?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'for_upstream' of https://git.kernel.org/pub/scm/virt/kvm/mst/qemu into...
[qemu/ar7.git] / blockdev.c
bloba32bafc07ac48b6d1d8e4050cfe864cb0358bdaa
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 "qemu/osdep.h"
34 #include "sysemu/block-backend.h"
35 #include "sysemu/blockdev.h"
36 #include "hw/block/block.h"
37 #include "block/blockjob.h"
38 #include "block/qdict.h"
39 #include "block/throttle-groups.h"
40 #include "monitor/monitor.h"
41 #include "qemu/error-report.h"
42 #include "qemu/option.h"
43 #include "qemu/qemu-print.h"
44 #include "qemu/config-file.h"
45 #include "qapi/qapi-commands-block.h"
46 #include "qapi/qapi-commands-transaction.h"
47 #include "qapi/qapi-visit-block-core.h"
48 #include "qapi/qmp/qdict.h"
49 #include "qapi/qmp/qnum.h"
50 #include "qapi/qmp/qstring.h"
51 #include "qapi/error.h"
52 #include "qapi/qmp/qerror.h"
53 #include "qapi/qmp/qlist.h"
54 #include "qapi/qobject-output-visitor.h"
55 #include "sysemu/sysemu.h"
56 #include "sysemu/iothread.h"
57 #include "block/block_int.h"
58 #include "block/trace.h"
59 #include "sysemu/runstate.h"
60 #include "sysemu/replay.h"
61 #include "qemu/cutils.h"
62 #include "qemu/help_option.h"
63 #include "qemu/main-loop.h"
64 #include "qemu/throttle-options.h"
65
66 /* Protected by BQL */
67 QTAILQ_HEAD(, BlockDriverState) monitor_bdrv_states =
68 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states);
69
70 void bdrv_set_monitor_owned(BlockDriverState *bs)
71 {
72 GLOBAL_STATE_CODE();
73 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
74 }
75
76 static const char *const if_name[IF_COUNT] = {
77 [IF_NONE] = "none",
78 [IF_IDE] = "ide",
79 [IF_SCSI] = "scsi",
80 [IF_FLOPPY] = "floppy",
81 [IF_PFLASH] = "pflash",
82 [IF_MTD] = "mtd",
83 [IF_SD] = "sd",
84 [IF_VIRTIO] = "virtio",
85 [IF_XEN] = "xen",
86 };
87
88 static int if_max_devs[IF_COUNT] = {
89 /*
90 * Do not change these numbers! They govern how drive option
91 * index maps to unit and bus. That mapping is ABI.
92 *
93 * All controllers used to implement if=T drives need to support
94 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
95 * Otherwise, some index values map to "impossible" bus, unit
96 * values.
97 *
98 * For instance, if you change [IF_SCSI] to 255, -drive
99 * if=scsi,index=12 no longer means bus=1,unit=5, but
100 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
101 * the drive can't be set up. Regression.
102 */
103 [IF_IDE] = 2,
104 [IF_SCSI] = 7,
105 };
106
107 /**
108 * Boards may call this to offer board-by-board overrides
109 * of the default, global values.
110 */
111 void override_max_devs(BlockInterfaceType type, int max_devs)
112 {
113 BlockBackend *blk;
114 DriveInfo *dinfo;
115
116 GLOBAL_STATE_CODE();
117
118 if (max_devs <= 0) {
119 return;
120 }
121
122 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
123 dinfo = blk_legacy_dinfo(blk);
124 if (dinfo->type == type) {
125 fprintf(stderr, "Cannot override units-per-bus property of"
126 " the %s interface, because a drive of that type has"
127 " already been added.\n", if_name[type]);
128 g_assert_not_reached();
129 }
130 }
131
132 if_max_devs[type] = max_devs;
133 }
134
135 /*
136 * We automatically delete the drive when a device using it gets
137 * unplugged. Questionable feature, but we can't just drop it.
138 * Device models call blockdev_mark_auto_del() to schedule the
139 * automatic deletion, and generic qdev code calls blockdev_auto_del()
140 * when deletion is actually safe.
141 */
142 void blockdev_mark_auto_del(BlockBackend *blk)
143 {
144 DriveInfo *dinfo = blk_legacy_dinfo(blk);
145 BlockJob *job;
146
147 GLOBAL_STATE_CODE();
148
149 if (!dinfo) {
150 return;
151 }
152
153 JOB_LOCK_GUARD();
154
155 for (job = block_job_next_locked(NULL); job;
156 job = block_job_next_locked(job)) {
157 if (block_job_has_bdrv(job, blk_bs(blk))) {
158 job_cancel_locked(&job->job, false);
159 }
160 }
161
162 dinfo->auto_del = 1;
163 }
164
165 void blockdev_auto_del(BlockBackend *blk)
166 {
167 DriveInfo *dinfo = blk_legacy_dinfo(blk);
168 GLOBAL_STATE_CODE();
169
170 if (dinfo && dinfo->auto_del) {
171 monitor_remove_blk(blk);
172 blk_unref(blk);
173 }
174 }
175
176 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
177 {
178 int max_devs = if_max_devs[type];
179 return max_devs ? index / max_devs : 0;
180 }
181
182 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
183 {
184 int max_devs = if_max_devs[type];
185 return max_devs ? index % max_devs : index;
186 }
187
188 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
189 const char *optstr)
190 {
191 QemuOpts *opts;
192
193 GLOBAL_STATE_CODE();
194
195 opts = qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
196 if (!opts) {
197 return NULL;
198 }
199 if (type != IF_DEFAULT) {
200 qemu_opt_set(opts, "if", if_name[type], &error_abort);
201 }
202 if (index >= 0) {
203 qemu_opt_set_number(opts, "index", index, &error_abort);
204 }
205 if (file)
206 qemu_opt_set(opts, "file", file, &error_abort);
207 return opts;
208 }
209
210 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
211 {
212 BlockBackend *blk;
213 DriveInfo *dinfo;
214
215 GLOBAL_STATE_CODE();
216
217 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
218 dinfo = blk_legacy_dinfo(blk);
219 if (dinfo && dinfo->type == type
220 && dinfo->bus == bus && dinfo->unit == unit) {
221 return dinfo;
222 }
223 }
224
225 return NULL;
226 }
227
228 /*
229 * Check board claimed all -drive that are meant to be claimed.
230 * Fatal error if any remain unclaimed.
231 */
232 void drive_check_orphaned(void)
233 {
234 BlockBackend *blk;
235 DriveInfo *dinfo;
236 Location loc;
237 bool orphans = false;
238
239 GLOBAL_STATE_CODE();
240
241 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
242 dinfo = blk_legacy_dinfo(blk);
243 /*
244 * Ignore default drives, because we create certain default
245 * drives unconditionally, then leave them unclaimed. Not the
246 * users fault.
247 * Ignore IF_VIRTIO, because it gets desugared into -device,
248 * so we can leave failing to -device.
249 * Ignore IF_NONE, because leaving unclaimed IF_NONE remains
250 * available for device_add is a feature.
251 */
252 if (dinfo->is_default || dinfo->type == IF_VIRTIO
253 || dinfo->type == IF_NONE) {
254 continue;
255 }
256 if (!blk_get_attached_dev(blk)) {
257 loc_push_none(&loc);
258 qemu_opts_loc_restore(dinfo->opts);
259 error_report("machine type does not support"
260 " if=%s,bus=%d,unit=%d",
261 if_name[dinfo->type], dinfo->bus, dinfo->unit);
262 loc_pop(&loc);
263 orphans = true;
264 }
265 }
266
267 if (orphans) {
268 exit(1);
269 }
270 }
271
272 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
273 {
274 GLOBAL_STATE_CODE();
275 return drive_get(type,
276 drive_index_to_bus_id(type, index),
277 drive_index_to_unit_id(type, index));
278 }
279
280 int drive_get_max_bus(BlockInterfaceType type)
281 {
282 int max_bus;
283 BlockBackend *blk;
284 DriveInfo *dinfo;
285
286 GLOBAL_STATE_CODE();
287
288 max_bus = -1;
289 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
290 dinfo = blk_legacy_dinfo(blk);
291 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
292 max_bus = dinfo->bus;
293 }
294 }
295 return max_bus;
296 }
297
298 static void bdrv_format_print(void *opaque, const char *name)
299 {
300 qemu_printf(" %s", name);
301 }
302
303 typedef struct {
304 QEMUBH *bh;
305 BlockDriverState *bs;
306 } BDRVPutRefBH;
307
308 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
309 {
310 if (!strcmp(buf, "ignore")) {
311 return BLOCKDEV_ON_ERROR_IGNORE;
312 } else if (!is_read && !strcmp(buf, "enospc")) {
313 return BLOCKDEV_ON_ERROR_ENOSPC;
314 } else if (!strcmp(buf, "stop")) {
315 return BLOCKDEV_ON_ERROR_STOP;
316 } else if (!strcmp(buf, "report")) {
317 return BLOCKDEV_ON_ERROR_REPORT;
318 } else {
319 error_setg(errp, "'%s' invalid %s error action",
320 buf, is_read ? "read" : "write");
321 return -1;
322 }
323 }
324
325 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
326 Error **errp)
327 {
328 const QListEntry *entry;
329 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
330 switch (qobject_type(entry->value)) {
331
332 case QTYPE_QSTRING: {
333 unsigned long long length;
334 const char *str = qstring_get_str(qobject_to(QString,
335 entry->value));
336 if (parse_uint_full(str, &length, 10) == 0 &&
337 length > 0 && length <= UINT_MAX) {
338 block_acct_add_interval(stats, (unsigned) length);
339 } else {
340 error_setg(errp, "Invalid interval length: %s", str);
341 return false;
342 }
343 break;
344 }
345
346 case QTYPE_QNUM: {
347 int64_t length = qnum_get_int(qobject_to(QNum, entry->value));
348
349 if (length > 0 && length <= UINT_MAX) {
350 block_acct_add_interval(stats, (unsigned) length);
351 } else {
352 error_setg(errp, "Invalid interval length: %" PRId64, length);
353 return false;
354 }
355 break;
356 }
357
358 default:
359 error_setg(errp, "The specification of stats-intervals is invalid");
360 return false;
361 }
362 }
363 return true;
364 }
365
366 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
367
368 /* All parameters but @opts are optional and may be set to NULL. */
369 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
370 const char **throttling_group, ThrottleConfig *throttle_cfg,
371 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
372 {
373 Error *local_error = NULL;
374 const char *aio;
375
376 if (bdrv_flags) {
377 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
378 *bdrv_flags |= BDRV_O_COPY_ON_READ;
379 }
380
381 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
382 if (bdrv_parse_aio(aio, bdrv_flags) < 0) {
383 error_setg(errp, "invalid aio option");
384 return;
385 }
386 }
387 }
388
389 /* disk I/O throttling */
390 if (throttling_group) {
391 *throttling_group = qemu_opt_get(opts, "throttling.group");
392 }
393
394 if (throttle_cfg) {
395 throttle_config_init(throttle_cfg);
396 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
397 qemu_opt_get_number(opts, "throttling.bps-total", 0);
398 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
399 qemu_opt_get_number(opts, "throttling.bps-read", 0);
400 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
401 qemu_opt_get_number(opts, "throttling.bps-write", 0);
402 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
403 qemu_opt_get_number(opts, "throttling.iops-total", 0);
404 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
405 qemu_opt_get_number(opts, "throttling.iops-read", 0);
406 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
407 qemu_opt_get_number(opts, "throttling.iops-write", 0);
408
409 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
410 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
411 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
412 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
413 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
414 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
415 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
416 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
417 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
418 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
419 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
420 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
421
422 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
423 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
424 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
425 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
426 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
427 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
428 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
429 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
430 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
431 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
432 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
433 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
434
435 throttle_cfg->op_size =
436 qemu_opt_get_number(opts, "throttling.iops-size", 0);
437
438 if (!throttle_is_valid(throttle_cfg, errp)) {
439 return;
440 }
441 }
442
443 if (detect_zeroes) {
444 *detect_zeroes =
445 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
446 qemu_opt_get(opts, "detect-zeroes"),
447 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
448 &local_error);
449 if (local_error) {
450 error_propagate(errp, local_error);
451 return;
452 }
453 }
454 }
455
456 static OnOffAuto account_get_opt(QemuOpts *opts, const char *name)
457 {
458 if (!qemu_opt_find(opts, name)) {
459 return ON_OFF_AUTO_AUTO;
460 }
461 if (qemu_opt_get_bool(opts, name, true)) {
462 return ON_OFF_AUTO_ON;
463 }
464 return ON_OFF_AUTO_OFF;
465 }
466
467 /* Takes the ownership of bs_opts */
468 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
469 Error **errp)
470 {
471 const char *buf;
472 int bdrv_flags = 0;
473 int on_read_error, on_write_error;
474 OnOffAuto account_invalid, account_failed;
475 bool writethrough, read_only;
476 BlockBackend *blk;
477 BlockDriverState *bs;
478 ThrottleConfig cfg;
479 int snapshot = 0;
480 Error *error = NULL;
481 QemuOpts *opts;
482 QDict *interval_dict = NULL;
483 QList *interval_list = NULL;
484 const char *id;
485 BlockdevDetectZeroesOptions detect_zeroes =
486 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
487 const char *throttling_group = NULL;
488
489 /* Check common options by copying from bs_opts to opts, all other options
490 * stay in bs_opts for processing by bdrv_open(). */
491 id = qdict_get_try_str(bs_opts, "id");
492 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, errp);
493 if (!opts) {
494 goto err_no_opts;
495 }
496
497 if (!qemu_opts_absorb_qdict(opts, bs_opts, errp)) {
498 goto early_err;
499 }
500
501 if (id) {
502 qdict_del(bs_opts, "id");
503 }
504
505 /* extract parameters */
506 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
507
508 account_invalid = account_get_opt(opts, "stats-account-invalid");
509 account_failed = account_get_opt(opts, "stats-account-failed");
510
511 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
512
513 id = qemu_opts_id(opts);
514
515 qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
516 qdict_array_split(interval_dict, &interval_list);
517
518 if (qdict_size(interval_dict) != 0) {
519 error_setg(errp, "Invalid option stats-intervals.%s",
520 qdict_first(interval_dict)->key);
521 goto early_err;
522 }
523
524 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
525 &detect_zeroes, &error);
526 if (error) {
527 error_propagate(errp, error);
528 goto early_err;
529 }
530
531 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
532 if (is_help_option(buf)) {
533 qemu_printf("Supported formats:");
534 bdrv_iterate_format(bdrv_format_print, NULL, false);
535 qemu_printf("\nSupported formats (read-only):");
536 bdrv_iterate_format(bdrv_format_print, NULL, true);
537 qemu_printf("\n");
538 goto early_err;
539 }
540
541 if (qdict_haskey(bs_opts, "driver")) {
542 error_setg(errp, "Cannot specify both 'driver' and 'format'");
543 goto early_err;
544 }
545 qdict_put_str(bs_opts, "driver", buf);
546 }
547
548 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
549 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
550 on_write_error = parse_block_error_action(buf, 0, &error);
551 if (error) {
552 error_propagate(errp, error);
553 goto early_err;
554 }
555 }
556
557 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
558 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
559 on_read_error = parse_block_error_action(buf, 1, &error);
560 if (error) {
561 error_propagate(errp, error);
562 goto early_err;
563 }
564 }
565
566 if (snapshot) {
567 bdrv_flags |= BDRV_O_SNAPSHOT;
568 }
569
570 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
571
572 /* init */
573 if ((!file || !*file) && !qdict_size(bs_opts)) {
574 BlockBackendRootState *blk_rs;
575
576 blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
577 blk_rs = blk_get_root_state(blk);
578 blk_rs->open_flags = bdrv_flags | (read_only ? 0 : BDRV_O_RDWR);
579 blk_rs->detect_zeroes = detect_zeroes;
580
581 qobject_unref(bs_opts);
582 } else {
583 if (file && !*file) {
584 file = NULL;
585 }
586
587 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
588 * with other callers) rather than what we want as the real defaults.
