?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
block/qcow2-bitmap: return status from qcow2_store_persistent_dirty_bitmaps
[qemu/kevin.git] / blockdev.c
blob65884a28265821c0d49f41b4a6049779d7034892
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/arch_init.h"
60 #include "sysemu/qtest.h"
61 #include "sysemu/runstate.h"
62 #include "sysemu/replay.h"
63 #include "qemu/cutils.h"
64 #include "qemu/help_option.h"
65 #include "qemu/main-loop.h"
66 #include "qemu/throttle-options.h"
67
68 QTAILQ_HEAD(, BlockDriverState) monitor_bdrv_states =
69 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states);
70
71 void bdrv_set_monitor_owned(BlockDriverState *bs)
72 {
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 if (max_devs <= 0) {
117 return;
118 }
119
120 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
121 dinfo = blk_legacy_dinfo(blk);
122 if (dinfo->type == type) {
123 fprintf(stderr, "Cannot override units-per-bus property of"
124 " the %s interface, because a drive of that type has"
125 " already been added.\n", if_name[type]);
126 g_assert_not_reached();
127 }
128 }
129
130 if_max_devs[type] = max_devs;
131 }
132
133 /*
134 * We automatically delete the drive when a device using it gets
135 * unplugged. Questionable feature, but we can't just drop it.
136 * Device models call blockdev_mark_auto_del() to schedule the
137 * automatic deletion, and generic qdev code calls blockdev_auto_del()
138 * when deletion is actually safe.
139 */
140 void blockdev_mark_auto_del(BlockBackend *blk)
141 {
142 DriveInfo *dinfo = blk_legacy_dinfo(blk);
143 BlockJob *job;
144
145 if (!dinfo) {
146 return;
147 }
148
149 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
150 if (block_job_has_bdrv(job, blk_bs(blk))) {
151 AioContext *aio_context = job->job.aio_context;
152 aio_context_acquire(aio_context);
153
154 job_cancel(&job->job, false);
155
156 aio_context_release(aio_context);
157 }
158 }
159
160 dinfo->auto_del = 1;
161 }
162
163 void blockdev_auto_del(BlockBackend *blk)
164 {
165 DriveInfo *dinfo = blk_legacy_dinfo(blk);
166
167 if (dinfo && dinfo->auto_del) {
168 monitor_remove_blk(blk);
169 blk_unref(blk);
170 }
171 }
172
173 /**
174 * Returns the current mapping of how many units per bus
175 * a particular interface can support.
176 *
177 * A positive integer indicates n units per bus.
178 * 0 implies the mapping has not been established.
179 * -1 indicates an invalid BlockInterfaceType was given.
180 */
181 int drive_get_max_devs(BlockInterfaceType type)
182 {
183 if (type >= IF_IDE && type < IF_COUNT) {
184 return if_max_devs[type];
185 }
186
187 return -1;
188 }
189
190 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
191 {
192 int max_devs = if_max_devs[type];
193 return max_devs ? index / max_devs : 0;
194 }
195
196 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
197 {
198 int max_devs = if_max_devs[type];
199 return max_devs ? index % max_devs : index;
200 }
201
202 QemuOpts *drive_def(const char *optstr)
203 {
204 return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
205 }
206
207 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
208 const char *optstr)
209 {
210 QemuOpts *opts;
211
212 opts = drive_def(optstr);
213 if (!opts) {
214 return NULL;
215 }
216 if (type != IF_DEFAULT) {
217 qemu_opt_set(opts, "if", if_name[type], &error_abort);
218 }
219 if (index >= 0) {
220 qemu_opt_set_number(opts, "index", index, &error_abort);
221 }
222 if (file)
223 qemu_opt_set(opts, "file", file, &error_abort);
224 return opts;
225 }
226
227 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
228 {
229 BlockBackend *blk;
230 DriveInfo *dinfo;
231
232 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
233 dinfo = blk_legacy_dinfo(blk);
234 if (dinfo && dinfo->type == type
235 && dinfo->bus == bus && dinfo->unit == unit) {
236 return dinfo;
237 }
238 }
239
240 return NULL;
241 }
242
243 void drive_mark_claimed_by_board(void)
244 {
245 BlockBackend *blk;
246 DriveInfo *dinfo;
247
248 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
249 dinfo = blk_legacy_dinfo(blk);
250 if (dinfo && blk_get_attached_dev(blk)) {
251 dinfo->claimed_by_board = true;
252 }
253 }
254 }
255
256 void drive_check_orphaned(void)
257 {
258 BlockBackend *blk;
259 DriveInfo *dinfo;
260 Location loc;
261 bool orphans = false;
262
263 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
264 dinfo = blk_legacy_dinfo(blk);
265 if (dinfo->is_default || dinfo->type == IF_NONE) {
266 continue;
267 }
268 if (!blk_get_attached_dev(blk)) {
269 loc_push_none(&loc);
270 qemu_opts_loc_restore(dinfo->opts);
271 error_report("machine type does not support"
272 " if=%s,bus=%d,unit=%d",
273 if_name[dinfo->type], dinfo->bus, dinfo->unit);
274 loc_pop(&loc);
275 orphans = true;
276 continue;
277 }
278 if (!dinfo->claimed_by_board && dinfo->type != IF_VIRTIO) {
279 loc_push_none(&loc);
280 qemu_opts_loc_restore(dinfo->opts);
281 warn_report("bogus if=%s is deprecated, use if=none",
282 if_name[dinfo->type]);
283 loc_pop(&loc);
284 }
285 }
286
287 if (orphans) {
288 exit(1);
289 }
290 }
291
292 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
293 {
294 return drive_get(type,
295 drive_index_to_bus_id(type, index),
296 drive_index_to_unit_id(type, index));
297 }
298
299 int drive_get_max_bus(BlockInterfaceType type)
300 {
301 int max_bus;
302 BlockBackend *blk;
303 DriveInfo *dinfo;
304
305 max_bus = -1;
306 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
307 dinfo = blk_legacy_dinfo(blk);
308 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
309 max_bus = dinfo->bus;
310 }
311 }
312 return max_bus;
313 }
314
315 /* Get a block device. This should only be used for single-drive devices
316 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
317 appropriate bus. */
318 DriveInfo *drive_get_next(BlockInterfaceType type)
319 {
320 static int next_block_unit[IF_COUNT];
321
322 return drive_get(type, 0, next_block_unit[type]++);
323 }
324
325 static void bdrv_format_print(void *opaque, const char *name)
326 {
327 qemu_printf(" %s", name);
328 }
329
330 typedef struct {
331 QEMUBH *bh;
332 BlockDriverState *bs;
333 } BDRVPutRefBH;
334
335 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
336 {
337 if (!strcmp(buf, "ignore")) {
338 return BLOCKDEV_ON_ERROR_IGNORE;
339 } else if (!is_read && !strcmp(buf, "enospc")) {
340 return BLOCKDEV_ON_ERROR_ENOSPC;
341 } else if (!strcmp(buf, "stop")) {
342 return BLOCKDEV_ON_ERROR_STOP;
343 } else if (!strcmp(buf, "report")) {
344 return BLOCKDEV_ON_ERROR_REPORT;
345 } else {
346 error_setg(errp, "'%s' invalid %s error action",
347 buf, is_read ? "read" : "write");
348 return -1;
349 }
350 }
351
352 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
353 Error **errp)
354 {
355 const QListEntry *entry;
356 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
357 switch (qobject_type(entry->value)) {
358
359 case QTYPE_QSTRING: {
360 unsigned long long length;
361 const char *str = qstring_get_str(qobject_to(QString,
362 entry->value));
363 if (parse_uint_full(str, &length, 10) == 0 &&
364 length > 0 && length <= UINT_MAX) {
365 block_acct_add_interval(stats, (unsigned) length);
366 } else {
367 error_setg(errp, "Invalid interval length: %s", str);
368 return false;
369 }
370 break;
371 }
372
373 case QTYPE_QNUM: {
374 int64_t length = qnum_get_int(qobject_to(QNum, entry->value));
375
376 if (length > 0 && length <= UINT_MAX) {
377 block_acct_add_interval(stats, (unsigned) length);
378 } else {
379 error_setg(errp, "Invalid interval length: %" PRId64, length);
380 return false;
381 }
382 break;
383 }
384
385 default:
386 error_setg(errp, "The specification of stats-intervals is invalid");
387 return false;
388 }
389 }
390 return true;
391 }
392
393 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
394
395 /* All parameters but @opts are optional and may be set to NULL. */
396 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
397 const char **throttling_group, ThrottleConfig *throttle_cfg,
398 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
399 {
400 Error *local_error = NULL;
401 const char *aio;
402
403 if (bdrv_flags) {
404 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
405 *bdrv_flags |= BDRV_O_COPY_ON_READ;
406 }
407
408 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
409 if (bdrv_parse_aio(aio, bdrv_flags) < 0) {
410 error_setg(errp, "invalid aio option");
411 return;
412 }
413 }
414 }
415
416 /* disk I/O throttling */
417 if (throttling_group) {
418 *throttling_group = qemu_opt_get(opts, "throttling.group");
419 }
420
421 if (throttle_cfg) {
422 throttle_config_init(throttle_cfg);
423 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
424 qemu_opt_get_number(opts, "throttling.bps-total", 0);
425 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
426 qemu_opt_get_number(opts, "throttling.bps-read", 0);
427 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
428 qemu_opt_get_number(opts, "throttling.bps-write", 0);
429 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
430 qemu_opt_get_number(opts, "throttling.iops-total", 0);
431 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
432 qemu_opt_get_number(opts, "throttling.iops-read", 0);
433 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
434 qemu_opt_get_number(opts, "throttling.iops-write", 0);
435
436 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
437 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
438 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
439 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
440 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
441 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
442 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
443 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
444 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
445 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
446 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
447 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
448
449 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
450 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
451 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
452 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
453 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
454 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
455 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
456 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
457 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
458 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
459 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
460 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
461
462 throttle_cfg->op_size =
463 qemu_opt_get_number(opts, "throttling.iops-size", 0);
464
465 if (!throttle_is_valid(throttle_cfg, errp)) {
466 return;
467 }
468 }
469
470 if (detect_zeroes) {
471 *detect_zeroes =
472 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
473 qemu_opt_get(opts, "detect-zeroes"),
474 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
475 &local_error);
476 if (local_error) {
477 error_propagate(errp, local_error);
478 return;
479 }
480 }
481 }
482
483 /* Takes the ownership of bs_opts */
484 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
485 Error **errp)
486 {
487 const char *buf;
488 int bdrv_flags = 0;
489 int on_read_error, on_write_error;
490 bool account_invalid, account_failed;
491 bool writethrough, read_only;
492 BlockBackend *blk;
493 BlockDriverState *bs;
494 ThrottleConfig cfg;
495 int snapshot = 0;
496 Error *error = NULL;
497 QemuOpts *opts;
498 QDict *interval_dict = NULL;
499 QList *interval_list = NULL;
500 const char *id;
501 BlockdevDetectZeroesOptions detect_zeroes =
502 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
503 const char *throttling_group = NULL;
504
505 /* Check common options by copying from bs_opts to opts, all other options
506 * stay in bs_opts for processing by bdrv_open(). */
507 id = qdict_get_try_str(bs_opts, "id");
508 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, errp);
509 if (!opts) {
510 goto err_no_opts;
511 }
512
513 if (!qemu_opts_absorb_qdict(opts, bs_opts, errp)) {
514 goto early_err;
515 }
516
517 if (id) {
518 qdict_del(bs_opts, "id");
519 }
520
521 /* extract parameters */
522 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
523
524 account_invalid = qemu_opt_get_bool(opts, "stats-account-invalid", true);
525 account_failed = qemu_opt_get_bool(opts, "stats-account-failed", true);
526
527 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
528
529 id = qemu_opts_id(opts);
530
531 qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
532 qdict_array_split(interval_dict, &interval_list);
533
534 if (qdict_size(interval_dict) != 0) {
535 error_setg(errp, "Invalid option stats-intervals.%s",
536 qdict_first(interval_dict)->key);
537 goto early_err;
538 }
539
540 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
541 &detect_zeroes, &error);
542 if (error) {
543 error_propagate(errp, error);
544 goto early_err;
545 }
546
547 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
548 if (is_help_option(buf)) {
549 qemu_printf("Supported formats:");
550 bdrv_iterate_format(bdrv_format_print, NULL, false);
551 qemu_printf("\nSupported formats (read-only):");
552 bdrv_iterate_format(bdrv_format_print, NULL, true);
553 qemu_printf("\n");
554 goto early_err;
555 }
556
557 if (qdict_haskey(bs_opts, "driver")) {
558 error_setg(errp, "Cannot specify both 'driver' and 'format'");
559 goto early_err;
560 }
561 qdict_put_str(bs_opts, "driver", buf);
562 }
563
564 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
565 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
566 on_write_error = parse_block_error_action(buf, 0, &error);
567 if (error) {
568 error_propagate(errp, error);
569 goto early_err;
570 }
571 }
572
573 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
574 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
575 on_read_error = parse_block_error_action(buf, 1, &error);
576 if (error) {
577 error_propagate(errp, error);
578 goto early_err;
579 }
580 }
581
582 if (snapshot) {
583 bdrv_flags |= BDRV_O_SNAPSHOT;
584 }
585
586 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
587
588 /* init */
589 if ((!file || !*file) && !qdict_size(bs_opts)) {
590 BlockBackendRootState *blk_rs;
591
592 blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
593 blk_rs = blk_get_root_state(blk);
594 blk_rs->open_flags = bdrv_flags;
595 blk_rs->read_only = read_only;
596 blk_rs->detect_zeroes = detect_zeroes;
597
598 qobject_unref(bs_opts);
599 } else {
600 if (file && !*file) {
601 file = NULL;
602 }
603
604 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
605 * with other callers) rather than what we want as the real defaults.
