]>
Commit | Line | Data |
---|---|---|
6e02c38d AL |
1 | /* |
2 | * Virtio Block Device | |
3 | * | |
4 | * Copyright IBM, Corp. 2007 | |
5 | * | |
6 | * Authors: | |
7 | * Anthony Liguori <aliguori@us.ibm.com> | |
8 | * | |
9 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
10 | * the COPYING file in the top-level directory. | |
11 | * | |
12 | */ | |
13 | ||
5a61cb60 | 14 | #include "qemu-common.h" |
d75d25e3 | 15 | #include "qemu-error.h" |
6d519a5f | 16 | #include "trace.h" |
2446333c | 17 | #include "blockdev.h" |
6e02c38d | 18 | #include "virtio-blk.h" |
5bb23927 | 19 | #include "scsi-defs.h" |
1063b8b1 CH |
20 | #ifdef __linux__ |
21 | # include <scsi/sg.h> | |
22 | #endif | |
6e02c38d AL |
23 | |
24 | typedef struct VirtIOBlock | |
25 | { | |
26 | VirtIODevice vdev; | |
27 | BlockDriverState *bs; | |
28 | VirtQueue *vq; | |
869a5c6d | 29 | void *rq; |
213189ab | 30 | QEMUBH *bh; |
9752c371 | 31 | BlockConf *conf; |
12c5674b | 32 | VirtIOBlkConf *blk; |
8cfacf07 | 33 | unsigned short sector_mask; |
9d0d3138 | 34 | DeviceState *qdev; |
6e02c38d AL |
35 | } VirtIOBlock; |
36 | ||
37 | static VirtIOBlock *to_virtio_blk(VirtIODevice *vdev) | |
38 | { | |
39 | return (VirtIOBlock *)vdev; | |
40 | } | |
41 | ||
42 | typedef struct VirtIOBlockReq | |
43 | { | |
44 | VirtIOBlock *dev; | |
45 | VirtQueueElement elem; | |
46 | struct virtio_blk_inhdr *in; | |
47 | struct virtio_blk_outhdr *out; | |
1063b8b1 | 48 | struct virtio_scsi_inhdr *scsi; |
d28a1b6e | 49 | QEMUIOVector qiov; |
869a5c6d | 50 | struct VirtIOBlockReq *next; |
a597e79c | 51 | BlockAcctCookie acct; |
6e02c38d AL |
52 | } VirtIOBlockReq; |
53 | ||
869a5c6d AL |
54 | static void virtio_blk_req_complete(VirtIOBlockReq *req, int status) |
55 | { | |
56 | VirtIOBlock *s = req->dev; | |
57 | ||
6d519a5f SH |
58 | trace_virtio_blk_req_complete(req, status); |
59 | ||
92e3c2a3 | 60 | stb_p(&req->in->status, status); |
d28a1b6e | 61 | virtqueue_push(s->vq, &req->elem, req->qiov.size + sizeof(*req->in)); |
869a5c6d | 62 | virtio_notify(&s->vdev, s->vq); |
869a5c6d AL |
63 | } |
64 | ||
f35d68f0 KW |
65 | static int virtio_blk_handle_rw_error(VirtIOBlockReq *req, int error, |
66 | int is_read) | |
869a5c6d | 67 | { |
abd7f68d | 68 | BlockErrorAction action = bdrv_get_on_error(req->dev->bs, is_read); |
869a5c6d AL |
69 | VirtIOBlock *s = req->dev; |
70 | ||
eaa6c85f | 71 | if (action == BLOCK_ERR_IGNORE) { |
329c0a48 | 72 | bdrv_emit_qmp_error_event(s->bs, BDRV_ACTION_IGNORE, is_read); |
869a5c6d | 73 | return 0; |
eaa6c85f | 74 | } |
869a5c6d AL |
75 | |
76 | if ((error == ENOSPC && action == BLOCK_ERR_STOP_ENOSPC) | |
77 | || action == BLOCK_ERR_STOP_ANY) { | |
78 | req->next = s->rq; | |
79 | s->rq = req; | |
329c0a48 | 80 | bdrv_emit_qmp_error_event(s->bs, BDRV_ACTION_STOP, is_read); |
0461d5a6 | 81 | vm_stop(RUN_STATE_IO_ERROR); |
af239a62 | 82 | bdrv_iostatus_set_err(s->bs, error); |
869a5c6d AL |
83 | } else { |
84 | virtio_blk_req_complete(req, VIRTIO_BLK_S_IOERR); | |
