: |
350 | mutex_unlock(&vblk->config_lock); | |
7a7c924c CH |
351 | } |
352 | ||
353 | static void virtblk_config_changed(struct virtio_device *vdev) | |
354 | { | |
355 | struct virtio_blk *vblk = vdev->priv; | |
356 | ||
357 | queue_work(virtblk_wq, &vblk->config_work); | |
358 | } | |
359 | ||
6abd6e5a AS |
360 | static int init_vq(struct virtio_blk *vblk) |
361 | { | |
362 | int err = 0; | |
363 | ||
364 | /* We expect one virtqueue, for output. */ | |
365 | vblk->vq = virtio_find_single_vq(vblk->vdev, blk_done, "requests"); | |
366 | if (IS_ERR(vblk->vq)) | |
367 | err = PTR_ERR(vblk->vq); | |
368 | ||
369 | return err; | |
370 | } | |
371 | ||
c0aa3e09 RM |
372 | /* |
373 | * Legacy naming scheme used for virtio devices. We are stuck with it for | |
374 | * virtio blk but don't ever use it for any new driver. | |
375 | */ | |
376 | static int virtblk_name_format(char *prefix, int index, char *buf, int buflen) | |
377 | { | |
378 | const int base = 'z' - 'a' + 1; | |
379 | char *begin = buf + strlen(prefix); | |
380 | char *end = buf + buflen; | |
381 | char *p; | |
382 | int unit; | |
383 | ||
384 | p = end - 1; | |
385 | *p = '\0'; | |
386 | unit = base; | |
387 | do { | |
388 | if (p == begin) | |
389 | return -EINVAL; | |
390 | *--p = 'a' + (index % unit); | |
391 | index = (index / unit) - 1; | |
392 | } while (index >= 0); | |
393 | ||
394 | memmove(begin, p, end - p); | |
395 | memcpy(buf, prefix, strlen(prefix)); | |
396 | ||
397 | return 0; | |
398 | } | |
399 | ||
98e94444 | 400 | static int __devinit virtblk_probe(struct virtio_device *vdev) |
e467cde2 RR |
401 | { |
402 | struct virtio_blk *vblk; | |
69740c8b | 403 | struct request_queue *q; |
5087a50e | 404 | int err, index; |
e467cde2 | 405 | u64 cap; |
69740c8b CH |
406 | u32 v, blk_size, sg_elems, opt_io_size; |
407 | u16 min_io_size; | |
408 | u8 physical_block_exp, alignment_offset; | |
e467cde2 | 409 | |
5087a50e MT |
410 | err = ida_simple_get(&vd_index_ida, 0, minor_to_index(1 << MINORBITS), |
411 | GFP_KERNEL); | |
412 | if (err < 0) | |
413 | goto out; | |
414 | index = err; | |
4f3bf19c | 415 | |
0864b79a RR |
416 | /* We need to know how many segments before we allocate. */ |
417 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX, | |
418 | offsetof(struct virtio_blk_config, seg_max), | |
419 | &sg_elems); | |
a5b365a6 CH |
420 | |
421 | /* We need at least one SG element, whatever they say. */ | |
422 | if (err || !sg_elems) | |
0864b79a RR |
423 | sg_elems = 1; |
424 | ||
425 | /* We need an extra sg elements at head and tail. */ | |
426 | sg_elems += 2; | |
427 | vdev->priv = vblk = kmalloc(sizeof(*vblk) + | |
428 | sizeof(vblk->sg[0]) * sg_elems, GFP_KERNEL); | |
e467cde2 RR |
429 | if (!vblk) { |
430 | err = -ENOMEM; | |
5087a50e | 431 | goto out_free_index; |
e467cde2 RR |
432 | } |
433 | ||
e467cde2 RR |
434 | spin_lock_init(&vblk->lock); |
435 | vblk->vdev = vdev; | |
0864b79a RR |
436 | vblk->sg_elems = sg_elems; |
437 | sg_init_table(vblk->sg, vblk->sg_elems); | |
4678d6f9 | 438 | mutex_init(&vblk->config_lock); |
7a7c924c | 439 | INIT_WORK(&vblk->config_work, virtblk_config_changed_work); |
4678d6f9 | 440 | vblk->config_enable = true; |
e467cde2 | 441 | |
6abd6e5a AS |
442 | err = init_vq(vblk); |
443 | if (err) | |
e467cde2 | 444 | goto out_free_vblk; |
e467cde2 RR |
445 | |