589 * Apply the defaults here instead. */
590 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
591 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
592 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
593 read_only ? "on" : "off");
594 qdict_set_default_str(bs_opts, BDRV_OPT_AUTO_READ_ONLY, "on");
595 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
596
597 if (runstate_check(RUN_STATE_INMIGRATE)) {
598 bdrv_flags |= BDRV_O_INACTIVE;
599 }
600
601 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
602 if (!blk) {
603 goto err_no_bs_opts;
604 }
605 bs = blk_bs(blk);
606
607 bs->detect_zeroes = detect_zeroes;
608
609 block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
610
611 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
612 blk_unref(blk);
613 blk = NULL;
614 goto err_no_bs_opts;
615 }
616 }
617
618 /* disk I/O throttling */
619 if (throttle_enabled(&cfg)) {
620 if (!throttling_group) {
621 throttling_group = id;
622 }
623 blk_io_limits_enable(blk, throttling_group);
624 blk_set_io_limits(blk, &cfg);
625 }
626
627 blk_set_enable_write_cache(blk, !writethrough);
628 blk_set_on_error(blk, on_read_error, on_write_error);
629
630 if (!monitor_add_blk(blk, id, errp)) {
631 blk_unref(blk);
632 blk = NULL;
633 goto err_no_bs_opts;
634 }
635
636 err_no_bs_opts:
637 qemu_opts_del(opts);
638 qobject_unref(interval_dict);
639 qobject_unref(interval_list);
640 return blk;
641
642 early_err:
643 qemu_opts_del(opts);
644 qobject_unref(interval_dict);
645 qobject_unref(interval_list);
646 err_no_opts:
647 qobject_unref(bs_opts);
648 return NULL;
649 }
650
651 /* Takes the ownership of bs_opts */
652 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
653 {
654 int bdrv_flags = 0;
655
656 GLOBAL_STATE_CODE();
657 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
658 * with other callers) rather than what we want as the real defaults.
659 * Apply the defaults here instead. */
660 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
661 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
662 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
663
664 if (runstate_check(RUN_STATE_INMIGRATE)) {
665 bdrv_flags |= BDRV_O_INACTIVE;
666 }
667
668 return bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
669 }
670
671 void blockdev_close_all_bdrv_states(void)
672 {
673 BlockDriverState *bs, *next_bs;
674
675 GLOBAL_STATE_CODE();
676 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
677 AioContext *ctx = bdrv_get_aio_context(bs);
678
679 aio_context_acquire(ctx);
680 bdrv_unref(bs);
681 aio_context_release(ctx);
682 }
683 }
684
685 /* Iterates over the list of monitor-owned BlockDriverStates */
686 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
687 {
688 GLOBAL_STATE_CODE();
689 return bs ? QTAILQ_NEXT(bs, monitor_list)
690 : QTAILQ_FIRST(&monitor_bdrv_states);
691 }
692
693 static bool qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
694 Error **errp)
695 {
696 const char *value;
697
698 value = qemu_opt_get(opts, from);
699 if (value) {
700 if (qemu_opt_find(opts, to)) {
701 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
702 "same time", to, from);
703 return false;
704 }
705 }
706
707 /* rename all items in opts */
708 while ((value = qemu_opt_get(opts, from))) {
709 qemu_opt_set(opts, to, value, &error_abort);
710 qemu_opt_unset(opts, from);
711 }
712 return true;
713 }
714
715 QemuOptsList qemu_legacy_drive_opts = {
716 .name = "drive",
717 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
718 .desc = {
719 {
720 .name = "bus",
721 .type = QEMU_OPT_NUMBER,
722 .help = "bus number",
723 },{
724 .name = "unit",
725 .type = QEMU_OPT_NUMBER,
726 .help = "unit number (i.e. lun for scsi)",
727 },{
728 .name = "index",
729 .type = QEMU_OPT_NUMBER,
730 .help = "index number",
731 },{
732 .name = "media",
733 .type = QEMU_OPT_STRING,
734 .help = "media type (disk, cdrom)",
735 },{
736 .name = "if",
737 .type = QEMU_OPT_STRING,
738 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
739 },{
740 .name = "file",
741 .type = QEMU_OPT_STRING,
742 .help = "file name",
743 },
744
745 /* Options that are passed on, but have special semantics with -drive */
746 {
747 .name = BDRV_OPT_READ_ONLY,
748 .type = QEMU_OPT_BOOL,
749 .help = "open drive file as read-only",
750 },{
751 .name = "rerror",
752 .type = QEMU_OPT_STRING,
753 .help = "read error action",
754 },{
755 .name = "werror",
756 .type = QEMU_OPT_STRING,
757 .help = "write error action",
758 },{
759 .name = "copy-on-read",
760 .type = QEMU_OPT_BOOL,
761 .help = "copy read data from backing file into image file",
762 },
763
764 { /* end of list */ }
765 },
766 };
767
768 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
769 Error **errp)
770 {
771 const char *value;
772 BlockBackend *blk;
773 DriveInfo *dinfo = NULL;
774 QDict *bs_opts;
775 QemuOpts *legacy_opts;
776 DriveMediaType media = MEDIA_DISK;
777 BlockInterfaceType type;
778 int max_devs, bus_id, unit_id, index;
779 const char *werror, *rerror;
780 bool read_only = false;
781 bool copy_on_read;
782 const char *filename;
783 int i;
784
785 GLOBAL_STATE_CODE();
786
787 /* Change legacy command line options into QMP ones */
788 static const struct {
789 const char *from;
790 const char *to;
791 } opt_renames[] = {
792 { "iops", "throttling.iops-total" },
793 { "iops_rd", "throttling.iops-read" },
794 { "iops_wr", "throttling.iops-write" },
795
796 { "bps", "throttling.bps-total" },
797 { "bps_rd", "throttling.bps-read" },
798 { "bps_wr", "throttling.bps-write" },
799
800 { "iops_max", "throttling.iops-total-max" },
801 { "iops_rd_max", "throttling.iops-read-max" },
802 { "iops_wr_max", "throttling.iops-write-max" },
803
804 { "bps_max", "throttling.bps-total-max" },
805 { "bps_rd_max", "throttling.bps-read-max" },
806 { "bps_wr_max", "throttling.bps-write-max" },
807
808 { "iops_size", "throttling.iops-size" },
809
810 { "group", "throttling.group" },
811
812 { "readonly", BDRV_OPT_READ_ONLY },
813 };
814
815 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
816 if (!qemu_opt_rename(all_opts, opt_renames[i].from,
817 opt_renames[i].to, errp)) {
818 return NULL;
819 }
820 }
821
822 value = qemu_opt_get(all_opts, "cache");
823 if (value) {
824 int flags = 0;
825 bool writethrough;
826
827 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
828 error_setg(errp, "invalid cache option");
829 return NULL;
830 }
831
832 /* Specific options take precedence */
833 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
834 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
835 !writethrough, &error_abort);
836 }
837 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
838 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
839 !!(flags & BDRV_O_NOCACHE), &error_abort);
840 }
841 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
842 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
843 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
844 }
845 qemu_opt_unset(all_opts, "cache");
846 }
847
848 /* Get a QDict for processing the options */
849 bs_opts = qdict_new();
850 qemu_opts_to_qdict(all_opts, bs_opts);
851
852 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
853 &error_abort);
854 if (!qemu_opts_absorb_qdict(legacy_opts, bs_opts, errp)) {
855 goto fail;
856 }
857
858 /* Media type */
859 value = qemu_opt_get(legacy_opts, "media");
860 if (value) {
861 if (!strcmp(value, "disk")) {
862 media = MEDIA_DISK;
863 } else if (!strcmp(value, "cdrom")) {
864 media = MEDIA_CDROM;
865 read_only = true;
866 } else {
867 error_setg(errp, "'%s' invalid media", value);
868 goto fail;
869 }
870 }
871
872 /* copy-on-read is disabled with a warning for read-only devices */
873 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
874 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
875
876 if (read_only && copy_on_read) {
877 warn_report("disabling copy-on-read on read-only drive");
878 copy_on_read = false;
879 }
880
881 qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
882 qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
883
884 /* Controller type */
885 value = qemu_opt_get(legacy_opts, "if");
886 if (value) {
887 for (type = 0;
888 type < IF_COUNT && strcmp(value, if_name[type]);
889 type++) {
890 }
891 if (type == IF_COUNT) {
892 error_setg(errp, "unsupported bus type '%s'", value);
893 goto fail;
894 }
895 } else {
896 type = block_default_type;
897 }
898
899 /* Device address specified by bus/unit or index.