a597e79c CH |
85 | bdrv_acct_done(s->bs, &req->acct); |
86 | g_free(req); | |
329c0a48 | 87 | bdrv_emit_qmp_error_event(s->bs, BDRV_ACTION_REPORT, is_read); |
869a5c6d AL |
88 | } |
89 | ||
90 | return 1; | |
91 | } | |
92 | ||
6e02c38d AL |
93 | static void virtio_blk_rw_complete(void *opaque, int ret) |
94 | { | |
95 | VirtIOBlockReq *req = opaque; | |
6e02c38d | 96 | |
6d519a5f SH |
97 | trace_virtio_blk_rw_complete(req, ret); |
98 | ||
f35d68f0 | 99 | if (ret) { |
92e3c2a3 | 100 | int is_read = !(ldl_p(&req->out->type) & VIRTIO_BLK_T_OUT); |
f35d68f0 | 101 | if (virtio_blk_handle_rw_error(req, -ret, is_read)) |
869a5c6d | 102 | return; |
6e02c38d AL |
103 | } |
104 | ||
f35d68f0 | 105 | virtio_blk_req_complete(req, VIRTIO_BLK_S_OK); |
a597e79c CH |
106 | bdrv_acct_done(req->dev->bs, &req->acct); |
107 | g_free(req); | |
869a5c6d | 108 | } |
6e02c38d | 109 | |
aa659be3 CH |
110 | static void virtio_blk_flush_complete(void *opaque, int ret) |
111 | { | |
112 | VirtIOBlockReq *req = opaque; | |
113 | ||
8c269b54 KW |
114 | if (ret) { |
115 | if (virtio_blk_handle_rw_error(req, -ret, 0)) { | |
116 | return; | |
117 | } | |
118 | } | |
119 | ||
120 | virtio_blk_req_complete(req, VIRTIO_BLK_S_OK); | |
a597e79c CH |
121 | bdrv_acct_done(req->dev->bs, &req->acct); |
122 | g_free(req); | |
aa659be3 CH |
123 | } |
124 | ||
869a5c6d AL |
125 | static VirtIOBlockReq *virtio_blk_alloc_request(VirtIOBlock *s) |
126 | { | |
7267c094 | 127 | VirtIOBlockReq *req = g_malloc(sizeof(*req)); |
487414f1 | 128 | req->dev = s; |
de6c8042 SH |
129 | req->qiov.size = 0; |
130 | req->next = NULL; | |
869a5c6d | 131 | return req; |
6e02c38d AL |
132 | } |
133 | ||
134 | static VirtIOBlockReq *virtio_blk_get_request(VirtIOBlock *s) | |
135 | { | |
869a5c6d | 136 | VirtIOBlockReq *req = virtio_blk_alloc_request(s); |
6e02c38d | 137 | |
869a5c6d AL |
138 | if (req != NULL) { |
139 | if (!virtqueue_pop(s->vq, &req->elem)) { | |
7267c094 | 140 | g_free(req); |
869a5c6d AL |
141 | return NULL; |
142 | } | |
6e02c38d AL |
143 | } |
144 | ||
145 | return req; | |
146 | } | |
147 | ||
1063b8b1 CH |
148 | static void virtio_blk_handle_scsi(VirtIOBlockReq *req) |
149 | { | |
4277906d | 150 | int ret; |
f34e73cd | 151 | int status = VIRTIO_BLK_S_OK; |
1063b8b1 CH |
152 | int i; |
153 | ||
154 | /* | |
155 | * We require at least one output segment each for the virtio_blk_outhdr | |
156 | * and the SCSI command block. | |
157 | * | |
158 | * We also at least require the virtio_blk_inhdr, the virtio_scsi_inhdr | |
159 | * and the sense buffer pointer in the input segments. | |
160 | */ | |
161 | if (req->elem.out_num < 2 || req->elem.in_num < 3) { | |
162 | virtio_blk_req_complete(req, VIRTIO_BLK_S_IOERR); | |
a597e79c | 163 | g_free(req); |
1063b8b1 CH |
164 | return; |
165 | } | |
166 | ||
167 | /* | |
f34e73cd PB |
168 | * The scsi inhdr is placed in the second-to-last input segment, just |
169 | * before the regular inhdr. | |
1063b8b1 | 170 | */ |
