]>
Commit | Line | Data |
---|---|---|
e467cde2 RR |
1 | //#define DEBUG |
2 | #include <linux/spinlock.h> | |
3 | #include <linux/blkdev.h> | |
4 | #include <linux/hdreg.h> | |
5 | #include <linux/virtio.h> | |
3ca4f5ca | 6 | #include <linux/virtio_ids.h> |
e467cde2 | 7 | #include <linux/virtio_blk.h> |
3d1266c7 JA |
8 | #include <linux/scatterlist.h> |
9 | ||
4f3bf19c | 10 | #define PART_BITS 4 |
e467cde2 | 11 | |
d50ed907 | 12 | static int major, index; |
4f3bf19c | 13 | |
e467cde2 RR |
14 | struct virtio_blk |
15 | { | |
16 | spinlock_t lock; | |
17 | ||
18 | struct virtio_device *vdev; | |
19 | struct virtqueue *vq; | |
20 | ||
21 | /* The disk structure for the kernel. */ | |
22 | struct gendisk *disk; | |
23 | ||
24 | /* Request tracking. */ | |
25 | struct list_head reqs; | |
26 | ||
27 | mempool_t *pool; | |
28 | ||
0864b79a RR |
29 | /* What host tells us, plus 2 for header & tailer. */ |
30 | unsigned int sg_elems; | |
31 | ||
e467cde2 | 32 | /* Scatterlist: can be too big for stack. */ |
0864b79a | 33 | struct scatterlist sg[/*sg_elems*/]; |
e467cde2 RR |
34 | }; |
35 | ||
36 | struct virtblk_req | |
37 | { | |
38 | struct list_head list; | |
39 | struct request *req; | |
40 | struct virtio_blk_outhdr out_hdr; | |
1cde26f9 | 41 | struct virtio_scsi_inhdr in_hdr; |
cb38fa23 | 42 | u8 status; |
e467cde2 RR |
43 | }; |
44 | ||
18445c4d | 45 | static void blk_done(struct virtqueue *vq) |
e467cde2 RR |
46 | { |
47 | struct virtio_blk *vblk = vq->vdev->priv; | |
48 | struct virtblk_req *vbr; | |
49 | unsigned int len; | |
50 | unsigned long flags; | |
51 | ||
52 | spin_lock_irqsave(&vblk->lock, flags); | |
53 | while ((vbr = vblk->vq->vq_ops->get_buf(vblk->vq, &len)) != NULL) { | |
8316982a | 54 | int error; |
1cde26f9 | 55 | |
cb38fa23 | 56 | switch (vbr->status) { |
e467cde2 | 57 | case VIRTIO_BLK_S_OK: |
8316982a | 58 | error = 0; |
e467cde2 RR |
59 | break; |
60 | case VIRTIO_BLK_S_UNSUPP: | |
8316982a | 61 | error = -ENOTTY; |
e467cde2 RR |
62 | break; |
63 | default: | |
8316982a | 64 | error = -EIO; |
e467cde2 RR |
65 | break; |
66 | } | |
67 | ||
1cde26f9 HR |
68 | if (blk_pc_request(vbr->req)) { |
69 | vbr->req->resid_len = vbr->in_hdr.residual; | |
70 | vbr->req->sense_len = vbr->in_hdr.sense_len; | |
71 | vbr->req->errors = vbr->in_hdr.errors; | |
72 | } | |
73 | ||
40cbbb78 | 74 | __blk_end_request_all(vbr->req, error); |
e467cde2 RR |
75 | list_del(&vbr->list); |
76 | mempool_free(vbr, vblk->pool); | |
77 | } | |
78 | /* In case queue is stopped waiting for more buffers. */ | |
79 | blk_start_queue(vblk->disk->queue); | |
80 | spin_unlock_irqrestore(&vblk->lock, flags); | |
e467cde2 RR |
81 | } |
82 | ||
83 | static bool do_req(struct request_queue *q, struct virtio_blk *vblk, | |
84 | struct request *req) | |
85 | { | |
1cde26f9 | 86 | unsigned long num, out = 0, in = 0; |
e467cde2 RR |