606 * Apply the defaults here instead. */
607 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
608 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
609 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
610 read_only ? "on" : "off");
611 qdict_set_default_str(bs_opts, BDRV_OPT_AUTO_READ_ONLY, "on");
612 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
613
614 if (runstate_check(RUN_STATE_INMIGRATE)) {
615 bdrv_flags |= BDRV_O_INACTIVE;
616 }
617
618 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
619 if (!blk) {
620 goto err_no_bs_opts;
621 }
622 bs = blk_bs(blk);
623
624 bs->detect_zeroes = detect_zeroes;
625
626 block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
627
628 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
629 blk_unref(blk);
630 blk = NULL;
631 goto err_no_bs_opts;
632 }
633 }
634
635 /* disk I/O throttling */
636 if (throttle_enabled(&cfg)) {
637 if (!throttling_group) {
638 throttling_group = id;
639 }
640 blk_io_limits_enable(blk, throttling_group);
641 blk_set_io_limits(blk, &cfg);
642 }
643
644 blk_set_enable_write_cache(blk, !writethrough);
645 blk_set_on_error(blk, on_read_error, on_write_error);
646
647 if (!monitor_add_blk(blk, id, errp)) {
648 blk_unref(blk);
649 blk = NULL;
650 goto err_no_bs_opts;
651 }
652
653 err_no_bs_opts:
654 qemu_opts_del(opts);
655 qobject_unref(interval_dict);
656 qobject_unref(interval_list);
657 return blk;
658
659 early_err:
660 qemu_opts_del(opts);
661 qobject_unref(interval_dict);
662 qobject_unref(interval_list);
663 err_no_opts:
664 qobject_unref(bs_opts);
665 return NULL;
666 }
667
668 /* Takes the ownership of bs_opts */
669 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
670 {
671 int bdrv_flags = 0;
672
673 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
674 * with other callers) rather than what we want as the real defaults.
675 * Apply the defaults here instead. */
676 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
677 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
678 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
679
680 if (runstate_check(RUN_STATE_INMIGRATE)) {
681 bdrv_flags |= BDRV_O_INACTIVE;
682 }
683
684 return bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
685 }
686
687 void blockdev_close_all_bdrv_states(void)
688 {
689 BlockDriverState *bs, *next_bs;
690
691 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
692 AioContext *ctx = bdrv_get_aio_context(bs);
693
694 aio_context_acquire(ctx);
695 bdrv_unref(bs);
696 aio_context_release(ctx);
697 }
698 }
699
700 /* Iterates over the list of monitor-owned BlockDriverStates */
701 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
702 {
703 return bs ? QTAILQ_NEXT(bs, monitor_list)
704 : QTAILQ_FIRST(&monitor_bdrv_states);
705 }
706
707 static bool qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
708 Error **errp)
709 {
710 const char *value;
711
712 value = qemu_opt_get(opts, from);
713 if (value) {
714 if (qemu_opt_find(opts, to)) {
715 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
716 "same time", to, from);
717 return false;
718 }
719 }
720
721 /* rename all items in opts */
722 while ((value = qemu_opt_get(opts, from))) {
723 qemu_opt_set(opts, to, value, &error_abort);
724 qemu_opt_unset(opts, from);
725 }
726 return true;
727 }
728
729 QemuOptsList qemu_legacy_drive_opts = {
730 .name = "drive",
731 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
732 .desc = {
733 {
734 .name = "bus",
735 .type = QEMU_OPT_NUMBER,
736 .help = "bus number",
737 },{
738 .name = "unit",
739 .type = QEMU_OPT_NUMBER,
740 .help = "unit number (i.e. lun for scsi)",
741 },{
742 .name = "index",
743 .type = QEMU_OPT_NUMBER,
744 .help = "index number",
745 },{
746 .name = "media",
747 .type = QEMU_OPT_STRING,
748 .help = "media type (disk, cdrom)",
749 },{
750 .name = "if",
751 .type = QEMU_OPT_STRING,
752 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
753 },{
754 .name = "file",
755 .type = QEMU_OPT_STRING,
756 .help = "file name",
757 },
758
759 /* Options that are passed on, but have special semantics with -drive */
760 {
761 .name = BDRV_OPT_READ_ONLY,
762 .type = QEMU_OPT_BOOL,
763 .help = "open drive file as read-only",
764 },{
765 .name = "rerror",
766 .type = QEMU_OPT_STRING,
767 .help = "read error action",
768 },{
769 .name = "werror",
770 .type = QEMU_OPT_STRING,
771 .help = "write error action",
772 },{
773 .name = "copy-on-read",
774 .type = QEMU_OPT_BOOL,
775 .help = "copy read data from backing file into image file",
776 },
777
778 { /* end of list */ }
779 },
780 };
781
782 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
783 Error **errp)
784 {
785 const char *value;
786 BlockBackend *blk;
787 DriveInfo *dinfo = NULL;
788 QDict *bs_opts;
789 QemuOpts *legacy_opts;
790 DriveMediaType media = MEDIA_DISK;
791 BlockInterfaceType type;
792 int max_devs, bus_id, unit_id, index;
793 const char *werror, *rerror;
794 bool read_only = false;
795 bool copy_on_read;
796 const char *filename;
797 int i;
798
799 /* Change legacy command line options into QMP ones */
800 static const struct {
801 const char *from;
802 const char *to;
803 } opt_renames[] = {
804 { "iops", "throttling.iops-total" },
805 { "iops_rd", "throttling.iops-read" },
806 { "iops_wr", "throttling.iops-write" },
807
808 { "bps", "throttling.bps-total" },
809 { "bps_rd", "throttling.bps-read" },
810 { "bps_wr", "throttling.bps-write" },
811
812 { "iops_max", "throttling.iops-total-max" },
813 { "iops_rd_max", "throttling.iops-read-max" },
814 { "iops_wr_max", "throttling.iops-write-max" },
815
816 { "bps_max", "throttling.bps-total-max" },
817 { "bps_rd_max", "throttling.bps-read-max" },
818 { "bps_wr_max", "throttling.bps-write-max" },
819
820 { "iops_size", "throttling.iops-size" },
821
822 { "group", "throttling.group" },
823
824 { "readonly", BDRV_OPT_READ_ONLY },
825 };
826
827 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
828 if (!qemu_opt_rename(all_opts, opt_renames[i].from,
829 opt_renames[i].to, errp)) {
830 return NULL;
831 }
832 }
833
834 value = qemu_opt_get(all_opts, "cache");
835 if (value) {
836 int flags = 0;
837 bool writethrough;
838
839 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
840 error_setg(errp, "invalid cache option");
841 return NULL;
842 }
843
844 /* Specific options take precedence */
845 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
846 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
847 !writethrough, &error_abort);
848 }
849 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
850 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
851 !!(flags & BDRV_O_NOCACHE), &error_abort);
852 }
853 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
854 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
855 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
856 }
857 qemu_opt_unset(all_opts, "cache");
858 }
859
860 /* Get a QDict for processing the options */
861 bs_opts = qdict_new();
862 qemu_opts_to_qdict(all_opts, bs_opts);
863
864 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
865 &error_abort);
866 if (!qemu_opts_absorb_qdict(legacy_opts, bs_opts, errp)) {
867 goto fail;
868 }
869
870 /* Media type */
871 value = qemu_opt_get(legacy_opts, "media");
872 if (value) {
873 if (!strcmp(value, "disk")) {
874 media = MEDIA_DISK;
875 } else if (!strcmp(value, "cdrom")) {
876 media = MEDIA_CDROM;
877 read_only = true;
878 } else {
879 error_setg(errp, "'%s' invalid media", value);
880 goto fail;
881 }
882 }
883
884 /* copy-on-read is disabled with a warning for read-only devices */
885 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
886 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
887
888 if (read_only && copy_on_read) {
889 warn_report("disabling copy-on-read on read-only drive");
890 copy_on_read = false;
891 }
892
893 qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
894 qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
895
896 /* Controller type */
897 value = qemu_opt_get(legacy_opts, "if");
898 if (value) {
899 for (type = 0;
900 type < IF_COUNT && strcmp(value, if_name[type]);
901 type++) {
902 }
903 if (type == IF_COUNT) {
904 error_setg(errp, "unsupported bus type '%s'", value);
905 goto fail;
906 }
907 } else {
908 type = block_default_type;
909 }
910
911 /* Device address specified by bus/unit or index.
912 * If none was specified, try to find the first free one. */
913 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
914 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
915 index = qemu_opt_get_number(legacy_opts, "index", -1);
916
917 max_devs = if_max_devs[type];
918
919 if (index != -1) {
920 if (bus_id != 0 || unit_id != -1) {
921 error_setg(errp, "index cannot be used with bus and unit");
922 goto fail;
923 }
924 bus_id = drive_index_to_bus_id(type, index);
925 unit_id = drive_index_to_unit_id(type, index);
926 }
927
928 if (unit_id == -1) {
929 unit_id = 0;
930 while (drive_get(type, bus_id, unit_id) != NULL) {
931 unit_id++;
932 if (max_devs && unit_id >= max_devs) {
933 unit_id -= max_devs;
934 bus_id++;
935 }
936 }
937 }
938
939 if (max_devs && unit_id >= max_devs) {
940 error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
941 goto fail;
942 }
943
944 if (drive_get(type, bus_id, unit_id) != NULL) {
945 error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
946 bus_id, unit_id, index);
947 goto fail;
948 }
949
950 /* no id supplied -> create one */
951 if (qemu_opts_id(all_opts) == NULL) {
952 char *new_id;
953 const char *mediastr = "";
954 if (type == IF_IDE || type == IF_SCSI) {
955 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
956 }
957 if (max_devs) {
958 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
959 mediastr, unit_id);
960 } else {
961 new_id = g_strdup_printf("%s%s%i", if_name[type],
962 mediastr, unit_id);
963 }
964 qdict_put_str(bs_opts, "id", new_id);
965 g_free(new_id);
966 }
967
968 /* Add virtio block device */
969 if (type == IF_VIRTIO) {
970 QemuOpts *devopts;
971 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
972 &error_abort);
973 if (arch_type == QEMU_ARCH_S390X) {
974 qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
975 } else {
976 qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
977 }
978 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
979 &error_abort);
980 }
981
982 filename = qemu_opt_get(legacy_opts, "file");
983
984 /* Check werror/rerror compatibility with if=... */
985 werror = qemu_opt_get(legacy_opts, "werror");
986 if (werror != NULL) {
987 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
988 type != IF_NONE) {
989 error_setg(errp, "werror is not supported by this bus type");
990 goto fail;
991 }
992 qdict_put_str(bs_opts, "werror", werror);
993 }
994
995 rerror = qemu_opt_get(legacy_opts, "rerror");
996 if (rerror != NULL) {
997 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
998 type != IF_NONE) {
999 error_setg(errp, "rerror is not supported by this bus type");
1000 goto fail;
1001 }
1002 qdict_put_str(bs_opts, "rerror", rerror);
1003 }
1004
1005 /* Actual block device init: Functionality shared with blockdev-add */
1006 blk = blockdev_init(filename, bs_opts, errp);
1007 bs_opts = NULL;
1008 if (!blk) {
1009 goto fail;
1010 }
1011
1012 /* Create legacy DriveInfo */
1013 dinfo = g_malloc0(sizeof(*dinfo));
1014 dinfo->opts = all_opts;
1015
1016 dinfo->type = type;
1017 dinfo->bus = bus_id;
1018 dinfo->unit = unit_id;
1019
1020 blk_set_legacy_dinfo(blk, dinfo);
1021
1022 switch(type) {
1023 case IF_IDE:
1024 case IF_SCSI:
1025 case IF_XEN:
1026 case IF_NONE:
1027 dinfo->media_cd = media == MEDIA_CDROM;
1028 break;
1029 default:
1030 break;
1031 }
1032
1033 fail:
1034 qemu_opts_del(legacy_opts);
1035 qobject_unref(bs_opts);
1036 return dinfo;
1037 }
1038
1039 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1040 {
1041 BlockDriverState *bs;
1042
1043 bs = bdrv_lookup_bs(name, name, errp);
1044 if (bs == NULL) {
1045 return NULL;
1046 }
1047
1048 if (!bdrv_is_root_node(bs)) {
1049 error_setg(errp, "Need a root block node");
1050 return NULL;
1051 }
1052
1053 if (!bdrv_is_inserted(bs)) {
1054 error_setg(errp, "Device has no medium");
1055 return NULL;
1056 }
1057
1058 return bs;
1059 }
1060
1061 static void blockdev_do_action(TransactionAction *action, Error **errp)
1062 {
1063 TransactionActionList list;
1064
1065 list.value = action;
1066 list.next = NULL;
1067 qmp_transaction(&list, false, NULL, errp);
1068 }
1069
1070 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1071 bool has_node_name, const char *node_name,
1072 const char *snapshot_file,
1073 bool has_snapshot_node_name,
1074 const char *snapshot_node_name,
1075 bool has_format, const char *format,
1076 bool has_mode, NewImageMode mode, Error **errp)
1077 {
1078 BlockdevSnapshotSync snapshot = {
1079 .has_device = has_device,
1080 .device = (char *) device,
1081 .has_node_name = has_node_name,
1082 .node_name = (char *) node_name,
1083 .snapshot_file = (char *) snapshot_file,
1084 .has_snapshot_node_name = has_snapshot_node_name,
1085 .snapshot_node_name = (char *) snapshot_node_name,
1086 .has_format = has_format,
1087 .format = (char *) format,
1088 .has_mode = has_mode,
1089 .mode = mode,
1090 };
1091 TransactionAction action = {
1092 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1093 .u.blockdev_snapshot_sync.data = &snapshot,
1094 };
1095 blockdev_do_action(&action, errp);
1096 }
1097
1098 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1099 Error **errp)
1100 {
1101 BlockdevSnapshot snapshot_data = {
1102 .node = (char *) node,
1103 .overlay = (char *) overlay
1104 };
1105 TransactionAction action = {
1106 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1107 .u.blockdev_snapshot.data = &snapshot_data,
1108 };
1109 blockdev_do_action(&action, errp);
1110 }
1111
1112 void qmp_blockdev_snapshot_internal_sync(const char *device,
1113 const char *name,
1114 Error **errp)
1115 {
1116 BlockdevSnapshotInternal snapshot = {
1117 .device = (char *) device,
1118 .name = (char *) name
1119 };
1120 TransactionAction action = {
1121 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1122 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1123 };
1124 blockdev_do_action(&action, errp);
1125 }
1126
1127 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1128 bool has_id,
1129 const char *id,
1130 bool has_name,
1131 const char *name,
1132 Error **errp)
1133 {
1134 BlockDriverState *bs;
1135 AioContext *aio_context;
1136 QEMUSnapshotInfo sn;
1137 Error *local_err = NULL;
1138 SnapshotInfo *info = NULL;
1139 int ret;
1140
1141 bs = qmp_get_root_bs(device, errp);
1142 if (!bs) {
1143 return NULL;
1144 }
1145 aio_context = bdrv_get_aio_context(bs);
1146 aio_context_acquire(aio_context);
1147
1148 if (!has_id) {
1149 id = NULL;
1150 }
1151
1152 if (!has_name) {
1153 name = NULL;
1154 }
1155
1156 if (!id && !name) {
1157 error_setg(errp, "Name or id must be provided");
1158 goto out_aio_context;
1159 }
1160
1161 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1162 goto out_aio_context;
1163 }
1164
1165 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1166 if (local_err) {
1167 error_propagate(errp, local_err);
1168 goto out_aio_context;
1169 }
1170 if (!ret) {
1171 error_setg(errp,
1172 "Snapshot with id '%s' and name '%s' does not exist on "
1173 "device '%s'",
1174 STR_OR_NULL(id), STR_OR_NULL(name), device);
1175 goto out_aio_context;
1176 }
1177
1178 bdrv_snapshot_delete(bs, id, name, &local_err);
1179 if (local_err) {
1180 error_propagate(errp, local_err);
1181 goto out_aio_context;
1182 }
1183
1184 aio_context_release(aio_context);
1185
1186 info = g_new0(SnapshotInfo, 1);
1187 info->id = g_strdup(sn.id_str);
1188 info->name = g_strdup(sn.name);
1189 info->date_nsec = sn.date_nsec;
1190 info->date_sec = sn.date_sec;
1191 info->vm_state_size = sn.vm_state_size;
1192 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1193 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1194 if (sn.icount != -1ULL) {
1195 info->icount = sn.icount;
1196 info->has_icount = true;
1197 }
1198
1199 return info;
1200
1201 out_aio_context:
1202 aio_context_release(aio_context);
1203 return NULL;
1204 }
1205
1206 /* New and old BlockDriverState structs for atomic group operations */
1207
1208 typedef struct BlkActionState BlkActionState;
1209
1210 /**
1211 * BlkActionOps:
1212 * Table of operations that define an Action.
