]>
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" |
1de7afc9 | 15 | #include "qemu/error-report.h" |
6d519a5f | 16 | #include "trace.h" |
9db1c0f7 | 17 | #include "hw/block-common.h" |
9c17d615 | 18 | #include "sysemu/blockdev.h" |
83c9f4ca | 19 | #include "hw/virtio-blk.h" |
83c9f4ca | 20 | #include "hw/scsi-defs.h" |
1063b8b1 CH |
21 | #ifdef __linux__ |
22 | # include <scsi/sg.h> | |
23 | #endif | |
6e02c38d | 24 | |
6e02c38d AL |
25 | static VirtIOBlock *to_virtio_blk(VirtIODevice *vdev) |
26 | { | |
27 | return (VirtIOBlock *)vdev; | |
28 | } | |
29 | ||
30 | typedef struct VirtIOBlockReq | |
31 | { | |
32 | VirtIOBlock *dev; | |
33 | VirtQueueElement elem; | |
34 | struct virtio_blk_inhdr *in; | |
35 | struct virtio_blk_outhdr *out; | |
1063b8b1 | 36 | struct virtio_scsi_inhdr *scsi; |
d28a1b6e | 37 | QEMUIOVector qiov; |
869a5c6d | 38 | struct VirtIOBlockReq *next; |
a597e79c | 39 | BlockAcctCookie acct; |
6e02c38d AL |
40 | } VirtIOBlockReq; |
41 | ||
869a5c6d AL |
42 | static void virtio_blk_req_complete(VirtIOBlockReq *req, int status) |
43 | { | |
44 | VirtIOBlock *s = req->dev; | |
45 | ||
6d519a5f SH |
46 | trace_virtio_blk_req_complete(req, status); |
47 | ||
92e3c2a3 | 48 | stb_p(&req->in->status, status); |
d28a1b6e | 49 | virtqueue_push(s->vq, &req->elem, req->qiov.size + sizeof(*req->in)); |
869a5c6d | 50 | virtio_notify(&s->vdev, s->vq); |
869a5c6d AL |
51 | } |
52 | ||
f35d68f0 | 53 | static int virtio_blk_handle_rw_error(VirtIOBlockReq *req, int error, |
1ceee0d5 | 54 | bool is_read) |
869a5c6d | 55 | { |
3e1caa5f | 56 | BlockErrorAction action = bdrv_get_error_action(req->dev->bs, is_read, error); |
869a5c6d AL |
57 | VirtIOBlock *s = req->dev; |
58 | ||
3e1caa5f | 59 | if (action == BDRV_ACTION_STOP) { |
869a5c6d AL |
60 | req->next = s->rq; |
61 | s->rq = req; | |
3e1caa5f | 62 | } else if (action == BDRV_ACTION_REPORT) { |
869a5c6d | 63 | virtio_blk_req_complete(req, VIRTIO_BLK_S_IOERR); |
a597e79c CH |
64 | bdrv_acct_done(s->bs, &req->acct); |
65 | g_free(req); | |
869a5c6d AL |
66 | } |
67 | ||
3e1caa5f PB |
68 | bdrv_error_action(s->bs, action, is_read, error); |
69 | return action != BDRV_ACTION_IGNORE; | |
869a5c6d AL |
70 | } |
71 | ||
6e02c38d AL |
72 | static void virtio_blk_rw_complete(void *opaque, int ret) |
73 | { | |
74 | VirtIOBlockReq *req = opaque; | |
6e02c38d | 75 | |
6d519a5f SH |
76 | trace_virtio_blk_rw_complete(req, ret); |
77 | ||
f35d68f0 | 78 | if (ret) { |
1ceee0d5 | 79 | bool is_read = !(ldl_p(&req->out->type) & VIRTIO_BLK_T_OUT); |
f35d68f0 | 80 | if (virtio_blk_handle_rw_error(req, -ret, is_read)) |
869a5c6d | 81 | return; |
6e02c38d AL |
82 | } |
83 | ||
f35d68f0 | 84 | virtio_blk_req_complete(req, VIRTIO_BLK_S_OK); |
a597e79c CH |
85 | bdrv_acct_done(req->dev->bs, &req->acct); |
86 | g_free(req); | |
869a5c6d | 87 | } |
6e02c38d | 88 | |
aa659be3 CH |
89 | static void virtio_blk_flush_complete(void *opaque, int ret) |
90 | { | |
91 | VirtIOBlockReq *req = opaque; | |
92 | ||
8c269b54 KW |
93 | if (ret) { |
94 | if (virtio_blk_handle_rw_error(req, -ret, 0)) { | |
95 | return; | |
96 | } | |
97 | } | |
98 | ||
99 | virtio_blk_req_complete(req, VIRTIO_BLK_S_OK); | |
a597e79c CH |
100 | bdrv_acct_done(req->dev->bs, &req->acct); |
101 | g_free(req); | |
aa659be3 CH |
