?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
nbd: server: Report holes for raw images
[qemu/kevin.git] / blockdev.c
blobcd438e60e35a0887abc9f6509609811c87284cef
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 bdrv_set_backing_hd(target_bs, source, &local_err);
1829 if (local_err) {
1830 error_propagate(errp, local_err);
1831 goto unref;
1832 }
1833 }
1834
1835 state->bs = bs;
1836
1837 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1838 bs, target_bs, aio_context,
1839 common->block_job_txn, errp);
1840
1841 unref:
1842 bdrv_unref(target_bs);
1843 out:
1844 aio_context_release(aio_context);
1845 }
1846
1847 static void drive_backup_commit(BlkActionState *common)
1848 {
1849 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1850 AioContext *aio_context;
1851
1852 aio_context = bdrv_get_aio_context(state->bs);
1853 aio_context_acquire(aio_context);
1854
1855 assert(state->job);
1856 job_start(&state->job->job);
1857
1858 aio_context_release(aio_context);
1859 }
1860
1861 static void drive_backup_abort(BlkActionState *common)
1862 {
1863 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1864
1865 if (state->job) {
1866 AioContext *aio_context;
1867
1868 aio_context = bdrv_get_aio_context(state->bs);
1869 aio_context_acquire(aio_context);
1870
1871 job_cancel_sync(&state->job->job);
1872
1873 aio_context_release(aio_context);
1874 }
1875 }
1876
1877 static void drive_backup_clean(BlkActionState *common)
1878 {
1879 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1880 AioContext *aio_context;
1881
1882 if (!state->bs) {
1883 return;
1884 }
1885
1886 aio_context = bdrv_get_aio_context(state->bs);
1887 aio_context_acquire(aio_context);
1888
1889 bdrv_drained_end(state->bs);
1890
1891 aio_context_release(aio_context);
1892 }
1893
1894 typedef struct BlockdevBackupState {
1895 BlkActionState common;
1896 BlockDriverState *bs;
1897 BlockJob *job;
1898 } BlockdevBackupState;
1899
1900 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1901 {
1902 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1903 BlockdevBackup *backup;
1904 BlockDriverState *bs;
1905 BlockDriverState *target_bs;
1906 AioContext *aio_context;
1907 AioContext *old_context;
1908 int ret;
1909
1910 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1911 backup = common->action->u.blockdev_backup.data;
1912
1913 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1914 if (!bs) {
1915 return;
1916 }
1917
1918 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1919 if (!target_bs) {
1920 return;
1921 }
1922
1923 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1924 aio_context = bdrv_get_aio_context(bs);
1925 old_context = bdrv_get_aio_context(target_bs);
1926 aio_context_acquire(old_context);
1927
1928 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1929 if (ret < 0) {
1930 aio_context_release(old_context);
1931 return;
1932 }
1933
1934 aio_context_release(old_context);
1935 aio_context_acquire(aio_context);
1936 state->bs = bs;
1937
1938 /* Paired with .clean() */
1939 bdrv_drained_begin(state->bs);
1940
1941 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1942 bs, target_bs, aio_context,
1943 common->block_job_txn, errp);
1944
1945 aio_context_release(aio_context);
1946 }
1947
1948 static void blockdev_backup_commit(BlkActionState *common)
1949 {
1950 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1951 AioContext *aio_context;
1952
1953 aio_context = bdrv_get_aio_context(state->bs);
1954 aio_context_acquire(aio_context);
1955
1956 assert(state->job);
1957 job_start(&state->job->job);
1958
1959 aio_context_release(aio_context);
1960 }
1961
1962 static void blockdev_backup_abort(BlkActionState *common)
1963 {
1964 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1965
1966 if (state->job) {
1967 AioContext *aio_context;
1968
1969 aio_context = bdrv_get_aio_context(state->bs);
1970 aio_context_acquire(aio_context);
