?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
configure: Prepend pixman and ftd flags to overrule system-provided ones
[qemu.git] / blockdev.c
blobe595910476b623ec25e6ea3d65aa53de8c89c95c
1 /*
2 * QEMU host block devices
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
6 * This work is licensed under the terms of the GNU GPL, version 2 or
7 * later. See the COPYING file in the top-level directory.
8 *
9 * This file incorporates work covered by the following copyright and
10 * permission notice:
11 *
12 * Copyright (c) 2003-2008 Fabrice Bellard
13 *
14 * Permission is hereby granted, free of charge, to any person obtaining a copy
15 * of this software and associated documentation files (the "Software"), to deal
16 * in the Software without restriction, including without limitation the rights
17 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
18 * copies of the Software, and to permit persons to whom the Software is
19 * furnished to do so, subject to the following conditions:
20 *
21 * The above copyright notice and this permission notice shall be included in
22 * all copies or substantial portions of the Software.
23 *
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
27 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
30 * THE SOFTWARE.
31 */
32
33 #include "sysemu/blockdev.h"
34 #include "hw/block/block.h"
35 #include "block/blockjob.h"
36 #include "monitor/monitor.h"
37 #include "qemu/option.h"
38 #include "qemu/config-file.h"
39 #include "qapi/qmp/types.h"
40 #include "qapi-visit.h"
41 #include "qapi/qmp-output-visitor.h"
42 #include "qapi/util.h"
43 #include "sysemu/sysemu.h"
44 #include "block/block_int.h"
45 #include "qmp-commands.h"
46 #include "trace.h"
47 #include "sysemu/arch_init.h"
48
49 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
50
51 static const char *const if_name[IF_COUNT] = {
52 [IF_NONE] = "none",
53 [IF_IDE] = "ide",
54 [IF_SCSI] = "scsi",
55 [IF_FLOPPY] = "floppy",
56 [IF_PFLASH] = "pflash",
57 [IF_MTD] = "mtd",
58 [IF_SD] = "sd",
59 [IF_VIRTIO] = "virtio",
60 [IF_XEN] = "xen",
61 };
62
63 static int if_max_devs[IF_COUNT] = {
64 /*
65 * Do not change these numbers! They govern how drive option
66 * index maps to unit and bus. That mapping is ABI.
67 *
68 * All controllers used to imlement if=T drives need to support
69 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
70 * Otherwise, some index values map to "impossible" bus, unit
71 * values.
72 *
73 * For instance, if you change [IF_SCSI] to 255, -drive
74 * if=scsi,index=12 no longer means bus=1,unit=5, but
75 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
76 * the drive can't be set up. Regression.
77 */
78 [IF_IDE] = 2,
79 [IF_SCSI] = 7,
80 };
81
82 /**
83 * Boards may call this to offer board-by-board overrides
84 * of the default, global values.
85 */
86 void override_max_devs(BlockInterfaceType type, int max_devs)
87 {
88 DriveInfo *dinfo;
89
90 if (max_devs <= 0) {
91 return;
92 }
93
94 QTAILQ_FOREACH(dinfo, &drives, next) {
95 if (dinfo->type == type) {
96 fprintf(stderr, "Cannot override units-per-bus property of"
97 " the %s interface, because a drive of that type has"
98 " already been added.\n", if_name[type]);
99 g_assert_not_reached();
100 }
101 }
102
103 if_max_devs[type] = max_devs;
104 }
105
106 /*
107 * We automatically delete the drive when a device using it gets
108 * unplugged. Questionable feature, but we can't just drop it.
109 * Device models call blockdev_mark_auto_del() to schedule the
110 * automatic deletion, and generic qdev code calls blockdev_auto_del()
111 * when deletion is actually safe.
112 */
113 void blockdev_mark_auto_del(BlockDriverState *bs)
114 {
115 DriveInfo *dinfo = drive_get_by_blockdev(bs);
116
117 if (dinfo && !dinfo->enable_auto_del) {
118 return;
119 }
120
121 if (bs->job) {
122 block_job_cancel(bs->job);
123 }
124 if (dinfo) {
125 dinfo->auto_del = 1;
126 }
127 }
128
129 void blockdev_auto_del(BlockDriverState *bs)
130 {
131 DriveInfo *dinfo = drive_get_by_blockdev(bs);
132
133 if (dinfo && dinfo->auto_del) {
134 drive_del(dinfo);
135 }
136 }
137
138 /**
139 * Returns the current mapping of how many units per bus
140 * a particular interface can support.
141 *
142 * A positive integer indicates n units per bus.
143 * 0 implies the mapping has not been established.
144 * -1 indicates an invalid BlockInterfaceType was given.
145 */
146 int drive_get_max_devs(BlockInterfaceType type)
147 {
148 if (type >= IF_IDE && type < IF_COUNT) {
149 return if_max_devs[type];
150 }
151
152 return -1;
153 }
154
155 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
156 {
157 int max_devs = if_max_devs[type];
158 return max_devs ? index / max_devs : 0;
159 }
160
161 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
162 {
163 int max_devs = if_max_devs[type];
164 return max_devs ? index % max_devs : index;
165 }
166
167 QemuOpts *drive_def(const char *optstr)
168 {
169 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
170 }
171
172 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
173 const char *optstr)
174 {
175 QemuOpts *opts;
176 char buf[32];
177
178 opts = drive_def(optstr);
179 if (!opts) {
180 return NULL;
181 }
182 if (type != IF_DEFAULT) {
183 qemu_opt_set(opts, "if", if_name[type]);
184 }
185 if (index >= 0) {
186 snprintf(buf, sizeof(buf), "%d", index);
187 qemu_opt_set(opts, "index", buf);
188 }
189 if (file)
190 qemu_opt_set(opts, "file", file);
191 return opts;
192 }
193
194 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
195 {
196 DriveInfo *dinfo;
197
198 /* seek interface, bus and unit */
199
200 QTAILQ_FOREACH(dinfo, &drives, next) {
201 if (dinfo->type == type &&
202 dinfo->bus == bus &&
203 dinfo->unit == unit)
204 return dinfo;
205 }
206
207 return NULL;
208 }
209
210 bool drive_check_orphaned(void)
211 {
212 DriveInfo *dinfo;
213 bool rs = false;
214
215 QTAILQ_FOREACH(dinfo, &drives, next) {
216 /* If dinfo->bdrv->dev is NULL, it has no device attached. */
217 /* Unless this is a default drive, this may be an oversight. */
218 if (!dinfo->bdrv->dev && !dinfo->is_default &&
219 dinfo->type != IF_NONE) {
220 fprintf(stderr, "Warning: Orphaned drive without device: "
221 "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
222 dinfo->id, dinfo->bdrv->filename, if_name[dinfo->type],
223 dinfo->bus, dinfo->unit);
224 rs = true;
225 }
226 }
227
228 return rs;
229 }
230
231 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
232 {
233 return drive_get(type,
234 drive_index_to_bus_id(type, index),
235 drive_index_to_unit_id(type, index));
236 }
237
238 int drive_get_max_bus(BlockInterfaceType type)
239 {
240 int max_bus;
241 DriveInfo *dinfo;
242
243 max_bus = -1;
244 QTAILQ_FOREACH(dinfo, &drives, next) {
245 if(dinfo->type == type &&
246 dinfo->bus > max_bus)
247 max_bus = dinfo->bus;
248 }
249 return max_bus;
250 }
251
252 /* Get a block device. This should only be used for single-drive devices
253 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
254 appropriate bus. */
255 DriveInfo *drive_get_next(BlockInterfaceType type)
256 {
257 static int next_block_unit[IF_COUNT];
258
259 return drive_get(type, 0, next_block_unit[type]++);
260 }
261
262 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
263 {
264 DriveInfo *dinfo;
265
266 QTAILQ_FOREACH(dinfo, &drives, next) {
267 if (dinfo->bdrv == bs) {
268 return dinfo;
269 }
270 }
271 return NULL;
272 }
273
274 static void bdrv_format_print(void *opaque, const char *name)
275 {
276 error_printf(" %s", name);
277 }
278
279 void drive_del(DriveInfo *dinfo)
280 {
281 bdrv_unref(dinfo->bdrv);
282 }
283
284 void drive_info_del(DriveInfo *dinfo)
285 {
286 if (!dinfo) {
287 return;
288 }
289 qemu_opts_del(dinfo->opts);
290 g_free(dinfo->id);
291 QTAILQ_REMOVE(&drives, dinfo, next);
292 g_free(dinfo->serial);
293 g_free(dinfo);
294 }
295
296 typedef struct {
297 QEMUBH *bh;
298 BlockDriverState *bs;
299 } BDRVPutRefBH;
300
301 static void bdrv_put_ref_bh(void *opaque)
302 {
303 BDRVPutRefBH *s = opaque;
304
305 bdrv_unref(s->bs);
306 qemu_bh_delete(s->bh);
307 g_free(s);
308 }
309
310 /*
311 * Release a BDS reference in a BH
312 *
313 * It is not safe to use bdrv_unref() from a callback function when the callers
314 * still need the BlockDriverState. In such cases we schedule a BH to release
315 * the reference.