87 | struct virtblk_req *vbr; |
88 | ||
89 | vbr = mempool_alloc(vblk->pool, GFP_ATOMIC); | |
90 | if (!vbr) | |
91 | /* When another request finishes we'll try again. */ | |
92 | return false; | |
93 | ||
94 | vbr->req = req; | |
95 | if (blk_fs_request(vbr->req)) { | |
96 | vbr->out_hdr.type = 0; | |
83096ebf | 97 | vbr->out_hdr.sector = blk_rq_pos(vbr->req); |
766ca442 | 98 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); |
e467cde2 RR |
99 | } else if (blk_pc_request(vbr->req)) { |
100 | vbr->out_hdr.type = VIRTIO_BLK_T_SCSI_CMD; | |
101 | vbr->out_hdr.sector = 0; | |
766ca442 | 102 | vbr->out_hdr.ioprio = req_get_ioprio(vbr->req); |
e467cde2 RR |
103 | } else { |
104 | /* We don't put anything else in the queue. */ | |
105 | BUG(); | |
106 | } | |
107 | ||
108 | if (blk_barrier_rq(vbr->req)) | |
109 | vbr->out_hdr.type |= VIRTIO_BLK_T_BARRIER; | |
110 | ||
1cde26f9 | 111 | sg_set_buf(&vblk->sg[out++], &vbr->out_hdr, sizeof(vbr->out_hdr)); |
e467cde2 | 112 | |
1cde26f9 HR |
113 | /* |
114 | * If this is a packet command we need a couple of additional headers. | |
115 | * Behind the normal outhdr we put a segment with the scsi command | |
116 | * block, and before the normal inhdr we put the sense data and the | |
117 | * inhdr with additional status information before the normal inhdr. | |
118 | */ | |
119 | if (blk_pc_request(vbr->req)) | |
120 | sg_set_buf(&vblk->sg[out++], vbr->req->cmd, vbr->req->cmd_len); | |
121 | ||
122 | num = blk_rq_map_sg(q, vbr->req, vblk->sg + out); | |
123 | ||
124 | if (blk_pc_request(vbr->req)) { | |
125 | sg_set_buf(&vblk->sg[num + out + in++], vbr->req->sense, 96); | |
126 | sg_set_buf(&vblk->sg[num + out + in++], &vbr->in_hdr, | |
127 | sizeof(vbr->in_hdr)); | |
128 | } | |
129 | ||
130 | sg_set_buf(&vblk->sg[num + out + in++], &vbr->status, | |
131 | sizeof(vbr->status)); | |
132 | ||
133 | if (num) { | |
134 | if (rq_data_dir(vbr->req) == WRITE) { | |
135 | vbr->out_hdr.type |= VIRTIO_BLK_T_OUT; | |
136 | out += num; | |
137 | } else { | |
138 | vbr->out_hdr.type |= VIRTIO_BLK_T_IN; | |
139 | in += num; | |
140 | } | |
e467cde2 RR |
141 | } |
142 | ||
3c1b27d5 | 143 | if (vblk->vq->vq_ops->add_buf(vblk->vq, vblk->sg, out, in, vbr) < 0) { |
e467cde2 RR |
144 | mempool_free(vbr, vblk->pool); |
145 | return false; | |
146 | } | |
147 | ||
148 | list_add_tail(&vbr->list, &vblk->reqs); | |
149 | return true; | |
150 | } | |
151 | ||
152 | static void do_virtblk_request(struct request_queue *q) | |
153 | { | |
6c3b46f7 | 154 | struct virtio_blk *vblk = q->queuedata; |
e467cde2 RR |
155 | struct request *req; |
156 | unsigned int issued = 0; | |
157 | ||
9934c8c0 | 158 | while ((req = blk_peek_request(q)) != NULL) { |
0864b79a | 159 | BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems); |
e467cde2 RR |
160 | |
161 | /* If this request fails, stop queue and wait for something to | |