1213 *
1214 * @instance_size: Size of state struct, in bytes.
1215 * @prepare: Prepare the work, must NOT be NULL.
1216 * @commit: Commit the changes, can be NULL.
1217 * @abort: Abort the changes on fail, can be NULL.
1218 * @clean: Clean up resources after all transaction actions have called
1219 * commit() or abort(). Can be NULL.
1220 *
1221 * Only prepare() may fail. In a single transaction, only one of commit() or
1222 * abort() will be called. clean() will always be called if it is present.
1223 */
1224 typedef struct BlkActionOps {
1225 size_t instance_size;
1226 void (*prepare)(BlkActionState *common, Error **errp);
1227 void (*commit)(BlkActionState *common);
1228 void (*abort)(BlkActionState *common);
1229 void (*clean)(BlkActionState *common);
1230 } BlkActionOps;
1231
1232 /**
1233 * BlkActionState:
1234 * Describes one Action's state within a Transaction.
1235 *
1236 * @action: QAPI-defined enum identifying which Action to perform.
1237 * @ops: Table of ActionOps this Action can perform.
1238 * @block_job_txn: Transaction which this action belongs to.
1239 * @entry: List membership for all Actions in this Transaction.
1240 *
1241 * This structure must be arranged as first member in a subclassed type,
1242 * assuming that the compiler will also arrange it to the same offsets as the
1243 * base class.
1244 */
1245 struct BlkActionState {
1246 TransactionAction *action;
1247 const BlkActionOps *ops;
1248 JobTxn *block_job_txn;
1249 TransactionProperties *txn_props;
1250 QTAILQ_ENTRY(BlkActionState) entry;
1251 };
1252
1253 /* internal snapshot private data */
1254 typedef struct InternalSnapshotState {
1255 BlkActionState common;
1256 BlockDriverState *bs;
1257 QEMUSnapshotInfo sn;
1258 bool created;
1259 } InternalSnapshotState;
1260
1261
1262 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1263 {
1264 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1265 error_setg(errp,
1266 "Action '%s' does not support Transaction property "
1267 "completion-mode = %s",
1268 TransactionActionKind_str(s->action->type),
1269 ActionCompletionMode_str(s->txn_props->completion_mode));
1270 return -1;
1271 }
1272 return 0;
1273 }
1274
1275 static void internal_snapshot_prepare(BlkActionState *common,
1276 Error **errp)
1277 {
1278 Error *local_err = NULL;
1279 const char *device;
1280 const char *name;
1281 BlockDriverState *bs;
1282 QEMUSnapshotInfo old_sn, *sn;
1283 bool ret;
1284 qemu_timeval tv;
1285 BlockdevSnapshotInternal *internal;
1286 InternalSnapshotState *state;
1287 AioContext *aio_context;
1288 int ret1;
1289
1290 g_assert(common->action->type ==
1291 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1292 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1293 state = DO_UPCAST(InternalSnapshotState, common, common);
1294
1295 /* 1. parse input */
1296 device = internal->device;
1297 name = internal->name;
1298
1299 /* 2. check for validation */
1300 if (action_check_completion_mode(common, errp) < 0) {
1301 return;
1302 }
1303
1304 bs = qmp_get_root_bs(device, errp);
1305 if (!bs) {
1306 return;
1307 }
1308
1309 aio_context = bdrv_get_aio_context(bs);
1310 aio_context_acquire(aio_context);
1311
1312 state->bs = bs;
1313
1314 /* Paired with .clean() */
1315 bdrv_drained_begin(bs);
1316
1317 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1318 goto out;
1319 }
1320
1321 if (bdrv_is_read_only(bs)) {
1322 error_setg(errp, "Device '%s' is read only", device);
1323 goto out;
1324 }
1325
1326 if (!bdrv_can_snapshot(bs)) {
1327 error_setg(errp, "Block format '%s' used by device '%s' "
1328 "does not support internal snapshots",
1329 bs->drv->format_name, device);
1330 goto out;
1331 }
1332
1333 if (!strlen(name)) {
1334 error_setg(errp, "Name is empty");
1335 goto out;
1336 }
1337
1338 /* check whether a snapshot with name exist */
1339 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1340 &local_err);
1341 if (local_err) {
1342 error_propagate(errp, local_err);
1343 goto out;
1344 } else if (ret) {
1345 error_setg(errp,
1346 "Snapshot with name '%s' already exists on device '%s'",
1347 name, device);
1348 goto out;
1349 }
1350
1351 /* 3. take the snapshot */
1352 sn = &state->sn;
1353 pstrcpy(sn->name, sizeof(sn->name), name);
1354 qemu_gettimeofday(&tv);
1355 sn->date_sec = tv.tv_sec;
1356 sn->date_nsec = tv.tv_usec * 1000;
1357 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1358 if (replay_mode != REPLAY_MODE_NONE) {
1359 sn->icount = replay_get_current_icount();
1360 } else {
1361 sn->icount = -1ULL;
1362 }
1363
1364 ret1 = bdrv_snapshot_create(bs, sn);
1365 if (ret1 < 0) {
1366 error_setg_errno(errp, -ret1,
1367 "Failed to create snapshot '%s' on device '%s'",
1368 name, device);
1369 goto out;
1370 }
1371
1372 /* 4. succeed, mark a snapshot is created */
1373 state->created = true;
1374
1375 out:
1376 aio_context_release(aio_context);
1377 }
1378
1379 static void internal_snapshot_abort(BlkActionState *common)
1380 {
1381 InternalSnapshotState *state =
1382 DO_UPCAST(InternalSnapshotState, common, common);
1383 BlockDriverState *bs = state->bs;
1384 QEMUSnapshotInfo *sn = &state->sn;
1385 AioContext *aio_context;
1386 Error *local_error = NULL;
1387
1388 if (!state->created) {
1389 return;
1390 }
1391
1392 aio_context = bdrv_get_aio_context(state->bs);
1393 aio_context_acquire(aio_context);
1394
1395 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1396 error_reportf_err(local_error,
1397 "Failed to delete snapshot with id '%s' and "
1398 "name '%s' on device '%s' in abort: ",
1399 sn->id_str, sn->name,
1400 bdrv_get_device_name(bs));
1401 }
1402
1403 aio_context_release(aio_context);
1404 }
1405
1406 static void internal_snapshot_clean(BlkActionState *common)
1407 {
1408 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1409 common, common);
1410 AioContext *aio_context;
1411
1412 if (!state->bs) {
1413 return;
1414 }
1415
1416 aio_context = bdrv_get_aio_context(state->bs);
1417 aio_context_acquire(aio_context);
1418
1419 bdrv_drained_end(state->bs);
1420
1421 aio_context_release(aio_context);
1422 }
1423
1424 /* external snapshot private data */
1425 typedef struct ExternalSnapshotState {
1426 BlkActionState common;
1427 BlockDriverState *old_bs;
1428 BlockDriverState *new_bs;
1429 bool overlay_appended;
1430 } ExternalSnapshotState;
1431
1432 static void external_snapshot_prepare(BlkActionState *common,
1433 Error **errp)
1434 {
1435 int ret;
1436 int flags = 0;
1437 QDict *options = NULL;
1438 Error *local_err = NULL;
1439 /* Device and node name of the image to generate the snapshot from */
1440 const char *device;
1441 const char *node_name;
1442 /* Reference to the new image (for 'blockdev-snapshot') */
1443 const char *snapshot_ref;
1444 /* File name of the new image (for 'blockdev-snapshot-sync') */
1445 const char *new_image_file;
1446 ExternalSnapshotState *state =
1447 DO_UPCAST(ExternalSnapshotState, common, common);
1448 TransactionAction *action = common->action;
1449 AioContext *aio_context;
1450 uint64_t perm, shared;
1451
1452 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1453 * purpose but a different set of parameters */
1454 switch (action->type) {
1455 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1456 {
1457 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1458 device = s->node;
1459 node_name = s->node;
1460 new_image_file = NULL;
1461 snapshot_ref = s->overlay;
1462 }
1463 break;
1464 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1465 {
1466 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1467 device = s->has_device ? s->device : NULL;
1468 node_name = s->has_node_name ? s->node_name : NULL;
1469 new_image_file = s->snapshot_file;
1470 snapshot_ref = NULL;
1471 }
1472 break;
1473 default:
1474 g_assert_not_reached();
1475 }
1476
1477 /* start processing */
1478 if (action_check_completion_mode(common, errp) < 0) {
1479 return;
1480 }
1481
1482 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1483 if (!state->old_bs) {
1484 return;
1485 }
1486
1487 aio_context = bdrv_get_aio_context(state->old_bs);
1488 aio_context_acquire(aio_context);
1489
1490 /* Paired with .clean() */
1491 bdrv_drained_begin(state->old_bs);
1492
1493 if (!bdrv_is_inserted(state->old_bs)) {
1494 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1495 goto out;
1496 }
1497
1498 if (bdrv_op_is_blocked(state->old_bs,
1499 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1500 goto out;
1501 }
1502
1503 if (!bdrv_is_read_only(state->old_bs)) {
1504 if (bdrv_flush(state->old_bs)) {
1505 error_setg(errp, QERR_IO_ERROR);
1506 goto out;
1507 }
1508 }
1509
1510 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1511 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1512 const char *format = s->has_format ? s->format : "qcow2";
1513 enum NewImageMode mode;
1514 const char *snapshot_node_name =
1515 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1516
1517 if (node_name && !snapshot_node_name) {
1518 error_setg(errp, "New overlay node name missing");
1519 goto out;
1520 }
1521
1522 if (snapshot_node_name &&
1523 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1524 error_setg(errp, "New overlay node name already in use");
1525 goto out;
1526 }
1527
1528 flags = state->old_bs->open_flags;
1529 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1530 flags |= BDRV_O_NO_BACKING;
1531
1532 /* create new image w/backing file */
1533 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1534 if (mode != NEW_IMAGE_MODE_EXISTING) {
1535 int64_t size = bdrv_getlength(state->old_bs);
1536 if (size < 0) {
1537 error_setg_errno(errp, -size, "bdrv_getlength failed");
1538 goto out;
1539 }
1540 bdrv_refresh_filename(state->old_bs);
1541 bdrv_img_create(new_image_file, format,
1542 state->old_bs->filename,
1543 state->old_bs->drv->format_name,
1544 NULL, size, flags, false, &local_err);
1545 if (local_err) {
1546 error_propagate(errp, local_err);
1547 goto out;
1548 }
1549 }
1550
1551 options = qdict_new();
1552 if (snapshot_node_name) {
1553 qdict_put_str(options, "node-name", snapshot_node_name);
1554 }
1555 qdict_put_str(options, "driver", format);
1556 }
1557
1558 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1559 errp);
1560 /* We will manually add the backing_hd field to the bs later */
1561 if (!state->new_bs) {
1562 goto out;
1563 }
1564
1565 /*
1566 * Allow attaching a backing file to an overlay that's already in use only
1567 * if the parents don't assume that they are already seeing a valid image.