316 */
317 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
318 {
319 BDRVPutRefBH *s;
320
321 s = g_new(BDRVPutRefBH, 1);
322 s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
323 s->bs = bs;
324 qemu_bh_schedule(s->bh);
325 }
326
327 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
328 {
329 if (!strcmp(buf, "ignore")) {
330 return BLOCKDEV_ON_ERROR_IGNORE;
331 } else if (!is_read && !strcmp(buf, "enospc")) {
332 return BLOCKDEV_ON_ERROR_ENOSPC;
333 } else if (!strcmp(buf, "stop")) {
334 return BLOCKDEV_ON_ERROR_STOP;
335 } else if (!strcmp(buf, "report")) {
336 return BLOCKDEV_ON_ERROR_REPORT;
337 } else {
338 error_setg(errp, "'%s' invalid %s error action",
339 buf, is_read ? "read" : "write");
340 return -1;
341 }
342 }
343
344 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
345 {
346 if (throttle_conflicting(cfg)) {
347 error_setg(errp, "bps/iops/max total values and read/write values"
348 " cannot be used at the same time");
349 return false;
350 }
351
352 if (!throttle_is_valid(cfg)) {
353 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
354 return false;
355 }
356
357 return true;
358 }
359
360 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
361
362 /* Takes the ownership of bs_opts */
363 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
364 Error **errp)
365 {
366 const char *buf;
367 int ro = 0;
368 int bdrv_flags = 0;
369 int on_read_error, on_write_error;
370 BlockDriverState *bs;
371 DriveInfo *dinfo;
372 ThrottleConfig cfg;
373 int snapshot = 0;
374 bool copy_on_read;
375 int ret;
376 Error *error = NULL;
377 QemuOpts *opts;
378 const char *id;
379 bool has_driver_specific_opts;
380 BlockdevDetectZeroesOptions detect_zeroes;
381 BlockDriver *drv = NULL;
382
383 /* Check common options by copying from bs_opts to opts, all other options
384 * stay in bs_opts for processing by bdrv_open(). */
385 id = qdict_get_try_str(bs_opts, "id");
386 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
387 if (error) {
388 error_propagate(errp, error);
389 goto err_no_opts;
390 }
391
392 qemu_opts_absorb_qdict(opts, bs_opts, &error);
393 if (error) {
394 error_propagate(errp, error);
395 goto early_err;
396 }
397
398 if (id) {
399 qdict_del(bs_opts, "id");
400 }
401
402 has_driver_specific_opts = !!qdict_size(bs_opts);
403
404 /* extract parameters */
405 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
406 ro = qemu_opt_get_bool(opts, "read-only", 0);
407 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
408
409 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
410 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
411 error_setg(errp, "invalid discard option");
412 goto early_err;
413 }
414 }
415
416 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
417 bdrv_flags |= BDRV_O_CACHE_WB;
418 }
419 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
420 bdrv_flags |= BDRV_O_NOCACHE;
421 }
422 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
423 bdrv_flags |= BDRV_O_NO_FLUSH;
424 }
425
426 #ifdef CONFIG_LINUX_AIO
427 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
428 if (!strcmp(buf, "native")) {
429 bdrv_flags |= BDRV_O_NATIVE_AIO;
430 } else if (!strcmp(buf, "threads")) {
431 /* this is the default */
432 } else {
433 error_setg(errp, "invalid aio option");
434 goto early_err;
435 }
436 }
437 #endif
438
439 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
440 if (is_help_option(buf)) {
441 error_printf("Supported formats:");
442 bdrv_iterate_format(bdrv_format_print, NULL);
443 error_printf("\n");
444 goto early_err;
445 }
446
447 drv = bdrv_find_format(buf);
448 if (!drv) {
449 error_setg(errp, "'%s' invalid format", buf);
450 goto early_err;
451 }
452 }
453
454 /* disk I/O throttling */
455 memset(&cfg, 0, sizeof(cfg));
456 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
457 qemu_opt_get_number(opts, "throttling.bps-total", 0);
458 cfg.buckets[THROTTLE_BPS_READ].avg =
459 qemu_opt_get_number(opts, "throttling.bps-read", 0);
460 cfg.buckets[THROTTLE_BPS_WRITE].avg =
461 qemu_opt_get_number(opts, "throttling.bps-write", 0);
462 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
463 qemu_opt_get_number(opts, "throttling.iops-total", 0);
464 cfg.buckets[THROTTLE_OPS_READ].avg =
465 qemu_opt_get_number(opts, "throttling.iops-read", 0);
466 cfg.buckets[THROTTLE_OPS_WRITE].avg =
467 qemu_opt_get_number(opts, "throttling.iops-write", 0);
468
469 cfg.buckets[THROTTLE_BPS_TOTAL].max =
470 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
471 cfg.buckets[THROTTLE_BPS_READ].max =
472 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
473 cfg.buckets[THROTTLE_BPS_WRITE].max =
474 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
475 cfg.buckets[THROTTLE_OPS_TOTAL].max =
476 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
477 cfg.buckets[THROTTLE_OPS_READ].max =
478 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
479 cfg.buckets[THROTTLE_OPS_WRITE].max =
480 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
481
482 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
483
484 if (!check_throttle_config(&cfg, &error)) {
485 error_propagate(errp, error);
486 goto early_err;
487 }
488
489 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
490 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
491 on_write_error = parse_block_error_action(buf, 0, &error);
492 if (error) {
493 error_propagate(errp, error);
494 goto early_err;
495 }
496 }
497
498 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
499 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
500 on_read_error = parse_block_error_action(buf, 1, &error);
501 if (error) {
502 error_propagate(errp, error);
503 goto early_err;
504 }
505 }
506
507 detect_zeroes =
508 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
509 qemu_opt_get(opts, "detect-zeroes"),
510 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
511 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
512 &error);
513 if (error) {
514 error_propagate(errp, error);
515 goto early_err;
516 }
517
518 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
519 !(bdrv_flags & BDRV_O_UNMAP)) {
520 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
521 "without setting discard operation to unmap");
522 goto early_err;
523 }
524
525 /* init */
526 bs = bdrv_new(qemu_opts_id(opts), errp);
527 if (!bs) {
528 goto early_err;
529 }
530 bs->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
531 bs->read_only = ro;
532 bs->detect_zeroes = detect_zeroes;
533
534 bdrv_set_on_error(bs, on_read_error, on_write_error);
535
536 /* disk I/O throttling */
537 if (throttle_enabled(&cfg)) {
538 bdrv_io_limits_enable(bs);
539 bdrv_set_io_limits(bs, &cfg);
540 }
541
542 dinfo = g_malloc0(sizeof(*dinfo));
543 dinfo->id = g_strdup(qemu_opts_id(opts));
544 dinfo->bdrv = bs;
545 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
546
547 if (!file || !*file) {
548 if (has_driver_specific_opts) {
549 file = NULL;
550 } else {
551 QDECREF(bs_opts);
552 qemu_opts_del(opts);
553 return dinfo;
554 }
555 }
556 if (snapshot) {
557 /* always use cache=unsafe with snapshot */
558 bdrv_flags &= ~BDRV_O_CACHE_MASK;
559 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
560 }
561
562 if (copy_on_read) {
563 bdrv_flags |= BDRV_O_COPY_ON_READ;
564 }
565
566 if (runstate_check(RUN_STATE_INMIGRATE)) {
567 bdrv_flags |= BDRV_O_INCOMING;
568 }
569
570 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
571
572 QINCREF(bs_opts);
573 ret = bdrv_open(&bs, file, NULL, bs_opts, bdrv_flags, drv, &error);
574 assert(bs == dinfo->bdrv);
575
576 if (ret < 0) {
577 error_setg(errp, "could not open disk image %s: %s",
578 file ?: dinfo->id, error_get_pretty(error));
579 error_free(error);
580 goto err;
581 }
582
583 if (bdrv_key_required(bs)) {
584 autostart = 0;
585 }
586
587 QDECREF(bs_opts);
588 qemu_opts_del(opts);
589
590 return dinfo;
591
592 err:
593 bdrv_unref(bs);
594 early_err:
595 qemu_opts_del(opts);
596 err_no_opts:
597 QDECREF(bs_opts);
598 return NULL;
599 }
600
601 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
602 Error **errp)
603 {
604 const char *value;
605
606 value = qemu_opt_get(opts, from);
607 if (value) {
608 if (qemu_opt_find(opts, to)) {
609 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
610 "same time", to, from);
611 return;
612 }
613 }
614
615 /* rename all items in opts */
616 while ((value = qemu_opt_get(opts, from))) {
617 qemu_opt_set(opts, to, value);
618 qemu_opt_unset(opts, from);
619 }
620 }
621
622 QemuOptsList qemu_legacy_drive_opts = {
623 .name = "drive",
624 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
625 .desc = {
626 {
627 .name = "bus",
628 .type = QEMU_OPT_NUMBER,
629 .help = "bus number",
630 },{
631 .name = "unit",
632 .type = QEMU_OPT_NUMBER,
633 .help = "unit number (i.e. lun for scsi)",
634 },{
635 .name = "index",
636 .type = QEMU_OPT_NUMBER,
637 .help = "index number",
638 },{
639 .name = "media",
640 .type = QEMU_OPT_STRING,
641 .help = "media type (disk, cdrom)",
642 },{
643 .name = "if",
644 .type = QEMU_OPT_STRING,
645 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
646 },{
647 .name = "cyls",
648 .type = QEMU_OPT_NUMBER,
649 .help = "number of cylinders (ide disk geometry)",
650 },{
651 .name = "heads",
652 .type = QEMU_OPT_NUMBER,
653 .help = "number of heads (ide disk geometry)",
654 },{
655 .name = "secs",
656 .type = QEMU_OPT_NUMBER,
657 .help = "number of sectors (ide disk geometry)",
658 },{
659 .name = "trans",
