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> |
ad71473d | 8 | #include <linux/interrupt.h> |
e467cde2 RR |
9 | #include <linux/virtio.h> |
10 | #include <linux/virtio_blk.h> | |
3d1266c7 | 11 | #include <linux/scatterlist.h> |
7a7c924c | 12 | #include <linux/string_helpers.h> |
6917f83f | 13 | #include <scsi/scsi_cmnd.h> |
5087a50e | 14 | #include <linux/idr.h> |
1cf7e9c6 | 15 | #include <linux/blk-mq.h> |
ad71473d | 16 | #include <linux/blk-mq-virtio.h> |
1cf7e9c6 | 17 | #include <linux/numa.h> |
3d1266c7 | 18 | |
4f3bf19c | 19 | #define PART_BITS 4 |
6a27b656 | 20 | #define VQ_NAME_LEN 16 |
1f23816b | 21 | #define MAX_DISCARD_SEGMENTS 256u |
e467cde2 | 22 | |
5087a50e MT |
23 | static int major; |
24 | static DEFINE_IDA(vd_index_ida); | |
25 | ||
2a647bfe | 26 | static struct workqueue_struct *virtblk_wq; |
4f3bf19c | 27 | |
6a27b656 ML |
28 | struct virtio_blk_vq { |
29 | struct virtqueue *vq; | |
30 | spinlock_t lock; | |
31 | char name[VQ_NAME_LEN]; | |
32 | } ____cacheline_aligned_in_smp; | |
33 | ||
bb6ec576 | 34 | struct virtio_blk { |
e467cde2 | 35 | struct virtio_device *vdev; |
e467cde2 RR |
36 | |
37 | /* The disk structure for the kernel. */ | |
38 | struct gendisk *disk; | |
39 | ||
24d2f903 CH |
40 | /* Block layer tags. */ |
41 | struct blk_mq_tag_set tag_set; | |
42 | ||
7a7c924c CH |
43 | /* Process context for config space updates */ |
44 | struct work_struct config_work; | |
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; | |
6a27b656 ML |
51 | |
52 | /* num of vqs */ | |
53 | int num_vqs; | |
54 | struct virtio_blk_vq *vqs; | |
e467cde2 RR |
55 | }; |
56 | ||
bb6ec576 | 57 | struct virtblk_req { |
97b50a65 CH |
58 | #ifdef CONFIG_VIRTIO_BLK_SCSI |
59 | struct scsi_request sreq; /* for SCSI passthrough, must be first */ | |
60 | u8 sense[SCSI_SENSE_BUFFERSIZE]; | |
1cde26f9 | 61 | struct virtio_scsi_inhdr in_hdr; |
97b50a65 CH |
62 | #endif |
63 | struct virtio_blk_outhdr out_hdr; | |
cb38fa23 | 64 | u8 status; |
a98755c5 | 65 | struct scatterlist sg[]; |
e467cde2 RR |
66 | }; |
67 | ||
2a842aca | 68 | static inline blk_status_t virtblk_result(struct virtblk_req *vbr) |
a98755c5 AH |
69 | { |
70 | switch (vbr->status) { | |
71 | case VIRTIO_BLK_S_OK: | |
2a842aca | 72 | return BLK_STS_OK; |
a98755c5 | 73 | case VIRTIO_BLK_S_UNSUPP: |
2a842aca | 74 | return BLK_STS_NOTSUPP; |
a98755c5 | 75 | default: |
2a842aca | 76 | return BLK_STS_IOERR; |
a98755c5 AH |
77 | } |
78 | } | |
79 | ||
97b50a65 CH |
80 | /* |
81 | * If this is a packet command we need a couple of additional headers. Behind | |
82 | * the normal outhdr we put a segment with the scsi command block, and before | |
83 | * the normal inhdr we put the sense data and the inhdr with additional status | |
84 | * information. | |
85 | */ | |
86 | #ifdef CONFIG_VIRTIO_BLK_SCSI | |
87 | static int virtblk_add_req_scsi(struct virtqueue *vq, struct virtblk_req *vbr, | |
88 | struct scatterlist *data_sg, bool have_data) | |
c85a1f91 | 89 | { |
20af3cfd | 90 | struct scatterlist hdr, status, cmd, sense, inhdr, *sgs[6]; |
8f39db9d PB |
91 | unsigned int num_out = 0, num_in = 0; |
92 | ||
93 | sg_init_one(&hdr, &vbr->out_hdr, sizeof(vbr->out_hdr)); | |
94 | sgs[num_out++] = &hdr; | |
97b50a65 CH |
95 | sg_init_one(&cmd, vbr->sreq.cmd, vbr->sreq.cmd_len); |
96 | sgs[num_out++] = &cmd; | |
97 | ||
98 | if (have_data) { | |
99 | if (vbr->out_hdr.type & cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_OUT)) | |
100 | sgs[num_out++] = data_sg; | |
101 | else | |
102 | sgs[num_out + num_in++] = data_sg; | |
103 | } | |
104 | ||
105 | sg_init_one(&sense, vbr->sense, SCSI_SENSE_BUFFERSIZE); | |
106 | sgs[num_out + num_in++] = &sense; | |
107 | sg_init_one(&inhdr, &vbr->in_hdr, sizeof(vbr->in_hdr)); | |
108 | sgs[num_out + num_in++] = &inhdr; | |
109 | sg_init_one(&status, &vbr->status, sizeof(vbr->status)); | |
110 | sgs[num_out + num_in++] = &status; | |
111 | ||
112 | return virtqueue_add_sgs(vq, sgs, num_out, num_in, vbr, GFP_ATOMIC); | |
113 | } | |
114 | ||
a1a6e62b | 115 | static inline void virtblk_scsi_request_done(struct request *req) |
97b50a65 CH |
116 | { |
117 | struct virtblk_req *vbr = blk_mq_rq_to_pdu(req); | |
118 | struct virtio_blk *vblk = req->q->queuedata; | |
119 | struct scsi_request *sreq = &vbr->sreq; | |
120 | ||
121 | sreq->resid_len = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.residual); | |
122 | sreq->sense_len = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.sense_len); | |
17d5363b | 123 | sreq->result = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.errors); |
97b50a65 CH |
124 | } |
125 | ||
126 | static int virtblk_ioctl(struct block_device *bdev, fmode_t mode, | |
127 | unsigned int cmd, unsigned long data) | |
128 | { | |
129 | struct gendisk *disk = bdev->bd_disk; | |
130 | struct virtio_blk *vblk = disk->private_data; | |
8f39db9d | 131 | |
20af3cfd | 132 | /* |
97b50a65 | 133 | * Only allow the generic SCSI ioctls if the host can support it. |
20af3cfd | 134 | */ |
97b50a65 CH |
135 | if (!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_SCSI)) |