102 | } |
103 | ||
869a5c6d AL |
104 | static VirtIOBlockReq *virtio_blk_alloc_request(VirtIOBlock *s) |
105 | { | |
7267c094 | 106 | VirtIOBlockReq *req = g_malloc(sizeof(*req)); |
487414f1 | 107 | req->dev = s; |
de6c8042 SH |
108 | req->qiov.size = 0; |
109 | req->next = NULL; | |
869a5c6d | 110 | return req; |
6e02c38d AL |
111 | } |
112 | ||
113 | static VirtIOBlockReq *virtio_blk_get_request(VirtIOBlock *s) | |
114 | { | |
869a5c6d | 115 | VirtIOBlockReq *req = virtio_blk_alloc_request(s); |
6e02c38d | 116 | |
869a5c6d AL |
117 | if (req != NULL) { |
118 | if (!virtqueue_pop(s->vq, &req->elem)) { | |
7267c094 | 119 | g_free(req); |
869a5c6d AL |
120 | return NULL; |
121 | } | |
6e02c38d AL |
122 | } |
123 | ||
124 | return req; | |
125 | } | |
126 | ||
1063b8b1 CH |
127 | static void virtio_blk_handle_scsi(VirtIOBlockReq *req) |
128 | { | |
47ce9ef7 | 129 | #ifdef __linux__ |
4277906d | 130 | int ret; |
1063b8b1 | 131 | int i; |
47ce9ef7 SW |
132 | #endif |
133 | int status = VIRTIO_BLK_S_OK; | |
1063b8b1 CH |
134 | |
135 | /* | |
136 | * We require at least one output segment each for the virtio_blk_outhdr | |
137 | * and the SCSI command block. | |
138 | * | |
139 | * We also at least require the virtio_blk_inhdr, the virtio_scsi_inhdr | |
140 | * and the sense buffer pointer in the input segments. | |
141 | */ | |
142 | if (req->elem.out_num < 2 || req->elem.in_num < 3) { | |
143 | virtio_blk_req_complete(req, VIRTIO_BLK_S_IOERR); | |
a597e79c | 144 | g_free(req); |
1063b8b1 CH |
145 | return; |
146 | } | |
147 | ||
148 | /* | |
f34e73cd PB |
149 | * The scsi inhdr is placed in the second-to-last input segment, just |
150 | * before the regular inhdr. | |
1063b8b1 | 151 | */ |
f34e73cd PB |
152 | req->scsi = (void *)req->elem.in_sg[req->elem.in_num - 2].iov_base; |
153 | ||
da3dcefa | 154 | if (!req->dev->blk.scsi) { |
f34e73cd PB |
155 | status = VIRTIO_BLK_S_UNSUPP; |
156 | goto fail; | |
1063b8b1 CH |
157 | } |
158 | ||
159 | /* | |
f34e73cd | 160 | * No support for bidirection commands yet. |
1063b8b1 | 161 | */ |
f34e73cd PB |
162 | if (req->elem.out_num > 2 && req->elem.in_num > 3) { |
163 | status = VIRTIO_BLK_S_UNSUPP; | |
164 | goto fail; | |
165 | } | |
1063b8b1 | 166 | |
f34e73cd PB |
167 | #ifdef __linux__ |
168 | struct sg_io_hdr hdr; | |
1063b8b1 CH |
169 | memset(&hdr, 0, sizeof(struct sg_io_hdr)); |
170 | hdr.interface_id = 'S'; | |
171 | hdr.cmd_len = req->elem.out_sg[1].iov_len; | |
172 | hdr.cmdp = req->elem.out_sg[1].iov_base; | |
173 | hdr.dxfer_len = 0; | |
174 | ||
175 | if (req->elem.out_num > 2) { | |
176 | /* | |
177 | * If there are more than the minimally required 2 output segments | |
178 | * there is write payload starting from the third iovec. | |
179 | */ | |
180 | hdr.dxfer_direction = SG_DXFER_TO_DEV; | |
181 | hdr.iovec_count = req->elem.out_num - 2; | |
182 | ||
183 | for (i = 0; i < hdr.iovec_count; i++) | |
184 | hdr.dxfer_len += req->elem.out_sg[i + 2].iov_len; | |
185 | ||
186 | hdr.dxferp = req->elem.out_sg + 2; | |
187 | ||
188 | } else if (req->elem.in_num > 3) { | |
189 | /* | |
190 | * If we have more than 3 input segments the guest wants to actually | |
191 | * read data. | |
192 | */ | |
193 | hdr.dxfer_direction = SG_DXFER_FROM_DEV; | |
194 | hdr.iovec_count = req->elem.in_num - 3; | |