660 .type = QEMU_OPT_STRING,
661 .help = "chs translation (auto, lba, none)",
662 },{
663 .name = "boot",
664 .type = QEMU_OPT_BOOL,
665 .help = "(deprecated, ignored)",
666 },{
667 .name = "addr",
668 .type = QEMU_OPT_STRING,
669 .help = "pci address (virtio only)",
670 },{
671 .name = "serial",
672 .type = QEMU_OPT_STRING,
673 .help = "disk serial number",
674 },{
675 .name = "file",
676 .type = QEMU_OPT_STRING,
677 .help = "file name",
678 },
679
680 /* Options that are passed on, but have special semantics with -drive */
681 {
682 .name = "read-only",
683 .type = QEMU_OPT_BOOL,
684 .help = "open drive file as read-only",
685 },{
686 .name = "rerror",
687 .type = QEMU_OPT_STRING,
688 .help = "read error action",
689 },{
690 .name = "werror",
691 .type = QEMU_OPT_STRING,
692 .help = "write error action",
693 },{
694 .name = "copy-on-read",
695 .type = QEMU_OPT_BOOL,
696 .help = "copy read data from backing file into image file",
697 },
698
699 { /* end of list */ }
700 },
701 };
702
703 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
704 {
705 const char *value;
706 DriveInfo *dinfo = NULL;
707 QDict *bs_opts;
708 QemuOpts *legacy_opts;
709 DriveMediaType media = MEDIA_DISK;
710 BlockInterfaceType type;
711 int cyls, heads, secs, translation;
712 int max_devs, bus_id, unit_id, index;
713 const char *devaddr;
714 const char *werror, *rerror;
715 bool read_only = false;
716 bool copy_on_read;
717 const char *serial;
718 const char *filename;
719 Error *local_err = NULL;
720 int i;
721
722 /* Change legacy command line options into QMP ones */
723 static const struct {
724 const char *from;
725 const char *to;
726 } opt_renames[] = {
727 { "iops", "throttling.iops-total" },
728 { "iops_rd", "throttling.iops-read" },
729 { "iops_wr", "throttling.iops-write" },
730
731 { "bps", "throttling.bps-total" },
732 { "bps_rd", "throttling.bps-read" },
733 { "bps_wr", "throttling.bps-write" },
734
735 { "iops_max", "throttling.iops-total-max" },
736 { "iops_rd_max", "throttling.iops-read-max" },
737 { "iops_wr_max", "throttling.iops-write-max" },
738
739 { "bps_max", "throttling.bps-total-max" },
740 { "bps_rd_max", "throttling.bps-read-max" },
741 { "bps_wr_max", "throttling.bps-write-max" },
742
743 { "iops_size", "throttling.iops-size" },
744
745 { "readonly", "read-only" },
746 };
747
748 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
749 qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
750 &local_err);
751 if (local_err) {
752 error_report("%s", error_get_pretty(local_err));
753 error_free(local_err);
754 return NULL;
755 }
756 }
757
758 value = qemu_opt_get(all_opts, "cache");
759 if (value) {
760 int flags = 0;
761
762 if (bdrv_parse_cache_flags(value, &flags) != 0) {
763 error_report("invalid cache option");
764 return NULL;
765 }
766
767 /* Specific options take precedence */
768 if (!qemu_opt_get(all_opts, "cache.writeback")) {
769 qemu_opt_set_bool(all_opts, "cache.writeback",
770 !!(flags & BDRV_O_CACHE_WB));
771 }
772 if (!qemu_opt_get(all_opts, "cache.direct")) {
773 qemu_opt_set_bool(all_opts, "cache.direct",
774 !!(flags & BDRV_O_NOCACHE));
775 }
776 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
777 qemu_opt_set_bool(all_opts, "cache.no-flush",
778 !!(flags & BDRV_O_NO_FLUSH));
779 }
780 qemu_opt_unset(all_opts, "cache");
781 }
782
783 /* Get a QDict for processing the options */
784 bs_opts = qdict_new();
785 qemu_opts_to_qdict(all_opts, bs_opts);
786
787 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
788 &error_abort);
789 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
790 if (local_err) {
791 error_report("%s", error_get_pretty(local_err));
792 error_free(local_err);
793 goto fail;
794 }
795
796 /* Deprecated option boot=[on|off] */
797 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
798 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
799 "ignored. Future versions will reject this parameter. Please "
800 "update your scripts.\n");
801 }
802
803 /* Media type */
804 value = qemu_opt_get(legacy_opts, "media");
805 if (value) {
806 if (!strcmp(value, "disk")) {
807 media = MEDIA_DISK;
808 } else if (!strcmp(value, "cdrom")) {
809 media = MEDIA_CDROM;
810 read_only = true;
811 } else {
812 error_report("'%s' invalid media", value);
813 goto fail;
814 }
815 }
816
817 /* copy-on-read is disabled with a warning for read-only devices */
818 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
819 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
820
821 if (read_only && copy_on_read) {
822 error_report("warning: disabling copy-on-read on read-only drive");
823 copy_on_read = false;
824 }
825
826 qdict_put(bs_opts, "read-only",
827 qstring_from_str(read_only ? "on" : "off"));
828 qdict_put(bs_opts, "copy-on-read",
829 qstring_from_str(copy_on_read ? "on" :"off"));
830
831 /* Controller type */
832 value = qemu_opt_get(legacy_opts, "if");
833 if (value) {
834 for (type = 0;
835 type < IF_COUNT && strcmp(value, if_name[type]);
836 type++) {
837 }
838 if (type == IF_COUNT) {
839 error_report("unsupported bus type '%s'", value);
840 goto fail;
841 }
842 } else {
843 type = block_default_type;
844 }
845
846 /* Geometry */
847 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
848 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
849 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
850
851 if (cyls || heads || secs) {
852 if (cyls < 1) {
853 error_report("invalid physical cyls number");
854 goto fail;
855 }
856 if (heads < 1) {
857 error_report("invalid physical heads number");
858 goto fail;
859 }
860 if (secs < 1) {
861 error_report("invalid physical secs number");
862 goto fail;
863 }
864 }
865
866 translation = BIOS_ATA_TRANSLATION_AUTO;
867 value = qemu_opt_get(legacy_opts, "trans");
868 if (value != NULL) {
869 if (!cyls) {
870 error_report("'%s' trans must be used with cyls, heads and secs",
871 value);
872 goto fail;
873 }
874 if (!strcmp(value, "none")) {
875 translation = BIOS_ATA_TRANSLATION_NONE;
876 } else if (!strcmp(value, "lba")) {
877 translation = BIOS_ATA_TRANSLATION_LBA;
878 } else if (!strcmp(value, "large")) {
879 translation = BIOS_ATA_TRANSLATION_LARGE;
880 } else if (!strcmp(value, "rechs")) {
881 translation = BIOS_ATA_TRANSLATION_RECHS;
882 } else if (!strcmp(value, "auto")) {
883 translation = BIOS_ATA_TRANSLATION_AUTO;
884 } else {
885 error_report("'%s' invalid translation type", value);
886 goto fail;
887 }
888 }
889
890 if (media == MEDIA_CDROM) {
891 if (cyls || secs || heads) {
892 error_report("CHS can't be set with media=cdrom");
893 goto fail;
894 }
895 }
896
897 /* Device address specified by bus/unit or index.
898 * If none was specified, try to find the first free one. */
899 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
900 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
901 index = qemu_opt_get_number(legacy_opts, "index", -1);
902
903 max_devs = if_max_devs[type];
904
905 if (index != -1) {
906 if (bus_id != 0 || unit_id != -1) {
907 error_report("index cannot be used with bus and unit");
908 goto fail;
909 }
910 bus_id = drive_index_to_bus_id(type, index);
911 unit_id = drive_index_to_unit_id(type, index);
912 }
913
914 if (unit_id == -1) {
915 unit_id = 0;
916 while (drive_get(type, bus_id, unit_id) != NULL) {
917 unit_id++;
918 if (max_devs && unit_id >= max_devs) {
919 unit_id -= max_devs;
920 bus_id++;
921 }
922 }
923 }
924
925 if (max_devs && unit_id >= max_devs) {
926 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
927 goto fail;
928 }
929
930 if (drive_get(type, bus_id, unit_id) != NULL) {
931 error_report("drive with bus=%d, unit=%d (index=%d) exists",
932 bus_id, unit_id, index);
933 goto fail;
934 }
935
936 /* Serial number */
937 serial = qemu_opt_get(legacy_opts, "serial");
938
939 /* no id supplied -> create one */
940 if (qemu_opts_id(all_opts) == NULL) {
941 char *new_id;
942 const char *mediastr = "";
943 if (type == IF_IDE || type == IF_SCSI) {
944 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
945 }
946 if (max_devs) {
947 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
948 mediastr, unit_id);
949 } else {
950 new_id = g_strdup_printf("%s%s%i", if_name[type],
951 mediastr, unit_id);
952 }
953 qdict_put(bs_opts, "id", qstring_from_str(new_id));
954 g_free(new_id);
955 }
956
957 /* Add virtio block device */
958 devaddr = qemu_opt_get(legacy_opts, "addr");
959 if (devaddr && type != IF_VIRTIO) {
960 error_report("addr is not supported by this bus type");
961 goto fail;
962 }
963
964 if (type == IF_VIRTIO) {
965 QemuOpts *devopts;
966 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
967 &error_abort);
968 if (arch_type == QEMU_ARCH_S390X) {
969 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
970 } else {
971 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
972 }
973 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
974 if (devaddr) {
975 qemu_opt_set(devopts, "addr", devaddr);
976 }
977 }
978
979 filename = qemu_opt_get(legacy_opts, "file");
980
981 /* Check werror/rerror compatibility with if=... */
982 werror = qemu_opt_get(legacy_opts, "werror");
983 if (werror != NULL) {
984 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
985 type != IF_NONE) {
986 error_report("werror is not supported by this bus type");
987 goto fail;
988 }
989 qdict_put(bs_opts, "werror", qstring_from_str(werror));
990 }
991
992 rerror = qemu_opt_get(legacy_opts, "rerror");
993 if (rerror != NULL) {
994 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