446 | vblk->pool = mempool_create_kmalloc_pool(1,sizeof(struct virtblk_req)); | |
447 | if (!vblk->pool) { | |
448 | err = -ENOMEM; | |
449 | goto out_free_vq; | |
450 | } | |
451 | ||
e467cde2 | 452 | /* FIXME: How many partitions? How long is a piece of string? */ |
4f3bf19c | 453 | vblk->disk = alloc_disk(1 << PART_BITS); |
e467cde2 RR |
454 | if (!vblk->disk) { |
455 | err = -ENOMEM; | |
4f3bf19c | 456 | goto out_mempool; |
e467cde2 RR |
457 | } |
458 | ||
69740c8b CH |
459 | q = vblk->disk->queue = blk_init_queue(do_virtblk_request, &vblk->lock); |
460 | if (!q) { | |
e467cde2 RR |
461 | err = -ENOMEM; |
462 | goto out_put_disk; | |
463 | } | |
464 | ||
69740c8b | 465 | q->queuedata = vblk; |
7d116b62 | 466 | |
c0aa3e09 | 467 | virtblk_name_format("vd", index, vblk->disk->disk_name, DISK_NAME_LEN); |
d50ed907 | 468 | |
e467cde2 | 469 | vblk->disk->major = major; |
d50ed907 | 470 | vblk->disk->first_minor = index_to_minor(index); |
e467cde2 RR |
471 | vblk->disk->private_data = vblk; |
472 | vblk->disk->fops = &virtblk_fops; | |
c4839346 | 473 | vblk->disk->driverfs_dev = &vdev->dev; |
5087a50e | 474 | vblk->index = index; |
4f3bf19c | 475 | |
02c42b7a | 476 | /* configure queue flush support */ |
4913efe4 TH |
477 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_FLUSH)) |
478 | blk_queue_flush(q, REQ_FLUSH); | |
e467cde2 | 479 | |
3ef53609 CB |
480 | /* If disk is read-only in the host, the guest should obey */ |
481 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO)) | |
482 | set_disk_ro(vblk->disk, 1); | |
483 | ||
a586d4f6 | 484 | /* Host must always specify the capacity. */ |
72e61eb4 RR |
485 | vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity), |
486 | &cap, sizeof(cap)); | |
e467cde2 RR |
487 | |
488 | /* If capacity is too big, truncate with warning. */ | |
489 | if ((sector_t)cap != cap) { | |
490 | dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n", | |
491 | (unsigned long long)cap); | |
492 | cap = (sector_t)-1; | |
493 | } | |
494 | set_capacity(vblk->disk, cap); | |
495 | ||
0864b79a | 496 | /* We can handle whatever the host told us to handle. */ |
ee714f2d | 497 | blk_queue_max_segments(q, vblk->sg_elems-2); |
0864b79a | 498 | |
4eff3cae | 499 | /* No need to bounce any requests */ |
69740c8b | 500 | blk_queue_bounce_limit(q, BLK_BOUNCE_ANY); |
4eff3cae | 501 | |
4b7f7e20 | 502 | /* No real sector limit. */ |
ee714f2d | 503 | blk_queue_max_hw_sectors(q, -1U); |
4b7f7e20 | 504 | |
a586d4f6 RR |
505 | /* Host can optionally specify maximum segment size and number of |
506 | * segments. */ | |
507 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX, | |
508 | offsetof(struct virtio_blk_config, size_max), | |
509 | &v); | |
e467cde2 | 510 | if (!err) |
69740c8b | 511 | blk_queue_max_segment_size(q, v); |
4b7f7e20 | 512 | else |
69740c8b | 513 | blk_queue_max_segment_size(q, -1U); |
e467cde2 | 514 | |
066f4d82 CB |
515 | /* Host can optionally specify the block size of the device */ |
516 | err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE, | |
517 | offsetof(struct virtio_blk_config, blk_size), | |
518 | &blk_size); | |
519 | if (!err) | |
69740c8b CH |
520 | blk_queue_logical_block_size(q, blk_size); |
521 | else | |
522 | blk_size = queue_logical_block_size(q); | |
523 | ||
524 | /* Use topology information if available */ | |