1971
1972 job_cancel_sync(&state->job->job);
1973
1974 aio_context_release(aio_context);
1975 }
1976 }
1977
1978 static void blockdev_backup_clean(BlkActionState *common)
1979 {
1980 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1981 AioContext *aio_context;
1982
1983 if (!state->bs) {
1984 return;
1985 }
1986
1987 aio_context = bdrv_get_aio_context(state->bs);
1988 aio_context_acquire(aio_context);
1989
1990 bdrv_drained_end(state->bs);
1991
1992 aio_context_release(aio_context);
1993 }
1994
1995 typedef struct BlockDirtyBitmapState {
1996 BlkActionState common;
1997 BdrvDirtyBitmap *bitmap;
1998 BlockDriverState *bs;
1999 HBitmap *backup;
2000 bool prepared;
2001 bool was_enabled;
2002 } BlockDirtyBitmapState;
2003
2004 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
2005 Error **errp)
2006 {
2007 Error *local_err = NULL;
2008 BlockDirtyBitmapAdd *action;
2009 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2010 common, common);
2011
2012 if (action_check_completion_mode(common, errp) < 0) {
2013 return;
2014 }
2015
2016 action = common->action->u.block_dirty_bitmap_add.data;
2017 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2018 qmp_block_dirty_bitmap_add(action->node, action->name,
2019 action->has_granularity, action->granularity,
2020 action->has_persistent, action->persistent,
2021 action->has_disabled, action->disabled,
2022 &local_err);
2023
2024 if (!local_err) {
2025 state->prepared = true;
2026 } else {
2027 error_propagate(errp, local_err);
2028 }
2029 }
2030
2031 static void block_dirty_bitmap_add_abort(BlkActionState *common)
2032 {
2033 BlockDirtyBitmapAdd *action;
2034 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2035 common, common);
2036
2037 action = common->action->u.block_dirty_bitmap_add.data;
2038 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2039 * then the node reference and bitmap name must have been valid.
2040 */
2041 if (state->prepared) {
2042 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2043 }
2044 }
2045
2046 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2047 Error **errp)
2048 {
2049 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2050 common, common);
2051 BlockDirtyBitmap *action;
2052
2053 if (action_check_completion_mode(common, errp) < 0) {
2054 return;
2055 }
2056
2057 action = common->action->u.block_dirty_bitmap_clear.data;
2058 state->bitmap = block_dirty_bitmap_lookup(action->node,
2059 action->name,
2060 &state->bs,
2061 errp);
2062 if (!state->bitmap) {
2063 return;
2064 }
2065
2066 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2067 return;
2068 }
2069
2070 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2071 }
2072
2073 static void block_dirty_bitmap_restore(BlkActionState *common)
2074 {
2075 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2076 common, common);
2077
2078 if (state->backup) {
2079 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2080 }
2081 }
2082
2083 static void block_dirty_bitmap_free_backup(BlkActionState *common)
2084 {
2085 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2086 common, common);
2087
2088 hbitmap_free(state->backup);
2089 }
2090
2091 static void block_dirty_bitmap_enable_prepare(BlkActionState *common,
2092 Error **errp)
2093 {
2094 BlockDirtyBitmap *action;
2095 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2096 common, common);
2097
2098 if (action_check_completion_mode(common, errp) < 0) {
2099 return;
2100 }
2101
2102 action = common->action->u.block_dirty_bitmap_enable.data;
2103 state->bitmap = block_dirty_bitmap_lookup(action->node,
2104 action->name,
2105 NULL,
2106 errp);
2107 if (!state->bitmap) {
2108 return;
2109 }
2110
2111 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2112 return;
2113 }
2114
2115 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2116 bdrv_enable_dirty_bitmap(state->bitmap);
2117 }
2118