995 type != IF_NONE) {
996 error_report("rerror is not supported by this bus type");
997 goto fail;
998 }
999 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
1000 }
1001
1002 /* Actual block device init: Functionality shared with blockdev-add */
1003 dinfo = blockdev_init(filename, bs_opts, &local_err);
1004 bs_opts = NULL;
1005 if (dinfo == NULL) {
1006 if (local_err) {
1007 error_report("%s", error_get_pretty(local_err));
1008 error_free(local_err);
1009 }
1010 goto fail;
1011 } else {
1012 assert(!local_err);
1013 }
1014
1015 /* Set legacy DriveInfo fields */
1016 dinfo->enable_auto_del = true;
1017 dinfo->opts = all_opts;
1018
1019 dinfo->cyls = cyls;
1020 dinfo->heads = heads;
1021 dinfo->secs = secs;
1022 dinfo->trans = translation;
1023
1024 dinfo->type = type;
1025 dinfo->bus = bus_id;
1026 dinfo->unit = unit_id;
1027 dinfo->devaddr = devaddr;
1028
1029 dinfo->serial = g_strdup(serial);
1030
1031 switch(type) {
1032 case IF_IDE:
1033 case IF_SCSI:
1034 case IF_XEN:
1035 case IF_NONE:
1036 dinfo->media_cd = media == MEDIA_CDROM;
1037 break;
1038 default:
1039 break;
1040 }
1041
1042 fail:
1043 qemu_opts_del(legacy_opts);
1044 QDECREF(bs_opts);
1045 return dinfo;
1046 }
1047
1048 void do_commit(Monitor *mon, const QDict *qdict)
1049 {
1050 const char *device = qdict_get_str(qdict, "device");
1051 BlockDriverState *bs;
1052 int ret;
1053
1054 if (!strcmp(device, "all")) {
1055 ret = bdrv_commit_all();
1056 } else {
1057 bs = bdrv_find(device);
1058 if (!bs) {
1059 monitor_printf(mon, "Device '%s' not found\n", device);
1060 return;
1061 }
1062 ret = bdrv_commit(bs);
1063 }
1064 if (ret < 0) {
1065 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1066 strerror(-ret));
1067 }
1068 }
1069
1070 static void blockdev_do_action(int kind, void *data, Error **errp)
1071 {
1072 TransactionAction action;
1073 TransactionActionList list;
1074
1075 action.kind = kind;
1076 action.data = data;
1077 list.value = &action;
1078 list.next = NULL;
1079 qmp_transaction(&list, errp);
1080 }
1081
1082 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1083 bool has_node_name, const char *node_name,
1084 const char *snapshot_file,
1085 bool has_snapshot_node_name,
1086 const char *snapshot_node_name,
1087 bool has_format, const char *format,
1088 bool has_mode, NewImageMode mode, Error **errp)
1089 {
1090 BlockdevSnapshot snapshot = {
1091 .has_device = has_device,
1092 .device = (char *) device,
1093 .has_node_name = has_node_name,
1094 .node_name = (char *) node_name,
1095 .snapshot_file = (char *) snapshot_file,
1096 .has_snapshot_node_name = has_snapshot_node_name,
1097 .snapshot_node_name = (char *) snapshot_node_name,
1098 .has_format = has_format,
1099 .format = (char *) format,
1100 .has_mode = has_mode,
1101 .mode = mode,
1102 };
1103 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1104 &snapshot, errp);
1105 }
1106
1107 void qmp_blockdev_snapshot_internal_sync(const char *device,
1108 const char *name,
1109 Error **errp)
1110 {
1111 BlockdevSnapshotInternal snapshot = {
1112 .device = (char *) device,
1113 .name = (char *) name
1114 };
1115
1116 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1117 &snapshot, errp);
1118 }
1119
1120 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1121 bool has_id,
1122 const char *id,
1123 bool has_name,
1124 const char *name,
1125 Error **errp)
1126 {
1127 BlockDriverState *bs = bdrv_find(device);
1128 QEMUSnapshotInfo sn;
1129 Error *local_err = NULL;
1130 SnapshotInfo *info = NULL;
1131 int ret;
1132
1133 if (!bs) {
1134 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1135 return NULL;
1136 }
1137
1138 if (!has_id) {
1139 id = NULL;
1140 }
1141
1142 if (!has_name) {
1143 name = NULL;
1144 }
1145
1146 if (!id && !name) {
1147 error_setg(errp, "Name or id must be provided");
1148 return NULL;
1149 }
1150
1151 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1152 if (local_err) {
1153 error_propagate(errp, local_err);
1154 return NULL;
1155 }
1156 if (!ret) {
1157 error_setg(errp,
1158 "Snapshot with id '%s' and name '%s' does not exist on "
1159 "device '%s'",
1160 STR_OR_NULL(id), STR_OR_NULL(name), device);
1161 return NULL;
1162 }
1163
1164 bdrv_snapshot_delete(bs, id, name, &local_err);
1165 if (local_err) {
1166 error_propagate(errp, local_err);
1167 return NULL;
1168 }
1169
1170 info = g_new0(SnapshotInfo, 1);
1171 info->id = g_strdup(sn.id_str);
1172 info->name = g_strdup(sn.name);
1173 info->date_nsec = sn.date_nsec;
1174 info->date_sec = sn.date_sec;
1175 info->vm_state_size = sn.vm_state_size;
1176 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1177 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1178
1179 return info;
1180 }
1181
1182 /* New and old BlockDriverState structs for group snapshots */
1183
1184 typedef struct BlkTransactionState BlkTransactionState;
1185
1186 /* Only prepare() may fail. In a single transaction, only one of commit() or
1187 abort() will be called, clean() will always be called if it present. */
1188 typedef struct BdrvActionOps {
1189 /* Size of state struct, in bytes. */
1190 size_t instance_size;
1191 /* Prepare the work, must NOT be NULL. */
1192 void (*prepare)(BlkTransactionState *common, Error **errp);
1193 /* Commit the changes, can be NULL. */
1194 void (*commit)(BlkTransactionState *common);
1195 /* Abort the changes on fail, can be NULL. */
1196 void (*abort)(BlkTransactionState *common);
1197 /* Clean up resource in the end, can be NULL. */
1198 void (*clean)(BlkTransactionState *common);
1199 } BdrvActionOps;
1200
1201 /*
1202 * This structure must be arranged as first member in child type, assuming
1203 * that compiler will also arrange it to the same address with parent instance.
1204 * Later it will be used in free().
1205 */
1206 struct BlkTransactionState {
1207 TransactionAction *action;
1208 const BdrvActionOps *ops;
1209 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1210 };
1211
1212 /* internal snapshot private data */
1213 typedef struct InternalSnapshotState {
1214 BlkTransactionState common;
1215 BlockDriverState *bs;
1216 QEMUSnapshotInfo sn;
1217 } InternalSnapshotState;
1218
1219 static void internal_snapshot_prepare(BlkTransactionState *common,
1220 Error **errp)
1221 {
1222 Error *local_err = NULL;
1223 const char *device;
1224 const char *name;
1225 BlockDriverState *bs;
1226 QEMUSnapshotInfo old_sn, *sn;
1227 bool ret;
1228 qemu_timeval tv;
1229 BlockdevSnapshotInternal *internal;
1230 InternalSnapshotState *state;
1231 int ret1;
1232
1233 g_assert(common->action->kind ==
1234 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1235 internal = common->action->blockdev_snapshot_internal_sync;
1236 state = DO_UPCAST(InternalSnapshotState, common, common);
1237
1238 /* 1. parse input */
1239 device = internal->device;
1240 name = internal->name;
1241
1242 /* 2. check for validation */
1243 bs = bdrv_find(device);
1244 if (!bs) {
1245 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1246 return;
1247 }
1248
1249 if (!bdrv_is_inserted(bs)) {
1250 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1251 return;
1252 }
1253
1254 if (bdrv_is_read_only(bs)) {
1255 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1256 return;
1257 }
1258
1259 if (!bdrv_can_snapshot(bs)) {
1260 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1261 bs->drv->format_name, device, "internal snapshot");
1262 return;
1263 }
1264
1265 if (!strlen(name)) {
1266 error_setg(errp, "Name is empty");
1267 return;
1268 }
1269
1270 /* check whether a snapshot with name exist */
1271 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1272 &local_err);
1273 if (local_err) {
1274 error_propagate(errp, local_err);
1275 return;
1276 } else if (ret) {
1277 error_setg(errp,
1278 "Snapshot with name '%s' already exists on device '%s'",
1279 name, device);
1280 return;
1281 }
1282
1283 /* 3. take the snapshot */
1284 sn = &state->sn;
1285 pstrcpy(sn->name, sizeof(sn->name), name);
1286 qemu_gettimeofday(&tv);
1287 sn->date_sec = tv.tv_sec;
1288 sn->date_nsec = tv.tv_usec * 1000;
1289 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1290
1291 ret1 = bdrv_snapshot_create(bs, sn);
1292 if (ret1 < 0) {
1293 error_setg_errno(errp, -ret1,
1294 "Failed to create snapshot '%s' on device '%s'",
1295 name, device);
1296 return;
1297 }
1298
1299 /* 4. succeed, mark a snapshot is created */
1300 state->bs = bs;
1301 }
1302
1303 static void internal_snapshot_abort(BlkTransactionState *common)
1304 {
1305 InternalSnapshotState *state =
1306 DO_UPCAST(InternalSnapshotState, common, common);
1307 BlockDriverState *bs = state->bs;
1308 QEMUSnapshotInfo *sn = &state->sn;
1309 Error *local_error = NULL;
1310
1311 if (!bs) {
1312 return;
1313 }
1314
1315 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1316 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1317 "device '%s' in abort: %s",
1318 sn->id_str,
1319 sn->name,
1320 bdrv_get_device_name(bs),
1321 error_get_pretty(local_error));
1322 error_free(local_error);
1323 }
1324 }
1325
1326 /* external snapshot private data */
1327 typedef struct ExternalSnapshotState {
1328 BlkTransactionState common;
1329 BlockDriverState *old_bs;
1330 BlockDriverState *new_bs;
1331 } ExternalSnapshotState;
1332
1333 static void external_snapshot_prepare(BlkTransactionState *common,
1334 Error **errp)
1335 {
1336 BlockDriver *drv;
1337 int flags, ret;
1338 QDict *options = NULL;
1339 Error *local_err = NULL;
1340 bool has_device = false;