136 | return -ENOTTY; | |
137 | ||
138 | return scsi_cmd_blk_ioctl(bdev, mode, cmd, | |
139 | (void __user *)data); | |
140 | } | |
141 | #else | |
142 | static inline int virtblk_add_req_scsi(struct virtqueue *vq, | |
143 | struct virtblk_req *vbr, struct scatterlist *data_sg, | |
144 | bool have_data) | |
145 | { | |
146 | return -EIO; | |
147 | } | |
a1a6e62b | 148 | static inline void virtblk_scsi_request_done(struct request *req) |
97b50a65 CH |
149 | { |
150 | } | |
151 | #define virtblk_ioctl NULL | |
152 | #endif /* CONFIG_VIRTIO_BLK_SCSI */ | |
153 | ||
154 | static int virtblk_add_req(struct virtqueue *vq, struct virtblk_req *vbr, | |
155 | struct scatterlist *data_sg, bool have_data) | |
156 | { | |
157 | struct scatterlist hdr, status, *sgs[3]; | |
158 | unsigned int num_out = 0, num_in = 0; | |
159 | ||
160 | sg_init_one(&hdr, &vbr->out_hdr, sizeof(vbr->out_hdr)); | |
161 | sgs[num_out++] = &hdr; | |
20af3cfd | 162 | |
0a11cc36 | 163 | if (have_data) { |
19c1c5a6 | 164 | if (vbr->out_hdr.type & cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_OUT)) |
20af3cfd | 165 | sgs[num_out++] = data_sg; |
8f39db9d | 166 | else |
20af3cfd PB |
167 | sgs[num_out + num_in++] = data_sg; |
168 | } | |
169 | ||
8f39db9d PB |
170 | sg_init_one(&status, &vbr->status, sizeof(vbr->status)); |
171 | sgs[num_out + num_in++] = &status; | |
172 | ||
173 | return virtqueue_add_sgs(vq, sgs, num_out, num_in, vbr, GFP_ATOMIC); | |
5ee21a52 PB |
174 | } |
175 | ||
1f23816b CL |
176 | static int virtblk_setup_discard_write_zeroes(struct request *req, bool unmap) |
177 | { | |
178 | unsigned short segments = blk_rq_nr_discard_segments(req); | |
179 | unsigned short n = 0; | |
180 | struct virtio_blk_discard_write_zeroes *range; | |
181 | struct bio *bio; | |
182 | u32 flags = 0; | |
183 | ||
184 | if (unmap) | |
185 | flags |= VIRTIO_BLK_WRITE_ZEROES_FLAG_UNMAP; | |
186 | ||
187 | range = kmalloc_array(segments, sizeof(*range), GFP_ATOMIC); | |
188 | if (!range) | |
189 | return -ENOMEM; | |
190 | ||
191 | __rq_for_each_bio(bio, req) { | |
192 | u64 sector = bio->bi_iter.bi_sector; | |
193 | u32 num_sectors = bio->bi_iter.bi_size >> SECTOR_SHIFT; | |
194 | ||
195 | range[n].flags = cpu_to_le32(flags); | |
196 | range[n].num_sectors = cpu_to_le32(num_sectors); | |
197 | range[n].sector = cpu_to_le64(sector); | |
198 | n++; | |
199 | } | |
200 | ||
201 | req->special_vec.bv_page = virt_to_page(range); | |
202 | req->special_vec.bv_offset = offset_in_page(range); | |
203 | req->special_vec.bv_len = sizeof(*range) * segments; | |
204 | req->rq_flags |= RQF_SPECIAL_PAYLOAD; | |
205 | ||
206 | return 0; | |
207 | } | |
208 | ||
5124c285 | 209 | static inline void virtblk_request_done(struct request *req) |
a98755c5 | 210 | { |
9d74e257 | 211 | struct virtblk_req *vbr = blk_mq_rq_to_pdu(req); |
a98755c5 | 212 | |
1f23816b CL |
213 | if (req->rq_flags & RQF_SPECIAL_PAYLOAD) { |
214 | kfree(page_address(req->special_vec.bv_page) + | |
215 | req->special_vec.bv_offset); | |
216 | } | |
217 | ||
aebf526b CH |
218 | switch (req_op(req)) { |
219 | case REQ_OP_SCSI_IN: | |
220 | case REQ_OP_SCSI_OUT: | |
a1a6e62b | 221 | virtblk_scsi_request_done(req); |
97b50a65 | 222 | break; |
a98755c5 AH |
223 | } |
224 | ||
d19633d5 | 225 | blk_mq_end_request(req, virtblk_result(vbr)); |
a98755c5 AH |
226 | } |
227 | ||
228 | static void virtblk_done(struct virtqueue *vq) | |
e467cde2 RR |
229 | { |
230 | struct virtio_blk *vblk = vq->vdev->priv; | |
1cf7e9c6 | 231 | bool req_done = false; |
6a27b656 | 232 | int qid = vq->index; |
e467cde2 | 233 | struct virtblk_req *vbr; |
e467cde2 | 234 | unsigned long flags; |
a98755c5 | 235 | unsigned int len; |
e467cde2 | 236 | |
6a27b656 | 237 | spin_lock_irqsave(&vblk->vqs[qid].lock, flags); |
bb811108 AH |
238 | do { |
239 | virtqueue_disable_cb(vq); | |
6a27b656 | 240 | while ((vbr = virtqueue_get_buf(vblk->vqs[qid].vq, &len)) != NULL) { |
85dada09 CH |
241 | struct request *req = blk_mq_rq_from_pdu(vbr); |
242 | ||
08e0029a | 243 | blk_mq_complete_request(req); |
1cf7e9c6 | 244 | req_done = true; |
33659ebb | 245 | } |
7f03b17d HG |
246 | if (unlikely(virtqueue_is_broken(vq))) |
247 | break; | |
bb811108 | 248 | } while (!virtqueue_enable_cb(vq)); |
1cf7e9c6 | 249 | |
e467cde2 | 250 | /* In case queue is stopped waiting for more buffers. */ |
a98755c5 | 251 | if (req_done) |
1b4a3258 | 252 | blk_mq_start_stopped_hw_queues(vblk->disk->queue, true); |
6a27b656 | 253 | spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags); |
a98755c5 AH |
254 | } |
255 | ||
944e7c87 JA |
256 | static void virtio_commit_rqs(struct blk_mq_hw_ctx *hctx) |
257 | { | |
258 | struct virtio_blk *vblk = hctx->queue->queuedata; | |
259 | struct virtio_blk_vq *vq = &vblk->vqs[hctx->queue_num]; | |
260 | bool kick; | |
261 | ||
262 | spin_lock_irq(&vq->lock); | |
263 | kick = virtqueue_kick_prepare(vq->vq); | |
264 | spin_unlock_irq(&vq->lock); | |
265 | ||
266 | if (kick) | |
267 | virtqueue_notify(vq->vq); | |
268 | } | |
269 | ||
fc17b653 | 270 | static blk_status_t virtio_queue_rq(struct blk_mq_hw_ctx *hctx, |