195 | for (i = 0; i < hdr.iovec_count; i++) | |
196 | hdr.dxfer_len += req->elem.in_sg[i].iov_len; | |
197 | ||
198 | hdr.dxferp = req->elem.in_sg; | |
1063b8b1 CH |
199 | } else { |
200 | /* | |
201 | * Some SCSI commands don't actually transfer any data. | |
202 | */ | |
203 | hdr.dxfer_direction = SG_DXFER_NONE; | |
204 | } | |
205 | ||
206 | hdr.sbp = req->elem.in_sg[req->elem.in_num - 3].iov_base; | |
207 | hdr.mx_sb_len = req->elem.in_sg[req->elem.in_num - 3].iov_len; | |
1063b8b1 CH |
208 | |
209 | ret = bdrv_ioctl(req->dev->bs, SG_IO, &hdr); | |
210 | if (ret) { | |
211 | status = VIRTIO_BLK_S_UNSUPP; | |
f34e73cd | 212 | goto fail; |
1063b8b1 CH |
213 | } |
214 | ||
5bb23927 PB |
215 | /* |
216 | * From SCSI-Generic-HOWTO: "Some lower level drivers (e.g. ide-scsi) | |
217 | * clear the masked_status field [hence status gets cleared too, see | |
218 | * block/scsi_ioctl.c] even when a CHECK_CONDITION or COMMAND_TERMINATED | |
219 | * status has occurred. However they do set DRIVER_SENSE in driver_status | |
220 | * field. Also a (sb_len_wr > 0) indicates there is a sense buffer. | |
221 | */ | |
222 | if (hdr.status == 0 && hdr.sb_len_wr > 0) { | |
223 | hdr.status = CHECK_CONDITION; | |
224 | } | |
225 | ||
226 | stl_p(&req->scsi->errors, | |
227 | hdr.status | (hdr.msg_status << 8) | | |
228 | (hdr.host_status << 16) | (hdr.driver_status << 24)); | |
92e3c2a3 AJ |
229 | stl_p(&req->scsi->residual, hdr.resid); |
230 | stl_p(&req->scsi->sense_len, hdr.sb_len_wr); | |
231 | stl_p(&req->scsi->data_len, hdr.dxfer_len); | |
1063b8b1 CH |
232 | |
233 | virtio_blk_req_complete(req, status); | |
a597e79c | 234 | g_free(req); |
730a9c53 | 235 | return; |
1063b8b1 | 236 | #else |
f34e73cd PB |
237 | abort(); |
238 | #endif | |
239 | ||
240 | fail: | |
241 | /* Just put anything nonzero so that the ioctl fails in the guest. */ | |
242 | stl_p(&req->scsi->errors, 255); | |
243 | virtio_blk_req_complete(req, status); | |
a597e79c | 244 | g_free(req); |
1063b8b1 | 245 | } |
1063b8b1 | 246 | |
c20fd872 CH |
247 | typedef struct MultiReqBuffer { |
248 | BlockRequest blkreq[32]; | |
249 | unsigned int num_writes; | |
250 | } MultiReqBuffer; | |
251 | ||
252 | static void virtio_submit_multiwrite(BlockDriverState *bs, MultiReqBuffer *mrb) | |
869a5c6d | 253 | { |
91553dcc | 254 | int i, ret; |
91553dcc | 255 | |
c20fd872 CH |
256 | if (!mrb->num_writes) { |
257 | return; | |
258 | } | |
259 | ||
260 | ret = bdrv_aio_multiwrite(bs, mrb->blkreq, mrb->num_writes); | |
91553dcc | 261 | if (ret != 0) { |
c20fd872 CH |
262 | for (i = 0; i < mrb->num_writes; i++) { |
263 | if (mrb->blkreq[i].error) { | |
264 | virtio_blk_rw_complete(mrb->blkreq[i].opaque, -EIO); | |
91553dcc KW |
265 | } |
266 | } | |
267 | } | |
c20fd872 CH |
268 | |
269 | mrb->num_writes = 0; | |
91553dcc | 270 | } |
87b245db | 271 | |
c20fd872 | 272 | static void virtio_blk_handle_flush(VirtIOBlockReq *req, MultiReqBuffer *mrb) |
aa659be3 | 273 | { |
a597e79c CH |
274 | bdrv_acct_start(req->dev->bs, &req->acct, 0, BDRV_ACCT_FLUSH); |
275 | ||
618fbb84 CH |
276 | /* |
277 | * Make sure all outstanding writes are posted to the backing device. | |
278 | */ | |
c20fd872 | 279 | virtio_submit_multiwrite(req->dev->bs, mrb); |