1341 const char *device;
1342 bool has_node_name = false;
1343 const char *node_name;
1344 bool has_snapshot_node_name = false;
1345 const char *snapshot_node_name;
1346 const char *new_image_file;
1347 const char *format = "qcow2";
1348 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1349 ExternalSnapshotState *state =
1350 DO_UPCAST(ExternalSnapshotState, common, common);
1351 TransactionAction *action = common->action;
1352
1353 /* get parameters */
1354 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1355
1356 has_device = action->blockdev_snapshot_sync->has_device;
1357 device = action->blockdev_snapshot_sync->device;
1358 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1359 node_name = action->blockdev_snapshot_sync->node_name;
1360 has_snapshot_node_name =
1361 action->blockdev_snapshot_sync->has_snapshot_node_name;
1362 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1363
1364 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1365 if (action->blockdev_snapshot_sync->has_format) {
1366 format = action->blockdev_snapshot_sync->format;
1367 }
1368 if (action->blockdev_snapshot_sync->has_mode) {
1369 mode = action->blockdev_snapshot_sync->mode;
1370 }
1371
1372 /* start processing */
1373 drv = bdrv_find_format(format);
1374 if (!drv) {
1375 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1376 return;
1377 }
1378
1379 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1380 has_node_name ? node_name : NULL,
1381 &local_err);
1382 if (local_err) {
1383 error_propagate(errp, local_err);
1384 return;
1385 }
1386
1387 if (has_node_name && !has_snapshot_node_name) {
1388 error_setg(errp, "New snapshot node name missing");
1389 return;
1390 }
1391
1392 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1393 error_setg(errp, "New snapshot node name already existing");
1394 return;
1395 }
1396
1397 if (!bdrv_is_inserted(state->old_bs)) {
1398 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1399 return;
1400 }
1401
1402 if (bdrv_op_is_blocked(state->old_bs,
1403 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1404 return;
1405 }
1406
1407 if (!bdrv_is_read_only(state->old_bs)) {
1408 if (bdrv_flush(state->old_bs)) {
1409 error_set(errp, QERR_IO_ERROR);
1410 return;
1411 }
1412 }
1413
1414 if (!bdrv_is_first_non_filter(state->old_bs)) {
1415 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1416 return;
1417 }
1418
1419 flags = state->old_bs->open_flags;
1420
1421 /* create new image w/backing file */
1422 if (mode != NEW_IMAGE_MODE_EXISTING) {
1423 bdrv_img_create(new_image_file, format,
1424 state->old_bs->filename,
1425 state->old_bs->drv->format_name,
1426 NULL, -1, flags, &local_err, false);
1427 if (local_err) {
1428 error_propagate(errp, local_err);
1429 return;
1430 }
1431 }
1432
1433 if (has_snapshot_node_name) {
1434 options = qdict_new();
1435 qdict_put(options, "node-name",
1436 qstring_from_str(snapshot_node_name));
1437 }
1438
1439 /* TODO Inherit bs->options or only take explicit options with an
1440 * extended QMP command? */
1441 assert(state->new_bs == NULL);
1442 ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1443 flags | BDRV_O_NO_BACKING, drv, &local_err);
1444 /* We will manually add the backing_hd field to the bs later */
1445 if (ret != 0) {
1446 error_propagate(errp, local_err);
1447 }
1448 }
1449
1450 static void external_snapshot_commit(BlkTransactionState *common)
1451 {
1452 ExternalSnapshotState *state =
1453 DO_UPCAST(ExternalSnapshotState, common, common);
1454
1455 /* This removes our old bs and adds the new bs */
1456 bdrv_append(state->new_bs, state->old_bs);
1457 /* We don't need (or want) to use the transactional
1458 * bdrv_reopen_multiple() across all the entries at once, because we
1459 * don't want to abort all of them if one of them fails the reopen */
1460 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1461 NULL);
1462 }
1463
1464 static void external_snapshot_abort(BlkTransactionState *common)
1465 {
1466 ExternalSnapshotState *state =
1467 DO_UPCAST(ExternalSnapshotState, common, common);
1468 if (state->new_bs) {
1469 bdrv_unref(state->new_bs);
1470 }
1471 }
1472
1473 typedef struct DriveBackupState {
1474 BlkTransactionState common;
1475 BlockDriverState *bs;
1476 BlockJob *job;
1477 } DriveBackupState;
1478
1479 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1480 {
1481 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1482 DriveBackup *backup;
1483 Error *local_err = NULL;
1484
1485 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1486 backup = common->action->drive_backup;
1487
1488 qmp_drive_backup(backup->device, backup->target,
1489 backup->has_format, backup->format,
1490 backup->sync,
1491 backup->has_mode, backup->mode,
1492 backup->has_speed, backup->speed,
1493 backup->has_on_source_error, backup->on_source_error,
1494 backup->has_on_target_error, backup->on_target_error,
1495 &local_err);
1496 if (local_err) {
1497 error_propagate(errp, local_err);
1498 state->bs = NULL;
1499 state->job = NULL;
1500 return;
1501 }
1502
1503 state->bs = bdrv_find(backup->device);
1504 state->job = state->bs->job;
1505 }
1506
1507 static void drive_backup_abort(BlkTransactionState *common)
1508 {
1509 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1510 BlockDriverState *bs = state->bs;
1511
1512 /* Only cancel if it's the job we started */
1513 if (bs && bs->job && bs->job == state->job) {
1514 block_job_cancel_sync(bs->job);
1515 }
1516 }
1517
1518 static void abort_prepare(BlkTransactionState *common, Error **errp)
1519 {
1520 error_setg(errp, "Transaction aborted using Abort action");
1521 }
1522
1523 static void abort_commit(BlkTransactionState *common)
1524 {
1525 g_assert_not_reached(); /* this action never succeeds */
1526 }
1527
1528 static const BdrvActionOps actions[] = {
1529 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1530 .instance_size = sizeof(ExternalSnapshotState),
1531 .prepare = external_snapshot_prepare,
1532 .commit = external_snapshot_commit,
1533 .abort = external_snapshot_abort,
1534 },
1535 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1536 .instance_size = sizeof(DriveBackupState),
1537 .prepare = drive_backup_prepare,
1538 .abort = drive_backup_abort,
1539 },
1540 [TRANSACTION_ACTION_KIND_ABORT] = {
1541 .instance_size = sizeof(BlkTransactionState),
1542 .prepare = abort_prepare,
1543 .commit = abort_commit,
1544 },
1545 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1546 .instance_size = sizeof(InternalSnapshotState),
1547 .prepare = internal_snapshot_prepare,
1548 .abort = internal_snapshot_abort,
1549 },
1550 };
1551
1552 /*
1553 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1554 * then we do not pivot any of the devices in the group, and abandon the
1555 * snapshots
1556 */
1557 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1558 {
1559 TransactionActionList *dev_entry = dev_list;
1560 BlkTransactionState *state, *next;
1561 Error *local_err = NULL;
1562
1563 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1564 QSIMPLEQ_INIT(&snap_bdrv_states);
1565
1566 /* drain all i/o before any snapshots */
1567 bdrv_drain_all();
1568
1569 /* We don't do anything in this loop that commits us to the snapshot */
1570 while (NULL != dev_entry) {
1571 TransactionAction *dev_info = NULL;
1572 const BdrvActionOps *ops;
1573
1574 dev_info = dev_entry->value;
1575 dev_entry = dev_entry->next;
1576
1577 assert(dev_info->kind < ARRAY_SIZE(actions));
1578
1579 ops = &actions[dev_info->kind];
1580 assert(ops->instance_size > 0);
1581
1582 state = g_malloc0(ops->instance_size);
1583 state->ops = ops;
1584 state->action = dev_info;
1585 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1586
1587 state->ops->prepare(state, &local_err);
1588 if (local_err) {
1589 error_propagate(errp, local_err);
1590 goto delete_and_fail;
1591 }
1592 }
1593
1594 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1595 if (state->ops->commit) {
1596 state->ops->commit(state);
1597 }
1598 }
1599
1600 /* success */
1601 goto exit;
1602
1603 delete_and_fail:
1604 /*
1605 * failure, and it is all-or-none; abandon each new bs, and keep using
1606 * the original bs for all images
1607 */
1608 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1609 if (state->ops->abort) {
1610 state->ops->abort(state);
1611 }
1612 }
1613 exit:
1614 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1615 if (state->ops->clean) {
1616 state->ops->clean(state);
1617 }
1618 g_free(state);
1619 }
1620 }
1621
1622
1623 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1624 {
1625 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1626 return;
1627 }
1628 if (!bdrv_dev_has_removable_media(bs)) {
1629 error_setg(errp, "Device '%s' is not removable",
1630 bdrv_get_device_name(bs));
1631 return;
1632 }
1633
1634 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1635 bdrv_dev_eject_request(bs, force);
1636 if (!force) {
1637 error_setg(errp, "Device '%s' is locked",
1638 bdrv_get_device_name(bs));
1639 return;
1640 }
1641 }
1642
1643 bdrv_close(bs);
1644 }
1645
1646 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1647 {
1648 BlockDriverState *bs;
1649
1650 bs = bdrv_find(device);
1651 if (!bs) {
1652 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1653 return;
1654 }
1655
1656 eject_device(bs, force, errp);
1657 }
1658
1659 void qmp_block_passwd(bool has_device, const char *device,
1660 bool has_node_name, const char *node_name,
1661 const char *password, Error **errp)