525 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
526 | offsetof(struct virtio_blk_config, physical_block_exp), | |
527 | &physical_block_exp); | |
528 | if (!err && physical_block_exp) | |
529 | blk_queue_physical_block_size(q, | |
530 | blk_size * (1 << physical_block_exp)); | |
531 | ||
532 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
533 | offsetof(struct virtio_blk_config, alignment_offset), | |
534 | &alignment_offset); | |
535 | if (!err && alignment_offset) | |
536 | blk_queue_alignment_offset(q, blk_size * alignment_offset); | |
537 | ||
538 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
539 | offsetof(struct virtio_blk_config, min_io_size), | |
540 | &min_io_size); | |
541 | if (!err && min_io_size) | |
542 | blk_queue_io_min(q, blk_size * min_io_size); | |
543 | ||
544 | err = virtio_config_val(vdev, VIRTIO_BLK_F_TOPOLOGY, | |
545 | offsetof(struct virtio_blk_config, opt_io_size), | |
546 | &opt_io_size); | |
547 | if (!err && opt_io_size) | |
548 | blk_queue_io_opt(q, blk_size * opt_io_size); | |
549 | ||
066f4d82 | 550 | |
e467cde2 | 551 | add_disk(vblk->disk); |
a5eb9e4f RH |
552 | err = device_create_file(disk_to_dev(vblk->disk), &dev_attr_serial); |
553 | if (err) | |
554 | goto out_del_disk; | |
555 | ||
e467cde2 RR |
556 | return 0; |
557 | ||
a5eb9e4f RH |
558 | out_del_disk: |
559 | del_gendisk(vblk->disk); | |
560 | blk_cleanup_queue(vblk->disk->queue); | |
e467cde2 RR |
561 | out_put_disk: |
562 | put_disk(vblk->disk); | |
e467cde2 RR |
563 | out_mempool: |
564 | mempool_destroy(vblk->pool); | |
565 | out_free_vq: | |
d2a7ddda | 566 | vdev->config->del_vqs(vdev); |
e467cde2 RR |
567 | out_free_vblk: |
568 | kfree(vblk); | |
5087a50e MT |
569 | out_free_index: |
570 | ida_simple_remove(&vd_index_ida, index); | |
e467cde2 RR |
571 | out: |
572 | return err; | |
573 | } | |
574 | ||
98e94444 | 575 | static void __devexit virtblk_remove(struct virtio_device *vdev) |
e467cde2 RR |
576 | { |
577 | struct virtio_blk *vblk = vdev->priv; | |
5087a50e | 578 | int index = vblk->index; |
e467cde2 | 579 | |
4678d6f9 MT |
580 | /* Prevent config work handler from accessing the device. */ |
581 | mutex_lock(&vblk->config_lock); | |
582 | vblk->config_enable = false; | |
583 | mutex_unlock(&vblk->config_lock); | |
7a7c924c | 584 | |
02e2b124 | 585 | del_gendisk(vblk->disk); |
483001c7 | 586 | blk_cleanup_queue(vblk->disk->queue); |
02e2b124 | 587 | |
6e5aa7ef RR |
588 | /* Stop all the virtqueues. */ |
589 | vdev->config->reset(vdev); | |
590 | ||
4678d6f9 MT |
591 | flush_work(&vblk->config_work); |
592 | ||
e467cde2 | 593 | put_disk(vblk->disk); |
e467cde2 | 594 | mempool_destroy(vblk->pool); |
d2a7ddda | 595 | vdev->config->del_vqs(vdev); |
e467cde2 | 596 | kfree(vblk); |
5087a50e | 597 | ida_simple_remove(&vd_index_ida, index); |
e467cde2 RR |
598 | } |
599 | ||
f8fb5bc2 AS |
600 | #ifdef CONFIG_PM |
601 | static int virtblk_freeze(struct virtio_device *vdev) | |
602 | { | |
603 | struct virtio_blk *vblk = vdev->priv; | |
604 | ||
605 | /* Ensure we don't receive any more interrupts */ | |
606 | vdev->config->reset(vdev); | |
607 | ||
608 | /* Prevent config work handler from accessing the device. */ | |
609 | mutex_lock(&vblk->config_lock); | |
610 | vblk->config_enable = false; | |
611 | mutex_unlock(&vblk->config_lock); | |