f34e73cd PB |
171 | req->scsi = (void *)req->elem.in_sg[req->elem.in_num - 2].iov_base; |
172 | ||
a6c5c84a | 173 | if (!req->dev->blk->scsi) { |
f34e73cd PB |
174 | status = VIRTIO_BLK_S_UNSUPP; |
175 | goto fail; | |
1063b8b1 CH |
176 | } |
177 | ||
178 | /* | |
f34e73cd | 179 | * No support for bidirection commands yet. |
1063b8b1 | 180 | */ |
f34e73cd PB |
181 | if (req->elem.out_num > 2 && req->elem.in_num > 3) { |
182 | status = VIRTIO_BLK_S_UNSUPP; | |
183 | goto fail; | |
184 | } | |
1063b8b1 | 185 | |
f34e73cd PB |
186 | #ifdef __linux__ |
187 | struct sg_io_hdr hdr; | |
1063b8b1 CH |
188 | memset(&hdr, 0, sizeof(struct sg_io_hdr)); |
189 | hdr.interface_id = 'S'; | |
190 | hdr.cmd_len = req->elem.out_sg[1].iov_len; | |
191 | hdr.cmdp = req->elem.out_sg[1].iov_base; | |
192 | hdr.dxfer_len = 0; | |
193 | ||
194 | if (req->elem.out_num > 2) { | |
195 | /* | |
196 | * If there are more than the minimally required 2 output segments | |
197 | * there is write payload starting from the third iovec. | |
198 | */ | |
199 | hdr.dxfer_direction = SG_DXFER_TO_DEV; | |
200 | hdr.iovec_count = req->elem.out_num - 2; | |
201 | ||
202 | for (i = 0; i < hdr.iovec_count; i++) | |
203 | hdr.dxfer_len += req->elem.out_sg[i + 2].iov_len; | |
204 | ||
205 | hdr.dxferp = req->elem.out_sg + 2; | |
206 | ||
207 | } else if (req->elem.in_num > 3) { | |
208 | /* | |
209 | * If we have more than 3 input segments the guest wants to actually | |
210 | * read data. | |
211 | */ | |
212 | hdr.dxfer_direction = SG_DXFER_FROM_DEV; | |
213 | hdr.iovec_count = req->elem.in_num - 3; | |
214 | for (i = 0; i < hdr.iovec_count; i++) | |
215 | hdr.dxfer_len += req->elem.in_sg[i].iov_len; | |
216 | ||
217 | hdr.dxferp = req->elem.in_sg; | |
1063b8b1 CH |
218 | } else { |
219 | /* | |
220 | * Some SCSI commands don't actually transfer any data. | |
221 | */ | |
222 | hdr.dxfer_direction = SG_DXFER_NONE; | |
223 | } | |
224 | ||
225 | hdr.sbp = req->elem.in_sg[req->elem.in_num - 3].iov_base; | |
226 | hdr.mx_sb_len = req->elem.in_sg[req->elem.in_num - 3].iov_len; | |
1063b8b1 CH |
227 | |
228 | ret = bdrv_ioctl(req->dev->bs, SG_IO, &hdr); | |
229 | if (ret) { | |
230 | status = VIRTIO_BLK_S_UNSUPP; | |
f34e73cd | 231 | goto fail; |
1063b8b1 CH |
232 | } |
233 | ||
5bb23927 PB |
234 | /* |
235 | * From SCSI-Generic-HOWTO: "Some lower level drivers (e.g. ide-scsi) | |
236 | * clear the masked_status field [hence status gets cleared too, see | |
237 | * block/scsi_ioctl.c] even when a CHECK_CONDITION or COMMAND_TERMINATED | |
238 | * status has occurred. However they do set DRIVER_SENSE in driver_status | |
239 | * field. Also a (sb_len_wr > 0) indicates there is a sense buffer. | |
240 | */ | |
241 | if (hdr.status == 0 && hdr.sb_len_wr > 0) { | |
242 | hdr.status = CHECK_CONDITION; | |
243 | } | |
244 | ||
245 | stl_p(&req->scsi->errors, | |
246 | hdr.status | (hdr.msg_status << 8) | | |
247 | (hdr.host_status << 16) | (hdr.driver_status << 24)); | |