1662 {
1663 Error *local_err = NULL;
1664 BlockDriverState *bs;
1665 int err;
1666
1667 bs = bdrv_lookup_bs(has_device ? device : NULL,
1668 has_node_name ? node_name : NULL,
1669 &local_err);
1670 if (local_err) {
1671 error_propagate(errp, local_err);
1672 return;
1673 }
1674
1675 err = bdrv_set_key(bs, password);
1676 if (err == -EINVAL) {
1677 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1678 return;
1679 } else if (err < 0) {
1680 error_set(errp, QERR_INVALID_PASSWORD);
1681 return;
1682 }
1683 }
1684
1685 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1686 int bdrv_flags, BlockDriver *drv,
1687 const char *password, Error **errp)
1688 {
1689 Error *local_err = NULL;
1690 int ret;
1691
1692 ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1693 if (ret < 0) {
1694 error_propagate(errp, local_err);
1695 return;
1696 }
1697
1698 if (bdrv_key_required(bs)) {
1699 if (password) {
1700 if (bdrv_set_key(bs, password) < 0) {
1701 error_set(errp, QERR_INVALID_PASSWORD);
1702 }
1703 } else {
1704 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1705 bdrv_get_encrypted_filename(bs));
1706 }
1707 } else if (password) {
1708 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1709 }
1710 }
1711
1712 void qmp_change_blockdev(const char *device, const char *filename,
1713 const char *format, Error **errp)
1714 {
1715 BlockDriverState *bs;
1716 BlockDriver *drv = NULL;
1717 int bdrv_flags;
1718 Error *err = NULL;
1719
1720 bs = bdrv_find(device);
1721 if (!bs) {
1722 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1723 return;
1724 }
1725
1726 if (format) {
1727 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1728 if (!drv) {
1729 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1730 return;
1731 }
1732 }
1733
1734 eject_device(bs, 0, &err);
1735 if (err) {
1736 error_propagate(errp, err);
1737 return;
1738 }
1739
1740 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1741 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1742
1743 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1744 }
1745
1746 /* throttling disk I/O limits */
1747 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1748 int64_t bps_wr,
1749 int64_t iops,
1750 int64_t iops_rd,
1751 int64_t iops_wr,
1752 bool has_bps_max,
1753 int64_t bps_max,
1754 bool has_bps_rd_max,
1755 int64_t bps_rd_max,
1756 bool has_bps_wr_max,
1757 int64_t bps_wr_max,
1758 bool has_iops_max,
1759 int64_t iops_max,
1760 bool has_iops_rd_max,
1761 int64_t iops_rd_max,
1762 bool has_iops_wr_max,
1763 int64_t iops_wr_max,
1764 bool has_iops_size,
1765 int64_t iops_size, Error **errp)
1766 {
1767 ThrottleConfig cfg;
1768 BlockDriverState *bs;
1769 AioContext *aio_context;
1770
1771 bs = bdrv_find(device);
1772 if (!bs) {
1773 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1774 return;
1775 }
1776
1777 memset(&cfg, 0, sizeof(cfg));
1778 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1779 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1780 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1781
1782 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1783 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1784 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1785
1786 if (has_bps_max) {
1787 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1788 }
1789 if (has_bps_rd_max) {
1790 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1791 }
1792 if (has_bps_wr_max) {
1793 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1794 }
1795 if (has_iops_max) {
1796 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1797 }
1798 if (has_iops_rd_max) {
1799 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1800 }
1801 if (has_iops_wr_max) {
1802 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1803 }
1804
1805 if (has_iops_size) {
1806 cfg.op_size = iops_size;
1807 }
1808
1809 if (!check_throttle_config(&cfg, errp)) {
1810 return;
1811 }
1812
1813 aio_context = bdrv_get_aio_context(bs);
1814 aio_context_acquire(aio_context);
1815
1816 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1817 bdrv_io_limits_enable(bs);
1818 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1819 bdrv_io_limits_disable(bs);
1820 }
1821
1822 if (bs->io_limits_enabled) {
1823 bdrv_set_io_limits(bs, &cfg);
1824 }
1825
1826 aio_context_release(aio_context);
1827 }
1828
1829 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1830 {
1831 const char *id = qdict_get_str(qdict, "id");
1832 BlockDriverState *bs;
1833 DriveInfo *dinfo;
1834 AioContext *aio_context;
1835 Error *local_err = NULL;
1836
1837 bs = bdrv_find(id);
1838 if (!bs) {
1839 error_report("Device '%s' not found", id);
1840 return -1;
1841 }
1842
1843 dinfo = drive_get_by_blockdev(bs);
1844 if (dinfo && !dinfo->enable_auto_del) {
1845 error_report("Deleting device added with blockdev-add"
1846 " is not supported");
1847 return -1;
1848 }
1849
1850 aio_context = bdrv_get_aio_context(bs);
1851 aio_context_acquire(aio_context);
1852
1853 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1854 error_report("%s", error_get_pretty(local_err));
1855 error_free(local_err);
1856 aio_context_release(aio_context);
1857 return -1;
1858 }
1859
1860 /* quiesce block driver; prevent further io */
1861 bdrv_drain_all();
1862 bdrv_flush(bs);
1863 bdrv_close(bs);
1864
1865 /* if we have a device attached to this BlockDriverState
1866 * then we need to make the drive anonymous until the device
1867 * can be removed. If this is a drive with no device backing
1868 * then we can just get rid of the block driver state right here.
1869 */
1870 if (bdrv_get_attached_dev(bs)) {
1871 bdrv_make_anon(bs);
1872
1873 /* Further I/O must not pause the guest */
1874 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1875 BLOCKDEV_ON_ERROR_REPORT);
1876 } else {
1877 drive_del(dinfo);
1878 }
1879
1880 aio_context_release(aio_context);
1881 return 0;
1882 }
1883
1884 void qmp_block_resize(bool has_device, const char *device,
1885 bool has_node_name, const char *node_name,
1886 int64_t size, Error **errp)
1887 {
1888 Error *local_err = NULL;
1889 BlockDriverState *bs;
1890 AioContext *aio_context;
1891 int ret;
1892
1893 bs = bdrv_lookup_bs(has_device ? device : NULL,
1894 has_node_name ? node_name : NULL,
1895 &local_err);
1896 if (local_err) {
1897 error_propagate(errp, local_err);
1898 return;
1899 }
1900
1901 aio_context = bdrv_get_aio_context(bs);
1902 aio_context_acquire(aio_context);
1903
1904 if (!bdrv_is_first_non_filter(bs)) {
1905 error_set(errp, QERR_FEATURE_DISABLED, "resize");
1906 goto out;
1907 }
1908
1909 if (size < 0) {
1910 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1911 goto out;
1912 }
1913
1914 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
1915 error_set(errp, QERR_DEVICE_IN_USE, device);
1916 goto out;
1917 }
1918
1919 /* complete all in-flight operations before resizing the device */
1920 bdrv_drain_all();
1921
1922 ret = bdrv_truncate(bs, size);
1923 switch (ret) {
1924 case 0:
1925 break;
1926 case -ENOMEDIUM:
1927 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1928 break;
1929 case -ENOTSUP:
1930 error_set(errp, QERR_UNSUPPORTED);
1931 break;
1932 case -EACCES:
1933 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1934 break;
1935 case -EBUSY:
1936 error_set(errp, QERR_DEVICE_IN_USE, device);
1937 break;
1938 default:
1939 error_setg_errno(errp, -ret, "Could not resize");
1940 break;
1941 }
1942
1943 out:
1944 aio_context_release(aio_context);
1945 }
1946
1947 static void block_job_cb(void *opaque, int ret)
1948 {
1949 BlockDriverState *bs = opaque;
1950 const char *msg = NULL;
1951
1952 trace_block_job_cb(bs, bs->job, ret);
1953
1954 assert(bs->job);
1955
1956 if (ret < 0) {
1957 msg = strerror(-ret);
1958 }
1959
1960 if (block_job_is_cancelled(bs->job)) {
1961 block_job_event_cancelled(bs->job);
1962 } else {
1963 block_job_event_completed(bs->job, msg);
1964 }
1965
1966 bdrv_put_ref_bh_schedule(bs);
1967 }
1968
1969 void qmp_block_stream(const char *device,
1970 bool has_base, const char *base,
1971 bool has_backing_file, const char *backing_file,
1972 bool has_speed, int64_t speed,
1973 bool has_on_error, BlockdevOnError on_error,
1974 Error **errp)
1975 {
1976 BlockDriverState *bs;
1977 BlockDriverState *base_bs = NULL;
1978 Error *local_err = NULL;
1979 const char *base_name = NULL;
1980
1981 if (!has_on_error) {
1982 on_error = BLOCKDEV_ON_ERROR_REPORT;
1983 }
1984
1985 bs = bdrv_find(device);
1986 if (!bs) {
1987 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1988 return;
1989 }
1990
1991 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
1992 return;
1993 }
1994
1995 if (has_base) {
1996 base_bs = bdrv_find_backing_image(bs, base);
1997 if (base_bs == NULL) {
1998 error_set(errp, QERR_BASE_NOT_FOUND, base);
1999 return;
2000 }
2001 base_name = base;
2002 }
2003
2004 /* if we are streaming the entire chain, the result will have no backing
2005 * file, and specifying one is therefore an error */
2006 if (base_bs == NULL && has_backing_file) {
2007 error_setg(errp, "backing file specified, but streaming the "
2008 "entire chain");
2009 return;
2010 }
2011
2012 /* backing_file string overrides base bs filename */
2013 base_name = has_backing_file ? backing_file : base_name;
2014
2015 stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2016 on_error, block_job_cb, bs, &local_err);
2017 if (local_err) {
2018 error_propagate(errp, local_err);
2019 return;
2020 }
2021
2022 trace_qmp_block_stream(bs, bs->job);
2023 }