612 | ||
613 | flush_work(&vblk->config_work); | |
614 | ||
615 | spin_lock_irq(vblk->disk->queue->queue_lock); | |
616 | blk_stop_queue(vblk->disk->queue); | |
617 | spin_unlock_irq(vblk->disk->queue->queue_lock); | |
618 | blk_sync_queue(vblk->disk->queue); | |
619 | ||
620 | vdev->config->del_vqs(vdev); | |
621 | return 0; | |
622 | } | |
623 | ||
624 | static int virtblk_restore(struct virtio_device *vdev) | |
625 | { | |
626 | struct virtio_blk *vblk = vdev->priv; | |
627 | int ret; | |
628 | ||
629 | vblk->config_enable = true; | |
630 | ret = init_vq(vdev->priv); | |
631 | if (!ret) { | |
632 | spin_lock_irq(vblk->disk->queue->queue_lock); | |
633 | blk_start_queue(vblk->disk->queue); | |
634 | spin_unlock_irq(vblk->disk->queue->queue_lock); | |
635 | } | |
636 | return ret; | |
637 | } | |
638 | #endif | |
639 | ||
47483e25 | 640 | static const struct virtio_device_id id_table[] = { |
e467cde2 RR |
641 | { VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID }, |
642 | { 0 }, | |
643 | }; | |
644 | ||
c45a6816 | 645 | static unsigned int features[] = { |
02c42b7a TH |
646 | VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY, |
647 | VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, VIRTIO_BLK_F_SCSI, | |
648 | VIRTIO_BLK_F_FLUSH, VIRTIO_BLK_F_TOPOLOGY | |
c45a6816 RR |
649 | }; |
650 | ||
4fbfff76 RM |
651 | /* |
652 | * virtio_blk causes spurious section mismatch warning by | |
653 | * simultaneously referring to a __devinit and a __devexit function. | |
654 | * Use __refdata to avoid this warning. | |
655 | */ | |
656 | static struct virtio_driver __refdata virtio_blk = { | |
7a7c924c CH |
657 | .feature_table = features, |
658 | .feature_table_size = ARRAY_SIZE(features), | |
659 | .driver.name = KBUILD_MODNAME, | |
660 | .driver.owner = THIS_MODULE, | |
661 | .id_table = id_table, | |
662 | .probe = virtblk_probe, | |
663 | .remove = __devexit_p(virtblk_remove), | |
664 | .config_changed = virtblk_config_changed, | |
f8fb5bc2 AS |
665 | #ifdef CONFIG_PM |
666 | .freeze = virtblk_freeze, | |
667 | .restore = virtblk_restore, | |
668 | #endif | |
e467cde2 RR |
669 | }; |
670 | ||
671 | static int __init init(void) | |
672 | { | |
7a7c924c CH |
673 | int error; |
674 | ||
675 | virtblk_wq = alloc_workqueue("virtio-blk", 0, 0); | |
676 | if (!virtblk_wq) | |
677 | return -ENOMEM; | |
678 | ||
4f3bf19c | 679 | major = register_blkdev(0, "virtblk"); |
7a7c924c CH |
680 | if (major < 0) { |
681 | error = major; | |
682 | goto out_destroy_workqueue; | |
683 | } | |
684 | ||
685 | error = register_virtio_driver(&virtio_blk); | |
686 | if (error) | |
687 | goto out_unregister_blkdev; | |
688 | return 0; | |
689 | ||
690 | out_unregister_blkdev: | |
691 | unregister_blkdev(major, "virtblk"); | |
692 | out_destroy_workqueue: | |
693 | destroy_workqueue(virtblk_wq); | |
694 | return error; | |
e467cde2 RR |
695 | } |
696 | ||
697 | static void __exit fini(void) | |
698 | { | |
4f3bf19c | 699 | unregister_blkdev(major, "virtblk"); |
e467cde2 | 700 | unregister_virtio_driver(&virtio_blk); |
7a7c924c | 701 | destroy_workqueue(virtblk_wq); |
e467cde2 RR |
702 | } |
703 | module_init(init); | |
704 | module_exit(fini); | |
705 | ||
706 | MODULE_DEVICE_TABLE(virtio, id_table); | |
707 | MODULE_DESCRIPTION("Virtio block driver"); | |
708 | MODULE_LICENSE("GPL"); |