2119 static void block_dirty_bitmap_enable_abort(BlkActionState *common)
2120 {
2121 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2122 common, common);
2123
2124 if (!state->was_enabled) {
2125 bdrv_disable_dirty_bitmap(state->bitmap);
2126 }
2127 }
2128
2129 static void block_dirty_bitmap_disable_prepare(BlkActionState *common,
2130 Error **errp)
2131 {
2132 BlockDirtyBitmap *action;
2133 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2134 common, common);
2135
2136 if (action_check_completion_mode(common, errp) < 0) {
2137 return;
2138 }
2139
2140 action = common->action->u.block_dirty_bitmap_disable.data;
2141 state->bitmap = block_dirty_bitmap_lookup(action->node,
2142 action->name,
2143 NULL,
2144 errp);
2145 if (!state->bitmap) {
2146 return;
2147 }
2148
2149 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2150 return;
2151 }
2152
2153 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2154 bdrv_disable_dirty_bitmap(state->bitmap);
2155 }
2156
2157 static void block_dirty_bitmap_disable_abort(BlkActionState *common)
2158 {
2159 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2160 common, common);
2161
2162 if (state->was_enabled) {
2163 bdrv_enable_dirty_bitmap(state->bitmap);
2164 }
2165 }
2166
2167 static void block_dirty_bitmap_merge_prepare(BlkActionState *common,
2168 Error **errp)
2169 {
2170 BlockDirtyBitmapMerge *action;
2171 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2172 common, common);
2173
2174 if (action_check_completion_mode(common, errp) < 0) {
2175 return;
2176 }
2177
2178 action = common->action->u.block_dirty_bitmap_merge.data;
2179
2180 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2181 action->bitmaps, &state->backup,
2182 errp);
2183 }
2184
2185 static void block_dirty_bitmap_remove_prepare(BlkActionState *common,
2186 Error **errp)
2187 {
2188 BlockDirtyBitmap *action;
2189 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2190 common, common);
2191
2192 if (action_check_completion_mode(common, errp) < 0) {
2193 return;
2194 }
2195
2196 action = common->action->u.block_dirty_bitmap_remove.data;
2197
2198 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2199 false, &state->bs, errp);
2200 if (state->bitmap) {
2201 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2202 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2203 }
2204 }
2205
2206 static void block_dirty_bitmap_remove_abort(BlkActionState *common)
2207 {
2208 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2209 common, common);
2210
2211 if (state->bitmap) {
2212 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2213 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2214 }
2215 }
2216
2217 static void block_dirty_bitmap_remove_commit(BlkActionState *common)
2218 {
2219 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2220 common, common);
2221
2222 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2223 bdrv_release_dirty_bitmap(state->bitmap);
2224 }
2225
2226 static void abort_prepare(BlkActionState *common, Error **errp)
2227 {
2228 error_setg(errp, "Transaction aborted using Abort action");
2229 }
2230
2231 static void abort_commit(BlkActionState *common)
2232 {
2233 g_assert_not_reached(); /* this action never succeeds */
2234 }
2235
2236 static const BlkActionOps actions[] = {
2237 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2238 .instance_size = sizeof(ExternalSnapshotState),
2239 .prepare = external_snapshot_prepare,
2240 .commit = external_snapshot_commit,
2241 .abort = external_snapshot_abort,
2242 .clean = external_snapshot_clean,
2243 },
2244 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2245 .instance_size = sizeof(ExternalSnapshotState),
2246 .prepare = external_snapshot_prepare,
2247 .commit = external_snapshot_commit,
2248 .abort = external_snapshot_abort,
2249 .clean = external_snapshot_clean,
2250 },