2024
2025 void qmp_block_commit(const char *device,
2026 bool has_base, const char *base,
2027 bool has_top, const char *top,
2028 bool has_backing_file, const char *backing_file,
2029 bool has_speed, int64_t speed,
2030 Error **errp)
2031 {
2032 BlockDriverState *bs;
2033 BlockDriverState *base_bs, *top_bs;
2034 Error *local_err = NULL;
2035 /* This will be part of the QMP command, if/when the
2036 * BlockdevOnError change for blkmirror makes it in
2037 */
2038 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2039
2040 if (!has_speed) {
2041 speed = 0;
2042 }
2043
2044 /* drain all i/o before commits */
2045 bdrv_drain_all();
2046
2047 /* Important Note:
2048 * libvirt relies on the DeviceNotFound error class in order to probe for
2049 * live commit feature versions; for this to work, we must make sure to
2050 * perform the device lookup before any generic errors that may occur in a
2051 * scenario in which all optional arguments are omitted. */
2052 bs = bdrv_find(device);
2053 if (!bs) {
2054 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2055 return;
2056 }
2057
2058 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, errp)) {
2059 return;
2060 }
2061
2062 /* default top_bs is the active layer */
2063 top_bs = bs;
2064
2065 if (has_top && top) {
2066 if (strcmp(bs->filename, top) != 0) {
2067 top_bs = bdrv_find_backing_image(bs, top);
2068 }
2069 }
2070
2071 if (top_bs == NULL) {
2072 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2073 return;
2074 }
2075
2076 if (has_base && base) {
2077 base_bs = bdrv_find_backing_image(top_bs, base);
2078 } else {
2079 base_bs = bdrv_find_base(top_bs);
2080 }
2081
2082 if (base_bs == NULL) {
2083 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2084 return;
2085 }
2086
2087 /* Do not allow attempts to commit an image into itself */
2088 if (top_bs == base_bs) {
2089 error_setg(errp, "cannot commit an image into itself");
2090 return;
2091 }
2092
2093 if (top_bs == bs) {
2094 if (has_backing_file) {
2095 error_setg(errp, "'backing-file' specified,"
2096 " but 'top' is the active layer");
2097 return;
2098 }
2099 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2100 bs, &local_err);
2101 } else {
2102 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2103 has_backing_file ? backing_file : NULL, &local_err);
2104 }
2105 if (local_err != NULL) {
2106 error_propagate(errp, local_err);
2107 return;
2108 }
2109 }
2110
2111 void qmp_drive_backup(const char *device, const char *target,
2112 bool has_format, const char *format,
2113 enum MirrorSyncMode sync,
2114 bool has_mode, enum NewImageMode mode,
2115 bool has_speed, int64_t speed,
2116 bool has_on_source_error, BlockdevOnError on_source_error,
2117 bool has_on_target_error, BlockdevOnError on_target_error,
2118 Error **errp)
2119 {
2120 BlockDriverState *bs;
2121 BlockDriverState *target_bs;
2122 BlockDriverState *source = NULL;
2123 BlockDriver *drv = NULL;
2124 Error *local_err = NULL;
2125 int flags;
2126 int64_t size;
2127 int ret;
2128
2129 if (!has_speed) {
2130 speed = 0;
2131 }
2132 if (!has_on_source_error) {
2133 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2134 }
2135 if (!has_on_target_error) {
2136 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2137 }
2138 if (!has_mode) {
2139 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2140 }
2141
2142 bs = bdrv_find(device);
2143 if (!bs) {
2144 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2145 return;
2146 }
2147
2148 if (!bdrv_is_inserted(bs)) {
2149 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2150 return;
2151 }
2152
2153 if (!has_format) {
2154 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2155 }
2156 if (format) {
2157 drv = bdrv_find_format(format);
2158 if (!drv) {
2159 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2160 return;
2161 }
2162 }
2163
2164 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2165 return;
2166 }
2167
2168 flags = bs->open_flags | BDRV_O_RDWR;
2169
2170 /* See if we have a backing HD we can use to create our new image
2171 * on top of. */
2172 if (sync == MIRROR_SYNC_MODE_TOP) {
2173 source = bs->backing_hd;
2174 if (!source) {
2175 sync = MIRROR_SYNC_MODE_FULL;
2176 }
2177 }
2178 if (sync == MIRROR_SYNC_MODE_NONE) {
2179 source = bs;
2180 }
2181
2182 size = bdrv_getlength(bs);
2183 if (size < 0) {
2184 error_setg_errno(errp, -size, "bdrv_getlength failed");
2185 return;
2186 }
2187
2188 if (mode != NEW_IMAGE_MODE_EXISTING) {
2189 assert(format && drv);
2190 if (source) {
2191 bdrv_img_create(target, format, source->filename,
2192 source->drv->format_name, NULL,
2193 size, flags, &local_err, false);
2194 } else {
2195 bdrv_img_create(target, format, NULL, NULL, NULL,
2196 size, flags, &local_err, false);
2197 }
2198 }
2199
2200 if (local_err) {
2201 error_propagate(errp, local_err);
2202 return;
2203 }
2204
2205 target_bs = NULL;
2206 ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2207 if (ret < 0) {
2208 error_propagate(errp, local_err);
2209 return;
2210 }
2211
2212 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2213 block_job_cb, bs, &local_err);
2214 if (local_err != NULL) {
2215 bdrv_unref(target_bs);
2216 error_propagate(errp, local_err);
2217 return;
2218 }
2219 }
2220
2221 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2222 {
2223 return bdrv_named_nodes_list();
2224 }
2225
2226 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2227
2228 void qmp_drive_mirror(const char *device, const char *target,
2229 bool has_format, const char *format,
2230 bool has_node_name, const char *node_name,
2231 bool has_replaces, const char *replaces,
2232 enum MirrorSyncMode sync,
2233 bool has_mode, enum NewImageMode mode,
2234 bool has_speed, int64_t speed,
2235 bool has_granularity, uint32_t granularity,
2236 bool has_buf_size, int64_t buf_size,
2237 bool has_on_source_error, BlockdevOnError on_source_error,
2238 bool has_on_target_error, BlockdevOnError on_target_error,
2239 Error **errp)
2240 {
2241 BlockDriverState *bs;
2242 BlockDriverState *source, *target_bs;
2243 BlockDriver *drv = NULL;
2244 Error *local_err = NULL;
2245 QDict *options = NULL;
2246 int flags;
2247 int64_t size;
2248 int ret;
2249
2250 if (!has_speed) {
2251 speed = 0;
2252 }
2253 if (!has_on_source_error) {
2254 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2255 }
2256 if (!has_on_target_error) {
2257 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2258 }
2259 if (!has_mode) {
2260 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2261 }
2262 if (!has_granularity) {
2263 granularity = 0;
2264 }
2265 if (!has_buf_size) {
2266 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2267 }
2268
2269 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2270 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2271 "a value in range [512B, 64MB]");
2272 return;
2273 }
2274 if (granularity & (granularity - 1)) {
2275 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity", "power of 2");
2276 return;
2277 }
2278
2279 bs = bdrv_find(device);
2280 if (!bs) {
2281 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2282 return;
2283 }
2284
2285 if (!bdrv_is_inserted(bs)) {
2286 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2287 return;
2288 }
2289
2290 if (!has_format) {
2291 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2292 }
2293 if (format) {
2294 drv = bdrv_find_format(format);
2295 if (!drv) {
2296 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2297 return;
2298 }
2299 }
2300
2301 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2302 return;
2303 }
2304
2305 flags = bs->open_flags | BDRV_O_RDWR;
2306 source = bs->backing_hd;
2307 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2308 sync = MIRROR_SYNC_MODE_FULL;
2309 }
2310 if (sync == MIRROR_SYNC_MODE_NONE) {
2311 source = bs;
2312 }
2313
2314 size = bdrv_getlength(bs);
2315 if (size < 0) {
2316 error_setg_errno(errp, -size, "bdrv_getlength failed");
2317 return;
2318 }
2319
2320 if (has_replaces) {
2321 BlockDriverState *to_replace_bs;
2322
2323 if (!has_node_name) {
2324 error_setg(errp, "a node-name must be provided when replacing a"
2325 " named node of the graph");
2326 return;
2327 }
2328
2329 to_replace_bs = check_to_replace_node(replaces, &local_err);
2330
2331 if (!to_replace_bs) {
2332 error_propagate(errp, local_err);
2333 return;
2334 }
2335
2336 if (size != bdrv_getlength(to_replace_bs)) {
2337 error_setg(errp, "cannot replace image with a mirror image of "
2338 "different size");
2339 return;
2340 }
2341 }
2342
2343 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2344 && mode != NEW_IMAGE_MODE_EXISTING)
2345 {
2346 /* create new image w/o backing file */
2347 assert(format && drv);
2348 bdrv_img_create(target, format,
2349 NULL, NULL, NULL, size, flags, &local_err, false);
2350 } else {
2351 switch (mode) {
2352 case NEW_IMAGE_MODE_EXISTING:
2353 break;
2354 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2355 /* create new image with backing file */
2356 bdrv_img_create(target, format,
2357 source->filename,
2358 source->drv->format_name,
2359 NULL, size, flags, &local_err, false);
2360 break;
2361 default:
2362 abort();
2363 }
2364 }
2365
2366 if (local_err) {
2367 error_propagate(errp, local_err);
2368 return;
2369 }
2370
2371 if (has_node_name) {
2372 options = qdict_new();
2373 qdict_put(options, "node-name", qstring_from_str(node_name));
2374 }
2375
2376 /* Mirroring takes care of copy-on-write using the source's backing
2377 * file.