900 * If none was specified, try to find the first free one. */
901 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
902 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
903 index = qemu_opt_get_number(legacy_opts, "index", -1);
904
905 max_devs = if_max_devs[type];
906
907 if (index != -1) {
908 if (bus_id != 0 || unit_id != -1) {
909 error_setg(errp, "index cannot be used with bus and unit");
910 goto fail;
911 }
912 bus_id = drive_index_to_bus_id(type, index);
913 unit_id = drive_index_to_unit_id(type, index);
914 }
915
916 if (unit_id == -1) {
917 unit_id = 0;
918 while (drive_get(type, bus_id, unit_id) != NULL) {
919 unit_id++;
920 if (max_devs && unit_id >= max_devs) {
921 unit_id -= max_devs;
922 bus_id++;
923 }
924 }
925 }
926
927 if (max_devs && unit_id >= max_devs) {
928 error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
929 goto fail;
930 }
931
932 if (drive_get(type, bus_id, unit_id) != NULL) {
933 error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
934 bus_id, unit_id, index);
935 goto fail;
936 }
937
938 /* no id supplied -> create one */
939 if (qemu_opts_id(all_opts) == NULL) {
940 char *new_id;
941 const char *mediastr = "";
942 if (type == IF_IDE || type == IF_SCSI) {
943 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
944 }
945 if (max_devs) {
946 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
947 mediastr, unit_id);
948 } else {
949 new_id = g_strdup_printf("%s%s%i", if_name[type],
950 mediastr, unit_id);
951 }
952 qdict_put_str(bs_opts, "id", new_id);
953 g_free(new_id);
954 }
955
956 /* Add virtio block device */
957 if (type == IF_VIRTIO) {
958 QemuOpts *devopts;
959 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
960 &error_abort);
961 qemu_opt_set(devopts, "driver", "virtio-blk", &error_abort);
962 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
963 &error_abort);
964 }
965
966 filename = qemu_opt_get(legacy_opts, "file");
967
968 /* Check werror/rerror compatibility with if=... */
969 werror = qemu_opt_get(legacy_opts, "werror");
970 if (werror != NULL) {
971 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
972 type != IF_NONE) {
973 error_setg(errp, "werror is not supported by this bus type");
974 goto fail;
975 }
976 qdict_put_str(bs_opts, "werror", werror);
977 }
978
979 rerror = qemu_opt_get(legacy_opts, "rerror");
980 if (rerror != NULL) {
981 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
982 type != IF_NONE) {
983 error_setg(errp, "rerror is not supported by this bus type");
984 goto fail;
985 }
986 qdict_put_str(bs_opts, "rerror", rerror);
987 }
988
989 /* Actual block device init: Functionality shared with blockdev-add */
990 blk = blockdev_init(filename, bs_opts, errp);
991 bs_opts = NULL;
992 if (!blk) {
993 goto fail;
994 }
995
996 /* Create legacy DriveInfo */
997 dinfo = g_malloc0(sizeof(*dinfo));
998 dinfo->opts = all_opts;
999
1000 dinfo->type = type;
1001 dinfo->bus = bus_id;
1002 dinfo->unit = unit_id;
1003
1004 blk_set_legacy_dinfo(blk, dinfo);
1005
1006 switch(type) {
1007 case IF_IDE:
1008 case IF_SCSI:
1009 case IF_XEN:
1010 case IF_NONE:
1011 dinfo->media_cd = media == MEDIA_CDROM;
1012 break;
1013 default:
1014 break;
1015 }
1016
1017 fail:
1018 qemu_opts_del(legacy_opts);
1019 qobject_unref(bs_opts);
1020 return dinfo;
1021 }
1022
1023 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1024 {
1025 BlockDriverState *bs;
1026
1027 bs = bdrv_lookup_bs(name, name, errp);
1028 if (bs == NULL) {
1029 return NULL;
1030 }
1031
1032 if (!bdrv_is_root_node(bs)) {
1033 error_setg(errp, "Need a root block node");
1034 return NULL;
1035 }
1036
1037 if (!bdrv_is_inserted(bs)) {
1038 error_setg(errp, "Device has no medium");
1039 return NULL;
1040 }
1041
1042 return bs;
1043 }
1044
1045 static void blockdev_do_action(TransactionAction *action, Error **errp)
1046 {
1047 TransactionActionList list;
1048
1049 list.value = action;
1050 list.next = NULL;
1051 qmp_transaction(&list, false, NULL, errp);
1052 }
1053
1054 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1055 bool has_node_name, const char *node_name,
1056 const char *snapshot_file,
1057 bool has_snapshot_node_name,
1058 const char *snapshot_node_name,
1059 bool has_format, const char *format,
1060 bool has_mode, NewImageMode mode, Error **errp)
1061 {
1062 BlockdevSnapshotSync snapshot = {
1063 .has_device = has_device,
1064 .device = (char *) device,
1065 .has_node_name = has_node_name,
1066 .node_name = (char *) node_name,
1067 .snapshot_file = (char *) snapshot_file,
1068 .has_snapshot_node_name = has_snapshot_node_name,
1069 .snapshot_node_name = (char *) snapshot_node_name,
1070 .has_format = has_format,
1071 .format = (char *) format,
1072 .has_mode = has_mode,
1073 .mode = mode,
1074 };
1075 TransactionAction action = {
1076 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1077 .u.blockdev_snapshot_sync.data = &snapshot,
1078 };
1079 blockdev_do_action(&action, errp);
1080 }
1081
1082 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1083 Error **errp)
1084 {
1085 BlockdevSnapshot snapshot_data = {
1086 .node = (char *) node,
1087 .overlay = (char *) overlay
1088 };
1089 TransactionAction action = {
1090 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1091 .u.blockdev_snapshot.data = &snapshot_data,
1092 };
1093 blockdev_do_action(&action, errp);
1094 }
1095
1096 void qmp_blockdev_snapshot_internal_sync(const char *device,
1097 const char *name,
1098 Error **errp)
1099 {
1100 BlockdevSnapshotInternal snapshot = {
1101 .device = (char *) device,
1102 .name = (char *) name
1103 };
1104 TransactionAction action = {
1105 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1106 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1107 };
1108 blockdev_do_action(&action, errp);
1109 }
1110
1111 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1112 bool has_id,
1113 const char *id,
1114 bool has_name,
1115 const char *name,
1116 Error **errp)
1117 {
1118 BlockDriverState *bs;
1119 AioContext *aio_context;
1120 QEMUSnapshotInfo sn;
1121 Error *local_err = NULL;
1122 SnapshotInfo *info = NULL;
1123 int ret;
1124
1125 bs = qmp_get_root_bs(device, errp);
1126 if (!bs) {
1127 return NULL;
1128 }
1129 aio_context = bdrv_get_aio_context(bs);
1130 aio_context_acquire(aio_context);
1131
1132 if (!has_id) {
1133 id = NULL;
1134 }
1135
1136 if (!has_name) {
1137 name = NULL;
1138 }
1139
1140 if (!id && !name) {
1141 error_setg(errp, "Name or id must be provided");
1142 goto out_aio_context;
1143 }
1144
1145 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1146 goto out_aio_context;
1147 }
1148
1149 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1150 if (local_err) {
1151 error_propagate(errp, local_err);
1152 goto out_aio_context;
1153 }
1154 if (!ret) {
1155 error_setg(errp,
1156 "Snapshot with id '%s' and name '%s' does not exist on "
1157 "device '%s'",
1158 STR_OR_NULL(id), STR_OR_NULL(name), device);
1159 goto out_aio_context;
1160 }
1161
1162 bdrv_snapshot_delete(bs, id, name, &local_err);
1163 if (local_err) {
1164 error_propagate(errp, local_err);
1165 goto out_aio_context;
1166 }
1167
1168 aio_context_release(aio_context);
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 if (sn.icount != -1ULL) {
1179 info->icount = sn.icount;
1180 info->has_icount = true;
1181 }
1182
1183 return info;
1184
1185 out_aio_context:
1186 aio_context_release(aio_context);
1187 return NULL;
1188 }
1189
1190 /* New and old BlockDriverState structs for atomic group operations */
1191
1192 typedef struct BlkActionState BlkActionState;
1193
1194 /**
1195 * BlkActionOps:
1196 * Table of operations that define an Action.
1197 *
1198 * @instance_size: Size of state struct, in bytes.
1199 * @prepare: Prepare the work, must NOT be NULL.
1200 * @commit: Commit the changes, can be NULL.
1201 * @abort: Abort the changes on fail, can be NULL.
1202 * @clean: Clean up resources after all transaction actions have called
1203 * commit() or abort(). Can be NULL.
1204 *
1205 * Only prepare() may fail. In a single transaction, only one of commit() or
1206 * abort() will be called. clean() will always be called if it is present.
1207 *
1208 * Always run under BQL.
1209 */
1210 typedef struct BlkActionOps {
1211 size_t instance_size;
1212 void (*prepare)(BlkActionState *common, Error **errp);
1213 void (*commit)(BlkActionState *common);
1214 void (*abort)(BlkActionState *common);
1215 void (*clean)(BlkActionState *common);
1216 } BlkActionOps;
1217
1218 /**
1219 * BlkActionState:
1220 * Describes one Action's state within a Transaction.
1221 *
1222 * @action: QAPI-defined enum identifying which Action to perform.
1223 * @ops: Table of ActionOps this Action can perform.
1224 * @block_job_txn: Transaction which this action belongs to.
1225 * @entry: List membership for all Actions in this Transaction.
1226 *
1227 * This structure must be arranged as first member in a subclassed type,
1228 * assuming that the compiler will also arrange it to the same offsets as the
1229 * base class.
1230 */
1231 struct BlkActionState {
1232 TransactionAction *action;
1233 const BlkActionOps *ops;
1234 JobTxn *block_job_txn;
1235 TransactionProperties *txn_props;
1236 QTAILQ_ENTRY(BlkActionState) entry;
1237 };
1238
1239 /* internal snapshot private data */
1240 typedef struct InternalSnapshotState {
1241 BlkActionState common;
1242 BlockDriverState *bs;
1243 QEMUSnapshotInfo sn;
1244 bool created;
1245 } InternalSnapshotState;
1246
1247
1248 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1249 {
1250 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1251 error_setg(errp,
1252 "Action '%s' does not support Transaction property "
1253 "completion-mode = %s",
1254 TransactionActionKind_str(s->action->type),
1255 ActionCompletionMode_str(s->txn_props->completion_mode));
1256 return -1;
1257 }
1258 return 0;
1259 }
1260
1261 static void internal_snapshot_prepare(BlkActionState *common,
1262 Error **errp)
1263 {
1264 Error *local_err = NULL;
1265 const char *device;
1266 const char *name;
1267 BlockDriverState *bs;
1268 QEMUSnapshotInfo old_sn, *sn;
1269 bool ret;
1270 int64_t rt;
1271 BlockdevSnapshotInternal *internal;
1272 InternalSnapshotState *state;
1273 AioContext *aio_context;
1274 int ret1;
1275
1276 g_assert(common->action->type ==
1277 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1278 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1279 state = DO_UPCAST(InternalSnapshotState, common, common);
1280
1281 /* 1. parse input */
1282 device = internal->device;
1283 name = internal->name;
1284
1285 /* 2. check for validation */
1286 if (action_check_completion_mode(common, errp) < 0) {
1287 return;
1288 }
1289
1290 bs = qmp_get_root_bs(device, errp);
1291 if (!bs) {
1292 return;
1293 }
1294
1295 aio_context = bdrv_get_aio_context(bs);
1296 aio_context_acquire(aio_context);
1297
1298 state->bs = bs;
1299
1300 /* Paired with .clean() */
1301 bdrv_drained_begin(bs);
1302
1303 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1304 goto out;
1305 }
1306
1307 if (bdrv_is_read_only(bs)) {
1308 error_setg(errp, "Device '%s' is read only", device);
1309 goto out;
1310 }
1311
1312 if (!bdrv_can_snapshot(bs)) {
1313 error_setg(errp, "Block format '%s' used by device '%s' "
1314 "does not support internal snapshots",
1315 bs->drv->format_name, device);
1316 goto out;
1317 }
1318
1319 if (!strlen(name)) {
1320 error_setg(errp, "Name is empty");
1321 goto out;
1322 }
1323
1324 /* check whether a snapshot with name exist */
1325 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1326 &local_err);
1327 if (local_err) {
1328 error_propagate(errp, local_err);
1329 goto out;
1330 } else if (ret) {
1331 error_setg(errp,
1332 "Snapshot with name '%s' already exists on device '%s'",
1333 name, device);
1334 goto out;
1335 }
1336
1337 /* 3. take the snapshot */
1338 sn = &state->sn;
1339 pstrcpy(sn->name, sizeof(sn->name), name);
1340 rt = g_get_real_time();
1341 sn->date_sec = rt / G_USEC_PER_SEC;
1342 sn->date_nsec = (rt % G_USEC_PER_SEC) * 1000;
1343 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1344 if (replay_mode != REPLAY_MODE_NONE) {
1345 sn->icount = replay_get_current_icount();
1346 } else {
1347 sn->icount = -1ULL;
1348 }
1349
1350 ret1 = bdrv_snapshot_create(bs, sn);
1351 if (ret1 < 0) {
1352 error_setg_errno(errp, -ret1,
1353 "Failed to create snapshot '%s' on device '%s'",
1354 name, device);
1355 goto out;
1356 }
1357
1358 /* 4. succeed, mark a snapshot is created */
1359 state->created = true;
1360
1361 out:
1362 aio_context_release(aio_context);
1363 }
1364
1365 static void internal_snapshot_abort(BlkActionState *common)
1366 {
1367 InternalSnapshotState *state =
1368 DO_UPCAST(InternalSnapshotState, common, common);
1369 BlockDriverState *bs = state->bs;
1370 QEMUSnapshotInfo *sn = &state->sn;
1371 AioContext *aio_context;
1372 Error *local_error = NULL;
1373
1374 if (!state->created) {
1375 return;
1376 }
1377
1378 aio_context = bdrv_get_aio_context(state->bs);
1379 aio_context_acquire(aio_context);
1380
1381 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1382 error_reportf_err(local_error,
1383 "Failed to delete snapshot with id '%s' and "
1384 "name '%s' on device '%s' in abort: ",
1385 sn->id_str, sn->name,
1386 bdrv_get_device_name(bs));
1387 }
1388
1389 aio_context_release(aio_context);
1390 }
1391
1392 static void internal_snapshot_clean(BlkActionState *common)
1393 {
1394 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1395 common, common);
1396 AioContext *aio_context;
1397
1398 if (!state->bs) {
1399 return;
1400 }
1401
1402 aio_context = bdrv_get_aio_context(state->bs);
1403 aio_context_acquire(aio_context);
1404
1405 bdrv_drained_end(state->bs);
1406
1407 aio_context_release(aio_context);
1408 }
1409
1410 /* external snapshot private data */
1411 typedef struct ExternalSnapshotState {
1412 BlkActionState common;
1413 BlockDriverState *old_bs;
1414 BlockDriverState *new_bs;
1415 bool overlay_appended;
1416 } ExternalSnapshotState;
1417
1418 static void external_snapshot_prepare(BlkActionState *common,
1419 Error **errp)
1420 {
1421 int ret;
1422 int flags = 0;
1423 QDict *options = NULL;
1424 Error *local_err = NULL;
1425 /* Device and node name of the image to generate the snapshot from */
1426 const char *device;
1427 const char *node_name;
1428 /* Reference to the new image (for 'blockdev-snapshot') */
1429 const char *snapshot_ref;
1430 /* File name of the new image (for 'blockdev-snapshot-sync') */
1431 const char *new_image_file;
1432 ExternalSnapshotState *state =
1433 DO_UPCAST(ExternalSnapshotState, common, common);
1434 TransactionAction *action = common->action;
1435 AioContext *aio_context;
1436 uint64_t perm, shared;
1437
1438 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1439 * purpose but a different set of parameters */
1440 switch (action->type) {
1441 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1442 {
1443 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1444 device = s->node;
1445 node_name = s->node;
1446 new_image_file = NULL;
1447 snapshot_ref = s->overlay;
1448 }
1449 break;
1450 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1451 {
1452 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1453 device = s->has_device ? s->device : NULL;
1454 node_name = s->has_node_name ? s->node_name : NULL;
1455 new_image_file = s->snapshot_file;
1456 snapshot_ref = NULL;
1457 }
1458 break;
1459 default:
1460 g_assert_not_reached();
1461 }
1462
1463 /* start processing */
1464 if (action_check_completion_mode(common, errp) < 0) {
1465 return;
1466 }
1467
1468 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1469 if (!state->old_bs) {
1470 return;
1471 }
1472
1473 aio_context = bdrv_get_aio_context(state->old_bs);
1474 aio_context_acquire(aio_context);
1475
1476 /* Paired with .clean() */
1477 bdrv_drained_begin(state->old_bs);
1478
1479 if (!bdrv_is_inserted(state->old_bs)) {
1480 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1481 goto out;
1482 }
1483
1484 if (bdrv_op_is_blocked(state->old_bs,
1485 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1486 goto out;
1487 }
1488
1489 if (!bdrv_is_read_only(state->old_bs)) {
1490 if (bdrv_flush(state->old_bs)) {
1491 error_setg(errp, QERR_IO_ERROR);
1492 goto out;
1493 }
1494 }
1495
1496 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1497 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1498 const char *format = s->has_format ? s->format : "qcow2";
1499 enum NewImageMode mode;
1500 const char *snapshot_node_name =
1501 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1502
1503 if (node_name && !snapshot_node_name) {
1504 error_setg(errp, "New overlay node-name missing");
1505 goto out;
1506 }
1507
1508 if (snapshot_node_name &&
1509 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1510 error_setg(errp, "New overlay node-name already in use");
1511 goto out;
1512 }
1513
1514 flags = state->old_bs->open_flags;
1515 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1516 flags |= BDRV_O_NO_BACKING;
1517
1518 /* create new image w/backing file */
1519 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1520 if (mode != NEW_IMAGE_MODE_EXISTING) {
1521 int64_t size = bdrv_getlength(state->old_bs);
1522 if (size < 0) {
1523 error_setg_errno(errp, -size, "bdrv_getlength failed");
1524 goto out;
1525 }
1526 bdrv_refresh_filename(state->old_bs);
1527 bdrv_img_create(new_image_file, format,
1528 state->old_bs->filename,
1529 state->old_bs->drv->format_name,
1530 NULL, size, flags, false, &local_err);
1531 if (local_err) {
1532 error_propagate(errp, local_err);
1533 goto out;
1534 }
1535 }
1536
1537 options = qdict_new();
1538 if (snapshot_node_name) {
1539 qdict_put_str(options, "node-name", snapshot_node_name);
1540 }
1541 qdict_put_str(options, "driver", format);
1542 }
1543
1544 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1545 errp);
1546 /* We will manually add the backing_hd field to the bs later */
1547 if (!state->new_bs) {
1548 goto out;
1549 }
1550
1551 /*
1552 * Allow attaching a backing file to an overlay that's already in use only
1553 * if the parents don't assume that they are already seeing a valid image.