ad54ae80 | 280 | bdrv_aio_flush(req->dev->bs, virtio_blk_flush_complete, req); |
aa659be3 CH |
281 | } |
282 | ||
c20fd872 | 283 | static void virtio_blk_handle_write(VirtIOBlockReq *req, MultiReqBuffer *mrb) |
91553dcc | 284 | { |
c20fd872 | 285 | BlockRequest *blkreq; |
92e3c2a3 | 286 | uint64_t sector; |
c20fd872 | 287 | |
92e3c2a3 | 288 | sector = ldq_p(&req->out->sector); |
6d519a5f | 289 | |
a597e79c CH |
290 | bdrv_acct_start(req->dev->bs, &req->acct, req->qiov.size, BDRV_ACCT_WRITE); |
291 | ||
92e3c2a3 AJ |
292 | trace_virtio_blk_handle_write(req, sector, req->qiov.size / 512); |
293 | ||
294 | if (sector & req->dev->sector_mask) { | |
8cfacf07 CH |
295 | virtio_blk_rw_complete(req, -EIO); |
296 | return; | |
297 | } | |
52c05023 CH |
298 | if (req->qiov.size % req->dev->conf->logical_block_size) { |
299 | virtio_blk_rw_complete(req, -EIO); | |
300 | return; | |
301 | } | |
8cfacf07 | 302 | |
c20fd872 CH |
303 | if (mrb->num_writes == 32) { |
304 | virtio_submit_multiwrite(req->dev->bs, mrb); | |
87b245db | 305 | } |
91553dcc | 306 | |
c20fd872 | 307 | blkreq = &mrb->blkreq[mrb->num_writes]; |
92e3c2a3 | 308 | blkreq->sector = sector; |
c20fd872 CH |
309 | blkreq->nb_sectors = req->qiov.size / BDRV_SECTOR_SIZE; |
310 | blkreq->qiov = &req->qiov; | |
311 | blkreq->cb = virtio_blk_rw_complete; | |
312 | blkreq->opaque = req; | |
313 | blkreq->error = 0; | |
91553dcc | 314 | |
c20fd872 | 315 | mrb->num_writes++; |
d28a1b6e | 316 | } |
869a5c6d | 317 | |
d28a1b6e AL |
318 | static void virtio_blk_handle_read(VirtIOBlockReq *req) |
319 | { | |
92e3c2a3 AJ |
320 | uint64_t sector; |
321 | ||
322 | sector = ldq_p(&req->out->sector); | |
87b245db | 323 | |
a597e79c CH |
324 | bdrv_acct_start(req->dev->bs, &req->acct, req->qiov.size, BDRV_ACCT_READ); |
325 | ||
81b6b9fa SH |
326 | trace_virtio_blk_handle_read(req, sector, req->qiov.size / 512); |
327 | ||
92e3c2a3 | 328 | if (sector & req->dev->sector_mask) { |
8cfacf07 CH |
329 | virtio_blk_rw_complete(req, -EIO); |
330 | return; | |
331 | } | |
52c05023 CH |
332 | if (req->qiov.size % req->dev->conf->logical_block_size) { |
333 | virtio_blk_rw_complete(req, -EIO); | |
334 | return; | |
335 | } | |
ad54ae80 PB |
336 | bdrv_aio_readv(req->dev->bs, sector, &req->qiov, |
337 | req->qiov.size / BDRV_SECTOR_SIZE, | |
338 | virtio_blk_rw_complete, req); | |
869a5c6d AL |
339 | } |
340 | ||
bc6694d4 KW |
341 | static void virtio_blk_handle_request(VirtIOBlockReq *req, |
342 | MultiReqBuffer *mrb) | |
343 | { | |
92e3c2a3 AJ |
344 | uint32_t type; |
345 | ||
bc6694d4 | 346 | if (req->elem.out_num < 1 || req->elem.in_num < 1) { |
870cef1d | 347 | error_report("virtio-blk missing headers"); |
bc6694d4 KW |
348 | exit(1); |
349 | } | |
350 | ||
351 | if (req->elem.out_sg[0].iov_len < sizeof(*req->out) || | |
352 | req->elem.in_sg[req->elem.in_num - 1].iov_len < sizeof(*req->in)) { | |
870cef1d | 353 | error_report("virtio-blk header not in correct element"); |
bc6694d4 KW |
354 | exit(1); |
355 | } | |
356 | ||
357 | req->out = (void *)req->elem.out_sg[0].iov_base; | |
358 | req->in = (void *)req->elem.in_sg[req->elem.in_num - 1].iov_base; | |
359 | ||
92e3c2a3 AJ |
360 | type = ldl_p(&req->out->type); |
361 | ||
362 | if (type & VIRTIO_BLK_T_FLUSH) { | |
c20fd872 | 363 | virtio_blk_handle_flush(req, mrb); |