2378 */
2379 target_bs = NULL;
2380 ret = bdrv_open(&target_bs, target, NULL, options,
2381 flags | BDRV_O_NO_BACKING, drv, &local_err);
2382 if (ret < 0) {
2383 error_propagate(errp, local_err);
2384 return;
2385 }
2386
2387 /* pass the node name to replace to mirror start since it's loose coupling
2388 * and will allow to check whether the node still exist at mirror completion
2389 */
2390 mirror_start(bs, target_bs,
2391 has_replaces ? replaces : NULL,
2392 speed, granularity, buf_size, sync,
2393 on_source_error, on_target_error,
2394 block_job_cb, bs, &local_err);
2395 if (local_err != NULL) {
2396 bdrv_unref(target_bs);
2397 error_propagate(errp, local_err);
2398 return;
2399 }
2400 }
2401
2402 static BlockJob *find_block_job(const char *device)
2403 {
2404 BlockDriverState *bs;
2405
2406 bs = bdrv_find(device);
2407 if (!bs || !bs->job) {
2408 return NULL;
2409 }
2410 return bs->job;
2411 }
2412
2413 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2414 {
2415 BlockJob *job = find_block_job(device);
2416
2417 if (!job) {
2418 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2419 return;
2420 }
2421
2422 block_job_set_speed(job, speed, errp);
2423 }
2424
2425 void qmp_block_job_cancel(const char *device,
2426 bool has_force, bool force, Error **errp)
2427 {
2428 BlockJob *job = find_block_job(device);
2429
2430 if (!has_force) {
2431 force = false;
2432 }
2433
2434 if (!job) {
2435 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2436 return;
2437 }
2438 if (job->paused && !force) {
2439 error_setg(errp, "The block job for device '%s' is currently paused",
2440 device);
2441 return;
2442 }
2443
2444 trace_qmp_block_job_cancel(job);
2445 block_job_cancel(job);
2446 }
2447
2448 void qmp_block_job_pause(const char *device, Error **errp)
2449 {
2450 BlockJob *job = find_block_job(device);
2451
2452 if (!job) {
2453 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2454 return;
2455 }
2456
2457 trace_qmp_block_job_pause(job);
2458 block_job_pause(job);
2459 }
2460
2461 void qmp_block_job_resume(const char *device, Error **errp)
2462 {
2463 BlockJob *job = find_block_job(device);
2464
2465 if (!job) {
2466 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2467 return;
2468 }
2469
2470 trace_qmp_block_job_resume(job);
2471 block_job_resume(job);
2472 }
2473
2474 void qmp_block_job_complete(const char *device, Error **errp)
2475 {
2476 BlockJob *job = find_block_job(device);
2477
2478 if (!job) {
2479 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2480 return;
2481 }
2482
2483 trace_qmp_block_job_complete(job);
2484 block_job_complete(job, errp);
2485 }
2486
2487 void qmp_change_backing_file(const char *device,
2488 const char *image_node_name,
2489 const char *backing_file,
2490 Error **errp)
2491 {
2492 BlockDriverState *bs = NULL;
2493 BlockDriverState *image_bs = NULL;
2494 Error *local_err = NULL;
2495 bool ro;
2496 int open_flags;
2497 int ret;
2498
2499 /* find the top layer BDS of the chain */
2500 bs = bdrv_find(device);
2501 if (!bs) {
2502 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2503 return;
2504 }
2505
2506 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
2507 if (local_err) {
2508 error_propagate(errp, local_err);
2509 return;
2510 }
2511
2512 if (!image_bs) {
2513 error_setg(errp, "image file not found");
2514 return;
2515 }
2516
2517 if (bdrv_find_base(image_bs) == image_bs) {
2518 error_setg(errp, "not allowing backing file change on an image "
2519 "without a backing file");
2520 return;
2521 }
2522
2523 /* even though we are not necessarily operating on bs, we need it to
2524 * determine if block ops are currently prohibited on the chain */
2525 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
2526 return;
2527 }
2528
2529 /* final sanity check */
2530 if (!bdrv_chain_contains(bs, image_bs)) {
2531 error_setg(errp, "'%s' and image file are not in the same chain",
2532 device);
2533 return;
2534 }
2535
2536 /* if not r/w, reopen to make r/w */
2537 open_flags = image_bs->open_flags;
2538 ro = bdrv_is_read_only(image_bs);
2539
2540 if (ro) {
2541 bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
2542 if (local_err) {
2543 error_propagate(errp, local_err);
2544 return;
2545 }
2546 }
2547
2548 ret = bdrv_change_backing_file(image_bs, backing_file,
2549 image_bs->drv ? image_bs->drv->format_name : "");
2550
2551 if (ret < 0) {
2552 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
2553 backing_file);
2554 /* don't exit here, so we can try to restore open flags if
2555 * appropriate */
2556 }
2557
2558 if (ro) {
2559 bdrv_reopen(image_bs, open_flags, &local_err);
2560 if (local_err) {
2561 error_propagate(errp, local_err); /* will preserve prior errp */
2562 }
2563 }
2564 }
2565
2566 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2567 {
2568 QmpOutputVisitor *ov = qmp_output_visitor_new();
2569 DriveInfo *dinfo;
2570 QObject *obj;
2571 QDict *qdict;
2572 Error *local_err = NULL;
2573
2574 /* Require an ID in the top level */
2575 if (!options->has_id) {
2576 error_setg(errp, "Block device needs an ID");
2577 goto fail;
2578 }
2579
2580 /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
2581 * cache.direct=false instead of silently switching to aio=threads, except
2582 * when called from drive_new().
2583 *
2584 * For now, simply forbidding the combination for all drivers will do. */
2585 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2586 bool direct = options->has_cache &&
2587 options->cache->has_direct &&
2588 options->cache->direct;
2589 if (!direct) {
2590 error_setg(errp, "aio=native requires cache.direct=true");
2591 goto fail;
2592 }
2593 }
2594
2595 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2596 &options, NULL, &local_err);
2597 if (local_err) {
2598 error_propagate(errp, local_err);
2599 goto fail;
2600 }
2601
2602 obj = qmp_output_get_qobject(ov);
2603 qdict = qobject_to_qdict(obj);
2604
2605 qdict_flatten(qdict);
2606
2607 dinfo = blockdev_init(NULL, qdict, &local_err);
2608 if (local_err) {
2609 error_propagate(errp, local_err);
2610 goto fail;
2611 }
2612
2613 if (bdrv_key_required(dinfo->bdrv)) {
2614 drive_del(dinfo);
2615 error_setg(errp, "blockdev-add doesn't support encrypted devices");
2616 goto fail;
2617 }
2618
2619 fail:
2620 qmp_output_visitor_cleanup(ov);
2621 }
2622
2623 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2624 {
2625 BlockJobInfoList **prev = opaque;
2626 BlockJob *job = bs->job;
2627
2628 if (job) {
2629 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2630 elem->value = block_job_query(bs->job);
2631 (*prev)->next = elem;
2632 *prev = elem;
2633 }
2634 }
2635
2636 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2637 {
2638 /* Dummy is a fake list element for holding the head pointer */
2639 BlockJobInfoList dummy = {};
2640 BlockJobInfoList *prev = &dummy;
2641 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2642 return dummy.next;
2643 }
2644
2645 QemuOptsList qemu_common_drive_opts = {
2646 .name = "drive",
2647 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2648 .desc = {
2649 {
2650 .name = "snapshot",
2651 .type = QEMU_OPT_BOOL,
2652 .help = "enable/disable snapshot mode",
2653 },{
2654 .name = "discard",
2655 .type = QEMU_OPT_STRING,
2656 .help = "discard operation (ignore/off, unmap/on)",
2657 },{
2658 .name = "cache.writeback",
2659 .type = QEMU_OPT_BOOL,
2660 .help = "enables writeback mode for any caches",
2661 },{
2662 .name = "cache.direct",
2663 .type = QEMU_OPT_BOOL,
2664 .help = "enables use of O_DIRECT (bypass the host page cache)",
2665 },{
2666 .name = "cache.no-flush",
2667 .type = QEMU_OPT_BOOL,
2668 .help = "ignore any flush requests for the device",
2669 },{
2670 .name = "aio",
2671 .type = QEMU_OPT_STRING,
2672 .help = "host AIO implementation (threads, native)",
2673 },{
2674 .name = "format",
2675 .type = QEMU_OPT_STRING,
2676 .help = "disk format (raw, qcow2, ...)",
2677 },{
2678 .name = "rerror",
2679 .type = QEMU_OPT_STRING,
2680 .help = "read error action",
2681 },{
2682 .name = "werror",
2683 .type = QEMU_OPT_STRING,
2684 .help = "write error action",
2685 },{
2686 .name = "read-only",
2687 .type = QEMU_OPT_BOOL,
2688 .help = "open drive file as read-only",
2689 },{
2690 .name = "throttling.iops-total",
2691 .type = QEMU_OPT_NUMBER,
2692 .help = "limit total I/O operations per second",
2693 },{
2694 .name = "throttling.iops-read",
2695 .type = QEMU_OPT_NUMBER,
2696 .help = "limit read operations per second",
2697 },{
2698 .name = "throttling.iops-write",
2699 .type = QEMU_OPT_NUMBER,
2700 .help = "limit write operations per second",
2701 },{
2702 .name = "throttling.bps-total",
2703 .type = QEMU_OPT_NUMBER,
2704 .help = "limit total bytes per second",
2705 },{
2706 .name = "throttling.bps-read",
2707 .type = QEMU_OPT_NUMBER,
2708 .help = "limit read bytes per second",
2709 },{
2710 .name = "throttling.bps-write",
2711 .type = QEMU_OPT_NUMBER,
2712 .help = "limit write bytes per second",
2713 },{
2714 .name = "throttling.iops-total-max",
2715 .type = QEMU_OPT_NUMBER,
2716 .help = "I/O operations burst",
2717 },{
2718 .name = "throttling.iops-read-max",
2719 .type = QEMU_OPT_NUMBER,
2720 .help = "I/O operations read burst",
2721 },{
2722 .name = "throttling.iops-write-max",
2723 .type = QEMU_OPT_NUMBER,
2724 .help = "I/O operations write burst",
2725 },{
2726 .name = "throttling.bps-total-max",
2727 .type = QEMU_OPT_NUMBER,
2728 .help = "total bytes burst",
2729 },{
2730 .name = "throttling.bps-read-max",
2731 .type = QEMU_OPT_NUMBER,
2732 .help = "total bytes read burst",
2733 },{
2734 .name = "throttling.bps-write-max",
2735 .type = QEMU_OPT_NUMBER,
2736 .help = "total bytes write burst",
2737 },{
2738 .name = "throttling.iops-size",
2739 .type = QEMU_OPT_NUMBER,
2740 .help = "when limiting by iops max size of an I/O in bytes",
2741 },{
2742 .name = "copy-on-read",
2743 .type = QEMU_OPT_BOOL,
2744 .help = "copy read data from backing file into image file",
2745 },{
2746 .name = "detect-zeroes",
2747 .type = QEMU_OPT_STRING,
2748 .help = "try to optimize zero writes (off, on, unmap)",
2749 },
2750 { /* end of list */ }
2751 },
2752 };
2753
2754 QemuOptsList qemu_drive_opts = {
2755 .name = "drive",
2756 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2757 .desc = {
2758 /*
2759 * no elements => accept any params
2760 * validation will happen later
2761 */
2762 { /* end of list */ }
2763 },
2764 };
QEmu