1554 * (Specifically, allow it as a mirror target, which is write-only access.)
1555 */
1556 bdrv_get_cumulative_perm(state->new_bs, &perm, &shared);
1557 if (perm & BLK_PERM_CONSISTENT_READ) {
1558 error_setg(errp, "The overlay is already in use");
1559 goto out;
1560 }
1561
1562 if (state->new_bs->drv->is_filter) {
1563 error_setg(errp, "Filters cannot be used as overlays");
1564 goto out;
1565 }
1566
1567 if (bdrv_cow_child(state->new_bs)) {
1568 error_setg(errp, "The overlay already has a backing image");
1569 goto out;
1570 }
1571
1572 if (!state->new_bs->drv->supports_backing) {
1573 error_setg(errp, "The overlay does not support backing images");
1574 goto out;
1575 }
1576
1577 ret = bdrv_append(state->new_bs, state->old_bs, errp);
1578 if (ret < 0) {
1579 goto out;
1580 }
1581 state->overlay_appended = true;
1582
1583 out:
1584 aio_context_release(aio_context);
1585 }
1586
1587 static void external_snapshot_commit(BlkActionState *common)
1588 {
1589 ExternalSnapshotState *state =
1590 DO_UPCAST(ExternalSnapshotState, common, common);
1591 AioContext *aio_context;
1592
1593 aio_context = bdrv_get_aio_context(state->old_bs);
1594 aio_context_acquire(aio_context);
1595
1596 /* We don't need (or want) to use the transactional
1597 * bdrv_reopen_multiple() across all the entries at once, because we
1598 * don't want to abort all of them if one of them fails the reopen */
1599 if (!qatomic_read(&state->old_bs->copy_on_read)) {
1600 bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1601 }
1602
1603 aio_context_release(aio_context);
1604 }
1605
1606 static void external_snapshot_abort(BlkActionState *common)
1607 {
1608 ExternalSnapshotState *state =
1609 DO_UPCAST(ExternalSnapshotState, common, common);
1610 if (state->new_bs) {
1611 if (state->overlay_appended) {
1612 AioContext *aio_context;
1613 AioContext *tmp_context;
1614 int ret;
1615
1616 aio_context = bdrv_get_aio_context(state->old_bs);
1617 aio_context_acquire(aio_context);
1618
1619 bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd()
1620 close state->old_bs; we need it */
1621 bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1622
1623 /*
1624 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1625 * the main AioContext. As we're still going to be using it, return
1626 * it to the AioContext it was before.
1627 */
1628 tmp_context = bdrv_get_aio_context(state->old_bs);
1629 if (aio_context != tmp_context) {
1630 aio_context_release(aio_context);
1631 aio_context_acquire(tmp_context);
1632
1633 ret = bdrv_try_set_aio_context(state->old_bs,
1634 aio_context, NULL);
1635 assert(ret == 0);
1636
1637 aio_context_release(tmp_context);
1638 aio_context_acquire(aio_context);
1639 }
1640
1641 bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1642 bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1643
1644 aio_context_release(aio_context);
1645 }
1646 }
1647 }
1648
1649 static void external_snapshot_clean(BlkActionState *common)
1650 {
1651 ExternalSnapshotState *state =
1652 DO_UPCAST(ExternalSnapshotState, common, common);
1653 AioContext *aio_context;
1654
1655 if (!state->old_bs) {
1656 return;
1657 }
1658
1659 aio_context = bdrv_get_aio_context(state->old_bs);
1660 aio_context_acquire(aio_context);
1661
1662 bdrv_drained_end(state->old_bs);
1663 bdrv_unref(state->new_bs);
1664
1665 aio_context_release(aio_context);
1666 }
1667
1668 typedef struct DriveBackupState {
1669 BlkActionState common;
1670 BlockDriverState *bs;
1671 BlockJob *job;
1672 } DriveBackupState;
1673
1674 static BlockJob *do_backup_common(BackupCommon *backup,
1675 BlockDriverState *bs,
1676 BlockDriverState *target_bs,
1677 AioContext *aio_context,
1678 JobTxn *txn, Error **errp);
1679
1680 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1681 {
1682 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1683 DriveBackup *backup;
1684 BlockDriverState *bs;
1685 BlockDriverState *target_bs;
1686 BlockDriverState *source = NULL;
1687 AioContext *aio_context;
1688 AioContext *old_context;
1689 QDict *options;
1690 Error *local_err = NULL;
1691 int flags;
1692 int64_t size;
1693 bool set_backing_hd = false;
1694 int ret;
1695
1696 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1697 backup = common->action->u.drive_backup.data;
1698
1699 if (!backup->has_mode) {
1700 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1701 }
1702
1703 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1704 if (!bs) {
1705 return;
1706 }
1707
1708 if (!bs->drv) {
1709 error_setg(errp, "Device has no medium");
1710 return;
1711 }
1712
1713 aio_context = bdrv_get_aio_context(bs);
1714 aio_context_acquire(aio_context);
1715
1716 state->bs = bs;
1717 /* Paired with .clean() */
1718 bdrv_drained_begin(bs);
1719
1720 if (!backup->has_format) {
1721 backup->format = backup->mode == NEW_IMAGE_MODE_EXISTING ?
1722 NULL : (char *) bs->drv->format_name;
1723 }
1724
1725 /* Early check to avoid creating target */
1726 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
1727 goto out;
1728 }
1729
1730 flags = bs->open_flags | BDRV_O_RDWR;
1731
1732 /*
1733 * See if we have a backing HD we can use to create our new image
1734 * on top of.
1735 */
1736 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
1737 /*
1738 * Backup will not replace the source by the target, so none
1739 * of the filters skipped here will be removed (in contrast to
1740 * mirror). Therefore, we can skip all of them when looking
1741 * for the first COW relationship.
1742 */
1743 source = bdrv_cow_bs(bdrv_skip_filters(bs));
1744 if (!source) {
1745 backup->sync = MIRROR_SYNC_MODE_FULL;
1746 }
1747 }
1748 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
1749 source = bs;
1750 flags |= BDRV_O_NO_BACKING;
1751 set_backing_hd = true;
1752 }
1753
1754 size = bdrv_getlength(bs);
1755 if (size < 0) {
1756 error_setg_errno(errp, -size, "bdrv_getlength failed");
1757 goto out;
1758 }
1759
1760 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
1761 assert(backup->format);
1762 if (source) {
1763 /* Implicit filters should not appear in the filename */
1764 BlockDriverState *explicit_backing =
1765 bdrv_skip_implicit_filters(source);
1766
1767 bdrv_refresh_filename(explicit_backing);
1768 bdrv_img_create(backup->target, backup->format,
1769 explicit_backing->filename,
1770 explicit_backing->drv->format_name, NULL,
1771 size, flags, false, &local_err);
1772 } else {
1773 bdrv_img_create(backup->target, backup->format, NULL, NULL, NULL,
1774 size, flags, false, &local_err);
1775 }
1776 }
1777
1778 if (local_err) {
1779 error_propagate(errp, local_err);
1780 goto out;
1781 }
1782
1783 options = qdict_new();
1784 qdict_put_str(options, "discard", "unmap");
1785 qdict_put_str(options, "detect-zeroes", "unmap");
1786 if (backup->format) {
1787 qdict_put_str(options, "driver", backup->format);
1788 }
1789
1790 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
1791 if (!target_bs) {
1792 goto out;
1793 }
1794
1795 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1796 old_context = bdrv_get_aio_context(target_bs);
1797 aio_context_release(aio_context);
1798 aio_context_acquire(old_context);
1799
1800 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1801 if (ret < 0) {
1802 bdrv_unref(target_bs);
1803 aio_context_release(old_context);
1804 return;
1805 }
1806
1807 aio_context_release(old_context);
1808 aio_context_acquire(aio_context);
1809
1810 if (set_backing_hd) {
1811 if (bdrv_set_backing_hd(target_bs, source, errp) < 0) {
1812 goto unref;
1813 }
1814 }
1815
1816 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1817 bs, target_bs, aio_context,
1818 common->block_job_txn, errp);
1819
1820 unref:
1821 bdrv_unref(target_bs);
1822 out:
1823 aio_context_release(aio_context);
1824 }
1825
1826 static void drive_backup_commit(BlkActionState *common)
1827 {
1828 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1829 AioContext *aio_context;
1830
1831 aio_context = bdrv_get_aio_context(state->bs);
1832 aio_context_acquire(aio_context);
1833
1834 assert(state->job);
1835 job_start(&state->job->job);
1836
1837 aio_context_release(aio_context);
1838 }
1839
1840 static void drive_backup_abort(BlkActionState *common)
1841 {
1842 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1843
1844 if (state->job) {
1845 job_cancel_sync(&state->job->job, true);
1846 }
1847 }
1848
1849 static void drive_backup_clean(BlkActionState *common)
1850 {
1851 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1852 AioContext *aio_context;
1853
1854 if (!state->bs) {
1855 return;
1856 }
1857
1858 aio_context = bdrv_get_aio_context(state->bs);
1859 aio_context_acquire(aio_context);
1860
1861 bdrv_drained_end(state->bs);
1862
1863 aio_context_release(aio_context);
1864 }
1865
1866 typedef struct BlockdevBackupState {
1867 BlkActionState common;
1868 BlockDriverState *bs;
1869 BlockJob *job;
1870 } BlockdevBackupState;
1871
1872 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1873 {
1874 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1875 BlockdevBackup *backup;
1876 BlockDriverState *bs;
1877 BlockDriverState *target_bs;
1878 AioContext *aio_context;
1879 AioContext *old_context;
1880 int ret;
1881
1882 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1883 backup = common->action->u.blockdev_backup.data;
1884
1885 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1886 if (!bs) {
1887 return;
1888 }
1889
1890 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1891 if (!target_bs) {
1892 return;
1893 }
1894
1895 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1896 aio_context = bdrv_get_aio_context(bs);
1897 old_context = bdrv_get_aio_context(target_bs);
1898 aio_context_acquire(old_context);
1899
1900 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1901 if (ret < 0) {
1902 aio_context_release(old_context);
1903 return;
1904 }
1905
1906 aio_context_release(old_context);
1907 aio_context_acquire(aio_context);
1908 state->bs = bs;
1909
1910 /* Paired with .clean() */
1911 bdrv_drained_begin(state->bs);
1912
1913 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1914 bs, target_bs, aio_context,
1915 common->block_job_txn, errp);
1916
1917 aio_context_release(aio_context);
1918 }
1919
1920 static void blockdev_backup_commit(BlkActionState *common)
1921 {
1922 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1923 AioContext *aio_context;
1924
1925 aio_context = bdrv_get_aio_context(state->bs);
1926 aio_context_acquire(aio_context);
1927
1928 assert(state->job);
1929 job_start(&state->job->job);
1930
1931 aio_context_release(aio_context);
1932 }
1933
1934 static void blockdev_backup_abort(BlkActionState *common)
1935 {
1936 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1937
1938 if (state->job) {
1939 job_cancel_sync(&state->job->job, true);
1940 }
1941 }
1942
1943 static void blockdev_backup_clean(BlkActionState *common)
1944 {
1945 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1946 AioContext *aio_context;
1947
1948 if (!state->bs) {
1949 return;
1950 }
1951
1952 aio_context = bdrv_get_aio_context(state->bs);
1953 aio_context_acquire(aio_context);
1954
1955 bdrv_drained_end(state->bs);
1956
1957 aio_context_release(aio_context);
1958 }
1959
1960 typedef struct BlockDirtyBitmapState {
1961 BlkActionState common;
1962 BdrvDirtyBitmap *bitmap;
1963 BlockDriverState *bs;
1964 HBitmap *backup;
1965 bool prepared;
1966 bool was_enabled;
1967 } BlockDirtyBitmapState;
1968
1969 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
1970 Error **errp)
1971 {
1972 Error *local_err = NULL;
1973 BlockDirtyBitmapAdd *action;
1974 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1975 common, common);
1976
1977 if (action_check_completion_mode(common, errp) < 0) {
1978 return;
1979 }
1980
1981 action = common->action->u.block_dirty_bitmap_add.data;
1982 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1983 qmp_block_dirty_bitmap_add(action->node, action->name,
1984 action->has_granularity, action->granularity,
1985 action->has_persistent, action->persistent,
1986 action->has_disabled, action->disabled,
1987 &local_err);
1988
1989 if (!local_err) {
1990 state->prepared = true;
1991 } else {
1992 error_propagate(errp, local_err);
1993 }
1994 }
1995
1996 static void block_dirty_bitmap_add_abort(BlkActionState *common)
1997 {
1998 BlockDirtyBitmapAdd *action;
1999 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2000 common, common);
2001
2002 action = common->action->u.block_dirty_bitmap_add.data;
2003 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2004 * then the node reference and bitmap name must have been valid.