92e3c2a3 | 364 | } else if (type & VIRTIO_BLK_T_SCSI_CMD) { |
bc6694d4 | 365 | virtio_blk_handle_scsi(req); |
92e3c2a3 | 366 | } else if (type & VIRTIO_BLK_T_GET_ID) { |
2930b313 | 367 | VirtIOBlock *s = req->dev; |
368 | ||
a8686a9b MA |
369 | /* |
370 | * NB: per existing s/n string convention the string is | |
371 | * terminated by '\0' only when shorter than buffer. | |
372 | */ | |
373 | strncpy(req->elem.in_sg[0].iov_base, | |
da3dcefa | 374 | s->blk.serial ? s->blk.serial : "", |
a8686a9b | 375 | MIN(req->elem.in_sg[0].iov_len, VIRTIO_BLK_ID_BYTES)); |
2930b313 | 376 | virtio_blk_req_complete(req, VIRTIO_BLK_S_OK); |
a597e79c | 377 | g_free(req); |
92e3c2a3 | 378 | } else if (type & VIRTIO_BLK_T_OUT) { |
bc6694d4 KW |
379 | qemu_iovec_init_external(&req->qiov, &req->elem.out_sg[1], |
380 | req->elem.out_num - 1); | |
c20fd872 | 381 | virtio_blk_handle_write(req, mrb); |
9e72c450 AZ |
382 | } else if (type == VIRTIO_BLK_T_IN || type == VIRTIO_BLK_T_BARRIER) { |
383 | /* VIRTIO_BLK_T_IN is 0, so we can't just & it. */ | |
bc6694d4 KW |
384 | qemu_iovec_init_external(&req->qiov, &req->elem.in_sg[0], |
385 | req->elem.in_num - 1); | |
386 | virtio_blk_handle_read(req); | |
9e72c450 AZ |
387 | } else { |
388 | virtio_blk_req_complete(req, VIRTIO_BLK_S_UNSUPP); | |
389 | g_free(req); | |
bc6694d4 KW |
390 | } |
391 | } | |
392 | ||
6e02c38d AL |
393 | static void virtio_blk_handle_output(VirtIODevice *vdev, VirtQueue *vq) |
394 | { | |
395 | VirtIOBlock *s = to_virtio_blk(vdev); | |
396 | VirtIOBlockReq *req; | |
bc6694d4 KW |
397 | MultiReqBuffer mrb = { |
398 | .num_writes = 0, | |
bc6694d4 | 399 | }; |
6e02c38d | 400 | |
392808b4 SH |
401 | #ifdef CONFIG_VIRTIO_BLK_DATA_PLANE |
402 | /* Some guests kick before setting VIRTIO_CONFIG_S_DRIVER_OK so start | |
403 | * dataplane here instead of waiting for .set_status(). | |
404 | */ | |
405 | if (s->dataplane) { | |
406 | virtio_blk_data_plane_start(s->dataplane); | |
407 | return; | |
408 | } | |
409 | #endif | |
410 | ||
6e02c38d | 411 | while ((req = virtio_blk_get_request(s))) { |
bc6694d4 | 412 | virtio_blk_handle_request(req, &mrb); |
6e02c38d | 413 | } |
91553dcc | 414 | |
c20fd872 | 415 | virtio_submit_multiwrite(s->bs, &mrb); |
91553dcc | 416 | |
6e02c38d AL |
417 | /* |
418 | * FIXME: Want to check for completions before returning to guest mode, | |
419 | * so cached reads and writes are reported as quickly as possible. But | |
420 | * that should be done in the generic block layer. | |
421 | */ | |
422 | } | |
423 | ||
213189ab | 424 | static void virtio_blk_dma_restart_bh(void *opaque) |
869a5c6d AL |
425 | { |
426 | VirtIOBlock *s = opaque; | |
427 | VirtIOBlockReq *req = s->rq; | |
f1b52868 KW |
428 | MultiReqBuffer mrb = { |
429 | .num_writes = 0, | |
f1b52868 | 430 | }; |
869a5c6d | 431 | |
213189ab MA |
432 | qemu_bh_delete(s->bh); |
433 | s->bh = NULL; | |
869a5c6d AL |
434 | |
435 | s->rq = NULL; | |
436 | ||
437 | while (req) { | |
f1b52868 | 438 | virtio_blk_handle_request(req, &mrb); |
869a5c6d AL |
439 | req = req->next; |
440 | } | |
f1b52868 | 441 | |
c20fd872 | 442 | virtio_submit_multiwrite(s->bs, &mrb); |
869a5c6d AL |
443 | } |
444 | ||
1dfb4dd9 LC |
445 | static void virtio_blk_dma_restart_cb(void *opaque, int running, |
446 | RunState state) | |
213189ab MA |