92e3c2a3 AJ |
248 | stl_p(&req->scsi->residual, hdr.resid); |
249 | stl_p(&req->scsi->sense_len, hdr.sb_len_wr); | |
250 | stl_p(&req->scsi->data_len, hdr.dxfer_len); | |
1063b8b1 CH |
251 | |
252 | virtio_blk_req_complete(req, status); | |
a597e79c | 253 | g_free(req); |
1063b8b1 | 254 | #else |
f34e73cd PB |
255 | abort(); |
256 | #endif | |
257 | ||
258 | fail: | |
259 | /* Just put anything nonzero so that the ioctl fails in the guest. */ | |
260 | stl_p(&req->scsi->errors, 255); | |
261 | virtio_blk_req_complete(req, status); | |
a597e79c | 262 | g_free(req); |
1063b8b1 | 263 | } |
1063b8b1 | 264 | |
c20fd872 CH |
265 | typedef struct MultiReqBuffer { |
266 | BlockRequest blkreq[32]; | |
267 | unsigned int num_writes; | |
268 | } MultiReqBuffer; | |
269 | ||
270 | static void virtio_submit_multiwrite(BlockDriverState *bs, MultiReqBuffer *mrb) | |
869a5c6d | 271 | { |
91553dcc | 272 | int i, ret; |
91553dcc | 273 | |
c20fd872 CH |
274 | if (!mrb->num_writes) { |
275 | return; | |
276 | } | |
277 | ||
278 | ret = bdrv_aio_multiwrite(bs, mrb->blkreq, mrb->num_writes); | |
91553dcc | 279 | if (ret != 0) { |
c20fd872 CH |
280 | for (i = 0; i < mrb->num_writes; i++) { |
281 | if (mrb->blkreq[i].error) { | |
282 | virtio_blk_rw_complete(mrb->blkreq[i].opaque, -EIO); | |
91553dcc KW |
283 | } |
284 | } | |
285 | } | |
c20fd872 CH |
286 | |
287 | mrb->num_writes = 0; | |
91553dcc | 288 | } |
87b245db | 289 | |
c20fd872 | 290 | static void virtio_blk_handle_flush(VirtIOBlockReq *req, MultiReqBuffer *mrb) |
aa659be3 | 291 | { |
a597e79c CH |
292 | bdrv_acct_start(req->dev->bs, &req->acct, 0, BDRV_ACCT_FLUSH); |
293 | ||
618fbb84 CH |
294 | /* |
295 | * Make sure all outstanding writes are posted to the backing device. | |
296 | */ | |
c20fd872 | 297 | virtio_submit_multiwrite(req->dev->bs, mrb); |
ad54ae80 | 298 | bdrv_aio_flush(req->dev->bs, virtio_blk_flush_complete, req); |
aa659be3 CH |
299 | } |
300 | ||
c20fd872 | 301 | static void virtio_blk_handle_write(VirtIOBlockReq *req, MultiReqBuffer *mrb) |
91553dcc | 302 | { |
c20fd872 | 303 | BlockRequest *blkreq; |
92e3c2a3 | 304 | uint64_t sector; |
c20fd872 | 305 | |
92e3c2a3 | 306 | sector = ldq_p(&req->out->sector); |
6d519a5f | 307 | |
a597e79c CH |
308 | bdrv_acct_start(req->dev->bs, &req->acct, req->qiov.size, BDRV_ACCT_WRITE); |
309 | ||
92e3c2a3 AJ |
310 | trace_virtio_blk_handle_write(req, sector, req->qiov.size / 512); |
311 | ||
312 | if (sector & req->dev->sector_mask) { | |
8cfacf07 CH |
313 | virtio_blk_rw_complete(req, -EIO); |
314 | return; | |
315 | } | |
52c05023 CH |
316 | if (req->qiov.size % req->dev->conf->logical_block_size) { |
317 | virtio_blk_rw_complete(req, -EIO); | |
318 | return; | |
319 | } | |
8cfacf07 | 320 | |
c20fd872 CH |
321 | if (mrb->num_writes == 32) { |
322 | virtio_submit_multiwrite(req->dev->bs, mrb); | |
87b245db | 323 | } |
91553dcc | 324 | |
c20fd872 | 325 | blkreq = &mrb->blkreq[mrb->num_writes]; |
92e3c2a3 | 326 | blkreq->sector = sector; |