2005 */
2006 if (state->prepared) {
2007 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2008 }
2009 }
2010
2011 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2012 Error **errp)
2013 {
2014 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2015 common, common);
2016 BlockDirtyBitmap *action;
2017
2018 if (action_check_completion_mode(common, errp) < 0) {
2019 return;
2020 }
2021
2022 action = common->action->u.block_dirty_bitmap_clear.data;
2023 state->bitmap = block_dirty_bitmap_lookup(action->node,
2024 action->name,
2025 &state->bs,
2026 errp);
2027 if (!state->bitmap) {
2028 return;
2029 }
2030
2031 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2032 return;
2033 }
2034
2035 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2036 }
2037
2038 static void block_dirty_bitmap_restore(BlkActionState *common)
2039 {
2040 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2041 common, common);
2042
2043 if (state->backup) {
2044 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2045 }
2046 }
2047
2048 static void block_dirty_bitmap_free_backup(BlkActionState *common)
2049 {
2050 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2051 common, common);
2052
2053 hbitmap_free(state->backup);
2054 }
2055
2056 static void block_dirty_bitmap_enable_prepare(BlkActionState *common,
2057 Error **errp)
2058 {
2059 BlockDirtyBitmap *action;
2060 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2061 common, common);
2062
2063 if (action_check_completion_mode(common, errp) < 0) {
2064 return;
2065 }
2066
2067 action = common->action->u.block_dirty_bitmap_enable.data;
2068 state->bitmap = block_dirty_bitmap_lookup(action->node,
2069 action->name,
2070 NULL,
2071 errp);
2072 if (!state->bitmap) {
2073 return;
2074 }
2075
2076 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2077 return;
2078 }
2079
2080 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2081 bdrv_enable_dirty_bitmap(state->bitmap);
2082 }
2083
2084 static void block_dirty_bitmap_enable_abort(BlkActionState *common)
2085 {
2086 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2087 common, common);
2088
2089 if (!state->was_enabled) {
2090 bdrv_disable_dirty_bitmap(state->bitmap);
2091 }
2092 }
2093
2094 static void block_dirty_bitmap_disable_prepare(BlkActionState *common,
2095 Error **errp)
2096 {
2097 BlockDirtyBitmap *action;
2098 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2099 common, common);
2100
2101 if (action_check_completion_mode(common, errp) < 0) {
2102 return;
2103 }
2104
2105 action = common->action->u.block_dirty_bitmap_disable.data;
2106 state->bitmap = block_dirty_bitmap_lookup(action->node,
2107 action->name,
2108 NULL,
2109 errp);
2110 if (!state->bitmap) {
2111 return;
2112 }
2113
2114 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2115 return;
2116 }
2117
2118 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2119 bdrv_disable_dirty_bitmap(state->bitmap);
2120 }
2121
2122 static void block_dirty_bitmap_disable_abort(BlkActionState *common)
2123 {
2124 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2125 common, common);
2126
2127 if (state->was_enabled) {
2128 bdrv_enable_dirty_bitmap(state->bitmap);
2129 }
2130 }
2131
2132 static void block_dirty_bitmap_merge_prepare(BlkActionState *common,
2133 Error **errp)
2134 {
2135 BlockDirtyBitmapMerge *action;
2136 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2137 common, common);
2138
2139 if (action_check_completion_mode(common, errp) < 0) {
2140 return;
2141 }
2142
2143 action = common->action->u.block_dirty_bitmap_merge.data;
2144
2145 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2146 action->bitmaps, &state->backup,
2147 errp);
2148 }
2149
2150 static void block_dirty_bitmap_remove_prepare(BlkActionState *common,
2151 Error **errp)
2152 {
2153 BlockDirtyBitmap *action;
2154 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2155 common, common);
2156
2157 if (action_check_completion_mode(common, errp) < 0) {
2158 return;
2159 }
2160
2161 action = common->action->u.block_dirty_bitmap_remove.data;
2162
2163 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2164 false, &state->bs, errp);
2165 if (state->bitmap) {
2166 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2167 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2168 }
2169 }
2170
2171 static void block_dirty_bitmap_remove_abort(BlkActionState *common)
2172 {
2173 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2174 common, common);
2175
2176 if (state->bitmap) {
2177 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2178 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2179 }
2180 }
2181
2182 static void block_dirty_bitmap_remove_commit(BlkActionState *common)
2183 {
2184 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2185 common, common);
2186
2187 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2188 bdrv_release_dirty_bitmap(state->bitmap);
2189 }
2190
2191 static void abort_prepare(BlkActionState *common, Error **errp)
2192 {
2193 error_setg(errp, "Transaction aborted using Abort action");
2194 }
2195
2196 static void abort_commit(BlkActionState *common)
2197 {
2198 g_assert_not_reached(); /* this action never succeeds */
2199 }
2200
2201 static const BlkActionOps actions[] = {
2202 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2203 .instance_size = sizeof(ExternalSnapshotState),
2204 .prepare = external_snapshot_prepare,
2205 .commit = external_snapshot_commit,
2206 .abort = external_snapshot_abort,
2207 .clean = external_snapshot_clean,
2208 },
2209 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2210 .instance_size = sizeof(ExternalSnapshotState),
2211 .prepare = external_snapshot_prepare,
2212 .commit = external_snapshot_commit,
2213 .abort = external_snapshot_abort,
2214 .clean = external_snapshot_clean,
2215 },
2216 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2217 .instance_size = sizeof(DriveBackupState),
2218 .prepare = drive_backup_prepare,
2219 .commit = drive_backup_commit,
2220 .abort = drive_backup_abort,
2221 .clean = drive_backup_clean,
2222 },
2223 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2224 .instance_size = sizeof(BlockdevBackupState),
2225 .prepare = blockdev_backup_prepare,
2226 .commit = blockdev_backup_commit,
2227 .abort = blockdev_backup_abort,
2228 .clean = blockdev_backup_clean,
2229 },
2230 [TRANSACTION_ACTION_KIND_ABORT] = {
2231 .instance_size = sizeof(BlkActionState),
2232 .prepare = abort_prepare,
2233 .commit = abort_commit,
2234 },
2235 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2236 .instance_size = sizeof(InternalSnapshotState),
2237 .prepare = internal_snapshot_prepare,
2238 .abort = internal_snapshot_abort,
2239 .clean = internal_snapshot_clean,
2240 },
2241 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2242 .instance_size = sizeof(BlockDirtyBitmapState),
2243 .prepare = block_dirty_bitmap_add_prepare,
2244 .abort = block_dirty_bitmap_add_abort,
2245 },
2246 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2247 .instance_size = sizeof(BlockDirtyBitmapState),
2248 .prepare = block_dirty_bitmap_clear_prepare,
2249 .commit = block_dirty_bitmap_free_backup,
2250 .abort = block_dirty_bitmap_restore,
2251 },
2252 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2253 .instance_size = sizeof(BlockDirtyBitmapState),
2254 .prepare = block_dirty_bitmap_enable_prepare,
2255 .abort = block_dirty_bitmap_enable_abort,
2256 },
2257 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2258 .instance_size = sizeof(BlockDirtyBitmapState),
2259 .prepare = block_dirty_bitmap_disable_prepare,
2260 .abort = block_dirty_bitmap_disable_abort,
2261 },
2262 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2263 .instance_size = sizeof(BlockDirtyBitmapState),
2264 .prepare = block_dirty_bitmap_merge_prepare,
2265 .commit = block_dirty_bitmap_free_backup,
2266 .abort = block_dirty_bitmap_restore,
2267 },
2268 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE] = {
2269 .instance_size = sizeof(BlockDirtyBitmapState),
2270 .prepare = block_dirty_bitmap_remove_prepare,
2271 .commit = block_dirty_bitmap_remove_commit,
2272 .abort = block_dirty_bitmap_remove_abort,
2273 },
2274 /* Where are transactions for MIRROR, COMMIT and STREAM?
2275 * Although these blockjobs use transaction callbacks like the backup job,
2276 * these jobs do not necessarily adhere to transaction semantics.
2277 * These jobs may not fully undo all of their actions on abort, nor do they
2278 * necessarily work in transactions with more than one job in them.
2279 */
2280 };
2281
2282 /**
2283 * Allocate a TransactionProperties structure if necessary, and fill
2284 * that structure with desired defaults if they are unset.
2285 */
2286 static TransactionProperties *get_transaction_properties(
2287 TransactionProperties *props)
2288 {
2289 if (!props) {
2290 props = g_new0(TransactionProperties, 1);
2291 }
2292
2293 if (!props->has_completion_mode) {
2294 props->has_completion_mode = true;
2295 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2296 }
2297
2298 return props;
2299 }
2300
2301 /*
2302 * 'Atomic' group operations. The operations are performed as a set, and if
2303 * any fail then we roll back all operations in the group.
2304 *
2305 * Always run under BQL.
2306 */
2307 void qmp_transaction(TransactionActionList *dev_list,
2308 bool has_props,
2309 struct TransactionProperties *props,
2310 Error **errp)
2311 {
2312 TransactionActionList *dev_entry = dev_list;
2313 JobTxn *block_job_txn = NULL;
2314 BlkActionState *state, *next;
2315 Error *local_err = NULL;
2316
2317 GLOBAL_STATE_CODE();
2318
2319 QTAILQ_HEAD(, BlkActionState) snap_bdrv_states;
2320 QTAILQ_INIT(&snap_bdrv_states);
2321
2322 /* Does this transaction get canceled as a group on failure?
2323 * If not, we don't really need to make a JobTxn.