162 | finish to restart it. */ | |
163 | if (!do_req(q, vblk, req)) { | |
164 | blk_stop_queue(q); | |
165 | break; | |
166 | } | |
9934c8c0 | 167 | blk_start_request(req); |
e467cde2 RR |
168 | issued++; |
169 | } | |
170 | ||
171 | if (issued) | |
172 | vblk->vq->vq_ops->kick(vblk->vq); | |
173 | } | |
174 | ||
1d589bb1 | 175 | /* return ATA identify data |
176 | */ | |
177 | static int virtblk_identify(struct gendisk *disk, void *argp) | |
178 | { | |
179 | struct virtio_blk *vblk = disk->private_data; | |
180 | void *opaque; | |
181 | int err = -ENOMEM; | |
182 | ||
183 | opaque = kmalloc(VIRTIO_BLK_ID_BYTES, GFP_KERNEL); | |
184 | if (!opaque) | |
185 | goto out; | |
186 | ||
187 | err = virtio_config_buf(vblk->vdev, VIRTIO_BLK_F_IDENTIFY, | |
188 | offsetof(struct virtio_blk_config, identify), opaque, | |
189 | VIRTIO_BLK_ID_BYTES); | |
190 | ||
191 | if (err) | |
192 | goto out_kfree; | |
193 | ||
194 | if (copy_to_user(argp, opaque, VIRTIO_BLK_ID_BYTES)) | |
195 | err = -EFAULT; | |
196 | ||
197 | out_kfree: | |
198 | kfree(opaque); | |
199 | out: | |
200 | return err; | |
201 | } | |
202 | ||
4e109852 | 203 | static int virtblk_ioctl(struct block_device *bdev, fmode_t mode, |
e467cde2 RR |
204 | unsigned cmd, unsigned long data) |
205 | { | |
1cde26f9 HR |
206 | struct gendisk *disk = bdev->bd_disk; |
207 | struct virtio_blk *vblk = disk->private_data; | |
1d589bb1 | 208 | void __user *argp = (void __user *)data; |
209 | ||
210 | if (cmd == HDIO_GET_IDENTITY) | |
211 | return virtblk_identify(disk, argp); | |
1cde26f9 HR |
212 | |
213 | /* | |
214 | * Only allow the generic SCSI ioctls if the host can support it. | |
215 | */ | |
216 | if (!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_SCSI)) | |
d9ecdea7 | 217 | return -ENOTTY; |
1cde26f9 | 218 | |
1d589bb1 | 219 | return scsi_cmd_ioctl(disk->queue, disk, mode, cmd, argp); |
e467cde2 RR |
220 | } |
221 | ||
135da0b0 CB |
222 | /* We provide getgeo only to please some old bootloader/partitioning tools */ |
223 | static int virtblk_getgeo(struct block_device *bd, struct hd_geometry *geo) | |
224 | { | |
48e4043d RH |
225 | struct virtio_blk *vblk = bd->bd_disk->private_data; |
226 | struct virtio_blk_geometry vgeo; | |
227 | int err; | |
228 | ||
229 | /* see if the host passed in geometry config */ | |
230 | err = virtio_config_val(vblk->vdev, VIRTIO_BLK_F_GEOMETRY, | |
231 | offsetof(struct virtio_blk_config, geometry), | |
232 | &vgeo); | |
233 | ||
234 | if (!err) { | |
235 | geo->heads = vgeo.heads; | |
236 | geo->sectors = vgeo.sectors; | |
237 | geo->cylinders = vgeo.cylinders; | |
238 | } else { | |
239 | /* some standard values, similar to sd */ | |
240 | geo->heads = 1 << 6; | |
241 | geo->sectors = 1 << 5; | |
242 | geo->cylinders = get_capacity(bd->bd_disk) >> 11; | |
243 | } | |
135da0b0 CB |
244 | return 0; |
245 | } | |
246 | ||