74c45052 | 271 | const struct blk_mq_queue_data *bd) |
e467cde2 | 272 | { |
1cf7e9c6 | 273 | struct virtio_blk *vblk = hctx->queue->queuedata; |
74c45052 | 274 | struct request *req = bd->rq; |
9d74e257 | 275 | struct virtblk_req *vbr = blk_mq_rq_to_pdu(req); |
1cf7e9c6 | 276 | unsigned long flags; |
20af3cfd | 277 | unsigned int num; |
6a27b656 | 278 | int qid = hctx->queue_num; |
5261b85e | 279 | int err; |
e8edca6f | 280 | bool notify = false; |
1f23816b | 281 | bool unmap = false; |
aebf526b | 282 | u32 type; |
e467cde2 | 283 | |
1cf7e9c6 | 284 | BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems); |
e467cde2 | 285 | |
aebf526b CH |
286 | switch (req_op(req)) { |
287 | case REQ_OP_READ: | |
288 | case REQ_OP_WRITE: | |
289 | type = 0; | |
290 | break; | |
291 | case REQ_OP_FLUSH: | |
292 | type = VIRTIO_BLK_T_FLUSH; | |
293 | break; | |
1f23816b CL |
294 | case REQ_OP_DISCARD: |
295 | type = VIRTIO_BLK_T_DISCARD; | |
296 | break; | |
297 | case REQ_OP_WRITE_ZEROES: | |
298 | type = VIRTIO_BLK_T_WRITE_ZEROES; | |
299 | unmap = !(req->cmd_flags & REQ_NOUNMAP); | |
300 | break; | |
aebf526b CH |
301 | case REQ_OP_SCSI_IN: |
302 | case REQ_OP_SCSI_OUT: | |
303 | type = VIRTIO_BLK_T_SCSI_CMD; | |
304 | break; | |
305 | case REQ_OP_DRV_IN: | |
306 | type = VIRTIO_BLK_T_GET_ID; | |
307 | break; | |
308 | default: | |
309 | WARN_ON_ONCE(1); | |
fc17b653 | 310 | return BLK_STS_IOERR; |
e467cde2 RR |
311 | } |
312 | ||
aebf526b CH |
313 | vbr->out_hdr.type = cpu_to_virtio32(vblk->vdev, type); |
314 | vbr->out_hdr.sector = type ? | |
315 | 0 : cpu_to_virtio64(vblk->vdev, blk_rq_pos(req)); | |
316 | vbr->out_hdr.ioprio = cpu_to_virtio32(vblk->vdev, req_get_ioprio(req)); | |
317 | ||
e2490073 CH |
318 | blk_mq_start_request(req); |
319 | ||
1f23816b CL |
320 | if (type == VIRTIO_BLK_T_DISCARD || type == VIRTIO_BLK_T_WRITE_ZEROES) { |
321 | err = virtblk_setup_discard_write_zeroes(req, unmap); | |
322 | if (err) | |
323 | return BLK_STS_RESOURCE; | |
324 | } | |
325 | ||
85dada09 | 326 | num = blk_rq_map_sg(hctx->queue, req, vbr->sg); |
1cde26f9 | 327 | if (num) { |
85dada09 | 328 | if (rq_data_dir(req) == WRITE) |
19c1c5a6 | 329 | vbr->out_hdr.type |= cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_OUT); |
20af3cfd | 330 | else |
19c1c5a6 | 331 | vbr->out_hdr.type |= cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_IN); |
e467cde2 RR |
332 | } |
333 | ||
6a27b656 | 334 | spin_lock_irqsave(&vblk->vqs[qid].lock, flags); |
6e9fe8dd | 335 | if (blk_rq_is_scsi(req)) |
97b50a65 CH |
336 | err = virtblk_add_req_scsi(vblk->vqs[qid].vq, vbr, vbr->sg, num); |
337 | else | |
338 | err = virtblk_add_req(vblk->vqs[qid].vq, vbr, vbr->sg, num); | |
5261b85e | 339 | if (err) { |
6a27b656 | 340 | virtqueue_kick(vblk->vqs[qid].vq); |
1cf7e9c6 | 341 | blk_mq_stop_hw_queue(hctx); |
6a27b656 | 342 | spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags); |
5261b85e RR |
343 | /* Out of mem doesn't actually happen, since we fall back |
344 | * to direct descriptors */ | |
345 | if (err == -ENOMEM || err == -ENOSPC) | |
86ff7c2a | 346 | return BLK_STS_DEV_RESOURCE; |
fc17b653 | 347 | return BLK_STS_IOERR; |
a98755c5 AH |
348 | } |
349 | ||
74c45052 | 350 | if (bd->last && virtqueue_kick_prepare(vblk->vqs[qid].vq)) |
e8edca6f | 351 | notify = true; |
6a27b656 | 352 | spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags); |
e8edca6f ML |
353 | |
354 | if (notify) | |
6a27b656 | 355 | virtqueue_notify(vblk->vqs[qid].vq); |
fc17b653 | 356 | return BLK_STS_OK; |
a98755c5 AH |
357 | } |
358 | ||
4cb2ea28 | 359 | /* return id (s/n) string for *disk to *id_str |
360 | */ | |
361 | static int virtblk_get_id(struct gendisk *disk, char *id_str) | |
362 | { | |
363 | struct virtio_blk *vblk = disk->private_data; | |
f9596695 | 364 | struct request_queue *q = vblk->disk->queue; |
4cb2ea28 | 365 | struct request *req; |
e4c4776d | 366 | int err; |
4cb2ea28 | 367 | |
ff005a06 | 368 | req = blk_get_request(q, REQ_OP_DRV_IN, 0); |
f9596695 | 369 | if (IS_ERR(req)) |
4cb2ea28 | 370 | return PTR_ERR(req); |
f9596695 CH |
371 | |
372 | err = blk_rq_map_kern(q, req, id_str, VIRTIO_BLK_ID_BYTES, GFP_KERNEL); | |
373 | if (err) | |
374 | goto out; | |
375 | ||
b7819b92 | 376 | blk_execute_rq(vblk->disk->queue, vblk->disk, req, false); |
2a842aca | 377 | err = blk_status_to_errno(virtblk_result(blk_mq_rq_to_pdu(req))); |
f9596695 | 378 | out: |
e4c4776d | 379 | blk_put_request(req); |
e4c4776d | 380 | return err; |
4cb2ea28 | 381 | } |
382 | ||
135da0b0 CB |
383 | /* We provide getgeo only to please some old bootloader/partitioning tools */ |
384 | static int virtblk_getgeo(struct block_device *bd, struct hd_geometry *geo) | |
385 | { | |
48e4043d | 386 | struct virtio_blk *vblk = bd->bd_disk->private_data; |
48e4043d RH |
387 | |
388 | /* see if the host passed in geometry config */ | |
855e0c52 RR |
389 | if (virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_GEOMETRY)) { |
390 | virtio_cread(vblk->vdev, struct virtio_blk_config, | |
391 | geometry.cylinders, &geo->cylinders); | |
392 | virtio_cread(vblk->vdev, struct virtio_blk_config, | |