2324 */
2325 props = get_transaction_properties(props);
2326 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2327 block_job_txn = job_txn_new();
2328 }
2329
2330 /* drain all i/o before any operations */
2331 bdrv_drain_all();
2332
2333 /* We don't do anything in this loop that commits us to the operations */
2334 while (NULL != dev_entry) {
2335 TransactionAction *dev_info = NULL;
2336 const BlkActionOps *ops;
2337
2338 dev_info = dev_entry->value;
2339 dev_entry = dev_entry->next;
2340
2341 assert(dev_info->type < ARRAY_SIZE(actions));
2342
2343 ops = &actions[dev_info->type];
2344 assert(ops->instance_size > 0);
2345
2346 state = g_malloc0(ops->instance_size);
2347 state->ops = ops;
2348 state->action = dev_info;
2349 state->block_job_txn = block_job_txn;
2350 state->txn_props = props;
2351 QTAILQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2352
2353 state->ops->prepare(state, &local_err);
2354 if (local_err) {
2355 error_propagate(errp, local_err);
2356 goto delete_and_fail;
2357 }
2358 }
2359
2360 QTAILQ_FOREACH(state, &snap_bdrv_states, entry) {
2361 if (state->ops->commit) {
2362 state->ops->commit(state);
2363 }
2364 }
2365
2366 /* success */
2367 goto exit;
2368
2369 delete_and_fail:
2370 /* failure, and it is all-or-none; roll back all operations */
2371 QTAILQ_FOREACH_REVERSE(state, &snap_bdrv_states, entry) {
2372 if (state->ops->abort) {
2373 state->ops->abort(state);
2374 }
2375 }
2376 exit:
2377 QTAILQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2378 if (state->ops->clean) {
2379 state->ops->clean(state);
2380 }
2381 g_free(state);
2382 }
2383 if (!has_props) {
2384 qapi_free_TransactionProperties(props);
2385 }
2386 job_txn_unref(block_job_txn);
2387 }
2388
2389 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2390 const char *name,
2391 Error **errp)
2392 {
2393 BdrvDirtyBitmap *bitmap;
2394 BlockDriverState *bs;
2395 BlockDirtyBitmapSha256 *ret = NULL;
2396 char *sha256;
2397
2398 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2399 if (!bitmap || !bs) {
2400 return NULL;
2401 }
2402
2403 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2404 if (sha256 == NULL) {
2405 return NULL;
2406 }
2407
2408 ret = g_new(BlockDirtyBitmapSha256, 1);
2409 ret->sha256 = sha256;
2410
2411 return ret;
2412 }
2413
2414 void coroutine_fn qmp_block_resize(bool has_device, const char *device,
2415 bool has_node_name, const char *node_name,
2416 int64_t size, Error **errp)
2417 {
2418 Error *local_err = NULL;
2419 BlockBackend *blk;
2420 BlockDriverState *bs;
2421 AioContext *old_ctx;
2422
2423 bs = bdrv_lookup_bs(has_device ? device : NULL,
2424 has_node_name ? node_name : NULL,
2425 &local_err);
2426 if (local_err) {
2427 error_propagate(errp, local_err);
2428 return;
2429 }
2430
2431 if (size < 0) {
2432 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2433 return;
2434 }
2435
2436 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2437 error_setg(errp, QERR_DEVICE_IN_USE, device);
2438 return;
2439 }
2440
2441 blk = blk_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2442 if (!blk) {
2443 return;
2444 }
2445
2446 bdrv_co_lock(bs);
2447 bdrv_drained_begin(bs);
2448 bdrv_co_unlock(bs);
2449
2450 old_ctx = bdrv_co_enter(bs);
2451 blk_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2452 bdrv_co_leave(bs, old_ctx);
2453
2454 bdrv_co_lock(bs);
2455 bdrv_drained_end(bs);
2456 blk_unref(blk);
2457 bdrv_co_unlock(bs);
2458 }
2459
2460 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
2461 bool has_base, const char *base,
2462 bool has_base_node, const char *base_node,
2463 bool has_backing_file, const char *backing_file,
2464 bool has_bottom, const char *bottom,
2465 bool has_speed, int64_t speed,
2466 bool has_on_error, BlockdevOnError on_error,
2467 bool has_filter_node_name, const char *filter_node_name,
2468 bool has_auto_finalize, bool auto_finalize,
2469 bool has_auto_dismiss, bool auto_dismiss,
2470 Error **errp)
2471 {
2472 BlockDriverState *bs, *iter, *iter_end;
2473 BlockDriverState *base_bs = NULL;
2474 BlockDriverState *bottom_bs = NULL;
2475 AioContext *aio_context;
2476 Error *local_err = NULL;
2477 int job_flags = JOB_DEFAULT;
2478
2479 if (has_base && has_base_node) {
2480 error_setg(errp, "'base' and 'base-node' cannot be specified "
2481 "at the same time");
2482 return;
2483 }
2484
2485 if (has_base && has_bottom) {
2486 error_setg(errp, "'base' and 'bottom' cannot be specified "
2487 "at the same time");
2488 return;
2489 }
2490
2491 if (has_bottom && has_base_node) {
2492 error_setg(errp, "'bottom' and 'base-node' cannot be specified "
2493 "at the same time");
2494 return;
2495 }
2496
2497 if (!has_on_error) {
2498 on_error = BLOCKDEV_ON_ERROR_REPORT;
2499 }
2500
2501 bs = bdrv_lookup_bs(device, device, errp);
2502 if (!bs) {
2503 return;
2504 }
2505
2506 aio_context = bdrv_get_aio_context(bs);
2507 aio_context_acquire(aio_context);
2508
2509 if (has_base) {
2510 base_bs = bdrv_find_backing_image(bs, base);
2511 if (base_bs == NULL) {
2512 error_setg(errp, "Can't find '%s' in the backing chain", base);
2513 goto out;
2514 }
2515 assert(bdrv_get_aio_context(base_bs) == aio_context);
2516 }
2517
2518 if (has_base_node) {
2519 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2520 if (!base_bs) {
2521 goto out;
2522 }
2523 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2524 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2525 base_node, device);
2526 goto out;
2527 }
2528 assert(bdrv_get_aio_context(base_bs) == aio_context);
2529 bdrv_refresh_filename(base_bs);
2530 }
2531
2532 if (has_bottom) {
2533 bottom_bs = bdrv_lookup_bs(NULL, bottom, errp);
2534 if (!bottom_bs) {
2535 goto out;
2536 }
2537 if (!bottom_bs->drv) {
2538 error_setg(errp, "Node '%s' is not open", bottom);
2539 goto out;
2540 }
2541 if (bottom_bs->drv->is_filter) {
2542 error_setg(errp, "Node '%s' is a filter, use a non-filter node "
2543 "as 'bottom'", bottom);
2544 goto out;
2545 }
2546 if (!bdrv_chain_contains(bs, bottom_bs)) {
2547 error_setg(errp, "Node '%s' is not in a chain starting from '%s'",
2548 bottom, device);
2549 goto out;
2550 }
2551 assert(bdrv_get_aio_context(bottom_bs) == aio_context);
2552 }
2553
2554 /*
2555 * Check for op blockers in the whole chain between bs and base (or bottom)
2556 */
2557 iter_end = has_bottom ? bdrv_filter_or_cow_bs(bottom_bs) : base_bs;
2558 for (iter = bs; iter && iter != iter_end;
2559 iter = bdrv_filter_or_cow_bs(iter))
2560 {
2561 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2562 goto out;
2563 }
2564 }
2565
2566 /* if we are streaming the entire chain, the result will have no backing
2567 * file, and specifying one is therefore an error */
2568 if (base_bs == NULL && has_backing_file) {
2569 error_setg(errp, "backing file specified, but streaming the "
2570 "entire chain");
2571 goto out;
2572 }
2573
2574 if (has_auto_finalize && !auto_finalize) {
2575 job_flags |= JOB_MANUAL_FINALIZE;
2576 }
2577 if (has_auto_dismiss && !auto_dismiss) {
2578 job_flags |= JOB_MANUAL_DISMISS;
2579 }
2580
2581 stream_start(has_job_id ? job_id : NULL, bs, base_bs, backing_file,
2582 bottom_bs, job_flags, has_speed ? speed : 0, on_error,
2583 filter_node_name, &local_err);
2584 if (local_err) {
2585 error_propagate(errp, local_err);
2586 goto out;
2587 }
2588
2589 trace_qmp_block_stream(bs);
2590
2591 out:
2592 aio_context_release(aio_context);
2593 }
2594
2595 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
2596 bool has_base_node, const char *base_node,
2597 bool has_base, const char *base,
2598 bool has_top_node, const char *top_node,
2599 bool has_top, const char *top,
2600 bool has_backing_file, const char *backing_file,
2601 bool has_speed, int64_t speed,
2602 bool has_on_error, BlockdevOnError on_error,
2603 bool has_filter_node_name, const char *filter_node_name,
2604 bool has_auto_finalize, bool auto_finalize,
2605 bool has_auto_dismiss, bool auto_dismiss,
2606 Error **errp)
2607 {
2608 BlockDriverState *bs;
2609 BlockDriverState *iter;
2610 BlockDriverState *base_bs, *top_bs;
2611 AioContext *aio_context;
2612 Error *local_err = NULL;
2613 int job_flags = JOB_DEFAULT;
2614 uint64_t top_perm, top_shared;
2615
2616 if (!has_speed) {
2617 speed = 0;
2618 }
2619 if (!has_on_error) {
2620 on_error = BLOCKDEV_ON_ERROR_REPORT;
2621 }
2622 if (!has_filter_node_name) {
2623 filter_node_name = NULL;
2624 }
2625 if (has_auto_finalize && !auto_finalize) {
2626 job_flags |= JOB_MANUAL_FINALIZE;
2627 }
2628 if (has_auto_dismiss && !auto_dismiss) {
2629 job_flags |= JOB_MANUAL_DISMISS;
2630 }
2631
2632 /* Important Note:
2633 * libvirt relies on the DeviceNotFound error class in order to probe for
2634 * live commit feature versions; for this to work, we must make sure to
2635 * perform the device lookup before any generic errors that may occur in a
2636 * scenario in which all optional arguments are omitted. */
2637 bs = qmp_get_root_bs(device, &local_err);
2638 if (!bs) {
2639 bs = bdrv_lookup_bs(device, device, NULL);
2640 if (!bs) {
2641 error_free(local_err);
2642 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2643 "Device '%s' not found", device);
2644 } else {
2645 error_propagate(errp, local_err);
2646 }
2647 return;
2648 }
2649
2650 aio_context = bdrv_get_aio_context(bs);
2651 aio_context_acquire(aio_context);
2652
2653 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2654 goto out;
2655 }
2656
2657 /* default top_bs is the active layer */
2658 top_bs = bs;
2659
2660 if (has_top_node && has_top) {
2661 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2662 goto out;
2663 } else if (has_top_node) {
2664 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2665 if (top_bs == NULL) {
2666 goto out;
2667 }
2668 if (!bdrv_chain_contains(bs, top_bs)) {
2669 error_setg(errp, "'%s' is not in this backing file chain",
2670 top_node);
2671 goto out;
2672 }
2673 } else if (has_top && top) {
2674 /* This strcmp() is just a shortcut, there is no need to
2675 * refresh @bs's filename. If it mismatches,
2676 * bdrv_find_backing_image() will do the refresh and may still
2677 * return @bs. */
2678 if (strcmp(bs->filename, top) != 0) {
2679 top_bs = bdrv_find_backing_image(bs, top);
2680 }
2681 }
2682
2683 if (top_bs == NULL) {
2684 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2685 goto out;
2686 }
2687
2688 assert(bdrv_get_aio_context(top_bs) == aio_context);
2689
2690 if (has_base_node && has_base) {
2691 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2692 goto out;
2693 } else if (has_base_node) {
2694 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2695 if (base_bs == NULL) {
2696 goto out;
2697 }
2698 if (!bdrv_chain_contains(top_bs, base_bs)) {
2699 error_setg(errp, "'%s' is not in this backing file chain",
2700 base_node);
2701 goto out;
2702 }
2703 } else if (has_base && base) {
2704 base_bs = bdrv_find_backing_image(top_bs, base);
2705 if (base_bs == NULL) {
2706 error_setg(errp, "Can't find '%s' in the backing chain", base);
2707 goto out;
2708 }
2709 } else {
2710 base_bs = bdrv_find_base(top_bs);
2711 if (base_bs == NULL) {
2712 error_setg(errp, "There is no backimg image");
2713 goto out;
2714 }
2715 }
2716
2717 assert(bdrv_get_aio_context(base_bs) == aio_context);
2718
2719 for (iter = top_bs; iter != bdrv_filter_or_cow_bs(base_bs);
2720 iter = bdrv_filter_or_cow_bs(iter))
2721 {
2722 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2723 goto out;
2724 }
2725 }
2726
2727 /* Do not allow attempts to commit an image into itself */
2728 if (top_bs == base_bs) {
2729 error_setg(errp, "cannot commit an image into itself");
2730 goto out;
2731 }
2732
2733 /*
2734 * Active commit is required if and only if someone has taken a
2735 * WRITE permission on the top node. Historically, we have always
2736 * used active commit for top nodes, so continue that practice
2737 * lest we possibly break clients that rely on this behavior, e.g.
2738 * to later attach this node to a writing parent.
2739 * (Active commit is never really wrong.)
2740 */
2741 bdrv_get_cumulative_perm(top_bs, &top_perm, &top_shared);
2742 if (top_perm & BLK_PERM_WRITE ||
2743 bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs))
2744 {
2745 if (has_backing_file) {
2746 if (bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs)) {
2747 error_setg(errp, "'backing-file' specified,"
2748 " but 'top' is the active layer");
2749 } else {
2750 error_setg(errp, "'backing-file' specified, but 'top' has a "
2751 "writer on it");
2752 }
2753 goto out;
2754 }
2755 if (!has_job_id) {
2756 /*
2757 * Emulate here what block_job_create() does, because it
2758 * is possible that @bs != @top_bs (the block job should
2759 * be named after @bs, even if @top_bs is the actual
2760 * source)
2761 */
2762 job_id = bdrv_get_device_name(bs);
2763 }
2764 commit_active_start(job_id, top_bs, base_bs, job_flags, speed, on_error,
2765 filter_node_name, NULL, NULL, false, &local_err);
2766 } else {
2767 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2768 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2769 goto out;
2770 }
2771 commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, job_flags,
2772 speed, on_error, has_backing_file ? backing_file : NULL,
2773 filter_node_name, &local_err);
2774 }
2775 if (local_err != NULL) {
2776 error_propagate(errp, local_err);
2777 goto out;
2778 }
2779
2780 out:
2781 aio_context_release(aio_context);
2782 }
2783
2784 /* Common QMP interface for drive-backup and blockdev-backup */
2785 static BlockJob *do_backup_common(BackupCommon *backup,
2786 BlockDriverState *bs,
2787 BlockDriverState *target_bs,
2788 AioContext *aio_context,
2789 JobTxn *txn, Error **errp)
2790 {
2791 BlockJob *job = NULL;
2792 BdrvDirtyBitmap *bmap = NULL;
2793 BackupPerf perf = { .max_workers = 64 };
2794 int job_flags = JOB_DEFAULT;
2795
2796 if (!backup->has_speed) {
2797 backup->speed = 0;
2798 }
2799 if (!backup->has_on_source_error) {
2800 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2801 }
2802 if (!backup->has_on_target_error) {
2803 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2804 }
2805 if (!backup->has_job_id) {
2806 backup->job_id = NULL;
2807 }
2808 if (!backup->has_auto_finalize) {
2809 backup->auto_finalize = true;
2810 }
2811 if (!backup->has_auto_dismiss) {
2812 backup->auto_dismiss = true;
2813 }
2814 if (!backup->has_compress) {
2815 backup->compress = false;
2816 }
2817
2818 if (backup->x_perf) {
2819 if (backup->x_perf->has_use_copy_range) {
2820 perf.use_copy_range = backup->x_perf->use_copy_range;
2821 }
2822 if (backup->x_perf->has_max_workers) {
2823 perf.max_workers = backup->x_perf->max_workers;
2824 }
2825 if (backup->x_perf->has_max_chunk) {
2826 perf.max_chunk = backup->x_perf->max_chunk;
2827 }
2828 }
2829
2830 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2831 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2832 /* done before desugaring 'incremental' to print the right message */
2833 if (!backup->has_bitmap) {
2834 error_setg(errp, "must provide a valid bitmap name for "
2835 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2836 return NULL;
2837 }
2838 }
2839
2840 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2841 if (backup->has_bitmap_mode &&
2842 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2843 error_setg(errp, "Bitmap sync mode must be '%s' "
2844 "when using sync mode '%s'",
2845 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2846 MirrorSyncMode_str(backup->sync));
2847 return NULL;
2848 }
2849 backup->has_bitmap_mode = true;
2850 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2851 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2852 }
2853
2854 if (backup->has_bitmap) {
2855 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2856 if (!bmap) {
2857 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2858 return NULL;
2859 }
2860 if (!backup->has_bitmap_mode) {
2861 error_setg(errp, "Bitmap sync mode must be given "
2862 "when providing a bitmap");
2863 return NULL;
2864 }
2865 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2866 return NULL;
2867 }
2868
2869 /* This does not produce a useful bitmap artifact: */
2870 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2871 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2872 " outputs", MirrorSyncMode_str(backup->sync));
2873 return NULL;
2874 }
2875
2876 /* If the bitmap isn't used for input or output, this is useless: */
2877 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2878 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2879 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2880 " when combined with sync mode '%s'",
2881 BitmapSyncMode_str(backup->bitmap_mode),
2882 MirrorSyncMode_str(backup->sync));
2883 return NULL;
2884 }
2885 }
2886
2887 if (!backup->has_bitmap && backup->has_bitmap_mode) {
2888 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2889 return NULL;
2890 }
2891
2892 if (!backup->auto_finalize) {
2893 job_flags |= JOB_MANUAL_FINALIZE;
2894 }
2895 if (!backup->auto_dismiss) {
2896 job_flags |= JOB_MANUAL_DISMISS;
2897 }
2898
2899 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2900 backup->sync, bmap, backup->bitmap_mode,
2901 backup->compress,
2902 backup->filter_node_name,
2903 &perf,
2904 backup->on_source_error,
2905 backup->on_target_error,
2906 job_flags, NULL, NULL, txn, errp);
2907 return job;
2908 }
2909
2910 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2911 {
2912 TransactionAction action = {
2913 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2914 .u.drive_backup.data = backup,
2915 };
2916 blockdev_do_action(&action, errp);
2917 }
2918
2919 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2920 bool flat,
2921 Error **errp)
2922 {
2923 bool return_flat = has_flat && flat;
2924
2925 return bdrv_named_nodes_list(return_flat, errp);
2926 }
2927
2928 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2929 {
2930 return bdrv_get_xdbg_block_graph(errp);
2931 }
2932
2933 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2934 {
2935 TransactionAction action = {
2936 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2937 .u.blockdev_backup.data = backup,
2938 };
2939 blockdev_do_action(&action, errp);
2940 }
2941
2942 /* Parameter check and block job starting for drive mirroring.