c20fd872 CH |
327 | blkreq->nb_sectors = req->qiov.size / BDRV_SECTOR_SIZE; |
328 | blkreq->qiov = &req->qiov; | |
329 | blkreq->cb = virtio_blk_rw_complete; | |
330 | blkreq->opaque = req; | |
331 | blkreq->error = 0; | |
91553dcc | 332 | |
c20fd872 | 333 | mrb->num_writes++; |
d28a1b6e | 334 | } |
869a5c6d | 335 | |
d28a1b6e AL |
336 | static void virtio_blk_handle_read(VirtIOBlockReq *req) |
337 | { | |
92e3c2a3 AJ |
338 | uint64_t sector; |
339 | ||
340 | sector = ldq_p(&req->out->sector); | |
87b245db | 341 | |
a597e79c CH |
342 | bdrv_acct_start(req->dev->bs, &req->acct, req->qiov.size, BDRV_ACCT_READ); |
343 | ||
81b6b9fa SH |
344 | trace_virtio_blk_handle_read(req, sector, req->qiov.size / 512); |
345 | ||
92e3c2a3 | 346 | if (sector & req->dev->sector_mask) { |
8cfacf07 CH |
347 | virtio_blk_rw_complete(req, -EIO); |
348 | return; | |
349 | } | |
52c05023 CH |
350 | if (req->qiov.size % req->dev->conf->logical_block_size) { |
351 | virtio_blk_rw_complete(req, -EIO); | |
352 | return; | |
353 | } | |
ad54ae80 PB |
354 | bdrv_aio_readv(req->dev->bs, sector, &req->qiov, |
355 | req->qiov.size / BDRV_SECTOR_SIZE, | |
356 | virtio_blk_rw_complete, req); | |
869a5c6d AL |
357 | } |
358 | ||
bc6694d4 KW |
359 | static void virtio_blk_handle_request(VirtIOBlockReq *req, |
360 | MultiReqBuffer *mrb) | |
361 | { | |
92e3c2a3 AJ |
362 | uint32_t type; |
363 | ||
bc6694d4 | 364 | if (req->elem.out_num < 1 || req->elem.in_num < 1) { |
870cef1d | 365 | error_report("virtio-blk missing headers"); |
bc6694d4 KW |
366 | exit(1); |
367 | } | |
368 | ||
369 | if (req->elem.out_sg[0].iov_len < sizeof(*req->out) || | |
370 | req->elem.in_sg[req->elem.in_num - 1].iov_len < sizeof(*req->in)) { | |
870cef1d | 371 | error_report("virtio-blk header not in correct element"); |
bc6694d4 KW |
372 | exit(1); |
373 | } | |
374 | ||
375 | req->out = (void *)req->elem.out_sg[0].iov_base; | |
376 | req->in = (void *)req->elem.in_sg[req->elem.in_num - 1].iov_base; | |
377 | ||
92e3c2a3 AJ |
378 | type = ldl_p(&req->out->type); |
379 | ||
380 | if (type & VIRTIO_BLK_T_FLUSH) { | |
c20fd872 | 381 | virtio_blk_handle_flush(req, mrb); |
92e3c2a3 | 382 | } else if (type & VIRTIO_BLK_T_SCSI_CMD) { |
bc6694d4 | 383 | virtio_blk_handle_scsi(req); |
92e3c2a3 | 384 | } else if (type & VIRTIO_BLK_T_GET_ID) { |
2930b313 | 385 | VirtIOBlock *s = req->dev; |
386 | ||
a8686a9b MA |
387 | /* |
388 | * NB: per existing s/n string convention the string is | |
389 | * terminated by '\0' only when shorter than buffer. | |
390 | */ | |
391 | strncpy(req->elem.in_sg[0].iov_base, | |
12c5674b | 392 | s->blk->serial ? s->blk->serial : "", |
a8686a9b | 393 | MIN(req->elem.in_sg[0].iov_len, VIRTIO_BLK_ID_BYTES)); |
2930b313 | 394 | virtio_blk_req_complete(req, VIRTIO_BLK_S_OK); |
a597e79c | 395 | g_free(req); |
92e3c2a3 | 396 | } else if (type & VIRTIO_BLK_T_OUT) { |
bc6694d4 KW |
397 | qemu_iovec_init_external(&req->qiov, &req->elem.out_sg[1], |