2251 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2252 .instance_size = sizeof(DriveBackupState),
2253 .prepare = drive_backup_prepare,
2254 .commit = drive_backup_commit,
2255 .abort = drive_backup_abort,
2256 .clean = drive_backup_clean,
2257 },
2258 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2259 .instance_size = sizeof(BlockdevBackupState),
2260 .prepare = blockdev_backup_prepare,
2261 .commit = blockdev_backup_commit,
2262 .abort = blockdev_backup_abort,
2263 .clean = blockdev_backup_clean,
2264 },
2265 [TRANSACTION_ACTION_KIND_ABORT] = {
2266 .instance_size = sizeof(BlkActionState),
2267 .prepare = abort_prepare,
2268 .commit = abort_commit,
2269 },
2270 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2271 .instance_size = sizeof(InternalSnapshotState),
2272 .prepare = internal_snapshot_prepare,
2273 .abort = internal_snapshot_abort,
2274 .clean = internal_snapshot_clean,
2275 },
2276 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2277 .instance_size = sizeof(BlockDirtyBitmapState),
2278 .prepare = block_dirty_bitmap_add_prepare,
2279 .abort = block_dirty_bitmap_add_abort,
2280 },
2281 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2282 .instance_size = sizeof(BlockDirtyBitmapState),
2283 .prepare = block_dirty_bitmap_clear_prepare,
2284 .commit = block_dirty_bitmap_free_backup,
2285 .abort = block_dirty_bitmap_restore,
2286 },
2287 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2288 .instance_size = sizeof(BlockDirtyBitmapState),
2289 .prepare = block_dirty_bitmap_enable_prepare,
2290 .abort = block_dirty_bitmap_enable_abort,
2291 },
2292 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2293 .instance_size = sizeof(BlockDirtyBitmapState),
2294 .prepare = block_dirty_bitmap_disable_prepare,
2295 .abort = block_dirty_bitmap_disable_abort,
2296 },
2297 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2298 .instance_size = sizeof(BlockDirtyBitmapState),
2299 .prepare = block_dirty_bitmap_merge_prepare,
2300 .commit = block_dirty_bitmap_free_backup,
2301 .abort = block_dirty_bitmap_restore,
2302 },
2303 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE] = {
2304 .instance_size = sizeof(BlockDirtyBitmapState),
2305 .prepare = block_dirty_bitmap_remove_prepare,
2306 .commit = block_dirty_bitmap_remove_commit,
2307 .abort = block_dirty_bitmap_remove_abort,
2308 },
2309 /* Where are transactions for MIRROR, COMMIT and STREAM?
2310 * Although these blockjobs use transaction callbacks like the backup job,
2311 * these jobs do not necessarily adhere to transaction semantics.
2312 * These jobs may not fully undo all of their actions on abort, nor do they
2313 * necessarily work in transactions with more than one job in them.
2314 */
2315 };
2316
2317 /**
2318 * Allocate a TransactionProperties structure if necessary, and fill
2319 * that structure with desired defaults if they are unset.
2320 */
2321 static TransactionProperties *get_transaction_properties(
2322 TransactionProperties *props)
2323 {
2324 if (!props) {
2325 props = g_new0(TransactionProperties, 1);
2326 }
2327
2328 if (!props->has_completion_mode) {
2329 props->has_completion_mode = true;
2330 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2331 }
2332
2333 return props;
2334 }
2335
2336 /*
2337 * 'Atomic' group operations. The operations are performed as a set, and if
2338 * any fail then we roll back all operations in the group.
2339 */
2340 void qmp_transaction(TransactionActionList *dev_list,
2341 bool has_props,
2342 struct TransactionProperties *props,
2343 Error **errp)
2344 {
2345 TransactionActionList *dev_entry = dev_list;
2346 JobTxn *block_job_txn = NULL;
2347 BlkActionState *state, *next;
2348 Error *local_err = NULL;
2349
2350 QTAILQ_HEAD(, BlkActionState) snap_bdrv_states;
2351 QTAILQ_INIT(&snap_bdrv_states);
2352
2353 /* Does this transaction get canceled as a group on failure?
2354 * If not, we don't really need to make a JobTxn.