e467cde2 | 247 | static struct block_device_operations virtblk_fops = { |
4e109852 | 248 | .locked_ioctl = virtblk_ioctl, |
135da0b0 CB |
249 | .owner = THIS_MODULE, |
250 | .getgeo = virtblk_getgeo, | |
e467cde2 RR |
251 | }; |
252 | ||
d50ed907 CB |
253 | static int index_to_minor(int index) |
254 | { | |
255 | return index << PART_BITS; | |
256 | } | |
257 | ||
98e94444 | 258 | static int __devinit virtblk_probe(struct virtio_device *vdev) |
e467cde2 RR |
259 | { |
260 | struct virtio_blk *vblk; | |
4f3bf19c | 261 | int err; |
e467cde2 RR |
262 | u64 cap; |
263 | u32 v; | |
0864b79a | 264 | u32 blk_size, sg_elems; |
e467cde2 | 265 | |
d50ed907 | 266 | if (index_to_minor(index) >= 1 << MINORBITS) |
4f3bf19c CB |
267 | return -ENOSPC; |
268 | ||
0864b79a RR |
269 | /* We need to know how many segments before we allocate. */ |
270 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX, | |
271 | offsetof(struct virtio_blk_config, seg_max), | |
272 | &sg_elems); | |
273 | if (err) | |
274 | sg_elems = 1; | |
275 | ||
276 | /* We need an extra sg elements at head and tail. */ | |
277 | sg_elems += 2; | |
278 | vdev->priv = vblk = kmalloc(sizeof(*vblk) + | |
279 | sizeof(vblk->sg[0]) * sg_elems, GFP_KERNEL); | |
e467cde2 RR |
280 | if (!vblk) { |
281 | err = -ENOMEM; | |
282 | goto out; | |
283 | } | |
284 | ||
285 | INIT_LIST_HEAD(&vblk->reqs); | |
286 | spin_lock_init(&vblk->lock); | |
287 | vblk->vdev = vdev; | |
0864b79a RR |
288 | vblk->sg_elems = sg_elems; |
289 | sg_init_table(vblk->sg, vblk->sg_elems); | |
e467cde2 RR |
290 | |
291 | /* We expect one virtqueue, for output. */ | |
d2a7ddda | 292 | vblk->vq = virtio_find_single_vq(vdev, blk_done, "requests"); |
e467cde2 RR |
293 | if (IS_ERR(vblk->vq)) { |
294 | err = PTR_ERR(vblk->vq); | |
295 | goto out_free_vblk; | |
296 | } | |
297 | ||
298 | vblk->pool = mempool_create_kmalloc_pool(1,sizeof(struct virtblk_req)); | |
299 | if (!vblk->pool) { | |
300 | err = -ENOMEM; | |
301 | goto out_free_vq; | |
302 | } | |
303 | ||
e467cde2 | 304 | /* FIXME: How many partitions? How long is a piece of string? */ |
4f3bf19c | 305 | vblk->disk = alloc_disk(1 << PART_BITS); |
e467cde2 RR |
306 | if (!vblk->disk) { |
307 | err = -ENOMEM; | |
4f3bf19c | 308 | goto out_mempool; |
e467cde2 RR |
309 | } |
310 | ||
311 | vblk->disk->queue = blk_init_queue(do_virtblk_request, &vblk->lock); | |
312 | if (!vblk->disk->queue) { | |
313 | err = -ENOMEM; | |
314 | goto out_put_disk; | |
315 | } | |
316 | ||
6c3b46f7 | 317 | vblk->disk->queue->queuedata = vblk; |
7d116b62 FLVC |
318 | queue_flag_set_unlocked(QUEUE_FLAG_VIRT, vblk->disk->queue); |
319 | ||
d50ed907 CB |
320 | if (index < 26) { |
321 | sprintf(vblk->disk->disk_name, "vd%c", 'a' + index % 26); | |
322 | } else if (index < (26 + 1) * 26) { | |
323 | sprintf(vblk->disk->disk_name, "vd%c%c", | |
324 | 'a' + index / 26 - 1, 'a' + index % 26); | |