1568 * (Specifically, allow it as a mirror target, which is write-only access.)
1569 */
1570 bdrv_get_cumulative_perm(state->new_bs, &perm, &shared);
1571 if (perm & BLK_PERM_CONSISTENT_READ) {
1572 error_setg(errp, "The overlay is already in use");
1573 goto out;
1574 }
1575
1576 if (state->new_bs->drv->is_filter) {
1577 error_setg(errp, "Filters cannot be used as overlays");
1578 goto out;
1579 }
1580
1581 if (bdrv_cow_child(state->new_bs)) {
1582 error_setg(errp, "The overlay already has a backing image");
1583 goto out;
1584 }
1585
1586 if (!state->new_bs->drv->supports_backing) {
1587 error_setg(errp, "The overlay does not support backing images");
1588 goto out;
1589 }
1590
1591 /* This removes our old bs and adds the new bs. This is an operation that
1592 * can fail, so we need to do it in .prepare; undoing it for abort is
1593 * always possible. */
1594 bdrv_ref(state->new_bs);
1595 ret = bdrv_append(state->new_bs, state->old_bs, errp);
1596 if (ret < 0) {
1597 goto out;
1598 }
1599 state->overlay_appended = true;
1600
1601 out:
1602 aio_context_release(aio_context);
1603 }
1604
1605 static void external_snapshot_commit(BlkActionState *common)
1606 {
1607 ExternalSnapshotState *state =
1608 DO_UPCAST(ExternalSnapshotState, common, common);
1609 AioContext *aio_context;
1610
1611 aio_context = bdrv_get_aio_context(state->old_bs);
1612 aio_context_acquire(aio_context);
1613
1614 /* We don't need (or want) to use the transactional
1615 * bdrv_reopen_multiple() across all the entries at once, because we
1616 * don't want to abort all of them if one of them fails the reopen */
1617 if (!qatomic_read(&state->old_bs->copy_on_read)) {
1618 bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1619 }
1620
1621 aio_context_release(aio_context);
1622 }
1623
1624 static void external_snapshot_abort(BlkActionState *common)
1625 {
1626 ExternalSnapshotState *state =
1627 DO_UPCAST(ExternalSnapshotState, common, common);
1628 if (state->new_bs) {
1629 if (state->overlay_appended) {
1630 AioContext *aio_context;
1631 AioContext *tmp_context;
1632 int ret;
1633
1634 aio_context = bdrv_get_aio_context(state->old_bs);
1635 aio_context_acquire(aio_context);
1636
1637 bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd()
1638 close state->old_bs; we need it */
1639 bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1640
1641 /*
1642 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1643 * the main AioContext. As we're still going to be using it, return
1644 * it to the AioContext it was before.
1645 */
1646 tmp_context = bdrv_get_aio_context(state->old_bs);
1647 if (aio_context != tmp_context) {
1648 aio_context_release(aio_context);
1649 aio_context_acquire(tmp_context);
1650
1651 ret = bdrv_try_set_aio_context(state->old_bs,
1652 aio_context, NULL);
1653 assert(ret == 0);
1654
1655 aio_context_release(tmp_context);
1656 aio_context_acquire(aio_context);
1657 }
1658
1659 bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1660 bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1661
1662 aio_context_release(aio_context);
1663 }
1664 }
1665 }
1666
1667 static void external_snapshot_clean(BlkActionState *common)
1668 {
1669 ExternalSnapshotState *state =
1670 DO_UPCAST(ExternalSnapshotState, common, common);
1671 AioContext *aio_context;
1672
1673 if (!state->old_bs) {
1674 return;
1675 }
1676
1677 aio_context = bdrv_get_aio_context(state->old_bs);
1678 aio_context_acquire(aio_context);
1679
1680 bdrv_drained_end(state->old_bs);
1681 bdrv_unref(state->new_bs);
1682
1683 aio_context_release(aio_context);
1684 }
1685
1686 typedef struct DriveBackupState {
1687 BlkActionState common;
1688 BlockDriverState *bs;
1689 BlockJob *job;
1690 } DriveBackupState;
1691
1692 static BlockJob *do_backup_common(BackupCommon *backup,
1693 BlockDriverState *bs,
1694 BlockDriverState *target_bs,
1695 AioContext *aio_context,
1696 JobTxn *txn, Error **errp);
1697
1698 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1699 {
1700 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1701 DriveBackup *backup;
1702 BlockDriverState *bs;
1703 BlockDriverState *target_bs;
1704 BlockDriverState *source = NULL;
1705 AioContext *aio_context;
1706 AioContext *old_context;
1707 QDict *options;
1708 Error *local_err = NULL;
1709 int flags;
1710 int64_t size;
1711 bool set_backing_hd = false;
1712 int ret;
1713
1714 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1715 backup = common->action->u.drive_backup.data;
1716
1717 if (!backup->has_mode) {
1718 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1719 }
1720
1721 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1722 if (!bs) {
1723 return;
1724 }
1725
1726 if (!bs->drv) {
1727 error_setg(errp, "Device has no medium");
1728 return;
1729 }
1730
1731 aio_context = bdrv_get_aio_context(bs);
1732 aio_context_acquire(aio_context);
1733
1734 /* Paired with .clean() */
1735 bdrv_drained_begin(bs);
1736
1737 if (!backup->has_format) {
1738 backup->format = backup->mode == NEW_IMAGE_MODE_EXISTING ?
1739 NULL : (char *) bs->drv->format_name;
1740 }
1741
1742 /* Early check to avoid creating target */
1743 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
1744 goto out;
1745 }
1746
1747 flags = bs->open_flags | BDRV_O_RDWR;
1748
1749 /*
1750 * See if we have a backing HD we can use to create our new image
1751 * on top of.
1752 */
1753 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
1754 /*
1755 * Backup will not replace the source by the target, so none
1756 * of the filters skipped here will be removed (in contrast to
1757 * mirror). Therefore, we can skip all of them when looking
1758 * for the first COW relationship.
1759 */
1760 source = bdrv_cow_bs(bdrv_skip_filters(bs));
1761 if (!source) {
1762 backup->sync = MIRROR_SYNC_MODE_FULL;
1763 }
1764 }
1765 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
1766 source = bs;
1767 flags |= BDRV_O_NO_BACKING;
1768 set_backing_hd = true;
1769 }
1770
1771 size = bdrv_getlength(bs);
1772 if (size < 0) {
1773 error_setg_errno(errp, -size, "bdrv_getlength failed");
1774 goto out;
1775 }
1776
1777 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
1778 assert(backup->format);
1779 if (source) {
1780 /* Implicit filters should not appear in the filename */
1781 BlockDriverState *explicit_backing =
1782 bdrv_skip_implicit_filters(source);
1783
1784 bdrv_refresh_filename(explicit_backing);
1785 bdrv_img_create(backup->target, backup->format,
1786 explicit_backing->filename,
1787 explicit_backing->drv->format_name, NULL,
1788 size, flags, false, &local_err);
1789 } else {
1790 bdrv_img_create(backup->target, backup->format, NULL, NULL, NULL,
1791 size, flags, false, &local_err);
1792 }
1793 }
1794
1795 if (local_err) {
1796 error_propagate(errp, local_err);
1797 goto out;
1798 }
1799
1800 options = qdict_new();
1801 qdict_put_str(options, "discard", "unmap");
1802 qdict_put_str(options, "detect-zeroes", "unmap");
1803 if (backup->format) {
1804 qdict_put_str(options, "driver", backup->format);
1805 }
1806
1807 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
1808 if (!target_bs) {
1809 goto out;
1810 }
1811
1812 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1813 old_context = bdrv_get_aio_context(target_bs);
1814 aio_context_release(aio_context);
1815 aio_context_acquire(old_context);
1816
1817 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1818 if (ret < 0) {
1819 bdrv_unref(target_bs);
1820 aio_context_release(old_context);
1821 return;
1822 }
1823
1824 aio_context_release(old_context);
1825 aio_context_acquire(aio_context);
1826
1827 if (set_backing_hd) {
1828 if (bdrv_set_backing_hd(target_bs, source, errp) < 0) {
1829 goto unref;
1830 }
1831 }
1832
1833 state->bs = bs;
1834
1835 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1836 bs, target_bs, aio_context,
1837 common->block_job_txn, errp);
1838
1839 unref:
1840 bdrv_unref(target_bs);
1841 out:
1842 aio_context_release(aio_context);
1843 }
1844
1845 static void drive_backup_commit(BlkActionState *common)
1846 {
1847 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1848 AioContext *aio_context;
1849
1850 aio_context = bdrv_get_aio_context(state->bs);
1851 aio_context_acquire(aio_context);
1852
1853 assert(state->job);
1854 job_start(&state->job->job);
1855
1856 aio_context_release(aio_context);
1857 }
1858
1859 static void drive_backup_abort(BlkActionState *common)
1860 {
1861 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1862
1863 if (state->job) {
1864 AioContext *aio_context;
1865
1866 aio_context = bdrv_get_aio_context(state->bs);
1867 aio_context_acquire(aio_context);
1868
1869 job_cancel_sync(&state->job->job);
1870
1871 aio_context_release(aio_context);
1872 }
1873 }
1874
1875 static void drive_backup_clean(BlkActionState *common)
1876 {
1877 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1878 AioContext *aio_context;
1879
1880 if (!state->bs) {
1881 return;
1882 }
1883
1884 aio_context = bdrv_get_aio_context(state->bs);
1885 aio_context_acquire(aio_context);
1886
1887 bdrv_drained_end(state->bs);
1888
1889 aio_context_release(aio_context);
1890 }
1891
1892 typedef struct BlockdevBackupState {
1893 BlkActionState common;
1894 BlockDriverState *bs;
1895 BlockJob *job;
1896 } BlockdevBackupState;
1897
1898 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1899 {
1900 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1901 BlockdevBackup *backup;
1902 BlockDriverState *bs;
1903 BlockDriverState *target_bs;
1904 AioContext *aio_context;
1905 AioContext *old_context;
1906 int ret;
1907
1908 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1909 backup = common->action->u.blockdev_backup.data;
1910
1911 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1912 if (!bs) {
1913 return;
1914 }
1915
1916 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1917 if (!target_bs) {
1918 return;
1919 }
1920
1921 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1922 aio_context = bdrv_get_aio_context(bs);
1923 old_context = bdrv_get_aio_context(target_bs);
1924 aio_context_acquire(old_context);
1925
1926 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1927 if (ret < 0) {
1928 aio_context_release(old_context);
1929 return;
1930 }
1931
1932 aio_context_release(old_context);
1933 aio_context_acquire(aio_context);
1934 state->bs = bs;
1935
1936 /* Paired with .clean() */
1937 bdrv_drained_begin(state->bs);
1938
1939 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1940 bs, target_bs, aio_context,
1941 common->block_job_txn, errp);
1942
1943 aio_context_release(aio_context);
1944 }
1945
1946 static void blockdev_backup_commit(BlkActionState *common)
1947 {
1948 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1949 AioContext *aio_context;
1950
1951 aio_context = bdrv_get_aio_context(state->bs);
1952 aio_context_acquire(aio_context);
1953
1954 assert(state->job);
1955 job_start(&state->job->job);
1956
1957 aio_context_release(aio_context);
1958 }
1959
1960 static void blockdev_backup_abort(BlkActionState *common)
1961 {
1962 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1963
1964 if (state->job) {
1965 AioContext *aio_context;
1966
1967 aio_context = bdrv_get_aio_context(state->bs);
1968 aio_context_acquire(aio_context);
1969
1970 job_cancel_sync(&state->job->job);
1971
1972 aio_context_release(aio_context);
1973 }
1974 }
1975
1976 static void blockdev_backup_clean(BlkActionState *common)
1977 {
1978 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1979 AioContext *aio_context;
1980
1981 if (!state->bs) {
1982 return;
1983 }
1984
1985 aio_context = bdrv_get_aio_context(state->bs);
1986 aio_context_acquire(aio_context);
1987
1988 bdrv_drained_end(state->bs);
1989
1990 aio_context_release(aio_context);
1991 }
1992
1993 typedef struct BlockDirtyBitmapState {
1994 BlkActionState common;
1995 BdrvDirtyBitmap *bitmap;
1996 BlockDriverState *bs;
1997 HBitmap *backup;
1998 bool prepared;
1999 bool was_enabled;
2000 } BlockDirtyBitmapState;
2001
2002 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
2003 Error **errp)
2004 {
2005 Error *local_err = NULL;
2006 BlockDirtyBitmapAdd *action;
2007 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2008 common, common);
2009
2010 if (action_check_completion_mode(common, errp) < 0) {
2011 return;
2012 }
2013
2014 action = common->action->u.block_dirty_bitmap_add.data;
2015 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2016 qmp_block_dirty_bitmap_add(action->node, action->name,
2017 action->has_granularity, action->granularity,
2018 action->has_persistent, action->persistent,
2019 action->has_disabled, action->disabled,
2020 &local_err);
2021
2022 if (!local_err) {
2023 state->prepared = true;
2024 } else {
2025 error_propagate(errp, local_err);
2026 }
2027 }
2028
2029 static void block_dirty_bitmap_add_abort(BlkActionState *common)
2030 {
2031 BlockDirtyBitmapAdd *action;
2032 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2033 common, common);
2034
2035 action = common->action->u.block_dirty_bitmap_add.data;
2036 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2037 * then the node reference and bitmap name must have been valid.