393 | geometry.heads, &geo->heads); | |
394 | virtio_cread(vblk->vdev, struct virtio_blk_config, | |
395 | geometry.sectors, &geo->sectors); | |
48e4043d RH |
396 | } else { |
397 | /* some standard values, similar to sd */ | |
398 | geo->heads = 1 << 6; | |
399 | geo->sectors = 1 << 5; | |
400 | geo->cylinders = get_capacity(bd->bd_disk) >> 11; | |
401 | } | |
135da0b0 CB |
402 | return 0; |
403 | } | |
404 | ||
83d5cde4 | 405 | static const struct block_device_operations virtblk_fops = { |
8a6cfeb6 | 406 | .ioctl = virtblk_ioctl, |
135da0b0 CB |
407 | .owner = THIS_MODULE, |
408 | .getgeo = virtblk_getgeo, | |
e467cde2 RR |
409 | }; |
410 | ||
d50ed907 CB |
411 | static int index_to_minor(int index) |
412 | { | |
413 | return index << PART_BITS; | |
414 | } | |
415 | ||
5087a50e MT |
416 | static int minor_to_index(int minor) |
417 | { | |
418 | return minor >> PART_BITS; | |
419 | } | |
420 | ||
e982c4d0 HR |
421 | static ssize_t serial_show(struct device *dev, |
422 | struct device_attribute *attr, char *buf) | |
a5eb9e4f RH |
423 | { |
424 | struct gendisk *disk = dev_to_disk(dev); | |
425 | int err; | |
426 | ||
427 | /* sysfs gives us a PAGE_SIZE buffer */ | |
428 | BUILD_BUG_ON(PAGE_SIZE < VIRTIO_BLK_ID_BYTES); | |
429 | ||
430 | buf[VIRTIO_BLK_ID_BYTES] = '\0'; | |
431 | err = virtblk_get_id(disk, buf); | |
432 | if (!err) | |
433 | return strlen(buf); | |
434 | ||
435 | if (err == -EIO) /* Unsupported? Make it empty. */ | |
436 | return 0; | |
437 | ||
438 | return err; | |
439 | } | |
393c525b | 440 | |
e982c4d0 | 441 | static DEVICE_ATTR_RO(serial); |
a5eb9e4f | 442 | |
daf2a501 SH |
443 | /* The queue's logical block size must be set before calling this */ |
444 | static void virtblk_update_capacity(struct virtio_blk *vblk, bool resize) | |
7a7c924c | 445 | { |
7a7c924c CH |
446 | struct virtio_device *vdev = vblk->vdev; |
447 | struct request_queue *q = vblk->disk->queue; | |
448 | char cap_str_2[10], cap_str_10[10]; | |
1046d304 | 449 | unsigned long long nblocks; |
b9f28d86 | 450 | u64 capacity; |
7a7c924c CH |
451 | |
452 | /* Host must always specify the capacity. */ | |
855e0c52 | 453 | virtio_cread(vdev, struct virtio_blk_config, capacity, &capacity); |
7a7c924c CH |
454 | |
455 | /* If capacity is too big, truncate with warning. */ | |
456 | if ((sector_t)capacity != capacity) { | |
457 | dev_warn(&vdev->dev, "Capacity %llu too large: truncating\n", | |
458 | (unsigned long long)capacity); | |
459 | capacity = (sector_t)-1; | |
460 | } | |
461 | ||
1046d304 SH |
462 | nblocks = DIV_ROUND_UP_ULL(capacity, queue_logical_block_size(q) >> 9); |
463 | ||
464 | string_get_size(nblocks, queue_logical_block_size(q), | |
b9f28d86 | 465 | STRING_UNITS_2, cap_str_2, sizeof(cap_str_2)); |
1046d304 | 466 | string_get_size(nblocks, queue_logical_block_size(q), |
b9f28d86 | 467 | STRING_UNITS_10, cap_str_10, sizeof(cap_str_10)); |
7a7c924c CH |
468 | |
469 | dev_notice(&vdev->dev, | |
daf2a501 SH |
470 | "[%s] %s%llu %d-byte logical blocks (%s/%s)\n", |
471 | vblk->disk->disk_name, | |
472 | resize ? "new size: " : "", | |
1046d304 SH |
473 | nblocks, |
474 | queue_logical_block_size(q), | |
475 | cap_str_10, | |
476 | cap_str_2); | |
7a7c924c CH |
477 | |
478 | set_capacity(vblk->disk, capacity); | |
daf2a501 SH |
479 | } |
480 | ||
481 | static void virtblk_config_changed_work(struct work_struct *work) | |
482 | { | |
483 | struct virtio_blk *vblk = | |
484 | container_of(work, struct virtio_blk, config_work); | |
485 | char *envp[] = { "RESIZE=1", NULL }; | |
486 | ||
487 | virtblk_update_capacity(vblk, true); | |
e9986f30 | 488 | revalidate_disk(vblk->disk); |
9d9598b8 | 489 | kobject_uevent_env(&disk_to_dev(vblk->disk)->kobj, KOBJ_CHANGE, envp); |
7a7c924c CH |
490 | } |
491 | ||
492 | static void virtblk_config_changed(struct virtio_device *vdev) | |
493 | { | |
494 | struct virtio_blk *vblk = vdev->priv; | |
495 | ||
496 | queue_work(virtblk_wq, &vblk->config_work); | |
497 | } | |
498 | ||
6abd6e5a AS |
499 | static int init_vq(struct virtio_blk *vblk) |
500 | { | |
2ff98449 | 501 | int err; |
6a27b656 ML |
502 | int i; |
503 | vq_callback_t **callbacks; | |
504 | const char **names; | |
505 | struct virtqueue **vqs; | |
506 | unsigned short num_vqs; | |
507 | struct virtio_device *vdev = vblk->vdev; | |
ad71473d | 508 | struct irq_affinity desc = { 0, }; |
6a27b656 ML |
509 | |
510 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_MQ, | |
511 | struct virtio_blk_config, num_queues, | |
512 | &num_vqs); | |
513 | if (err) | |
514 | num_vqs = 1; | |
515 | ||
668866b6 | 516 | vblk->vqs = kmalloc_array(num_vqs, sizeof(*vblk->vqs), GFP_KERNEL); |
347a5293 MH |
517 | if (!vblk->vqs) |
518 | return -ENOMEM; | |
6a27b656 | 519 | |
668866b6 ME |
520 | names = kmalloc_array(num_vqs, sizeof(*names), GFP_KERNEL); |
521 | callbacks = kmalloc_array(num_vqs, sizeof(*callbacks), GFP_KERNEL); | |
522 | vqs = kmalloc_array(num_vqs, sizeof(*vqs), GFP_KERNEL); | |
347a5293 MH |
523 | if (!names || !callbacks || !vqs) { |
524 | err = -ENOMEM; | |
525 | goto out; | |
526 | } | |
6abd6e5a | 527 | |