447 | { |
448 | VirtIOBlock *s = opaque; | |
449 | ||
392808b4 | 450 | if (!running) { |
213189ab | 451 | return; |
392808b4 | 452 | } |
213189ab MA |
453 | |
454 | if (!s->bh) { | |
455 | s->bh = qemu_bh_new(virtio_blk_dma_restart_bh, s); | |
456 | qemu_bh_schedule(s->bh); | |
457 | } | |
458 | } | |
459 | ||
6e02c38d AL |
460 | static void virtio_blk_reset(VirtIODevice *vdev) |
461 | { | |
392808b4 SH |
462 | #ifdef CONFIG_VIRTIO_BLK_DATA_PLANE |
463 | VirtIOBlock *s = to_virtio_blk(vdev); | |
464 | ||
465 | if (s->dataplane) { | |
466 | virtio_blk_data_plane_stop(s->dataplane); | |
467 | } | |
468 | #endif | |
469 | ||
6e02c38d AL |
470 | /* |
471 | * This should cancel pending requests, but can't do nicely until there | |
472 | * are per-device request lists. | |
473 | */ | |
922453bc | 474 | bdrv_drain_all(); |
6e02c38d AL |
475 | } |
476 | ||
bf011293 | 477 | /* coalesce internal state, copy to pci i/o region 0 |
478 | */ | |
6e02c38d AL |
479 | static void virtio_blk_update_config(VirtIODevice *vdev, uint8_t *config) |
480 | { | |
481 | VirtIOBlock *s = to_virtio_blk(vdev); | |
482 | struct virtio_blk_config blkcfg; | |
483 | uint64_t capacity; | |
3a395142 | 484 | int blk_size = s->conf->logical_block_size; |
6e02c38d AL |
485 | |
486 | bdrv_get_geometry(s->bs, &capacity); | |
5c5dafdc | 487 | memset(&blkcfg, 0, sizeof(blkcfg)); |
6e02c38d AL |
488 | stq_raw(&blkcfg.capacity, capacity); |
489 | stl_raw(&blkcfg.seg_max, 128 - 2); | |
e63e7fde | 490 | stw_raw(&blkcfg.cylinders, s->conf->cyls); |
3a395142 PB |
491 | stl_raw(&blkcfg.blk_size, blk_size); |
492 | stw_raw(&blkcfg.min_io_size, s->conf->min_io_size / blk_size); | |
493 | stw_raw(&blkcfg.opt_io_size, s->conf->opt_io_size / blk_size); | |
e63e7fde | 494 | blkcfg.heads = s->conf->heads; |
136be99e CB |
495 | /* |
496 | * We must ensure that the block device capacity is a multiple of | |
497 | * the logical block size. If that is not the case, lets use | |
498 | * sector_mask to adopt the geometry to have a correct picture. | |
499 | * For those devices where the capacity is ok for the given geometry | |
500 | * we dont touch the sector value of the geometry, since some devices | |
501 | * (like s390 dasd) need a specific value. Here the capacity is already | |
502 | * cyls*heads*secs*blk_size and the sector value is not block size | |
503 | * divided by 512 - instead it is the amount of blk_size blocks | |
504 | * per track (cylinder). | |
505 | */ | |
e63e7fde MA |
506 | if (bdrv_getlength(s->bs) / s->conf->heads / s->conf->secs % blk_size) { |
507 | blkcfg.sectors = s->conf->secs & ~s->sector_mask; | |
136be99e | 508 | } else { |
e63e7fde | 509 | blkcfg.sectors = s->conf->secs; |
136be99e | 510 | } |
c7085da7 | 511 | blkcfg.size_max = 0; |
9752c371 CH |
512 | blkcfg.physical_block_exp = get_physical_block_exp(s->conf); |
513 | blkcfg.alignment_offset = 0; | |
13e3dce0 | 514 | blkcfg.wce = bdrv_enable_write_cache(s->bs); |
37d5ddd6 | 515 | memcpy(config, &blkcfg, sizeof(struct virtio_blk_config)); |
6e02c38d AL |
516 | } |
517 | ||
13e3dce0 PB |
518 | static void virtio_blk_set_config(VirtIODevice *vdev, const uint8_t *config) |
519 | { | |
520 | VirtIOBlock *s = to_virtio_blk(vdev); | |
521 | struct virtio_blk_config blkcfg; | |
522 | ||