2943 * Caller should hold @device and @target's aio context (must be the same).
2944 **/
2945 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2946 BlockDriverState *target,
2947 bool has_replaces, const char *replaces,
2948 enum MirrorSyncMode sync,
2949 BlockMirrorBackingMode backing_mode,
2950 bool zero_target,
2951 bool has_speed, int64_t speed,
2952 bool has_granularity, uint32_t granularity,
2953 bool has_buf_size, int64_t buf_size,
2954 bool has_on_source_error,
2955 BlockdevOnError on_source_error,
2956 bool has_on_target_error,
2957 BlockdevOnError on_target_error,
2958 bool has_unmap, bool unmap,
2959 bool has_filter_node_name,
2960 const char *filter_node_name,
2961 bool has_copy_mode, MirrorCopyMode copy_mode,
2962 bool has_auto_finalize, bool auto_finalize,
2963 bool has_auto_dismiss, bool auto_dismiss,
2964 Error **errp)
2965 {
2966 BlockDriverState *unfiltered_bs;
2967 int job_flags = JOB_DEFAULT;
2968
2969 if (!has_speed) {
2970 speed = 0;
2971 }
2972 if (!has_on_source_error) {
2973 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2974 }
2975 if (!has_on_target_error) {
2976 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2977 }
2978 if (!has_granularity) {
2979 granularity = 0;
2980 }
2981 if (!has_buf_size) {
2982 buf_size = 0;
2983 }
2984 if (!has_unmap) {
2985 unmap = true;
2986 }
2987 if (!has_filter_node_name) {
2988 filter_node_name = NULL;
2989 }
2990 if (!has_copy_mode) {
2991 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
2992 }
2993 if (has_auto_finalize && !auto_finalize) {
2994 job_flags |= JOB_MANUAL_FINALIZE;
2995 }
2996 if (has_auto_dismiss && !auto_dismiss) {
2997 job_flags |= JOB_MANUAL_DISMISS;
2998 }
2999
3000 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3001 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3002 "a value in range [512B, 64MB]");
3003 return;
3004 }
3005 if (granularity & (granularity - 1)) {
3006 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3007 "a power of 2");
3008 return;
3009 }
3010
3011 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3012 return;
3013 }
3014 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3015 return;
3016 }
3017
3018 if (!bdrv_backing_chain_next(bs) && sync == MIRROR_SYNC_MODE_TOP) {
3019 sync = MIRROR_SYNC_MODE_FULL;
3020 }
3021
3022 if (!has_replaces) {
3023 /* We want to mirror from @bs, but keep implicit filters on top */
3024 unfiltered_bs = bdrv_skip_implicit_filters(bs);
3025 if (unfiltered_bs != bs) {
3026 replaces = unfiltered_bs->node_name;
3027 has_replaces = true;
3028 }
3029 }
3030
3031 if (has_replaces) {
3032 BlockDriverState *to_replace_bs;
3033 AioContext *replace_aio_context;
3034 int64_t bs_size, replace_size;
3035
3036 bs_size = bdrv_getlength(bs);
3037 if (bs_size < 0) {
3038 error_setg_errno(errp, -bs_size, "Failed to query device's size");
3039 return;
3040 }
3041
3042 to_replace_bs = check_to_replace_node(bs, replaces, errp);
3043 if (!to_replace_bs) {
3044 return;
3045 }
3046
3047 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3048 aio_context_acquire(replace_aio_context);
3049 replace_size = bdrv_getlength(to_replace_bs);
3050 aio_context_release(replace_aio_context);
3051
3052 if (replace_size < 0) {
3053 error_setg_errno(errp, -replace_size,
3054 "Failed to query the replacement node's size");
3055 return;
3056 }
3057 if (bs_size != replace_size) {
3058 error_setg(errp, "cannot replace image with a mirror image of "
3059 "different size");
3060 return;
3061 }
3062 }
3063
3064 /* pass the node name to replace to mirror start since it's loose coupling
3065 * and will allow to check whether the node still exist at mirror completion
3066 */
3067 mirror_start(job_id, bs, target,
3068 has_replaces ? replaces : NULL, job_flags,
3069 speed, granularity, buf_size, sync, backing_mode, zero_target,
3070 on_source_error, on_target_error, unmap, filter_node_name,
3071 copy_mode, errp);
3072 }
3073
3074 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3075 {
3076 BlockDriverState *bs;
3077 BlockDriverState *target_backing_bs, *target_bs;
3078 AioContext *aio_context;
3079 AioContext *old_context;
3080 BlockMirrorBackingMode backing_mode;
3081 Error *local_err = NULL;
3082 QDict *options = NULL;
3083 int flags;
3084 int64_t size;
3085 const char *format = arg->format;
3086 bool zero_target;
3087 int ret;
3088
3089 bs = qmp_get_root_bs(arg->device, errp);
3090 if (!bs) {
3091 return;
3092 }
3093
3094 /* Early check to avoid creating target */
3095 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3096 return;
3097 }
3098
3099 aio_context = bdrv_get_aio_context(bs);
3100 aio_context_acquire(aio_context);
3101
3102 if (!arg->has_mode) {
3103 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3104 }
3105
3106 if (!arg->has_format) {
3107 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3108 ? NULL : bs->drv->format_name);
3109 }
3110
3111 flags = bs->open_flags | BDRV_O_RDWR;
3112 target_backing_bs = bdrv_cow_bs(bdrv_skip_filters(bs));
3113 if (!target_backing_bs && arg->sync == MIRROR_SYNC_MODE_TOP) {
3114 arg->sync = MIRROR_SYNC_MODE_FULL;
3115 }
3116 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3117 target_backing_bs = bs;
3118 }
3119
3120 size = bdrv_getlength(bs);
3121 if (size < 0) {
3122 error_setg_errno(errp, -size, "bdrv_getlength failed");
3123 goto out;
3124 }
3125
3126 if (arg->has_replaces) {
3127 if (!arg->has_node_name) {
3128 error_setg(errp, "a node-name must be provided when replacing a"
3129 " named node of the graph");
3130 goto out;
3131 }
3132 }
3133
3134 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3135 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3136 } else {
3137 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3138 }
3139
3140 /* Don't open backing image in create() */
3141 flags |= BDRV_O_NO_BACKING;
3142
3143 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !target_backing_bs)
3144 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3145 {
3146 /* create new image w/o backing file */
3147 assert(format);
3148 bdrv_img_create(arg->target, format,
3149 NULL, NULL, NULL, size, flags, false, &local_err);
3150 } else {
3151 /* Implicit filters should not appear in the filename */
3152 BlockDriverState *explicit_backing =
3153 bdrv_skip_implicit_filters(target_backing_bs);
3154
3155 switch (arg->mode) {
3156 case NEW_IMAGE_MODE_EXISTING:
3157 break;
3158 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3159 /* create new image with backing file */
3160 bdrv_refresh_filename(explicit_backing);
3161 bdrv_img_create(arg->target, format,
3162 explicit_backing->filename,
3163 explicit_backing->drv->format_name,
3164 NULL, size, flags, false, &local_err);
3165 break;
3166 default:
3167 abort();
3168 }
3169 }
3170
3171 if (local_err) {
3172 error_propagate(errp, local_err);
3173 goto out;
3174 }
3175
3176 options = qdict_new();
3177 if (arg->has_node_name) {
3178 qdict_put_str(options, "node-name", arg->node_name);
3179 }
3180 if (format) {
3181 qdict_put_str(options, "driver", format);
3182 }
3183
3184 /* Mirroring takes care of copy-on-write using the source's backing
3185 * file.
3186 */
3187 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3188 if (!target_bs) {
3189 goto out;
3190 }
3191
3192 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3193 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3194 !bdrv_has_zero_init(target_bs)));
3195
3196
3197 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3198 old_context = bdrv_get_aio_context(target_bs);
3199 aio_context_release(aio_context);
3200 aio_context_acquire(old_context);
3201
3202 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3203 if (ret < 0) {
3204 bdrv_unref(target_bs);
3205 aio_context_release(old_context);
3206 return;
3207 }
3208
3209 aio_context_release(old_context);
3210 aio_context_acquire(aio_context);
3211
3212 blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3213 arg->has_replaces, arg->replaces, arg->sync,
3214 backing_mode, zero_target,
3215 arg->has_speed, arg->speed,
3216 arg->has_granularity, arg->granularity,
3217 arg->has_buf_size, arg->buf_size,
3218 arg->has_on_source_error, arg->on_source_error,
3219 arg->has_on_target_error, arg->on_target_error,
3220 arg->has_unmap, arg->unmap,
3221 false, NULL,
3222 arg->has_copy_mode, arg->copy_mode,
3223 arg->has_auto_finalize, arg->auto_finalize,
3224 arg->has_auto_dismiss, arg->auto_dismiss,
3225 errp);
3226 bdrv_unref(target_bs);
3227 out:
3228 aio_context_release(aio_context);
3229 }
3230
3231 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3232 const char *device, const char *target,
3233 bool has_replaces, const char *replaces,
3234 MirrorSyncMode sync,
3235 bool has_speed, int64_t speed,
3236 bool has_granularity, uint32_t granularity,
3237 bool has_buf_size, int64_t buf_size,
3238 bool has_on_source_error,
3239 BlockdevOnError on_source_error,
3240 bool has_on_target_error,
3241 BlockdevOnError on_target_error,
3242 bool has_filter_node_name,
3243 const char *filter_node_name,
3244 bool has_copy_mode, MirrorCopyMode copy_mode,
3245 bool has_auto_finalize, bool auto_finalize,
3246 bool has_auto_dismiss, bool auto_dismiss,
3247 Error **errp)
3248 {
3249 BlockDriverState *bs;
3250 BlockDriverState *target_bs;
3251 AioContext *aio_context;
3252 AioContext *old_context;
3253 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3254 bool zero_target;
3255 int ret;
3256
3257 bs = qmp_get_root_bs(device, errp);
3258 if (!bs) {
3259 return;
3260 }
3261
3262 target_bs = bdrv_lookup_bs(target, target, errp);
3263 if (!target_bs) {
3264 return;
3265 }
3266
3267 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3268
3269 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3270 old_context = bdrv_get_aio_context(target_bs);
3271 aio_context = bdrv_get_aio_context(bs);
3272 aio_context_acquire(old_context);
3273
3274 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3275
3276 aio_context_release(old_context);
3277 aio_context_acquire(aio_context);
3278
3279 if (ret < 0) {
3280 goto out;
3281 }
3282
3283 blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
3284 has_replaces, replaces, sync, backing_mode,
3285 zero_target, has_speed, speed,
3286 has_granularity, granularity,
3287 has_buf_size, buf_size,
3288 has_on_source_error, on_source_error,
3289 has_on_target_error, on_target_error,
3290 true, true,
3291 has_filter_node_name, filter_node_name,
3292 has_copy_mode, copy_mode,
3293 has_auto_finalize, auto_finalize,
3294 has_auto_dismiss, auto_dismiss,
3295 errp);
3296 out:
3297 aio_context_release(aio_context);
3298 }
3299
3300 /*
3301 * Get a block job using its ID. Called with job_mutex held.