6a27b656 ML |
528 | for (i = 0; i < num_vqs; i++) { |
529 | callbacks[i] = virtblk_done; | |
530 | snprintf(vblk->vqs[i].name, VQ_NAME_LEN, "req.%d", i); | |
531 | names[i] = vblk->vqs[i].name; | |
532 | } | |
533 | ||
534 | /* Discover virtqueues and write information to configuration. */ | |
9b2bbdb2 | 535 | err = virtio_find_vqs(vdev, num_vqs, vqs, callbacks, names, &desc); |
6a27b656 | 536 | if (err) |
347a5293 | 537 | goto out; |
6abd6e5a | 538 | |
6a27b656 ML |
539 | for (i = 0; i < num_vqs; i++) { |
540 | spin_lock_init(&vblk->vqs[i].lock); | |
541 | vblk->vqs[i].vq = vqs[i]; | |
542 | } | |
543 | vblk->num_vqs = num_vqs; | |
544 | ||
347a5293 | 545 | out: |
6a27b656 | 546 | kfree(vqs); |
6a27b656 | 547 | kfree(callbacks); |
6a27b656 | 548 | kfree(names); |
6a27b656 ML |
549 | if (err) |
550 | kfree(vblk->vqs); | |
6abd6e5a AS |
551 | return err; |
552 | } | |
553 | ||
c0aa3e09 RM |
554 | /* |
555 | * Legacy naming scheme used for virtio devices. We are stuck with it for | |
556 | * virtio blk but don't ever use it for any new driver. | |
557 | */ | |
558 | static int virtblk_name_format(char *prefix, int index, char *buf, int buflen) | |
559 | { | |
560 | const int base = 'z' - 'a' + 1; | |
561 | char *begin = buf + strlen(prefix); | |
562 | char *end = buf + buflen; | |
563 | char *p; | |
564 | int unit; | |
565 | ||
566 | p = end - 1; | |
567 | *p = '\0'; | |
568 | unit = base; | |
569 | do { | |
570 | if (p == begin) | |
571 | return -EINVAL; | |
572 | *--p = 'a' + (index % unit); | |
573 | index = (index / unit) - 1; | |
574 | } while (index >= 0); | |
575 | ||
576 | memmove(begin, p, end - p); | |
577 | memcpy(buf, prefix, strlen(prefix)); | |
578 | ||
579 | return 0; | |
580 | } | |
581 | ||
cd5d5038 PB |
582 | static int virtblk_get_cache_mode(struct virtio_device *vdev) |
583 | { | |
584 | u8 writeback; | |
585 | int err; | |
586 | ||
855e0c52 RR |
587 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_CONFIG_WCE, |
588 | struct virtio_blk_config, wce, | |
589 | &writeback); | |
592002f5 MT |
590 | |
591 | /* | |
592 | * If WCE is not configurable and flush is not available, | |
593 | * assume no writeback cache is in use. | |
594 | */ | |
cd5d5038 | 595 | if (err) |
592002f5 | 596 | writeback = virtio_has_feature(vdev, VIRTIO_BLK_F_FLUSH); |
cd5d5038 PB |
597 | |
598 | return writeback; | |
599 | } | |
600 | ||
601 | static void virtblk_update_cache_mode(struct virtio_device *vdev) | |
602 | { | |
603 | u8 writeback = virtblk_get_cache_mode(vdev); | |
604 | struct virtio_blk *vblk = vdev->priv; | |
605 | ||
ad9126ac | 606 | blk_queue_write_cache(vblk->disk->queue, writeback, false); |
cd5d5038 PB |
607 | revalidate_disk(vblk->disk); |
608 | } | |
609 | ||
610 | static const char *const virtblk_cache_types[] = { | |
611 | "write through", "write back" | |
612 | }; | |
613 | ||
614 | static ssize_t | |
e982c4d0 HR |
615 | cache_type_store(struct device *dev, struct device_attribute *attr, |
616 | const char *buf, size_t count) | |
cd5d5038 PB |
617 | { |
618 | struct gendisk *disk = dev_to_disk(dev); | |
619 | struct virtio_blk *vblk = disk->private_data; | |
620 | struct virtio_device *vdev = vblk->vdev; | |
621 | int i; | |
cd5d5038 PB |
622 | |
623 | BUG_ON(!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_CONFIG_WCE)); | |
f53d5aa0 | 624 | i = sysfs_match_string(virtblk_cache_types, buf); |
cd5d5038 | 625 | if (i < 0) |
f53d5aa0 | 626 | return i; |
cd5d5038 | 627 | |
855e0c52 | 628 | virtio_cwrite8(vdev, offsetof(struct virtio_blk_config, wce), i); |
cd5d5038 PB |
629 | virtblk_update_cache_mode(vdev); |
630 | return count; | |
631 | } | |
632 | ||
633 | static ssize_t | |
e982c4d0 | 634 | cache_type_show(struct device *dev, struct device_attribute *attr, char *buf) |
cd5d5038 PB |
635 | { |
636 | struct gendisk *disk = dev_to_disk(dev); | |
637 | struct virtio_blk *vblk = disk->private_data; | |
638 | u8 writeback = virtblk_get_cache_mode(vblk->vdev); | |
639 | ||
640 | BUG_ON(writeback >= ARRAY_SIZE(virtblk_cache_types)); | |
641 | return snprintf(buf, 40, "%s\n", virtblk_cache_types[writeback]); | |
642 | } | |
643 | ||
e982c4d0 HR |
644 | static DEVICE_ATTR_RW(cache_type); |
645 | ||
646 | static struct attribute *virtblk_attrs[] = { | |
647 | &dev_attr_serial.attr, | |
648 | &dev_attr_cache_type.attr, | |
649 | NULL, | |
650 | }; | |
651 | ||
652 | static umode_t virtblk_attrs_are_visible(struct kobject *kobj, | |
653 | struct attribute *a, int n) | |
654 | { | |
655 | struct device *dev = container_of(kobj, struct device, kobj); | |
656 | struct gendisk *disk = dev_to_disk(dev); | |
657 | struct virtio_blk *vblk = disk->private_data; | |
658 | struct virtio_device *vdev = vblk->vdev; | |
659 | ||
660 | if (a == &dev_attr_cache_type.attr && | |
661 | !virtio_has_feature(vdev, VIRTIO_BLK_F_CONFIG_WCE)) | |
662 | return S_IRUGO; | |
663 | ||
664 | return a->mode; | |
665 | } | |
666 | ||
667 | static const struct attribute_group virtblk_attr_group = { | |
668 | .attrs = virtblk_attrs, | |
669 | .is_visible = virtblk_attrs_are_visible, | |
670 | }; | |
671 | ||