2038 */
2039 if (state->prepared) {
2040 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2041 }
2042 }
2043
2044 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2045 Error **errp)
2046 {
2047 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2048 common, common);
2049 BlockDirtyBitmap *action;
2050
2051 if (action_check_completion_mode(common, errp) < 0) {
2052 return;
2053 }
2054
2055 action = common->action->u.block_dirty_bitmap_clear.data;
2056 state->bitmap = block_dirty_bitmap_lookup(action->node,
2057 action->name,
2058 &state->bs,
2059 errp);
2060 if (!state->bitmap) {
2061 return;
2062 }
2063
2064 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2065 return;
2066 }
2067
2068 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2069 }
2070
2071 static void block_dirty_bitmap_restore(BlkActionState *common)
2072 {
2073 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2074 common, common);
2075
2076 if (state->backup) {
2077 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2078 }
2079 }
2080
2081 static void block_dirty_bitmap_free_backup(BlkActionState *common)
2082 {
2083 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2084 common, common);
2085
2086 hbitmap_free(state->backup);
2087 }
2088
2089 static void block_dirty_bitmap_enable_prepare(BlkActionState *common,
2090 Error **errp)
2091 {
2092 BlockDirtyBitmap *action;
2093 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2094 common, common);
2095
2096 if (action_check_completion_mode(common, errp) < 0) {
2097 return;
2098 }
2099
2100 action = common->action->u.block_dirty_bitmap_enable.data;
2101 state->bitmap = block_dirty_bitmap_lookup(action->node,
2102 action->name,
2103 NULL,
2104 errp);
2105 if (!state->bitmap) {
2106 return;
2107 }
2108
2109 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2110 return;
2111 }
2112
2113 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2114 bdrv_enable_dirty_bitmap(state->bitmap);
2115 }
2116
2117 static void block_dirty_bitmap_enable_abort(BlkActionState *common)
2118 {
2119 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2120 common, common);
2121
2122 if (!state->was_enabled) {
2123 bdrv_disable_dirty_bitmap(state->bitmap);
2124 }
2125 }
2126
2127 static void block_dirty_bitmap_disable_prepare(BlkActionState *common,
2128 Error **errp)
2129 {
2130 BlockDirtyBitmap *action;
2131 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2132 common, common);
2133
2134 if (action_check_completion_mode(common, errp) < 0) {
2135 return;
2136 }
2137
2138 action = common->action->u.block_dirty_bitmap_disable.data;
2139 state->bitmap = block_dirty_bitmap_lookup(action->node,
2140 action->name,
2141 NULL,
2142 errp);
2143 if (!state->bitmap) {
2144 return;
2145 }
2146
2147 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2148 return;
2149 }
2150
2151 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2152 bdrv_disable_dirty_bitmap(state->bitmap);
2153 }
2154
2155 static void block_dirty_bitmap_disable_abort(BlkActionState *common)
2156 {
2157 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2158 common, common);
2159
2160 if (state->was_enabled) {
2161 bdrv_enable_dirty_bitmap(state->bitmap);
2162 }
2163 }
2164
2165 static void block_dirty_bitmap_merge_prepare(BlkActionState *common,
2166 Error **errp)
2167 {
2168 BlockDirtyBitmapMerge *action;
2169 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2170 common, common);
2171
2172 if (action_check_completion_mode(common, errp) < 0) {
2173 return;
2174 }
2175
2176 action = common->action->u.block_dirty_bitmap_merge.data;
2177
2178 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2179 action->bitmaps, &state->backup,
2180 errp);
2181 }
2182
2183 static void block_dirty_bitmap_remove_prepare(BlkActionState *common,
2184 Error **errp)
2185 {
2186 BlockDirtyBitmap *action;
2187 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2188 common, common);
2189
2190 if (action_check_completion_mode(common, errp) < 0) {
2191 return;
2192 }
2193
2194 action = common->action->u.block_dirty_bitmap_remove.data;
2195
2196 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2197 false, &state->bs, errp);
2198 if (state->bitmap) {
2199 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2200 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2201 }
2202 }
2203
2204 static void block_dirty_bitmap_remove_abort(BlkActionState *common)
2205 {
2206 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2207 common, common);
2208
2209 if (state->bitmap) {
2210 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2211 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2212 }
2213 }
2214
2215 static void block_dirty_bitmap_remove_commit(BlkActionState *common)
2216 {
2217 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2218 common, common);
2219
2220 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2221 bdrv_release_dirty_bitmap(state->bitmap);
2222 }
2223
2224 static void abort_prepare(BlkActionState *common, Error **errp)
2225 {
2226 error_setg(errp, "Transaction aborted using Abort action");
2227 }
2228
2229 static void abort_commit(BlkActionState *common)
2230 {
2231 g_assert_not_reached(); /* this action never succeeds */
2232 }
2233
2234 static const BlkActionOps actions[] = {
2235 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2236 .instance_size = sizeof(ExternalSnapshotState),
2237 .prepare = external_snapshot_prepare,
2238 .commit = external_snapshot_commit,
2239 .abort = external_snapshot_abort,
2240 .clean = external_snapshot_clean,
2241 },
2242 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2243 .instance_size = sizeof(ExternalSnapshotState),
2244 .prepare = external_snapshot_prepare,
2245 .commit = external_snapshot_commit,
2246 .abort = external_snapshot_abort,
2247 .clean = external_snapshot_clean,
2248 },
2249 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2250 .instance_size = sizeof(DriveBackupState),
2251 .prepare = drive_backup_prepare,
2252 .commit = drive_backup_commit,
2253 .abort = drive_backup_abort,
2254 .clean = drive_backup_clean,
2255 },
2256 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2257 .instance_size = sizeof(BlockdevBackupState),
2258 .prepare = blockdev_backup_prepare,
2259 .commit = blockdev_backup_commit,
2260 .abort = blockdev_backup_abort,
2261 .clean = blockdev_backup_clean,
2262 },
2263 [TRANSACTION_ACTION_KIND_ABORT] = {
2264 .instance_size = sizeof(BlkActionState),
2265 .prepare = abort_prepare,
2266 .commit = abort_commit,
2267 },
2268 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2269 .instance_size = sizeof(InternalSnapshotState),
2270 .prepare = internal_snapshot_prepare,
2271 .abort = internal_snapshot_abort,
2272 .clean = internal_snapshot_clean,
2273 },
2274 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2275 .instance_size = sizeof(BlockDirtyBitmapState),
2276 .prepare = block_dirty_bitmap_add_prepare,
2277 .abort = block_dirty_bitmap_add_abort,
2278 },
2279 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2280 .instance_size = sizeof(BlockDirtyBitmapState),
2281 .prepare = block_dirty_bitmap_clear_prepare,
2282 .commit = block_dirty_bitmap_free_backup,
2283 .abort = block_dirty_bitmap_restore,
2284 },
2285 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2286 .instance_size = sizeof(BlockDirtyBitmapState),
2287 .prepare = block_dirty_bitmap_enable_prepare,
2288 .abort = block_dirty_bitmap_enable_abort,
2289 },
2290 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2291 .instance_size = sizeof(BlockDirtyBitmapState),
2292 .prepare = block_dirty_bitmap_disable_prepare,
2293 .abort = block_dirty_bitmap_disable_abort,
2294 },
2295 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2296 .instance_size = sizeof(BlockDirtyBitmapState),
2297 .prepare = block_dirty_bitmap_merge_prepare,
2298 .commit = block_dirty_bitmap_free_backup,
2299 .abort = block_dirty_bitmap_restore,
2300 },
2301 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE] = {
2302 .instance_size = sizeof(BlockDirtyBitmapState),
2303 .prepare = block_dirty_bitmap_remove_prepare,
2304 .commit = block_dirty_bitmap_remove_commit,
2305 .abort = block_dirty_bitmap_remove_abort,
2306 },
2307 /* Where are transactions for MIRROR, COMMIT and STREAM?
2308 * Although these blockjobs use transaction callbacks like the backup job,
2309 * these jobs do not necessarily adhere to transaction semantics.
2310 * These jobs may not fully undo all of their actions on abort, nor do they
2311 * necessarily work in transactions with more than one job in them.
2312 */
2313 };
2314
2315 /**
2316 * Allocate a TransactionProperties structure if necessary, and fill
2317 * that structure with desired defaults if they are unset.
2318 */
2319 static TransactionProperties *get_transaction_properties(
2320 TransactionProperties *props)
2321 {
2322 if (!props) {
2323 props = g_new0(TransactionProperties, 1);
2324 }
2325
2326 if (!props->has_completion_mode) {
2327 props->has_completion_mode = true;
2328 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2329 }
2330
2331 return props;
2332 }
2333
2334 /*
2335 * 'Atomic' group operations. The operations are performed as a set, and if
2336 * any fail then we roll back all operations in the group.
2337 */
2338 void qmp_transaction(TransactionActionList *dev_list,
2339 bool has_props,
2340 struct TransactionProperties *props,
2341 Error **errp)
2342 {
2343 TransactionActionList *dev_entry = dev_list;
2344 JobTxn *block_job_txn = NULL;
2345 BlkActionState *state, *next;
2346 Error *local_err = NULL;
2347
2348 QTAILQ_HEAD(, BlkActionState) snap_bdrv_states;
2349 QTAILQ_INIT(&snap_bdrv_states);
2350
2351 /* Does this transaction get canceled as a group on failure?
2352 * If not, we don't really need to make a JobTxn.
2353 */
2354 props = get_transaction_properties(props);
2355 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2356 block_job_txn = job_txn_new();
2357 }
2358
2359 /* drain all i/o before any operations */
2360 bdrv_drain_all();
2361
2362 /* We don't do anything in this loop that commits us to the operations */
2363 while (NULL != dev_entry) {
2364 TransactionAction *dev_info = NULL;
2365 const BlkActionOps *ops;
2366
2367 dev_info = dev_entry->value;
2368 dev_entry = dev_entry->next;
2369
2370 assert(dev_info->type < ARRAY_SIZE(actions));
2371
2372 ops = &actions[dev_info->type];
2373 assert(ops->instance_size > 0);
2374
2375 state = g_malloc0(ops->instance_size);
2376 state->ops = ops;
2377 state->action = dev_info;
2378 state->block_job_txn = block_job_txn;
2379 state->txn_props = props;
2380 QTAILQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2381
2382 state->ops->prepare(state, &local_err);
2383 if (local_err) {
2384 error_propagate(errp, local_err);
2385 goto delete_and_fail;
2386 }
2387 }
2388
2389 QTAILQ_FOREACH(state, &snap_bdrv_states, entry) {
2390 if (state->ops->commit) {
2391 state->ops->commit(state);
2392 }
2393 }
2394
2395 /* success */
2396 goto exit;
2397
2398 delete_and_fail:
2399 /* failure, and it is all-or-none; roll back all operations */
2400 QTAILQ_FOREACH_REVERSE(state, &snap_bdrv_states, entry) {
2401 if (state->ops->abort) {
2402 state->ops->abort(state);
2403 }
2404 }
2405 exit:
2406 QTAILQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2407 if (state->ops->clean) {
2408 state->ops->clean(state);
2409 }
2410 g_free(state);
2411 }
2412 if (!has_props) {
2413 qapi_free_TransactionProperties(props);
2414 }
2415 job_txn_unref(block_job_txn);
2416 }
2417
2418 void qmp_block_passwd(bool has_device, const char *device,
2419 bool has_node_name, const char *node_name,
2420 const char *password, Error **errp)
2421 {
2422 error_setg(errp,
2423 "Setting block passwords directly is no longer supported");
2424 }
2425
2426 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2427 const char *name,
2428 Error **errp)
2429 {
2430 BdrvDirtyBitmap *bitmap;
2431 BlockDriverState *bs;
2432 BlockDirtyBitmapSha256 *ret = NULL;
2433 char *sha256;
2434
2435 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2436 if (!bitmap || !bs) {
2437 return NULL;
2438 }
2439
2440 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2441 if (sha256 == NULL) {
2442 return NULL;
2443 }
2444
2445 ret = g_new(BlockDirtyBitmapSha256, 1);
2446 ret->sha256 = sha256;
2447
2448 return ret;
2449 }
2450
2451 void coroutine_fn qmp_block_resize(bool has_device, const char *device,
2452 bool has_node_name, const char *node_name,
2453 int64_t size, Error **errp)
2454 {
2455 Error *local_err = NULL;
2456 BlockBackend *blk;
2457 BlockDriverState *bs;
2458 AioContext *old_ctx;
2459
2460 bs = bdrv_lookup_bs(has_device ? device : NULL,
2461 has_node_name ? node_name : NULL,
2462 &local_err);
2463 if (local_err) {
2464 error_propagate(errp, local_err);
2465 return;
2466 }
2467
2468 if (size < 0) {
2469 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2470 return;
2471 }
2472
2473 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2474 error_setg(errp, QERR_DEVICE_IN_USE, device);
2475 return;
2476 }
2477
2478 blk = blk_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2479 if (!blk) {
2480 return;
2481 }
2482
2483 bdrv_co_lock(bs);
2484 bdrv_drained_begin(bs);
2485 bdrv_co_unlock(bs);
2486
2487 old_ctx = bdrv_co_enter(bs);
2488 blk_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2489 bdrv_co_leave(bs, old_ctx);
2490
2491 bdrv_co_lock(bs);
2492 bdrv_drained_end(bs);
2493 blk_unref(blk);
2494 bdrv_co_unlock(bs);
2495 }
2496
2497 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
2498 bool has_base, const char *base,
2499 bool has_base_node, const char *base_node,
2500 bool has_backing_file, const char *backing_file,
2501 bool has_bottom, const char *bottom,
2502 bool has_speed, int64_t speed,
2503 bool has_on_error, BlockdevOnError on_error,
2504 bool has_filter_node_name, const char *filter_node_name,
2505 bool has_auto_finalize, bool auto_finalize,
2506 bool has_auto_dismiss, bool auto_dismiss,
2507 Error **errp)
2508 {
2509 BlockDriverState *bs, *iter, *iter_end;
2510 BlockDriverState *base_bs = NULL;
2511 BlockDriverState *bottom_bs = NULL;
2512 AioContext *aio_context;