2355 */
2356 props = get_transaction_properties(props);
2357 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2358 block_job_txn = job_txn_new();
2359 }
2360
2361 /* drain all i/o before any operations */
2362 bdrv_drain_all();
2363
2364 /* We don't do anything in this loop that commits us to the operations */
2365 while (NULL != dev_entry) {
2366 TransactionAction *dev_info = NULL;
2367 const BlkActionOps *ops;
2368
2369 dev_info = dev_entry->value;
2370 dev_entry = dev_entry->next;
2371
2372 assert(dev_info->type < ARRAY_SIZE(actions));
2373
2374 ops = &actions[dev_info->type];
2375 assert(ops->instance_size > 0);
2376
2377 state = g_malloc0(ops->instance_size);
2378 state->ops = ops;
2379 state->action = dev_info;
2380 state->block_job_txn = block_job_txn;
2381 state->txn_props = props;
2382 QTAILQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2383
2384 state->ops->prepare(state, &local_err);
2385 if (local_err) {
2386 error_propagate(errp, local_err);
2387 goto delete_and_fail;
2388 }
2389 }
2390
2391 QTAILQ_FOREACH(state, &snap_bdrv_states, entry) {
2392 if (state->ops->commit) {
2393 state->ops->commit(state);
2394 }
2395 }
2396
2397 /* success */
2398 goto exit;
2399
2400 delete_and_fail:
2401 /* failure, and it is all-or-none; roll back all operations */
2402 QTAILQ_FOREACH_REVERSE(state, &snap_bdrv_states, entry) {
2403 if (state->ops->abort) {
2404 state->ops->abort(state);
2405 }
2406 }
2407 exit:
2408 QTAILQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2409 if (state->ops->clean) {
2410 state->ops->clean(state);
2411 }
2412 g_free(state);
2413 }
2414 if (!has_props) {
2415 qapi_free_TransactionProperties(props);
2416 }
2417 job_txn_unref(block_job_txn);
2418 }
2419
2420 void qmp_block_passwd(bool has_device, const char *device,
2421 bool has_node_name, const char *node_name,
2422 const char *password, Error **errp)
2423 {
2424 error_setg(errp,
2425 "Setting block passwords directly is no longer supported");
2426 }
2427
2428 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2429 const char *name,
2430 Error **errp)
2431 {
2432 BdrvDirtyBitmap *bitmap;
2433 BlockDriverState *bs;
2434 BlockDirtyBitmapSha256 *ret = NULL;
2435 char *sha256;
2436
2437 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2438 if (!bitmap || !bs) {
2439 return NULL;
2440 }
2441
2442 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2443 if (sha256 == NULL) {
2444 return NULL;
2445 }
2446
2447 ret = g_new(BlockDirtyBitmapSha256, 1);
2448 ret->sha256 = sha256;
2449
2450 return ret;
2451 }
2452
2453 void coroutine_fn qmp_block_resize(bool has_device, const char *device,
2454 bool has_node_name, const char *node_name,
2455 int64_t size, Error **errp)
2456 {
2457 Error *local_err = NULL;
2458 BlockBackend *blk;
2459 BlockDriverState *bs;
2460 AioContext *old_ctx;
2461
2462 bs = bdrv_lookup_bs(has_device ? device : NULL,
2463 has_node_name ? node_name : NULL,
2464 &local_err);
2465 if (local_err) {
2466 error_propagate(errp, local_err);
2467 return;
2468 }
2469
2470 if (size < 0) {
2471 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2472 return;
2473 }
2474
2475 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2476 error_setg(errp, QERR_DEVICE_IN_USE, device);
2477 return;
2478 }
2479
2480 blk = blk_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2481 if (!blk) {
2482 return;
2483 }
2484
2485 bdrv_co_lock(bs);
2486 bdrv_drained_begin(bs);
2487 bdrv_co_unlock(bs);
2488
2489 old_ctx = bdrv_co_enter(bs);
2490 blk_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2491 bdrv_co_leave(bs, old_ctx);
2492
2493 bdrv_co_lock(bs);
2494 bdrv_drained_end(bs);
2495 blk_unref(blk);
2496 bdrv_co_unlock(bs);
2497 }
2498
2499 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
2500 bool has_base, const char *base,
2501 bool has_base_node, const char *base_node,
2502 bool has_backing_file, const char *backing_file,
2503 bool has_bottom, const char *bottom,
2504 bool has_speed, int64_t speed,
2505 bool has_on_error, BlockdevOnError on_error,
2506 bool has_filter_node_name, const char *filter_node_name,
2507 bool has_auto_finalize, bool auto_finalize,