672 | static const struct attribute_group *virtblk_attr_groups[] = { | |
673 | &virtblk_attr_group, | |
674 | NULL, | |
675 | }; | |
cd5d5038 | 676 | |
d6296d39 CH |
677 | static int virtblk_init_request(struct blk_mq_tag_set *set, struct request *rq, |
678 | unsigned int hctx_idx, unsigned int numa_node) | |
e9b267d9 | 679 | { |
d6296d39 | 680 | struct virtio_blk *vblk = set->driver_data; |
e9b267d9 CH |
681 | struct virtblk_req *vbr = blk_mq_rq_to_pdu(rq); |
682 | ||
97b50a65 | 683 | #ifdef CONFIG_VIRTIO_BLK_SCSI |
82ed4db4 | 684 | vbr->sreq.sense = vbr->sense; |
97b50a65 | 685 | #endif |
e9b267d9 CH |
686 | sg_init_table(vbr->sg, vblk->sg_elems); |
687 | return 0; | |
688 | } | |
689 | ||
ad71473d CH |
690 | static int virtblk_map_queues(struct blk_mq_tag_set *set) |
691 | { | |
692 | struct virtio_blk *vblk = set->driver_data; | |
693 | ||
ed76e329 | 694 | return blk_mq_virtio_map_queues(&set->map[0], vblk->vdev, 0); |
ad71473d CH |
695 | } |
696 | ||
efea2abc BVA |
697 | #ifdef CONFIG_VIRTIO_BLK_SCSI |
698 | static void virtblk_initialize_rq(struct request *req) | |
699 | { | |
700 | struct virtblk_req *vbr = blk_mq_rq_to_pdu(req); | |
701 | ||
702 | scsi_req_init(&vbr->sreq); | |
703 | } | |
704 | #endif | |
705 | ||
f363b089 | 706 | static const struct blk_mq_ops virtio_mq_ops = { |
1cf7e9c6 | 707 | .queue_rq = virtio_queue_rq, |
944e7c87 | 708 | .commit_rqs = virtio_commit_rqs, |
5124c285 | 709 | .complete = virtblk_request_done, |
24d2f903 | 710 | .init_request = virtblk_init_request, |
efea2abc BVA |
711 | #ifdef CONFIG_VIRTIO_BLK_SCSI |
712 | .initialize_rq_fn = virtblk_initialize_rq, | |
713 | #endif | |
ad71473d | 714 | .map_queues = virtblk_map_queues, |
1cf7e9c6 JA |
715 | }; |
716 | ||
24d2f903 CH |
717 | static unsigned int virtblk_queue_depth; |
718 | module_param_named(queue_depth, virtblk_queue_depth, uint, 0444); | |
1cf7e9c6 | 719 | |
8d85fce7 | 720 | static int virtblk_probe(struct virtio_device *vdev) |
e467cde2 RR |
721 | { |
722 | struct virtio_blk *vblk; | |
69740c8b | 723 | struct request_queue *q; |
5087a50e | 724 | int err, index; |
a98755c5 | 725 | |
69740c8b CH |
726 | u32 v, blk_size, sg_elems, opt_io_size; |
727 | u16 min_io_size; | |
728 | u8 physical_block_exp, alignment_offset; | |
e467cde2 | 729 | |
a4379fd8 MT |
730 | if (!vdev->config->get) { |
731 | dev_err(&vdev->dev, "%s failure: config access disabled\n", | |
732 | __func__); | |
733 | return -EINVAL; | |
734 | } | |
735 | ||
5087a50e MT |
736 | err = ida_simple_get(&vd_index_ida, 0, minor_to_index(1 << MINORBITS), |
737 | GFP_KERNEL); | |
738 | if (err < 0) | |
739 | goto out; | |
740 | index = err; | |
4f3bf19c | 741 | |
0864b79a | 742 | /* We need to know how many segments before we allocate. */ |
855e0c52 RR |
743 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_SEG_MAX, |
744 | struct virtio_blk_config, seg_max, | |
745 | &sg_elems); | |
a5b365a6 CH |
746 | |
747 | /* We need at least one SG element, whatever they say. */ | |
748 | if (err || !sg_elems) | |
0864b79a RR |
749 | sg_elems = 1; |
750 | ||
751 | /* We need an extra sg elements at head and tail. */ | |
752 | sg_elems += 2; | |
1cf7e9c6 | 753 | vdev->priv = vblk = kmalloc(sizeof(*vblk), GFP_KERNEL); |
e467cde2 RR |
754 | if (!vblk) { |
755 | err = -ENOMEM; | |
5087a50e | 756 | goto out_free_index; |
e467cde2 RR |
757 | } |
758 | ||
e467cde2 | 759 | vblk->vdev = vdev; |
0864b79a | 760 | vblk->sg_elems = sg_elems; |
a98755c5 | 761 | |
7a7c924c | 762 | INIT_WORK(&vblk->config_work, virtblk_config_changed_work); |
e467cde2 | 763 | |
6abd6e5a AS |
764 | err = init_vq(vblk); |
765 | if (err) | |
e467cde2 | 766 | goto out_free_vblk; |
e467cde2 | 767 | |
e467cde2 | 768 | /* FIXME: How many partitions? How long is a piece of string? */ |
4f3bf19c | 769 | vblk->disk = alloc_disk(1 << PART_BITS); |
e467cde2 RR |
770 | if (!vblk->disk) { |
771 | err = -ENOMEM; | |
1cf7e9c6 | 772 | goto out_free_vq; |
e467cde2 RR |
773 | } |
774 | ||
fc4324b4 | 775 | /* Default queue sizing is to fill the ring. */ |
24d2f903 | 776 | if (!virtblk_queue_depth) { |
6a27b656 | 777 | virtblk_queue_depth = vblk->vqs[0].vq->num_free; |
fc4324b4 RR |
778 | /* ... but without indirect descs, we use 2 descs per req */ |
779 | if (!virtio_has_feature(vdev, VIRTIO_RING_F_INDIRECT_DESC)) | |
24d2f903 | 780 | virtblk_queue_depth /= 2; |
fc4324b4 | 781 | } |
24d2f903 CH |
782 | |
783 | memset(&vblk->tag_set, 0, sizeof(vblk->tag_set)); | |
784 | vblk->tag_set.ops = &virtio_mq_ops; | |
24d2f903 CH |
785 | vblk->tag_set.queue_depth = virtblk_queue_depth; |
786 | vblk->tag_set.numa_node = NUMA_NO_NODE; | |
787 | vblk->tag_set.flags = BLK_MQ_F_SHOULD_MERGE; | |
788 | vblk->tag_set.cmd_size = | |
1cf7e9c6 JA |
789 | sizeof(struct virtblk_req) + |
790 | sizeof(struct scatterlist) * sg_elems; | |
24d2f903 | 791 | vblk->tag_set.driver_data = vblk; |
6a27b656 | 792 | vblk->tag_set.nr_hw_queues = vblk->num_vqs; |
1cf7e9c6 | 793 | |
24d2f903 CH |
794 | err = blk_mq_alloc_tag_set(&vblk->tag_set); |
795 | if (err) | |
796 | goto out_put_disk; | |
797 | ||
6bf6b0aa | 798 | q = blk_mq_init_queue(&vblk->tag_set); |