2513 Error *local_err = NULL;
2514 int job_flags = JOB_DEFAULT;
2515
2516 if (has_base && has_base_node) {
2517 error_setg(errp, "'base' and 'base-node' cannot be specified "
2518 "at the same time");
2519 return;
2520 }
2521
2522 if (has_base && has_bottom) {
2523 error_setg(errp, "'base' and 'bottom' cannot be specified "
2524 "at the same time");
2525 return;
2526 }
2527
2528 if (has_bottom && has_base_node) {
2529 error_setg(errp, "'bottom' and 'base-node' cannot be specified "
2530 "at the same time");
2531 return;
2532 }
2533
2534 if (!has_on_error) {
2535 on_error = BLOCKDEV_ON_ERROR_REPORT;
2536 }
2537
2538 bs = bdrv_lookup_bs(device, device, errp);
2539 if (!bs) {
2540 return;
2541 }
2542
2543 aio_context = bdrv_get_aio_context(bs);
2544 aio_context_acquire(aio_context);
2545
2546 if (has_base) {
2547 base_bs = bdrv_find_backing_image(bs, base);
2548 if (base_bs == NULL) {
2549 error_setg(errp, "Can't find '%s' in the backing chain", base);
2550 goto out;
2551 }
2552 assert(bdrv_get_aio_context(base_bs) == aio_context);
2553 }
2554
2555 if (has_base_node) {
2556 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2557 if (!base_bs) {
2558 goto out;
2559 }
2560 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2561 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2562 base_node, device);
2563 goto out;
2564 }
2565 assert(bdrv_get_aio_context(base_bs) == aio_context);
2566 bdrv_refresh_filename(base_bs);
2567 }
2568
2569 if (has_bottom) {
2570 bottom_bs = bdrv_lookup_bs(NULL, bottom, errp);
2571 if (!bottom_bs) {
2572 goto out;
2573 }
2574 if (!bottom_bs->drv) {
2575 error_setg(errp, "Node '%s' is not open", bottom);
2576 goto out;
2577 }
2578 if (bottom_bs->drv->is_filter) {
2579 error_setg(errp, "Node '%s' is a filter, use a non-filter node "
2580 "as 'bottom'", bottom);
2581 goto out;
2582 }
2583 if (!bdrv_chain_contains(bs, bottom_bs)) {
2584 error_setg(errp, "Node '%s' is not in a chain starting from '%s'",
2585 bottom, device);
2586 goto out;
2587 }
2588 assert(bdrv_get_aio_context(bottom_bs) == aio_context);
2589 }
2590
2591 /*
2592 * Check for op blockers in the whole chain between bs and base (or bottom)
2593 */
2594 iter_end = has_bottom ? bdrv_filter_or_cow_bs(bottom_bs) : base_bs;
2595 for (iter = bs; iter && iter != iter_end;
2596 iter = bdrv_filter_or_cow_bs(iter))
2597 {
2598 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2599 goto out;
2600 }
2601 }
2602
2603 /* if we are streaming the entire chain, the result will have no backing
2604 * file, and specifying one is therefore an error */
2605 if (base_bs == NULL && has_backing_file) {
2606 error_setg(errp, "backing file specified, but streaming the "
2607 "entire chain");
2608 goto out;
2609 }
2610
2611 if (has_auto_finalize && !auto_finalize) {
2612 job_flags |= JOB_MANUAL_FINALIZE;
2613 }
2614 if (has_auto_dismiss && !auto_dismiss) {
2615 job_flags |= JOB_MANUAL_DISMISS;
2616 }
2617
2618 stream_start(has_job_id ? job_id : NULL, bs, base_bs, backing_file,
2619 bottom_bs, job_flags, has_speed ? speed : 0, on_error,
2620 filter_node_name, &local_err);
2621 if (local_err) {
2622 error_propagate(errp, local_err);
2623 goto out;
2624 }
2625
2626 trace_qmp_block_stream(bs);
2627
2628 out:
2629 aio_context_release(aio_context);
2630 }
2631
2632 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
2633 bool has_base_node, const char *base_node,
2634 bool has_base, const char *base,
2635 bool has_top_node, const char *top_node,
2636 bool has_top, const char *top,
2637 bool has_backing_file, const char *backing_file,
2638 bool has_speed, int64_t speed,
2639 bool has_on_error, BlockdevOnError on_error,
2640 bool has_filter_node_name, const char *filter_node_name,
2641 bool has_auto_finalize, bool auto_finalize,
2642 bool has_auto_dismiss, bool auto_dismiss,
2643 Error **errp)
2644 {
2645 BlockDriverState *bs;
2646 BlockDriverState *iter;
2647 BlockDriverState *base_bs, *top_bs;
2648 AioContext *aio_context;
2649 Error *local_err = NULL;
2650 int job_flags = JOB_DEFAULT;
2651 uint64_t top_perm, top_shared;
2652
2653 if (!has_speed) {
2654 speed = 0;
2655 }
2656 if (!has_on_error) {
2657 on_error = BLOCKDEV_ON_ERROR_REPORT;
2658 }
2659 if (!has_filter_node_name) {
2660 filter_node_name = NULL;
2661 }
2662 if (has_auto_finalize && !auto_finalize) {
2663 job_flags |= JOB_MANUAL_FINALIZE;
2664 }
2665 if (has_auto_dismiss && !auto_dismiss) {
2666 job_flags |= JOB_MANUAL_DISMISS;
2667 }
2668
2669 /* Important Note:
2670 * libvirt relies on the DeviceNotFound error class in order to probe for
2671 * live commit feature versions; for this to work, we must make sure to
2672 * perform the device lookup before any generic errors that may occur in a
2673 * scenario in which all optional arguments are omitted. */
2674 bs = qmp_get_root_bs(device, &local_err);
2675 if (!bs) {
2676 bs = bdrv_lookup_bs(device, device, NULL);
2677 if (!bs) {
2678 error_free(local_err);
2679 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2680 "Device '%s' not found", device);
2681 } else {
2682 error_propagate(errp, local_err);
2683 }
2684 return;
2685 }
2686
2687 aio_context = bdrv_get_aio_context(bs);
2688 aio_context_acquire(aio_context);
2689
2690 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2691 goto out;
2692 }
2693
2694 /* default top_bs is the active layer */
2695 top_bs = bs;
2696
2697 if (has_top_node && has_top) {
2698 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2699 goto out;
2700 } else if (has_top_node) {
2701 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2702 if (top_bs == NULL) {
2703 goto out;
2704 }
2705 if (!bdrv_chain_contains(bs, top_bs)) {
2706 error_setg(errp, "'%s' is not in this backing file chain",
2707 top_node);
2708 goto out;
2709 }
2710 } else if (has_top && top) {
2711 /* This strcmp() is just a shortcut, there is no need to
2712 * refresh @bs's filename. If it mismatches,
2713 * bdrv_find_backing_image() will do the refresh and may still
2714 * return @bs. */
2715 if (strcmp(bs->filename, top) != 0) {
2716 top_bs = bdrv_find_backing_image(bs, top);
2717 }
2718 }
2719
2720 if (top_bs == NULL) {
2721 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2722 goto out;
2723 }
2724
2725 assert(bdrv_get_aio_context(top_bs) == aio_context);
2726
2727 if (has_base_node && has_base) {
2728 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2729 goto out;
2730 } else if (has_base_node) {
2731 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2732 if (base_bs == NULL) {
2733 goto out;
2734 }
2735 if (!bdrv_chain_contains(top_bs, base_bs)) {
2736 error_setg(errp, "'%s' is not in this backing file chain",
2737 base_node);
2738 goto out;
2739 }
2740 } else if (has_base && base) {
2741 base_bs = bdrv_find_backing_image(top_bs, base);
2742 if (base_bs == NULL) {
2743 error_setg(errp, "Can't find '%s' in the backing chain", base);
2744 goto out;
2745 }
2746 } else {
2747 base_bs = bdrv_find_base(top_bs);
2748 if (base_bs == NULL) {
2749 error_setg(errp, "There is no backimg image");
2750 goto out;
2751 }
2752 }
2753
2754 assert(bdrv_get_aio_context(base_bs) == aio_context);
2755
2756 for (iter = top_bs; iter != bdrv_filter_or_cow_bs(base_bs);
2757 iter = bdrv_filter_or_cow_bs(iter))
2758 {
2759 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2760 goto out;
2761 }
2762 }
2763
2764 /* Do not allow attempts to commit an image into itself */
2765 if (top_bs == base_bs) {
2766 error_setg(errp, "cannot commit an image into itself");
2767 goto out;
2768 }
2769
2770 /*
2771 * Active commit is required if and only if someone has taken a
2772 * WRITE permission on the top node. Historically, we have always
2773 * used active commit for top nodes, so continue that practice
2774 * lest we possibly break clients that rely on this behavior, e.g.
2775 * to later attach this node to a writing parent.
2776 * (Active commit is never really wrong.)
2777 */
2778 bdrv_get_cumulative_perm(top_bs, &top_perm, &top_shared);
2779 if (top_perm & BLK_PERM_WRITE ||
2780 bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs))
2781 {
2782 if (has_backing_file) {
2783 if (bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs)) {
2784 error_setg(errp, "'backing-file' specified,"
2785 " but 'top' is the active layer");
2786 } else {
2787 error_setg(errp, "'backing-file' specified, but 'top' has a "
2788 "writer on it");
2789 }
2790 goto out;
2791 }
2792 if (!has_job_id) {
2793 /*
2794 * Emulate here what block_job_create() does, because it
2795 * is possible that @bs != @top_bs (the block job should
2796 * be named after @bs, even if @top_bs is the actual
2797 * source)
2798 */
2799 job_id = bdrv_get_device_name(bs);
2800 }
2801 commit_active_start(job_id, top_bs, base_bs, job_flags, speed, on_error,
2802 filter_node_name, NULL, NULL, false, &local_err);
2803 } else {
2804 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2805 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2806 goto out;
2807 }
2808 commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, job_flags,
2809 speed, on_error, has_backing_file ? backing_file : NULL,
2810 filter_node_name, &local_err);
2811 }
2812 if (local_err != NULL) {
2813 error_propagate(errp, local_err);
2814 goto out;
2815 }
2816
2817 out:
2818 aio_context_release(aio_context);
2819 }
2820
2821 /* Common QMP interface for drive-backup and blockdev-backup */
2822 static BlockJob *do_backup_common(BackupCommon *backup,
2823 BlockDriverState *bs,
2824 BlockDriverState *target_bs,
2825 AioContext *aio_context,
2826 JobTxn *txn, Error **errp)
2827 {
2828 BlockJob *job = NULL;
2829 BdrvDirtyBitmap *bmap = NULL;
2830 BackupPerf perf = { .max_workers = 64 };
2831 int job_flags = JOB_DEFAULT;
2832
2833 if (!backup->has_speed) {
2834 backup->speed = 0;
2835 }
2836 if (!backup->has_on_source_error) {
2837 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2838 }
2839 if (!backup->has_on_target_error) {
2840 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2841 }
2842 if (!backup->has_job_id) {
2843 backup->job_id = NULL;
2844 }
2845 if (!backup->has_auto_finalize) {
2846 backup->auto_finalize = true;
2847 }
2848 if (!backup->has_auto_dismiss) {
2849 backup->auto_dismiss = true;
2850 }
2851 if (!backup->has_compress) {
2852 backup->compress = false;
2853 }
2854
2855 if (backup->x_perf) {
2856 if (backup->x_perf->has_use_copy_range) {
2857 perf.use_copy_range = backup->x_perf->use_copy_range;
2858 }
2859 if (backup->x_perf->has_max_workers) {
2860 perf.max_workers = backup->x_perf->max_workers;
2861 }
2862 if (backup->x_perf->has_max_chunk) {
2863 perf.max_chunk = backup->x_perf->max_chunk;
2864 }
2865 }
2866
2867 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2868 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2869 /* done before desugaring 'incremental' to print the right message */
2870 if (!backup->has_bitmap) {
2871 error_setg(errp, "must provide a valid bitmap name for "
2872 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2873 return NULL;
2874 }
2875 }
2876
2877 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2878 if (backup->has_bitmap_mode &&
2879 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2880 error_setg(errp, "Bitmap sync mode must be '%s' "
2881 "when using sync mode '%s'",
2882 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2883 MirrorSyncMode_str(backup->sync));
2884 return NULL;
2885 }
2886 backup->has_bitmap_mode = true;
2887 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2888 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2889 }
2890
2891 if (backup->has_bitmap) {
2892 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2893 if (!bmap) {
2894 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2895 return NULL;
2896 }
2897 if (!backup->has_bitmap_mode) {
2898 error_setg(errp, "Bitmap sync mode must be given "
2899 "when providing a bitmap");
2900 return NULL;
2901 }
2902 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2903 return NULL;
2904 }
2905
2906 /* This does not produce a useful bitmap artifact: */
2907 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2908 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2909 " outputs", MirrorSyncMode_str(backup->sync));
2910 return NULL;
2911 }
2912
2913 /* If the bitmap isn't used for input or output, this is useless: */
2914 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2915 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2916 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2917 " when combined with sync mode '%s'",
2918 BitmapSyncMode_str(backup->bitmap_mode),
2919 MirrorSyncMode_str(backup->sync));
2920 return NULL;
2921 }
2922 }
2923
2924 if (!backup->has_bitmap && backup->has_bitmap_mode) {
2925 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2926 return NULL;
2927 }
2928
2929 if (!backup->auto_finalize) {
2930 job_flags |= JOB_MANUAL_FINALIZE;
2931 }
2932 if (!backup->auto_dismiss) {
2933 job_flags |= JOB_MANUAL_DISMISS;
2934 }
2935
2936 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2937 backup->sync, bmap, backup->bitmap_mode,
2938 backup->compress,
2939 backup->filter_node_name,
2940 &perf,
2941 backup->on_source_error,
2942 backup->on_target_error,
2943 job_flags, NULL, NULL, txn, errp);
2944 return job;
2945 }
2946
2947 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2948 {
2949 TransactionAction action = {
2950 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2951 .u.drive_backup.data = backup,
2952 };
2953 blockdev_do_action(&action, errp);
2954 }
2955
2956 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2957 bool flat,
2958 Error **errp)
2959 {
2960 bool return_flat = has_flat && flat;
2961
2962 return bdrv_named_nodes_list(return_flat, errp);
2963 }
2964
2965 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2966 {
2967 return bdrv_get_xdbg_block_graph(errp);
2968 }
2969
2970 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2971 {
2972 TransactionAction action = {
2973 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2974 .u.blockdev_backup.data = backup,
2975 };
2976 blockdev_do_action(&action, errp);
2977 }
2978
2979 /* Parameter check and block job starting for drive mirroring.