325 | } else { | |
326 | const unsigned int m1 = (index / 26 - 1) / 26 - 1; | |
327 | const unsigned int m2 = (index / 26 - 1) % 26; | |
328 | const unsigned int m3 = index % 26; | |
329 | sprintf(vblk->disk->disk_name, "vd%c%c%c", | |
330 | 'a' + m1, 'a' + m2, 'a' + m3); | |
331 | } | |
332 | ||
e467cde2 | 333 | vblk->disk->major = major; |
d50ed907 | 334 | vblk->disk->first_minor = index_to_minor(index); |
e467cde2 RR |
335 | vblk->disk->private_data = vblk; |
336 | vblk->disk->fops = &virtblk_fops; | |
c4839346 | 337 | vblk->disk->driverfs_dev = &vdev->dev; |
d50ed907 | 338 | index++; |
4f3bf19c | 339 | |
e467cde2 | 340 | /* If barriers are supported, tell block layer that queue is ordered */ |
c45a6816 | 341 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_BARRIER)) |
e467cde2 RR |
342 | blk_queue_ordered(vblk->disk->queue, QUEUE_ORDERED_TAG, NULL); |
343 | ||
3ef53609 CB |
344 | /* If disk is read-only in the host, the guest should obey */ |
345 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO)) | |
346 | set_disk_ro(vblk->disk, 1); | |
347 | ||
a586d4f6 | 348 | /* Host must always specify the capacity. */ |
72e61eb4 RR |
349 | vdev->config->get(vdev, offsetof(struct virtio_blk_config, capacity), |
350 | &cap, sizeof(cap)); | |
e467cde2 RR |
351 | |
352 | /* If capacity is too big, truncate with warning. */ | |
353 | if ((sector_t)cap != cap) { | |
354 | dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n", | |
355 | (unsigned long long)cap); | |
356 | cap = (sector_t)-1; | |
357 | } | |
358 | set_capacity(vblk->disk, cap); | |
359 | ||
0864b79a RR |
360 | /* We can handle whatever the host told us to handle. */ |
361 | blk_queue_max_phys_segments(vblk->disk->queue, vblk->sg_elems-2); | |
362 | blk_queue_max_hw_segments(vblk->disk->queue, vblk->sg_elems-2); | |
363 | ||
4eff3cae CH |
364 | /* No need to bounce any requests */ |
365 | blk_queue_bounce_limit(vblk->disk->queue, BLK_BOUNCE_ANY); | |
366 | ||
4b7f7e20 RR |
367 | /* No real sector limit. */ |
368 | blk_queue_max_sectors(vblk->disk->queue, -1U); | |
369 | ||
a586d4f6 RR |
370 | /* Host can optionally specify maximum segment size and number of |
371 | * segments. */ | |
372 | err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX, | |
373 | offsetof(struct virtio_blk_config, size_max), | |
374 | &v); | |
e467cde2 RR |
375 | if (!err) |
376 | blk_queue_max_segment_size(vblk->disk->queue, v); | |
4b7f7e20 | 377 | else |
b194aee9 | 378 | blk_queue_max_segment_size(vblk->disk->queue, -1U); |
e467cde2 | 379 | |
066f4d82 CB |
380 | /* Host can optionally specify the block size of the device */ |
381 | err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE, | |
382 | offsetof(struct virtio_blk_config, blk_size), | |
383 | &blk_size); | |
384 | if (!err) | |
e1defc4f | 385 | blk_queue_logical_block_size(vblk->disk->queue, blk_size); |