35b489d3 | 799 | if (IS_ERR(q)) { |
e467cde2 | 800 | err = -ENOMEM; |
24d2f903 | 801 | goto out_free_tags; |
e467cde2 | 802 | } |
6bf6b0aa | 803 | vblk->disk->queue = q; |
e467cde2 | 804 | |
69740c8b | 805 | q->queuedata = vblk; |
7d116b62 | 806 | |
c0aa3e09 | 807 | virtblk_name_format("vd", index, vblk->disk->disk_name, DISK_NAME_LEN); |
d50ed907 | 808 | |
e467cde2 | 809 | vblk->disk->major = major; |
d50ed907 | 810 | vblk->disk->first_minor = index_to_minor(index); |
e467cde2 RR |
811 | vblk->disk->private_data = vblk; |
812 | vblk->disk->fops = &virtblk_fops; | |
5fa3142d | 813 | vblk->disk->flags |= GENHD_FL_EXT_DEVT; |
5087a50e | 814 | vblk->index = index; |
4f3bf19c | 815 | |
02c42b7a | 816 | /* configure queue flush support */ |
cd5d5038 | 817 | virtblk_update_cache_mode(vdev); |
e467cde2 | 818 | |
3ef53609 CB |
819 | /* If disk is read-only in the host, the guest should obey */ |
820 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO)) | |
821 | set_disk_ro(vblk->disk, 1); | |
822 | ||
0864b79a | 823 | /* We can handle whatever the host told us to handle. */ |
ee714f2d | 824 | blk_queue_max_segments(q, vblk->sg_elems-2); |
0864b79a | 825 | |
4b7f7e20 | 826 | /* No real sector limit. */ |
ee714f2d | 827 | blk_queue_max_hw_sectors(q, -1U); |
4b7f7e20 | 828 | |
a586d4f6 RR |
829 | /* Host can optionally specify maximum segment size and number of |
830 | * segments. */ | |
855e0c52 RR |
831 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_SIZE_MAX, |
832 | struct virtio_blk_config, size_max, &v); | |
e467cde2 | 833 | if (!err) |
69740c8b | 834 | blk_queue_max_segment_size(q, v); |
4b7f7e20 | 835 | else |
69740c8b | 836 | blk_queue_max_segment_size(q, -1U); |
e467cde2 | 837 | |
066f4d82 | 838 | /* Host can optionally specify the block size of the device */ |
855e0c52 RR |
839 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_BLK_SIZE, |
840 | struct virtio_blk_config, blk_size, | |
841 | &blk_size); | |
066f4d82 | 842 | if (!err) |
69740c8b CH |
843 | blk_queue_logical_block_size(q, blk_size); |
844 | else | |
845 | blk_size = queue_logical_block_size(q); | |
846 | ||
847 | /* Use topology information if available */ | |
855e0c52 RR |
848 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY, |
849 | struct virtio_blk_config, physical_block_exp, | |
850 | &physical_block_exp); | |
69740c8b CH |
851 | if (!err && physical_block_exp) |
852 | blk_queue_physical_block_size(q, | |
853 | blk_size * (1 << physical_block_exp)); | |
854 | ||
855e0c52 RR |
855 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY, |
856 | struct virtio_blk_config, alignment_offset, | |
857 | &alignment_offset); | |
69740c8b CH |
858 | if (!err && alignment_offset) |
859 | blk_queue_alignment_offset(q, blk_size * alignment_offset); | |
860 | ||
855e0c52 RR |
861 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY, |
862 | struct virtio_blk_config, min_io_size, | |
863 | &min_io_size); | |
69740c8b CH |
864 | if (!err && min_io_size) |
865 | blk_queue_io_min(q, blk_size * min_io_size); | |
866 | ||
855e0c52 RR |
867 | err = virtio_cread_feature(vdev, VIRTIO_BLK_F_TOPOLOGY, |
868 | struct virtio_blk_config, opt_io_size, | |
869 | &opt_io_size); | |
69740c8b CH |
870 | if (!err && opt_io_size) |
871 | blk_queue_io_opt(q, blk_size * opt_io_size); | |
872 | ||
1f23816b CL |
873 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_DISCARD)) { |
874 | q->limits.discard_granularity = blk_size; | |
875 | ||
876 | virtio_cread(vdev, struct virtio_blk_config, | |
877 | discard_sector_alignment, &v); | |
878 | q->limits.discard_alignment = v ? v << SECTOR_SHIFT : 0; | |
879 | ||
880 | virtio_cread(vdev, struct virtio_blk_config, | |
881 | max_discard_sectors, &v); | |
882 | blk_queue_max_discard_sectors(q, v ? v : UINT_MAX); | |
883 | ||
884 | virtio_cread(vdev, struct virtio_blk_config, max_discard_seg, | |
885 | &v); | |
886 | blk_queue_max_discard_segments(q, | |
887 | min_not_zero(v, | |
888 | MAX_DISCARD_SEGMENTS)); | |
889 | ||
890 | blk_queue_flag_set(QUEUE_FLAG_DISCARD, q); | |
891 | } | |
892 | ||
893 | if (virtio_has_feature(vdev, VIRTIO_BLK_F_WRITE_ZEROES)) { | |
894 | virtio_cread(vdev, struct virtio_blk_config, | |
895 | max_write_zeroes_sectors, &v); | |
896 | blk_queue_max_write_zeroes_sectors(q, v ? v : UINT_MAX); | |
897 | } | |
898 | ||
daf2a501 | 899 | virtblk_update_capacity(vblk, false); |
7a11370e MT |
900 | virtio_device_ready(vdev); |
901 | ||
e982c4d0 | 902 | device_add_disk(&vdev->dev, vblk->disk, virtblk_attr_groups); |
e467cde2 RR |
903 | return 0; |
904 | ||
24d2f903 CH |
905 | out_free_tags: |
906 | blk_mq_free_tag_set(&vblk->tag_set); | |
e467cde2 RR |
907 | out_put_disk: |
908 | put_disk(vblk->disk); | |
e467cde2 | 909 | out_free_vq: |
d2a7ddda | 910 | vdev->config->del_vqs(vdev); |
e467cde2 RR |
911 | out_free_vblk: |
912 | kfree(vblk); | |
5087a50e MT |
913 | out_free_index: |
914 | ida_simple_remove(&vd_index_ida, index); | |
e467cde2 RR |
915 | out: |
916 | return err; | |
917 | } | |
918 | ||
8d85fce7 | 919 | static void virtblk_remove(struct virtio_device *vdev) |