2980 * Caller should hold @device and @target's aio context (must be the same).
2981 **/
2982 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2983 BlockDriverState *target,
2984 bool has_replaces, const char *replaces,
2985 enum MirrorSyncMode sync,
2986 BlockMirrorBackingMode backing_mode,
2987 bool zero_target,
2988 bool has_speed, int64_t speed,
2989 bool has_granularity, uint32_t granularity,
2990 bool has_buf_size, int64_t buf_size,
2991 bool has_on_source_error,
2992 BlockdevOnError on_source_error,
2993 bool has_on_target_error,
2994 BlockdevOnError on_target_error,
2995 bool has_unmap, bool unmap,
2996 bool has_filter_node_name,
2997 const char *filter_node_name,
2998 bool has_copy_mode, MirrorCopyMode copy_mode,
2999 bool has_auto_finalize, bool auto_finalize,
3000 bool has_auto_dismiss, bool auto_dismiss,
3001 Error **errp)
3002 {
3003 BlockDriverState *unfiltered_bs;
3004 int job_flags = JOB_DEFAULT;
3005
3006 if (!has_speed) {
3007 speed = 0;
3008 }
3009 if (!has_on_source_error) {
3010 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3011 }
3012 if (!has_on_target_error) {
3013 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3014 }
3015 if (!has_granularity) {
3016 granularity = 0;
3017 }
3018 if (!has_buf_size) {
3019 buf_size = 0;
3020 }
3021 if (!has_unmap) {
3022 unmap = true;
3023 }
3024 if (!has_filter_node_name) {
3025 filter_node_name = NULL;
3026 }
3027 if (!has_copy_mode) {
3028 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
3029 }
3030 if (has_auto_finalize && !auto_finalize) {
3031 job_flags |= JOB_MANUAL_FINALIZE;
3032 }
3033 if (has_auto_dismiss && !auto_dismiss) {
3034 job_flags |= JOB_MANUAL_DISMISS;
3035 }
3036
3037 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3038 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3039 "a value in range [512B, 64MB]");
3040 return;
3041 }
3042 if (granularity & (granularity - 1)) {
3043 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3044 "a power of 2");
3045 return;
3046 }
3047
3048 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3049 return;
3050 }
3051 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3052 return;
3053 }
3054
3055 if (!bdrv_backing_chain_next(bs) && sync == MIRROR_SYNC_MODE_TOP) {
3056 sync = MIRROR_SYNC_MODE_FULL;
3057 }
3058
3059 if (!has_replaces) {
3060 /* We want to mirror from @bs, but keep implicit filters on top */
3061 unfiltered_bs = bdrv_skip_implicit_filters(bs);
3062 if (unfiltered_bs != bs) {
3063 replaces = unfiltered_bs->node_name;
3064 has_replaces = true;
3065 }
3066 }
3067
3068 if (has_replaces) {
3069 BlockDriverState *to_replace_bs;
3070 AioContext *replace_aio_context;
3071 int64_t bs_size, replace_size;
3072
3073 bs_size = bdrv_getlength(bs);
3074 if (bs_size < 0) {
3075 error_setg_errno(errp, -bs_size, "Failed to query device's size");
3076 return;
3077 }
3078
3079 to_replace_bs = check_to_replace_node(bs, replaces, errp);
3080 if (!to_replace_bs) {
3081 return;
3082 }
3083
3084 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3085 aio_context_acquire(replace_aio_context);
3086 replace_size = bdrv_getlength(to_replace_bs);
3087 aio_context_release(replace_aio_context);
3088
3089 if (replace_size < 0) {
3090 error_setg_errno(errp, -replace_size,
3091 "Failed to query the replacement node's size");
3092 return;
3093 }
3094 if (bs_size != replace_size) {
3095 error_setg(errp, "cannot replace image with a mirror image of "
3096 "different size");
3097 return;
3098 }
3099 }
3100
3101 /* pass the node name to replace to mirror start since it's loose coupling
3102 * and will allow to check whether the node still exist at mirror completion
3103 */
3104 mirror_start(job_id, bs, target,
3105 has_replaces ? replaces : NULL, job_flags,
3106 speed, granularity, buf_size, sync, backing_mode, zero_target,
3107 on_source_error, on_target_error, unmap, filter_node_name,
3108 copy_mode, errp);
3109 }
3110
3111 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3112 {
3113 BlockDriverState *bs;
3114 BlockDriverState *target_backing_bs, *target_bs;
3115 AioContext *aio_context;
3116 AioContext *old_context;
3117 BlockMirrorBackingMode backing_mode;
3118 Error *local_err = NULL;
3119 QDict *options = NULL;
3120 int flags;
3121 int64_t size;
3122 const char *format = arg->format;
3123 bool zero_target;
3124 int ret;
3125
3126 bs = qmp_get_root_bs(arg->device, errp);
3127 if (!bs) {
3128 return;
3129 }
3130
3131 /* Early check to avoid creating target */
3132 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3133 return;
3134 }
3135
3136 aio_context = bdrv_get_aio_context(bs);
3137 aio_context_acquire(aio_context);
3138
3139 if (!arg->has_mode) {
3140 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3141 }
3142
3143 if (!arg->has_format) {
3144 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3145 ? NULL : bs->drv->format_name);
3146 }
3147
3148 flags = bs->open_flags | BDRV_O_RDWR;
3149 target_backing_bs = bdrv_cow_bs(bdrv_skip_filters(bs));
3150 if (!target_backing_bs && arg->sync == MIRROR_SYNC_MODE_TOP) {
3151 arg->sync = MIRROR_SYNC_MODE_FULL;
3152 }
3153 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3154 target_backing_bs = bs;
3155 }
3156
3157 size = bdrv_getlength(bs);
3158 if (size < 0) {
3159 error_setg_errno(errp, -size, "bdrv_getlength failed");
3160 goto out;
3161 }
3162
3163 if (arg->has_replaces) {
3164 if (!arg->has_node_name) {
3165 error_setg(errp, "a node-name must be provided when replacing a"
3166 " named node of the graph");
3167 goto out;
3168 }
3169 }
3170
3171 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3172 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3173 } else {
3174 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3175 }
3176
3177 /* Don't open backing image in create() */
3178 flags |= BDRV_O_NO_BACKING;
3179
3180 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !target_backing_bs)
3181 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3182 {
3183 /* create new image w/o backing file */
3184 assert(format);
3185 bdrv_img_create(arg->target, format,
3186 NULL, NULL, NULL, size, flags, false, &local_err);
3187 } else {
3188 /* Implicit filters should not appear in the filename */
3189 BlockDriverState *explicit_backing =
3190 bdrv_skip_implicit_filters(target_backing_bs);
3191
3192 switch (arg->mode) {
3193 case NEW_IMAGE_MODE_EXISTING:
3194 break;
3195 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3196 /* create new image with backing file */
3197 bdrv_refresh_filename(explicit_backing);
3198 bdrv_img_create(arg->target, format,
3199 explicit_backing->filename,
3200 explicit_backing->drv->format_name,
3201 NULL, size, flags, false, &local_err);
3202 break;
3203 default:
3204 abort();
3205 }
3206 }
3207
3208 if (local_err) {
3209 error_propagate(errp, local_err);
3210 goto out;
3211 }
3212
3213 options = qdict_new();
3214 if (arg->has_node_name) {
3215 qdict_put_str(options, "node-name", arg->node_name);
3216 }
3217 if (format) {
3218 qdict_put_str(options, "driver", format);
3219 }
3220
3221 /* Mirroring takes care of copy-on-write using the source's backing
3222 * file.
3223 */
3224 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3225 if (!target_bs) {
3226 goto out;
3227 }
3228
3229 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3230 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3231 !bdrv_has_zero_init(target_bs)));
3232
3233
3234 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3235 old_context = bdrv_get_aio_context(target_bs);
3236 aio_context_release(aio_context);
3237 aio_context_acquire(old_context);
3238
3239 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3240 if (ret < 0) {
3241 bdrv_unref(target_bs);
3242 aio_context_release(old_context);
3243 return;
3244 }
3245
3246 aio_context_release(old_context);
3247 aio_context_acquire(aio_context);
3248
3249 blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3250 arg->has_replaces, arg->replaces, arg->sync,
3251 backing_mode, zero_target,
3252 arg->has_speed, arg->speed,
3253 arg->has_granularity, arg->granularity,
3254 arg->has_buf_size, arg->buf_size,
3255 arg->has_on_source_error, arg->on_source_error,
3256 arg->has_on_target_error, arg->on_target_error,
3257 arg->has_unmap, arg->unmap,
3258 false, NULL,
3259 arg->has_copy_mode, arg->copy_mode,
3260 arg->has_auto_finalize, arg->auto_finalize,
3261 arg->has_auto_dismiss, arg->auto_dismiss,
3262 errp);
3263 bdrv_unref(target_bs);
3264 out:
3265 aio_context_release(aio_context);
3266 }
3267
3268 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3269 const char *device, const char *target,
3270 bool has_replaces, const char *replaces,
3271 MirrorSyncMode sync,
3272 bool has_speed, int64_t speed,
3273 bool has_granularity, uint32_t granularity,
3274 bool has_buf_size, int64_t buf_size,
3275 bool has_on_source_error,
3276 BlockdevOnError on_source_error,
3277 bool has_on_target_error,
3278 BlockdevOnError on_target_error,
3279 bool has_filter_node_name,
3280 const char *filter_node_name,
3281 bool has_copy_mode, MirrorCopyMode copy_mode,
3282 bool has_auto_finalize, bool auto_finalize,
3283 bool has_auto_dismiss, bool auto_dismiss,
3284 Error **errp)
3285 {
3286 BlockDriverState *bs;
3287 BlockDriverState *target_bs;
3288 AioContext *aio_context;
3289 AioContext *old_context;
3290 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3291 bool zero_target;
3292 int ret;
3293
3294 bs = qmp_get_root_bs(device, errp);
3295 if (!bs) {
3296 return;
3297 }
3298
3299 target_bs = bdrv_lookup_bs(target, target, errp);
3300 if (!target_bs) {
3301 return;
3302 }
3303
3304 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3305
3306 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3307 old_context = bdrv_get_aio_context(target_bs);
3308 aio_context = bdrv_get_aio_context(bs);
3309 aio_context_acquire(old_context);
3310
3311 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3312
3313 aio_context_release(old_context);
3314 aio_context_acquire(aio_context);
3315
3316 if (ret < 0) {
3317 goto out;
3318 }
3319
3320 blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
3321 has_replaces, replaces, sync, backing_mode,
3322 zero_target, has_speed, speed,
3323 has_granularity, granularity,
3324 has_buf_size, buf_size,
3325 has_on_source_error, on_source_error,
3326 has_on_target_error, on_target_error,
3327 true, true,
3328 has_filter_node_name, filter_node_name,
3329 has_copy_mode, copy_mode,
3330 has_auto_finalize, auto_finalize,
3331 has_auto_dismiss, auto_dismiss,
3332 errp);
3333 out:
3334 aio_context_release(aio_context);
3335 }
3336
3337 /* Get a block job using its ID and acquire its AioContext */
3338 static BlockJob *find_block_job(const char *id, AioContext **aio_context,
3339 Error **errp)
3340 {
3341 BlockJob *job;
3342
3343 assert(id != NULL);
3344
3345 *aio_context = NULL;
3346
3347 job = block_job_get(id);
3348
3349 if (!job) {
3350 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3351 "Block job '%s' not found", id);
3352 return NULL;
3353 }
3354
3355 *aio_context = blk_get_aio_context(job->blk);
3356 aio_context_acquire(*aio_context);
3357
3358 return job;
3359 }
3360
3361 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3362 {
3363 AioContext *aio_context;
3364 BlockJob *job = find_block_job(device, &aio_context, errp);
3365
3366 if (!job) {
3367 return;
3368 }
3369
3370 block_job_set_speed(job, speed, errp);
3371 aio_context_release(aio_context);
3372 }
3373
3374 void qmp_block_job_cancel(const char *device,
3375 bool has_force, bool force, Error **errp)
3376 {
3377 AioContext *aio_context;
3378 BlockJob *job = find_block_job(device, &aio_context, errp);
3379
3380 if (!job) {
3381 return;
3382 }
3383
3384 if (!has_force) {
3385 force = false;
3386 }
3387
3388 if (job_user_paused(&job->job) && !force) {
3389 error_setg(errp, "The block job for device '%s' is currently paused",
3390 device);
3391 goto out;
3392 }
3393
3394 trace_qmp_block_job_cancel(job);
3395 job_user_cancel(&job->job, force, errp);
3396 out:
3397 aio_context_release(aio_context);
3398 }
3399
3400 void qmp_block_job_pause(const char *device, Error **errp)
3401 {
3402 AioContext *aio_context;
3403 BlockJob *job = find_block_job(device, &aio_context, errp);
3404
3405 if (!job) {
3406 return;
3407 }
3408
3409 trace_qmp_block_job_pause(job);
3410 job_user_pause(&job->job, errp);
3411 aio_context_release(aio_context);
3412 }
3413
3414 void qmp_block_job_resume(const char *device, Error **errp)
3415 {
3416 AioContext *aio_context;
3417 BlockJob *job = find_block_job(device, &aio_context, errp);
3418
3419 if (!job) {
3420 return;
3421 }
3422
3423 trace_qmp_block_job_resume(job);
3424 job_user_resume(&job->job, errp);
3425 aio_context_release(aio_context);
3426 }
3427
3428 void qmp_block_job_complete(const char *device, Error **errp)
3429 {
3430 AioContext *aio_context;
3431 BlockJob *job = find_block_job(device, &aio_context, errp);
3432
3433 if (!job) {
3434 return;
3435 }
3436
3437 trace_qmp_block_job_complete(job);
3438 job_complete(&job->job, errp);
3439 aio_context_release(aio_context);
3440 }
3441
3442 void qmp_block_job_finalize(const char *id, Error **errp)
3443 {
3444 AioContext *aio_context;
3445 BlockJob *job = find_block_job(id, &aio_context, errp);
3446
3447 if (!job) {
3448 return;
3449 }
3450
3451 trace_qmp_block_job_finalize(job);
3452 job_ref(&job->job);
3453 job_finalize(&job->job, errp);
3454
3455 /*
3456 * Job's context might have changed via job_finalize (and job_txn_apply
3457 * automatically acquires the new one), so make sure we release the correct
3458 * one.