398 | req->elem.out_num - 1); | |
c20fd872 | 399 | virtio_blk_handle_write(req, mrb); |
bc6694d4 KW |
400 | } else { |
401 | qemu_iovec_init_external(&req->qiov, &req->elem.in_sg[0], | |
402 | req->elem.in_num - 1); | |
403 | virtio_blk_handle_read(req); | |
404 | } | |
405 | } | |
406 | ||
6e02c38d AL |
407 | static void virtio_blk_handle_output(VirtIODevice *vdev, VirtQueue *vq) |
408 | { | |
409 | VirtIOBlock *s = to_virtio_blk(vdev); | |
410 | VirtIOBlockReq *req; | |
bc6694d4 KW |
411 | MultiReqBuffer mrb = { |
412 | .num_writes = 0, | |
bc6694d4 | 413 | }; |
6e02c38d AL |
414 | |
415 | while ((req = virtio_blk_get_request(s))) { | |
bc6694d4 | 416 | virtio_blk_handle_request(req, &mrb); |
6e02c38d | 417 | } |
91553dcc | 418 | |
c20fd872 | 419 | virtio_submit_multiwrite(s->bs, &mrb); |
91553dcc | 420 | |
6e02c38d AL |
421 | /* |
422 | * FIXME: Want to check for completions before returning to guest mode, | |
423 | * so cached reads and writes are reported as quickly as possible. But | |
424 | * that should be done in the generic block layer. | |
425 | */ | |
426 | } | |
427 | ||
213189ab | 428 | static void virtio_blk_dma_restart_bh(void *opaque) |
869a5c6d AL |
429 | { |
430 | VirtIOBlock *s = opaque; | |
431 | VirtIOBlockReq *req = s->rq; | |
f1b52868 KW |
432 | MultiReqBuffer mrb = { |
433 | .num_writes = 0, | |
f1b52868 | 434 | }; |
869a5c6d | 435 | |
213189ab MA |
436 | qemu_bh_delete(s->bh); |
437 | s->bh = NULL; | |
869a5c6d AL |
438 | |
439 | s->rq = NULL; | |
440 | ||
441 | while (req) { | |
f1b52868 | 442 | virtio_blk_handle_request(req, &mrb); |
869a5c6d AL |
443 | req = req->next; |
444 | } | |
f1b52868 | 445 | |
c20fd872 | 446 | virtio_submit_multiwrite(s->bs, &mrb); |
869a5c6d AL |
447 | } |
448 | ||
1dfb4dd9 LC |
449 | static void virtio_blk_dma_restart_cb(void *opaque, int running, |
450 | RunState state) | |
213189ab MA |
451 | { |
452 | VirtIOBlock *s = opaque; | |
453 | ||
454 | if (!running) | |
455 | return; | |
456 | ||
457 | if (!s->bh) { | |
458 | s->bh = qemu_bh_new(virtio_blk_dma_restart_bh, s); | |
459 | qemu_bh_schedule(s->bh); | |
460 | } | |
461 | } | |
462 | ||
6e02c38d AL |
463 | static void virtio_blk_reset(VirtIODevice *vdev) |
464 | { | |
465 | /* | |
466 | * This should cancel pending requests, but can't do nicely until there | |
467 | * are per-device request lists. | |
468 | */ | |
922453bc | 469 | bdrv_drain_all(); |
6e02c38d AL |
470 | } |
471 | ||
bf011293 | 472 | /* coalesce internal state, copy to pci i/o region 0 |
473 | */ | |
6e02c38d AL |
474 | static void virtio_blk_update_config(VirtIODevice *vdev, uint8_t *config) |
475 | { | |
476 | VirtIOBlock *s = to_virtio_blk(vdev); | |
477 | struct virtio_blk_config blkcfg; | |
478 | uint64_t capacity; | |
479 | int cylinders, heads, secs; | |
3a395142 | 480 | int blk_size = s->conf->logical_block_size; |
6e02c38d AL |
481 | |
482 | bdrv_get_geometry(s->bs, &capacity); | |
483 | bdrv_get_geometry_hint(s->bs, &cylinders, &heads, &secs); | |