e467cde2 RR |
920 | { |
921 | struct virtio_blk *vblk = vdev->priv; | |
5087a50e | 922 | int index = vblk->index; |
f4953fe6 | 923 | int refc; |
e467cde2 | 924 | |
cc74f719 MT |
925 | /* Make sure no work handler is accessing the device. */ |
926 | flush_work(&vblk->config_work); | |
7a7c924c | 927 | |
02e2b124 | 928 | del_gendisk(vblk->disk); |
483001c7 | 929 | blk_cleanup_queue(vblk->disk->queue); |
02e2b124 | 930 | |
24d2f903 CH |
931 | blk_mq_free_tag_set(&vblk->tag_set); |
932 | ||
6e5aa7ef RR |
933 | /* Stop all the virtqueues. */ |
934 | vdev->config->reset(vdev); | |
935 | ||
2c935bc5 | 936 | refc = kref_read(&disk_to_dev(vblk->disk)->kobj.kref); |
e467cde2 | 937 | put_disk(vblk->disk); |
d2a7ddda | 938 | vdev->config->del_vqs(vdev); |
6a27b656 | 939 | kfree(vblk->vqs); |
e467cde2 | 940 | kfree(vblk); |
f4953fe6 AG |
941 | |
942 | /* Only free device id if we don't have any users */ | |
943 | if (refc == 1) | |
944 | ida_simple_remove(&vd_index_ida, index); | |
e467cde2 RR |
945 | } |
946 | ||
89107000 | 947 | #ifdef CONFIG_PM_SLEEP |
f8fb5bc2 AS |
948 | static int virtblk_freeze(struct virtio_device *vdev) |
949 | { | |
950 | struct virtio_blk *vblk = vdev->priv; | |
951 | ||
952 | /* Ensure we don't receive any more interrupts */ | |
953 | vdev->config->reset(vdev); | |
954 | ||
cc74f719 | 955 | /* Make sure no work handler is accessing the device. */ |
f8fb5bc2 AS |
956 | flush_work(&vblk->config_work); |
957 | ||
9b3e9905 | 958 | blk_mq_quiesce_queue(vblk->disk->queue); |
f8fb5bc2 AS |
959 | |
960 | vdev->config->del_vqs(vdev); | |
961 | return 0; | |
962 | } | |
963 | ||
964 | static int virtblk_restore(struct virtio_device *vdev) | |
965 | { | |
966 | struct virtio_blk *vblk = vdev->priv; | |
967 | int ret; | |
968 | ||
f8fb5bc2 | 969 | ret = init_vq(vdev->priv); |
6d62c37f MT |
970 | if (ret) |
971 | return ret; | |
972 | ||
973 | virtio_device_ready(vdev); | |
1cf7e9c6 | 974 | |
9b3e9905 | 975 | blk_mq_unquiesce_queue(vblk->disk->queue); |
6d62c37f | 976 | return 0; |
f8fb5bc2 AS |
977 | } |
978 | #endif | |
979 | ||
47483e25 | 980 | static const struct virtio_device_id id_table[] = { |
e467cde2 RR |
981 | { VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID }, |
982 | { 0 }, | |
983 | }; | |
984 | ||
19c1c5a6 | 985 | static unsigned int features_legacy[] = { |
02c42b7a | 986 | VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY, |
97b50a65 CH |
987 | VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, |
988 | #ifdef CONFIG_VIRTIO_BLK_SCSI | |
989 | VIRTIO_BLK_F_SCSI, | |
990 | #endif | |
592002f5 | 991 | VIRTIO_BLK_F_FLUSH, VIRTIO_BLK_F_TOPOLOGY, VIRTIO_BLK_F_CONFIG_WCE, |
1f23816b | 992 | VIRTIO_BLK_F_MQ, VIRTIO_BLK_F_DISCARD, VIRTIO_BLK_F_WRITE_ZEROES, |
19c1c5a6 MT |
993 | } |
994 | ; | |
995 | static unsigned int features[] = { | |
996 | VIRTIO_BLK_F_SEG_MAX, VIRTIO_BLK_F_SIZE_MAX, VIRTIO_BLK_F_GEOMETRY, | |
997 | VIRTIO_BLK_F_RO, VIRTIO_BLK_F_BLK_SIZE, | |
592002f5 | 998 | VIRTIO_BLK_F_FLUSH, VIRTIO_BLK_F_TOPOLOGY, VIRTIO_BLK_F_CONFIG_WCE, |
1f23816b | 999 | VIRTIO_BLK_F_MQ, VIRTIO_BLK_F_DISCARD, VIRTIO_BLK_F_WRITE_ZEROES, |
c45a6816 RR |
1000 | }; |
1001 | ||
8d85fce7 | 1002 | static struct virtio_driver virtio_blk = { |
19c1c5a6 MT |
1003 | .feature_table = features, |
1004 | .feature_table_size = ARRAY_SIZE(features), | |
1005 | .feature_table_legacy = features_legacy, | |
1006 | .feature_table_size_legacy = ARRAY_SIZE(features_legacy), | |
1007 | .driver.name = KBUILD_MODNAME, | |
1008 | .driver.owner = THIS_MODULE, | |
1009 | .id_table = id_table, | |
1010 | .probe = virtblk_probe, | |
1011 | .remove = virtblk_remove, | |
1012 | .config_changed = virtblk_config_changed, | |
89107000 | 1013 | #ifdef CONFIG_PM_SLEEP |
19c1c5a6 MT |
1014 | .freeze = virtblk_freeze, |
1015 | .restore = virtblk_restore, | |
f8fb5bc2 | 1016 | #endif |
e467cde2 RR |
1017 | }; |
1018 | ||
1019 | static int __init init(void) | |
1020 | { | |
7a7c924c CH |
1021 | int error; |
1022 | ||
1023 | virtblk_wq = alloc_workqueue("virtio-blk", 0, 0); | |
1024 | if (!virtblk_wq) | |
1025 | return -ENOMEM; | |
1026 | ||
4f3bf19c | 1027 | major = register_blkdev(0, "virtblk"); |
7a7c924c CH |
1028 | if (major < 0) { |
1029 | error = major; | |
1030 | goto out_destroy_workqueue; | |
1031 | } | |
1032 | ||
1033 | error = register_virtio_driver(&virtio_blk); | |
1034 | if (error) | |
1035 | goto out_unregister_blkdev; | |
1036 | return 0; | |
1037 | ||
1038 | out_unregister_blkdev: | |
1039 | unregister_blkdev(major, "virtblk"); | |
1040 | out_destroy_workqueue: | |
1041 | destroy_workqueue(virtblk_wq); | |
1042 | return error; | |
e467cde2 RR |
1043 | } |
1044 | ||
1045 | static void __exit fini(void) | |
1046 | { | |
1047 | unregister_virtio_driver(&virtio_blk); | |
38f37b57 | 1048 | unregister_blkdev(major, "virtblk"); |
7a7c924c | 1049 | destroy_workqueue(virtblk_wq); |
e467cde2 RR |
1050 | } |
1051 | module_init(init); | |
1052 | module_exit(fini); | |
1053 | ||
1054 | MODULE_DEVICE_TABLE(virtio, id_table); | |
1055 | MODULE_DESCRIPTION("Virtio block driver"); | |
1056 | MODULE_LICENSE("GPL"); |