066f4d82 | 386 | |
e467cde2 RR |
387 | add_disk(vblk->disk); |
388 | return 0; | |
389 | ||
390 | out_put_disk: | |
391 | put_disk(vblk->disk); | |
e467cde2 RR |
392 | out_mempool: |
393 | mempool_destroy(vblk->pool); | |
394 | out_free_vq: | |
d2a7ddda | 395 | vdev->config->del_vqs(vdev); |
e467cde2 RR |
396 | out_free_vblk: |
397 | kfree(vblk); | |
398 | out: | |
399 | return err; | |
400 | } | |
401 | ||
98e94444 | 402 | static void __devexit virtblk_remove(struct virtio_device *vdev) |
e467cde2 RR |
403 | { |
404 | struct virtio_blk *vblk = vdev->priv; | |
e467cde2 | 405 | |
6e5aa7ef | 406 | /* Nothing should be pending. */ |
e467cde2 | 407 | BUG_ON(!list_empty(&vblk->reqs)); |
6e5aa7ef RR |
408 | |
409 | /* Stop all the virtqueues. */ | |
410 | vdev->config->reset(vdev); | |
411 | ||
ac9d463a | 412 | del_gendisk(vblk->disk); |
e467cde2 RR |
413 | blk_cleanup_queue(vblk->disk->queue); |
414 | put_disk(vblk->disk); | |
e467cde2 | 415 | mempool_destroy(vblk->pool); |
d2a7ddda | 416 | vdev->config->del_vqs(vdev); |
e467cde2 RR |
417 | kfree(vblk); |
418 | } | |
419 | ||
420 | static struct virtio_device_id id_table[] = { | |
421 | { VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID }, | |
422 | { 0 }, | |
423 | }; | |
424 | ||
c45a6816 RR |
425 | static unsigned int features[] = { |
426 | VIRTIO_BLK_F_BARRIER, VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, | |
066f4d82 | 427 | VIRTIO_BLK_F_GEOMETRY, VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, |
1d589bb1 | 428 | VIRTIO_BLK_F_SCSI, VIRTIO_BLK_F_IDENTIFY |
c45a6816 RR |
429 | }; |
430 | ||
4fbfff76 RM |
431 | /* |
432 | * virtio_blk causes spurious section mismatch warning by | |
433 | * simultaneously referring to a __devinit and a __devexit function. | |
434 | * Use __refdata to avoid this warning. | |
435 | */ | |
436 | static struct virtio_driver __refdata virtio_blk = { | |
c45a6816 RR |
437 | .feature_table = features, |
438 | .feature_table_size = ARRAY_SIZE(features), | |
e467cde2 RR |
439 | .driver.name = KBUILD_MODNAME, |
440 | .driver.owner = THIS_MODULE, | |
441 | .id_table = id_table, | |
442 | .probe = virtblk_probe, | |
443 | .remove = __devexit_p(virtblk_remove), | |
444 | }; | |
445 | ||
446 | static int __init init(void) | |
447 | { | |
4f3bf19c CB |
448 | major = register_blkdev(0, "virtblk"); |
449 | if (major < 0) | |
450 | return major; | |
e467cde2 RR |
451 | return register_virtio_driver(&virtio_blk); |
452 | } | |
453 | ||
454 | static void __exit fini(void) | |
455 | { | |
4f3bf19c | 456 | unregister_blkdev(major, "virtblk"); |
e467cde2 RR |
457 | unregister_virtio_driver(&virtio_blk); |
458 | } | |
459 | module_init(init); | |
460 | module_exit(fini); | |
461 | ||
462 | MODULE_DEVICE_TABLE(virtio, id_table); | |
463 | MODULE_DESCRIPTION("Virtio block driver"); | |
464 | MODULE_LICENSE("GPL"); |