5c5dafdc | 484 | memset(&blkcfg, 0, sizeof(blkcfg)); |
6e02c38d AL |
485 | stq_raw(&blkcfg.capacity, capacity); |
486 | stl_raw(&blkcfg.seg_max, 128 - 2); | |
487 | stw_raw(&blkcfg.cylinders, cylinders); | |
3a395142 PB |
488 | stl_raw(&blkcfg.blk_size, blk_size); |
489 | stw_raw(&blkcfg.min_io_size, s->conf->min_io_size / blk_size); | |
490 | stw_raw(&blkcfg.opt_io_size, s->conf->opt_io_size / blk_size); | |
6e02c38d | 491 | blkcfg.heads = heads; |
8cfacf07 | 492 | blkcfg.sectors = secs & ~s->sector_mask; |
c7085da7 | 493 | blkcfg.size_max = 0; |
9752c371 CH |
494 | blkcfg.physical_block_exp = get_physical_block_exp(s->conf); |
495 | blkcfg.alignment_offset = 0; | |
37d5ddd6 | 496 | memcpy(config, &blkcfg, sizeof(struct virtio_blk_config)); |
6e02c38d AL |
497 | } |
498 | ||
8172539d | 499 | static uint32_t virtio_blk_get_features(VirtIODevice *vdev, uint32_t features) |
6e02c38d | 500 | { |
bf011293 | 501 | VirtIOBlock *s = to_virtio_blk(vdev); |
1063b8b1 CH |
502 | |
503 | features |= (1 << VIRTIO_BLK_F_SEG_MAX); | |
504 | features |= (1 << VIRTIO_BLK_F_GEOMETRY); | |
9752c371 | 505 | features |= (1 << VIRTIO_BLK_F_TOPOLOGY); |
8cfacf07 | 506 | features |= (1 << VIRTIO_BLK_F_BLK_SIZE); |
a6c5c84a | 507 | features |= (1 << VIRTIO_BLK_F_SCSI); |
aa659be3 CH |
508 | |
509 | if (bdrv_enable_write_cache(s->bs)) | |
510 | features |= (1 << VIRTIO_BLK_F_WCACHE); | |
c79662f7 NS |
511 | |
512 | if (bdrv_is_read_only(s->bs)) | |
513 | features |= 1 << VIRTIO_BLK_F_RO; | |
1063b8b1 CH |
514 | |
515 | return features; | |
6e02c38d AL |
516 | } |
517 | ||
518 | static void virtio_blk_save(QEMUFile *f, void *opaque) | |
519 | { | |
520 | VirtIOBlock *s = opaque; | |
869a5c6d AL |
521 | VirtIOBlockReq *req = s->rq; |
522 | ||
6e02c38d | 523 | virtio_save(&s->vdev, f); |
869a5c6d AL |
524 | |
525 | while (req) { | |
526 | qemu_put_sbyte(f, 1); | |
527 | qemu_put_buffer(f, (unsigned char*)&req->elem, sizeof(req->elem)); | |
528 | req = req->next; | |
529 | } | |
530 | qemu_put_sbyte(f, 0); | |
6e02c38d AL |
531 | } |
532 | ||
533 | static int virtio_blk_load(QEMUFile *f, void *opaque, int version_id) | |
534 | { | |
535 | VirtIOBlock *s = opaque; | |
2a633c46 | 536 | int ret; |
6e02c38d | 537 | |
869a5c6d | 538 | if (version_id != 2) |
6e02c38d AL |
539 | return -EINVAL; |
540 | ||
2a633c46 OW |
541 | ret = virtio_load(&s->vdev, f); |
542 | if (ret) { | |
543 | return ret; | |
544 | } | |
545 | ||
869a5c6d AL |
546 | while (qemu_get_sbyte(f)) { |
547 | VirtIOBlockReq *req = virtio_blk_alloc_request(s); | |
548 | qemu_get_buffer(f, (unsigned char*)&req->elem, sizeof(req->elem)); | |
549 | req->next = s->rq; | |
20a81e4d | 550 | s->rq = req; |
b6a4805b KW |
551 | |
552 | virtqueue_map_sg(req->elem.in_sg, req->elem.in_addr, | |
553 | req->elem.in_num, 1); | |
554 | virtqueue_map_sg(req->elem.out_sg, req->elem.out_addr, | |
555 | req->elem.out_num, 0); | |
869a5c6d | 556 | } |
6e02c38d AL |
557 | |
558 | return 0; | |
559 | } | |
560 | ||