523 | memcpy(&blkcfg, config, sizeof(blkcfg)); | |
524 | bdrv_set_enable_write_cache(s->bs, blkcfg.wce != 0); | |
525 | } | |
526 | ||
8172539d | 527 | static uint32_t virtio_blk_get_features(VirtIODevice *vdev, uint32_t features) |
6e02c38d | 528 | { |
bf011293 | 529 | VirtIOBlock *s = to_virtio_blk(vdev); |
1063b8b1 CH |
530 | |
531 | features |= (1 << VIRTIO_BLK_F_SEG_MAX); | |
532 | features |= (1 << VIRTIO_BLK_F_GEOMETRY); | |
9752c371 | 533 | features |= (1 << VIRTIO_BLK_F_TOPOLOGY); |
8cfacf07 | 534 | features |= (1 << VIRTIO_BLK_F_BLK_SIZE); |
a6c5c84a | 535 | features |= (1 << VIRTIO_BLK_F_SCSI); |
aa659be3 | 536 | |
da3dcefa | 537 | if (s->blk.config_wce) { |
8a873ba7 SH |
538 | features |= (1 << VIRTIO_BLK_F_CONFIG_WCE); |
539 | } | |
aa659be3 | 540 | if (bdrv_enable_write_cache(s->bs)) |
13e3dce0 PB |
541 | features |= (1 << VIRTIO_BLK_F_WCE); |
542 | ||
c79662f7 NS |
543 | if (bdrv_is_read_only(s->bs)) |
544 | features |= 1 << VIRTIO_BLK_F_RO; | |
1063b8b1 CH |
545 | |
546 | return features; | |
6e02c38d AL |
547 | } |
548 | ||
9315cbfd PB |
549 | static void virtio_blk_set_status(VirtIODevice *vdev, uint8_t status) |
550 | { | |
551 | VirtIOBlock *s = to_virtio_blk(vdev); | |
552 | uint32_t features; | |
553 | ||
392808b4 | 554 | #ifdef CONFIG_VIRTIO_BLK_DATA_PLANE |
cf139388 SH |
555 | if (s->dataplane && !(status & (VIRTIO_CONFIG_S_DRIVER | |
556 | VIRTIO_CONFIG_S_DRIVER_OK))) { | |
392808b4 SH |
557 | virtio_blk_data_plane_stop(s->dataplane); |
558 | } | |
559 | #endif | |
560 | ||
9315cbfd PB |
561 | if (!(status & VIRTIO_CONFIG_S_DRIVER_OK)) { |
562 | return; | |
563 | } | |
564 | ||
565 | features = vdev->guest_features; | |
566 | bdrv_set_enable_write_cache(s->bs, !!(features & (1 << VIRTIO_BLK_F_WCE))); | |
567 | } | |
568 | ||
6e02c38d AL |
569 | static void virtio_blk_save(QEMUFile *f, void *opaque) |
570 | { | |
571 | VirtIOBlock *s = opaque; | |
869a5c6d AL |
572 | VirtIOBlockReq *req = s->rq; |
573 | ||
6e02c38d | 574 | virtio_save(&s->vdev, f); |
869a5c6d AL |
575 | |
576 | while (req) { | |
577 | qemu_put_sbyte(f, 1); | |
578 | qemu_put_buffer(f, (unsigned char*)&req->elem, sizeof(req->elem)); | |
579 | req = req->next; | |
580 | } | |
581 | qemu_put_sbyte(f, 0); | |
6e02c38d AL |
582 | } |
583 | ||
584 | static int virtio_blk_load(QEMUFile *f, void *opaque, int version_id) | |
585 | { | |
586 | VirtIOBlock *s = opaque; | |
2a633c46 | 587 | int ret; |
6e02c38d | 588 | |
869a5c6d | 589 | if (version_id != 2) |
6e02c38d AL |
590 | return -EINVAL; |
591 | ||
2a633c46 OW |
592 | ret = virtio_load(&s->vdev, f); |
593 | if (ret) { | |
594 | return ret; | |
595 | } | |
596 | ||
869a5c6d AL |
597 | while (qemu_get_sbyte(f)) { |
598 | VirtIOBlockReq *req = virtio_blk_alloc_request(s); | |
599 | qemu_get_buffer(f, (unsigned char*)&req->elem, sizeof(req->elem)); | |
600 | req->next = s->rq; | |
20a81e4d | 601 | s->rq = req; |
b6a4805b KW |
602 | |
603 | virtqueue_map_sg(req->elem.in_sg, req->elem.in_addr, | |
604 | req->elem.in_num, 1); | |
605 | virtqueue_map_sg(req->elem.out_sg, req->elem.out_addr, | |
606 | req->elem.out_num, 0); | |
869a5c6d | 607 | } |
6e02c38d AL |
608 | |
609 | return 0; | |
610 | } | |
611 | ||
145feb17 | 612 | static void virtio_blk_resize(void *opaque) |
e5051fc7 CH |
613 | { |
614 | VirtIOBlock *s = opaque; | |