3459 */
3460 aio_context = blk_get_aio_context(job->blk);
3461 job_unref(&job->job);
3462 aio_context_release(aio_context);
3463 }
3464
3465 void qmp_block_job_dismiss(const char *id, Error **errp)
3466 {
3467 AioContext *aio_context;
3468 BlockJob *bjob = find_block_job(id, &aio_context, errp);
3469 Job *job;
3470
3471 if (!bjob) {
3472 return;
3473 }
3474
3475 trace_qmp_block_job_dismiss(bjob);
3476 job = &bjob->job;
3477 job_dismiss(&job, errp);
3478 aio_context_release(aio_context);
3479 }
3480
3481 void qmp_change_backing_file(const char *device,
3482 const char *image_node_name,
3483 const char *backing_file,
3484 Error **errp)
3485 {
3486 BlockDriverState *bs = NULL;
3487 AioContext *aio_context;
3488 BlockDriverState *image_bs = NULL;
3489 Error *local_err = NULL;
3490 bool ro;
3491 int ret;
3492
3493 bs = qmp_get_root_bs(device, errp);
3494 if (!bs) {
3495 return;
3496 }
3497
3498 aio_context = bdrv_get_aio_context(bs);
3499 aio_context_acquire(aio_context);
3500
3501 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3502 if (local_err) {
3503 error_propagate(errp, local_err);
3504 goto out;
3505 }
3506
3507 if (!image_bs) {
3508 error_setg(errp, "image file not found");
3509 goto out;
3510 }
3511
3512 if (bdrv_find_base(image_bs) == image_bs) {
3513 error_setg(errp, "not allowing backing file change on an image "
3514 "without a backing file");
3515 goto out;
3516 }
3517
3518 /* even though we are not necessarily operating on bs, we need it to
3519 * determine if block ops are currently prohibited on the chain */
3520 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3521 goto out;
3522 }
3523
3524 /* final sanity check */
3525 if (!bdrv_chain_contains(bs, image_bs)) {
3526 error_setg(errp, "'%s' and image file are not in the same chain",
3527 device);
3528 goto out;
3529 }
3530
3531 /* if not r/w, reopen to make r/w */
3532 ro = bdrv_is_read_only(image_bs);
3533
3534 if (ro) {
3535 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3536 goto out;
3537 }
3538 }
3539
3540 ret = bdrv_change_backing_file(image_bs, backing_file,
3541 image_bs->drv ? image_bs->drv->format_name : "",
3542 false);
3543
3544 if (ret < 0) {
3545 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3546 backing_file);
3547 /* don't exit here, so we can try to restore open flags if
3548 * appropriate */
3549 }
3550
3551 if (ro) {
3552 bdrv_reopen_set_read_only(image_bs, true, errp);
3553 }
3554
3555 out:
3556 aio_context_release(aio_context);
3557 }
3558
3559 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3560 {
3561 BlockDriverState *bs;
3562 QObject *obj;
3563 Visitor *v = qobject_output_visitor_new(&obj);
3564 QDict *qdict;
3565
3566 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3567 visit_complete(v, &obj);
3568 qdict = qobject_to(QDict, obj);
3569
3570 qdict_flatten(qdict);
3571
3572 if (!qdict_get_try_str(qdict, "node-name")) {
3573 error_setg(errp, "'node-name' must be specified for the root node");
3574 goto fail;
3575 }
3576
3577 bs = bds_tree_init(qdict, errp);
3578 if (!bs) {
3579 goto fail;
3580 }
3581
3582 bdrv_set_monitor_owned(bs);
3583
3584 fail:
3585 visit_free(v);
3586 }
3587
3588 void qmp_x_blockdev_reopen(BlockdevOptions *options, Error **errp)
3589 {
3590 BlockDriverState *bs;
3591 AioContext *ctx;
3592 QObject *obj;
3593 Visitor *v = qobject_output_visitor_new(&obj);
3594 BlockReopenQueue *queue;
3595 QDict *qdict;
3596
3597 /* Check for the selected node name */
3598 if (!options->has_node_name) {
3599 error_setg(errp, "Node name not specified");
3600 goto fail;
3601 }
3602
3603 bs = bdrv_find_node(options->node_name);
3604 if (!bs) {
3605 error_setg(errp, "Cannot find node named '%s'", options->node_name);
3606 goto fail;
3607 }
3608
3609 /* Put all options in a QDict and flatten it */
3610 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3611 visit_complete(v, &obj);
3612 qdict = qobject_to(QDict, obj);
3613
3614 qdict_flatten(qdict);
3615
3616 /* Perform the reopen operation */
3617 ctx = bdrv_get_aio_context(bs);
3618 aio_context_acquire(ctx);
3619 bdrv_subtree_drained_begin(bs);
3620 queue = bdrv_reopen_queue(NULL, bs, qdict, false);
3621 bdrv_reopen_multiple(queue, errp);
3622 bdrv_subtree_drained_end(bs);
3623 aio_context_release(ctx);
3624
3625 fail:
3626 visit_free(v);
3627 }
3628
3629 void qmp_blockdev_del(const char *node_name, Error **errp)
3630 {
3631 AioContext *aio_context;
3632 BlockDriverState *bs;
3633
3634 bs = bdrv_find_node(node_name);
3635 if (!bs) {
3636 error_setg(errp, "Cannot find node %s", node_name);
3637 return;
3638 }
3639 if (bdrv_has_blk(bs)) {
3640 error_setg(errp, "Node %s is in use", node_name);
3641 return;
3642 }
3643 aio_context = bdrv_get_aio_context(bs);
3644 aio_context_acquire(aio_context);
3645
3646 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3647 goto out;
3648 }
3649
3650 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3651 error_setg(errp, "Node %s is not owned by the monitor",
3652 bs->node_name);
3653 goto out;
3654 }
3655
3656 if (bs->refcnt > 1) {
3657 error_setg(errp, "Block device %s is in use",
3658 bdrv_get_device_or_node_name(bs));
3659 goto out;
3660 }
3661
3662 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3663 bdrv_unref(bs);
3664
3665 out:
3666 aio_context_release(aio_context);
3667 }
3668
3669 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3670 const char *child_name)
3671 {
3672 BdrvChild *child;
3673
3674 QLIST_FOREACH(child, &parent_bs->children, next) {
3675 if (strcmp(child->name, child_name) == 0) {
3676 return child;
3677 }
3678 }
3679
3680 return NULL;
3681 }
3682
3683 void qmp_x_blockdev_change(const char *parent, bool has_child,
3684 const char *child, bool has_node,
3685 const char *node, Error **errp)
3686 {
3687 BlockDriverState *parent_bs, *new_bs = NULL;
3688 BdrvChild *p_child;
3689
3690 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3691 if (!parent_bs) {
3692 return;
3693 }
3694
3695 if (has_child == has_node) {
3696 if (has_child) {
3697 error_setg(errp, "The parameters child and node are in conflict");
3698 } else {
3699 error_setg(errp, "Either child or node must be specified");
3700 }
3701 return;
3702 }
3703
3704 if (has_child) {
3705 p_child = bdrv_find_child(parent_bs, child);
3706 if (!p_child) {
3707 error_setg(errp, "Node '%s' does not have child '%s'",
3708 parent, child);
3709 return;
3710 }
3711 bdrv_del_child(parent_bs, p_child, errp);
3712 }
3713
3714 if (has_node) {
3715 new_bs = bdrv_find_node(node);
3716 if (!new_bs) {
3717 error_setg(errp, "Node '%s' not found", node);
3718 return;
3719 }
3720 bdrv_add_child(parent_bs, new_bs, errp);
3721 }
3722 }
3723
3724 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3725 {
3726 BlockJobInfoList *head = NULL, **tail = &head;
3727 BlockJob *job;
3728
3729 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
3730 BlockJobInfo *value;
3731 AioContext *aio_context;
3732
3733 if (block_job_is_internal(job)) {
3734 continue;
3735 }
3736 aio_context = blk_get_aio_context(job->blk);
3737 aio_context_acquire(aio_context);
3738 value = block_job_query(job, errp);
3739 aio_context_release(aio_context);
3740 if (!value) {
3741 qapi_free_BlockJobInfoList(head);
3742 return NULL;
3743 }
3744 QAPI_LIST_APPEND(tail, value);
3745 }
3746
3747 return head;
3748 }
3749
3750 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3751 bool has_force, bool force, Error **errp)
3752 {
3753 AioContext *old_context;
3754 AioContext *new_context;
3755 BlockDriverState *bs;
3756
3757 bs = bdrv_find_node(node_name);
3758 if (!bs) {
3759 error_setg(errp, "Cannot find node %s", node_name);
3760 return;
3761 }
3762
3763 /* Protects against accidents. */
3764 if (!(has_force && force) && bdrv_has_blk(bs)) {
3765 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3766 "be in use (use force=true to override this check)",
3767 node_name);
3768 return;
3769 }
3770
3771 if (iothread->type == QTYPE_QSTRING) {
3772 IOThread *obj = iothread_by_id(iothread->u.s);
3773 if (!obj) {
3774 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3775 return;
3776 }
3777
3778 new_context = iothread_get_aio_context(obj);
3779 } else {
3780 new_context = qemu_get_aio_context();
3781 }
3782
3783 old_context = bdrv_get_aio_context(bs);
3784 aio_context_acquire(old_context);
3785
3786 bdrv_try_set_aio_context(bs, new_context, errp);
3787
3788 aio_context_release(old_context);
3789 }
3790
3791 QemuOptsList qemu_common_drive_opts = {
3792 .name = "drive",
3793 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3794 .desc = {
3795 {
3796 .name = "snapshot",
3797 .type = QEMU_OPT_BOOL,
3798 .help = "enable/disable snapshot mode",
3799 },{
3800 .name = "aio",
3801 .type = QEMU_OPT_STRING,
3802 .help = "host AIO implementation (threads, native, io_uring)",
3803 },{
3804 .name = BDRV_OPT_CACHE_WB,
3805 .type = QEMU_OPT_BOOL,
3806 .help = "Enable writeback mode",
3807 },{
3808 .name = "format",
3809 .type = QEMU_OPT_STRING,
3810 .help = "disk format (raw, qcow2, ...)",
3811 },{
3812 .name = "rerror",
3813 .type = QEMU_OPT_STRING,
3814 .help = "read error action",
3815 },{
3816 .name = "werror",
3817 .type = QEMU_OPT_STRING,
3818 .help = "write error action",
3819 },{
3820 .name = BDRV_OPT_READ_ONLY,
3821 .type = QEMU_OPT_BOOL,
3822 .help = "open drive file as read-only",
3823 },
3824
3825 THROTTLE_OPTS,
3826
3827 {
3828 .name = "throttling.group",
3829 .type = QEMU_OPT_STRING,
3830 .help = "name of the block throttling group",
3831 },{
3832 .name = "copy-on-read",
3833 .type = QEMU_OPT_BOOL,
3834 .help = "copy read data from backing file into image file",
3835 },{
3836 .name = "detect-zeroes",
3837 .type = QEMU_OPT_STRING,
3838 .help = "try to optimize zero writes (off, on, unmap)",
3839 },{
3840 .name = "stats-account-invalid",
3841 .type = QEMU_OPT_BOOL,
3842 .help = "whether to account for invalid I/O operations "
3843 "in the statistics",
3844 },{
3845 .name = "stats-account-failed",
3846 .type = QEMU_OPT_BOOL,
3847 .help = "whether to account for failed I/O operations "
3848 "in the statistics",
3849 },
3850 { /* end of list */ }
3851 },
3852 };
3853
3854 QemuOptsList qemu_drive_opts = {
3855 .name = "drive",
3856 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3857 .desc = {
3858 /*
3859 * no elements => accept any params
3860 * validation will happen later
3861 */
3862 { /* end of list */ }
3863 },
3864 };
Random patches