3302 */
3303 static BlockJob *find_block_job_locked(const char *id, Error **errp)
3304 {
3305 BlockJob *job;
3306
3307 assert(id != NULL);
3308
3309 job = block_job_get_locked(id);
3310
3311 if (!job) {
3312 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3313 "Block job '%s' not found", id);
3314 return NULL;
3315 }
3316
3317 return job;
3318 }
3319
3320 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3321 {
3322 BlockJob *job;
3323
3324 JOB_LOCK_GUARD();
3325 job = find_block_job_locked(device, errp);
3326
3327 if (!job) {
3328 return;
3329 }
3330
3331 block_job_set_speed_locked(job, speed, errp);
3332 }
3333
3334 void qmp_block_job_cancel(const char *device,
3335 bool has_force, bool force, Error **errp)
3336 {
3337 BlockJob *job;
3338
3339 JOB_LOCK_GUARD();
3340 job = find_block_job_locked(device, errp);
3341
3342 if (!job) {
3343 return;
3344 }
3345
3346 if (!has_force) {
3347 force = false;
3348 }
3349
3350 if (job_user_paused_locked(&job->job) && !force) {
3351 error_setg(errp, "The block job for device '%s' is currently paused",
3352 device);
3353 return;
3354 }
3355
3356 trace_qmp_block_job_cancel(job);
3357 job_user_cancel_locked(&job->job, force, errp);
3358 }
3359
3360 void qmp_block_job_pause(const char *device, Error **errp)
3361 {
3362 BlockJob *job;
3363
3364 JOB_LOCK_GUARD();
3365 job = find_block_job_locked(device, errp);
3366
3367 if (!job) {
3368 return;
3369 }
3370
3371 trace_qmp_block_job_pause(job);
3372 job_user_pause_locked(&job->job, errp);
3373 }
3374
3375 void qmp_block_job_resume(const char *device, Error **errp)
3376 {
3377 BlockJob *job;
3378
3379 JOB_LOCK_GUARD();
3380 job = find_block_job_locked(device, errp);
3381
3382 if (!job) {
3383 return;
3384 }
3385
3386 trace_qmp_block_job_resume(job);
3387 job_user_resume_locked(&job->job, errp);
3388 }
3389
3390 void qmp_block_job_complete(const char *device, Error **errp)
3391 {
3392 BlockJob *job;
3393
3394 JOB_LOCK_GUARD();
3395 job = find_block_job_locked(device, errp);
3396
3397 if (!job) {
3398 return;
3399 }
3400
3401 trace_qmp_block_job_complete(job);
3402 job_complete_locked(&job->job, errp);
3403 }
3404
3405 void qmp_block_job_finalize(const char *id, Error **errp)
3406 {
3407 BlockJob *job;
3408
3409 JOB_LOCK_GUARD();
3410 job = find_block_job_locked(id, errp);
3411
3412 if (!job) {
3413 return;
3414 }
3415
3416 trace_qmp_block_job_finalize(job);
3417 job_ref_locked(&job->job);
3418 job_finalize_locked(&job->job, errp);
3419
3420 job_unref_locked(&job->job);
3421 }
3422
3423 void qmp_block_job_dismiss(const char *id, Error **errp)
3424 {
3425 BlockJob *bjob;
3426 Job *job;
3427
3428 JOB_LOCK_GUARD();
3429 bjob = find_block_job_locked(id, errp);
3430
3431 if (!bjob) {
3432 return;
3433 }
3434
3435 trace_qmp_block_job_dismiss(bjob);
3436 job = &bjob->job;
3437 job_dismiss_locked(&job, errp);
3438 }
3439
3440 void qmp_change_backing_file(const char *device,
3441 const char *image_node_name,
3442 const char *backing_file,
3443 Error **errp)
3444 {
3445 BlockDriverState *bs = NULL;
3446 AioContext *aio_context;
3447 BlockDriverState *image_bs = NULL;
3448 Error *local_err = NULL;
3449 bool ro;
3450 int ret;
3451
3452 bs = qmp_get_root_bs(device, errp);
3453 if (!bs) {
3454 return;
3455 }
3456
3457 aio_context = bdrv_get_aio_context(bs);
3458 aio_context_acquire(aio_context);
3459
3460 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3461 if (local_err) {
3462 error_propagate(errp, local_err);
3463 goto out;
3464 }
3465
3466 if (!image_bs) {
3467 error_setg(errp, "image file not found");
3468 goto out;
3469 }
3470
3471 if (bdrv_find_base(image_bs) == image_bs) {
3472 error_setg(errp, "not allowing backing file change on an image "
3473 "without a backing file");
3474 goto out;
3475 }
3476
3477 /* even though we are not necessarily operating on bs, we need it to
3478 * determine if block ops are currently prohibited on the chain */
3479 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3480 goto out;
3481 }
3482
3483 /* final sanity check */
3484 if (!bdrv_chain_contains(bs, image_bs)) {
3485 error_setg(errp, "'%s' and image file are not in the same chain",
3486 device);
3487 goto out;
3488 }
3489
3490 /* if not r/w, reopen to make r/w */
3491 ro = bdrv_is_read_only(image_bs);
3492
3493 if (ro) {
3494 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3495 goto out;
3496 }
3497 }
3498
3499 ret = bdrv_change_backing_file(image_bs, backing_file,
3500 image_bs->drv ? image_bs->drv->format_name : "",
3501 false);
3502
3503 if (ret < 0) {
3504 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3505 backing_file);
3506 /* don't exit here, so we can try to restore open flags if
3507 * appropriate */
3508 }
3509
3510 if (ro) {
3511 bdrv_reopen_set_read_only(image_bs, true, errp);
3512 }
3513
3514 out:
3515 aio_context_release(aio_context);
3516 }
3517
3518 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3519 {
3520 BlockDriverState *bs;
3521 QObject *obj;
3522 Visitor *v = qobject_output_visitor_new(&obj);
3523 QDict *qdict;
3524
3525 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3526 visit_complete(v, &obj);
3527 qdict = qobject_to(QDict, obj);
3528
3529 qdict_flatten(qdict);
3530
3531 if (!qdict_get_try_str(qdict, "node-name")) {
3532 error_setg(errp, "'node-name' must be specified for the root node");
3533 goto fail;
3534 }
3535
3536 bs = bds_tree_init(qdict, errp);
3537 if (!bs) {
3538 goto fail;
3539 }
3540
3541 bdrv_set_monitor_owned(bs);
3542
3543 fail:
3544 visit_free(v);
3545 }
3546
3547 void qmp_blockdev_reopen(BlockdevOptionsList *reopen_list, Error **errp)
3548 {
3549 BlockReopenQueue *queue = NULL;
3550 GSList *drained = NULL;
3551 GSList *p;
3552
3553 /* Add each one of the BDS that we want to reopen to the queue */
3554 for (; reopen_list != NULL; reopen_list = reopen_list->next) {
3555 BlockdevOptions *options = reopen_list->value;
3556 BlockDriverState *bs;
3557 AioContext *ctx;
3558 QObject *obj;
3559 Visitor *v;
3560 QDict *qdict;
3561
3562 /* Check for the selected node name */
3563 if (!options->has_node_name) {
3564 error_setg(errp, "node-name not specified");
3565 goto fail;
3566 }
3567
3568 bs = bdrv_find_node(options->node_name);
3569 if (!bs) {
3570 error_setg(errp, "Failed to find node with node-name='%s'",
3571 options->node_name);
3572 goto fail;
3573 }
3574
3575 /* Put all options in a QDict and flatten it */
3576 v = qobject_output_visitor_new(&obj);
3577 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3578 visit_complete(v, &obj);
3579 visit_free(v);
3580
3581 qdict = qobject_to(QDict, obj);
3582
3583 qdict_flatten(qdict);
3584
3585 ctx = bdrv_get_aio_context(bs);
3586 aio_context_acquire(ctx);
3587
3588 bdrv_subtree_drained_begin(bs);
3589 queue = bdrv_reopen_queue(queue, bs, qdict, false);
3590 drained = g_slist_prepend(drained, bs);
3591
3592 aio_context_release(ctx);
3593 }
3594
3595 /* Perform the reopen operation */
3596 bdrv_reopen_multiple(queue, errp);
3597 queue = NULL;
3598
3599 fail:
3600 bdrv_reopen_queue_free(queue);
3601 for (p = drained; p; p = p->next) {
3602 BlockDriverState *bs = p->data;
3603 AioContext *ctx = bdrv_get_aio_context(bs);
3604
3605 aio_context_acquire(ctx);
3606 bdrv_subtree_drained_end(bs);
3607 aio_context_release(ctx);
3608 }
3609 g_slist_free(drained);
3610 }
3611
3612 void qmp_blockdev_del(const char *node_name, Error **errp)
3613 {
3614 AioContext *aio_context;
3615 BlockDriverState *bs;
3616
3617 GLOBAL_STATE_CODE();
3618
3619 bs = bdrv_find_node(node_name);
3620 if (!bs) {
3621 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3622 return;
3623 }
3624 if (bdrv_has_blk(bs)) {
3625 error_setg(errp, "Node %s is in use", node_name);
3626 return;
3627 }
3628 aio_context = bdrv_get_aio_context(bs);
3629 aio_context_acquire(aio_context);
3630
3631 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3632 goto out;
3633 }
3634
3635 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3636 error_setg(errp, "Node %s is not owned by the monitor",
3637 bs->node_name);
3638 goto out;
3639 }
3640
3641 if (bs->refcnt > 1) {
3642 error_setg(errp, "Block device %s is in use",
3643 bdrv_get_device_or_node_name(bs));
3644 goto out;
3645 }
3646
3647 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3648 bdrv_unref(bs);
3649
3650 out:
3651 aio_context_release(aio_context);
3652 }
3653
3654 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3655 const char *child_name)
3656 {
3657 BdrvChild *child;
3658
3659 QLIST_FOREACH(child, &parent_bs->children, next) {
3660 if (strcmp(child->name, child_name) == 0) {
3661 return child;
3662 }
3663 }
3664
3665 return NULL;
3666 }
3667
3668 void qmp_x_blockdev_change(const char *parent, bool has_child,
3669 const char *child, bool has_node,
3670 const char *node, Error **errp)
3671 {
3672 BlockDriverState *parent_bs, *new_bs = NULL;
3673 BdrvChild *p_child;
3674
3675 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3676 if (!parent_bs) {
3677 return;
3678 }
3679
3680 if (has_child == has_node) {
3681 if (has_child) {
3682 error_setg(errp, "The parameters child and node are in conflict");
3683 } else {
3684 error_setg(errp, "Either child or node must be specified");
3685 }
3686 return;
3687 }
3688
3689 if (has_child) {
3690 p_child = bdrv_find_child(parent_bs, child);
3691 if (!p_child) {
3692 error_setg(errp, "Node '%s' does not have child '%s'",
3693 parent, child);
3694 return;
3695 }
3696 bdrv_del_child(parent_bs, p_child, errp);
3697 }
3698
3699 if (has_node) {
3700 new_bs = bdrv_find_node(node);
3701 if (!new_bs) {
3702 error_setg(errp, "Node '%s' not found", node);
3703 return;
3704 }
3705 bdrv_add_child(parent_bs, new_bs, errp);
3706 }
3707 }
3708
3709 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3710 {
3711 BlockJobInfoList *head = NULL, **tail = &head;
3712 BlockJob *job;
3713
3714 JOB_LOCK_GUARD();
3715
3716 for (job = block_job_next_locked(NULL); job;
3717 job = block_job_next_locked(job)) {
3718 BlockJobInfo *value;
3719
3720 if (block_job_is_internal(job)) {
3721 continue;
3722 }
3723 value = block_job_query_locked(job, errp);
3724 if (!value) {
3725 qapi_free_BlockJobInfoList(head);
3726 return NULL;
3727 }
3728 QAPI_LIST_APPEND(tail, value);
3729 }
3730
3731 return head;
3732 }
3733
3734 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3735 bool has_force, bool force, Error **errp)
3736 {
3737 AioContext *old_context;
3738 AioContext *new_context;
3739 BlockDriverState *bs;
3740
3741 bs = bdrv_find_node(node_name);
3742 if (!bs) {
3743 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3744 return;
3745 }
3746
3747 /* Protects against accidents. */
3748 if (!(has_force && force) && bdrv_has_blk(bs)) {
3749 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3750 "be in use (use force=true to override this check)",
3751 node_name);
3752 return;
3753 }
3754
3755 if (iothread->type == QTYPE_QSTRING) {
3756 IOThread *obj = iothread_by_id(iothread->u.s);
3757 if (!obj) {
3758 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3759 return;
3760 }
3761
3762 new_context = iothread_get_aio_context(obj);
3763 } else {
3764 new_context = qemu_get_aio_context();
3765 }
3766
3767 old_context = bdrv_get_aio_context(bs);
3768 aio_context_acquire(old_context);
3769
3770 bdrv_try_set_aio_context(bs, new_context, errp);
3771
3772 aio_context_release(old_context);
3773 }
3774
3775 QemuOptsList qemu_common_drive_opts = {
3776 .name = "drive",
3777 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3778 .desc = {
3779 {
3780 .name = "snapshot",
3781 .type = QEMU_OPT_BOOL,
3782 .help = "enable/disable snapshot mode",
3783 },{
3784 .name = "aio",
3785 .type = QEMU_OPT_STRING,
3786 .help = "host AIO implementation (threads, native, io_uring)",
3787 },{
3788 .name = BDRV_OPT_CACHE_WB,
3789 .type = QEMU_OPT_BOOL,
3790 .help = "Enable writeback mode",
3791 },{
3792 .name = "format",
3793 .type = QEMU_OPT_STRING,
3794 .help = "disk format (raw, qcow2, ...)",
3795 },{
3796 .name = "rerror",
3797 .type = QEMU_OPT_STRING,
3798 .help = "read error action",
3799 },{
3800 .name = "werror",
3801 .type = QEMU_OPT_STRING,
3802 .help = "write error action",
3803 },{
3804 .name = BDRV_OPT_READ_ONLY,
3805 .type = QEMU_OPT_BOOL,
3806 .help = "open drive file as read-only",
3807 },
3808
3809 THROTTLE_OPTS,
3810
3811 {
3812 .name = "throttling.group",
3813 .type = QEMU_OPT_STRING,
3814 .help = "name of the block throttling group",
3815 },{
3816 .name = "copy-on-read",
3817 .type = QEMU_OPT_BOOL,
3818 .help = "copy read data from backing file into image file",
3819 },{
3820 .name = "detect-zeroes",
3821 .type = QEMU_OPT_STRING,
3822 .help = "try to optimize zero writes (off, on, unmap)",
3823 },{
3824 .name = "stats-account-invalid",
3825 .type = QEMU_OPT_BOOL,
3826 .help = "whether to account for invalid I/O operations "
3827 "in the statistics",
3828 },{
3829 .name = "stats-account-failed",
3830 .type = QEMU_OPT_BOOL,
3831 .help = "whether to account for failed I/O operations "
3832 "in the statistics",
3833 },
3834 { /* end of list */ }
3835 },
3836 };
3837
3838 QemuOptsList qemu_drive_opts = {
3839 .name = "drive",
3840 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3841 .desc = {
3842 /*
3843 * no elements => accept any params
3844 * validation will happen later
3845 */
3846 { /* end of list */ }
3847 },
3848 };
AR7 emulation for QEMU