615 | ||
145feb17 | 616 | virtio_notify_config(&s->vdev); |
e5051fc7 CH |
617 | } |
618 | ||
0e49de52 | 619 | static const BlockDevOps virtio_block_ops = { |
145feb17 | 620 | .resize_cb = virtio_blk_resize, |
0e49de52 MA |
621 | }; |
622 | ||
12c5674b | 623 | VirtIODevice *virtio_blk_init(DeviceState *dev, VirtIOBlkConf *blk) |
6e02c38d AL |
624 | { |
625 | VirtIOBlock *s; | |
6e02c38d | 626 | static int virtio_blk_id; |
cf21e106 | 627 | |
12c5674b | 628 | if (!blk->conf.bs) { |
6a84cb1f | 629 | error_report("drive property not set"); |
d75d25e3 MA |
630 | return NULL; |
631 | } | |
12c5674b | 632 | if (!bdrv_is_inserted(blk->conf.bs)) { |
98f28ad7 MA |
633 | error_report("Device needs media, but drive is empty"); |
634 | return NULL; | |
635 | } | |
d75d25e3 | 636 | |
911525db | 637 | blkconf_serial(&blk->conf, &blk->serial); |
b7eb0c9f MA |
638 | if (blkconf_geometry(&blk->conf, NULL, 65535, 255, 255) < 0) { |
639 | return NULL; | |
640 | } | |
a8686a9b | 641 | |
53c25cea | 642 | s = (VirtIOBlock *)virtio_common_init("virtio-blk", VIRTIO_ID_BLOCK, |
37d5ddd6 | 643 | sizeof(struct virtio_blk_config), |
53c25cea | 644 | sizeof(VirtIOBlock)); |
6e02c38d AL |
645 | |
646 | s->vdev.get_config = virtio_blk_update_config; | |
13e3dce0 | 647 | s->vdev.set_config = virtio_blk_set_config; |
6e02c38d | 648 | s->vdev.get_features = virtio_blk_get_features; |
9315cbfd | 649 | s->vdev.set_status = virtio_blk_set_status; |
6e02c38d | 650 | s->vdev.reset = virtio_blk_reset; |
12c5674b PB |
651 | s->bs = blk->conf.bs; |
652 | s->conf = &blk->conf; | |
da3dcefa | 653 | memcpy(&(s->blk), blk, sizeof(struct VirtIOBlkConf)); |
869a5c6d | 654 | s->rq = NULL; |
1573a35d | 655 | s->sector_mask = (s->conf->logical_block_size / BDRV_SECTOR_SIZE) - 1; |
e63e7fde | 656 | |
6e02c38d | 657 | s->vq = virtio_add_queue(&s->vdev, 128, virtio_blk_handle_output); |
392808b4 SH |
658 | #ifdef CONFIG_VIRTIO_BLK_DATA_PLANE |
659 | if (!virtio_blk_data_plane_create(&s->vdev, blk, &s->dataplane)) { | |
660 | virtio_cleanup(&s->vdev); | |
661 | return NULL; | |
662 | } | |
663 | #endif | |
6e02c38d | 664 | |
69b302b2 | 665 | s->change = qemu_add_vm_change_state_handler(virtio_blk_dma_restart_cb, s); |
9d0d3138 | 666 | s->qdev = dev; |
0be71e32 | 667 | register_savevm(dev, "virtio-blk", virtio_blk_id++, 2, |
6e02c38d | 668 | virtio_blk_save, virtio_blk_load, s); |
0e49de52 | 669 | bdrv_set_dev_ops(s->bs, &virtio_block_ops, s); |
12c5674b | 670 | bdrv_set_buffer_alignment(s->bs, s->conf->logical_block_size); |
6e02c38d | 671 | |
af239a62 | 672 | bdrv_iostatus_enable(s->bs); |
12c5674b | 673 | add_boot_device_path(s->conf->bootindex, dev, "/disk@0,0"); |
1ca4d09a | 674 | |
53c25cea | 675 | return &s->vdev; |
6e02c38d | 676 | } |
9d0d3138 AW |
677 | |
678 | void virtio_blk_exit(VirtIODevice *vdev) | |
679 | { | |
680 | VirtIOBlock *s = to_virtio_blk(vdev); | |
392808b4 SH |
681 | |
682 | #ifdef CONFIG_VIRTIO_BLK_DATA_PLANE | |
683 | virtio_blk_data_plane_destroy(s->dataplane); | |
684 | s->dataplane = NULL; | |
685 | #endif | |
69b302b2 | 686 | qemu_del_vm_change_state_handler(s->change); |
9d0d3138 | 687 | unregister_savevm(s->qdev, "virtio-blk", s); |
0e47931b | 688 | blockdev_mark_auto_del(s->bs); |
d92551f2 | 689 | virtio_cleanup(vdev); |
9d0d3138 | 690 | } |