145feb17 | 561 | static void virtio_blk_resize(void *opaque) |
e5051fc7 CH |
562 | { |
563 | VirtIOBlock *s = opaque; | |
564 | ||
145feb17 | 565 | virtio_notify_config(&s->vdev); |
e5051fc7 CH |
566 | } |
567 | ||
0e49de52 | 568 | static const BlockDevOps virtio_block_ops = { |
145feb17 | 569 | .resize_cb = virtio_blk_resize, |
0e49de52 MA |
570 | }; |
571 | ||
12c5674b | 572 | VirtIODevice *virtio_blk_init(DeviceState *dev, VirtIOBlkConf *blk) |
6e02c38d AL |
573 | { |
574 | VirtIOBlock *s; | |
575 | int cylinders, heads, secs; | |
576 | static int virtio_blk_id; | |
2930b313 | 577 | DriveInfo *dinfo; |
cf21e106 | 578 | |
12c5674b | 579 | if (!blk->conf.bs) { |
6a84cb1f | 580 | error_report("drive property not set"); |
d75d25e3 MA |
581 | return NULL; |
582 | } | |
12c5674b | 583 | if (!bdrv_is_inserted(blk->conf.bs)) { |
98f28ad7 MA |
584 | error_report("Device needs media, but drive is empty"); |
585 | return NULL; | |
586 | } | |
d75d25e3 | 587 | |
12c5674b | 588 | if (!blk->serial) { |
a8686a9b | 589 | /* try to fall back to value set with legacy -drive serial=... */ |
12c5674b | 590 | dinfo = drive_get_by_blockdev(blk->conf.bs); |
a8686a9b | 591 | if (*dinfo->serial) { |
12c5674b | 592 | blk->serial = strdup(dinfo->serial); |
a8686a9b MA |
593 | } |
594 | } | |
595 | ||
53c25cea | 596 | s = (VirtIOBlock *)virtio_common_init("virtio-blk", VIRTIO_ID_BLOCK, |
37d5ddd6 | 597 | sizeof(struct virtio_blk_config), |
53c25cea | 598 | sizeof(VirtIOBlock)); |
6e02c38d AL |
599 | |
600 | s->vdev.get_config = virtio_blk_update_config; | |
601 | s->vdev.get_features = virtio_blk_get_features; | |
602 | s->vdev.reset = virtio_blk_reset; | |
12c5674b PB |
603 | s->bs = blk->conf.bs; |
604 | s->conf = &blk->conf; | |
605 | s->blk = blk; | |
869a5c6d | 606 | s->rq = NULL; |
1573a35d | 607 | s->sector_mask = (s->conf->logical_block_size / BDRV_SECTOR_SIZE) - 1; |
6e02c38d | 608 | bdrv_guess_geometry(s->bs, &cylinders, &heads, &secs); |
6e02c38d AL |
609 | |
610 | s->vq = virtio_add_queue(&s->vdev, 128, virtio_blk_handle_output); | |
611 | ||
869a5c6d | 612 | qemu_add_vm_change_state_handler(virtio_blk_dma_restart_cb, s); |
9d0d3138 | 613 | s->qdev = dev; |
0be71e32 | 614 | register_savevm(dev, "virtio-blk", virtio_blk_id++, 2, |
6e02c38d | 615 | virtio_blk_save, virtio_blk_load, s); |
0e49de52 | 616 | bdrv_set_dev_ops(s->bs, &virtio_block_ops, s); |
12c5674b | 617 | bdrv_set_buffer_alignment(s->bs, s->conf->logical_block_size); |
6e02c38d | 618 | |
af239a62 | 619 | bdrv_iostatus_enable(s->bs); |
12c5674b | 620 | add_boot_device_path(s->conf->bootindex, dev, "/disk@0,0"); |
1ca4d09a | 621 | |
53c25cea | 622 | return &s->vdev; |
6e02c38d | 623 | } |
9d0d3138 AW |
624 | |
625 | void virtio_blk_exit(VirtIODevice *vdev) | |
626 | { | |
627 | VirtIOBlock *s = to_virtio_blk(vdev); | |
628 | unregister_savevm(s->qdev, "virtio-blk", s); | |
0e47931b | 629 | blockdev_mark_auto_del(s->bs); |
d92551f2 | 630 | virtio_cleanup(vdev); |
9d0d3138 | 631 | } |