2508 bool has_auto_dismiss, bool auto_dismiss,
2509 Error **errp)
2510 {
2511 BlockDriverState *bs, *iter, *iter_end;
2512 BlockDriverState *base_bs = NULL;
2513 BlockDriverState *bottom_bs = NULL;
2514 AioContext *aio_context;
2515 Error *local_err = NULL;
2516 int job_flags = JOB_DEFAULT;
2517
2518 if (has_base && has_base_node) {
2519 error_setg(errp, "'base' and 'base-node' cannot be specified "
2520 "at the same time");
2521 return;
2522 }
2523
2524 if (has_base && has_bottom) {
2525 error_setg(errp, "'base' and 'bottom' cannot be specified "
2526 "at the same time");
2527 return;
2528 }
2529
2530 if (has_bottom && has_base_node) {
2531 error_setg(errp, "'bottom' and 'base-node' cannot be specified "
2532 "at the same time");
2533 return;
2534 }
2535
2536 if (!has_on_error) {
2537 on_error = BLOCKDEV_ON_ERROR_REPORT;
2538 }
2539
2540 bs = bdrv_lookup_bs(device, device, errp);
2541 if (!bs) {
2542 return;
2543 }
2544
2545 aio_context = bdrv_get_aio_context(bs);
2546 aio_context_acquire(aio_context);
2547
2548 if (has_base) {
2549 base_bs = bdrv_find_backing_image(bs, base);
2550 if (base_bs == NULL) {
2551 error_setg(errp, "Can't find '%s' in the backing chain", base);
2552 goto out;
2553 }
2554 assert(bdrv_get_aio_context(base_bs) == aio_context);
2555 }
2556
2557 if (has_base_node) {
2558 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2559 if (!base_bs) {
2560 goto out;
2561 }
2562 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2563 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2564 base_node, device);
2565 goto out;
2566 }
2567 assert(bdrv_get_aio_context(base_bs) == aio_context);
2568 bdrv_refresh_filename(base_bs);
2569 }
2570
2571 if (has_bottom) {
2572 bottom_bs = bdrv_lookup_bs(NULL, bottom, errp);
2573 if (!bottom_bs) {
2574 goto out;
2575 }
2576 if (!bottom_bs->drv) {
2577 error_setg(errp, "Node '%s' is not open", bottom);
2578 goto out;
2579 }
2580 if (bottom_bs->drv->is_filter) {
2581 error_setg(errp, "Node '%s' is a filter, use a non-filter node "
2582 "as 'bottom'", bottom);
2583 goto out;
2584 }
2585 if (!bdrv_chain_contains(bs, bottom_bs)) {
2586 error_setg(errp, "Node '%s' is not in a chain starting from '%s'",
2587 bottom, device);
2588 goto out;
2589 }
2590 assert(bdrv_get_aio_context(bottom_bs) == aio_context);
2591 }
2592
2593 /*
2594 * Check for op blockers in the whole chain between bs and base (or bottom)
2595 */
2596 iter_end = has_bottom ? bdrv_filter_or_cow_bs(bottom_bs) : base_bs;
2597 for (iter = bs; iter && iter != iter_end;
2598 iter = bdrv_filter_or_cow_bs(iter))
2599 {
2600 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2601 goto out;
2602 }
2603 }
2604
2605 /* if we are streaming the entire chain, the result will have no backing
2606 * file, and specifying one is therefore an error */
2607 if (base_bs == NULL && has_backing_file) {
2608 error_setg(errp, "backing file specified, but streaming the "
2609 "entire chain");
2610 goto out;
2611 }
2612
2613 if (has_auto_finalize && !auto_finalize) {
2614 job_flags |= JOB_MANUAL_FINALIZE;
2615 }
2616 if (has_auto_dismiss && !auto_dismiss) {
2617 job_flags |= JOB_MANUAL_DISMISS;
2618 }
2619
2620 stream_start(has_job_id ? job_id : NULL, bs, base_bs, backing_file,
2621 bottom_bs, job_flags, has_speed ? speed : 0, on_error,
2622 filter_node_name, &local_err);
2623 if (local_err) {
2624 error_propagate(errp, local_err);
2625 goto out;
2626 }
2627
2628 trace_qmp_block_stream(bs);
2629
2630 out:
2631 aio_context_release(aio_context);
2632 }
2633
2634 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
2635 bool has_base_node, const char *base_node,
2636 bool has_base, const char *base,
2637 bool has_top_node, const char *top_node,
2638 bool has_top, const char *top,
2639 bool has_backing_file, const char *backing_file,
2640 bool has_speed, int64_t speed,
2641 bool has_on_error, BlockdevOnError on_error,
2642 bool has_filter_node_name, const char *filter_node_name,
2643 bool has_auto_finalize, bool auto_finalize,
2644 bool has_auto_dismiss, bool auto_dismiss,
2645 Error **errp)
2646 {
2647 BlockDriverState *bs;
2648 BlockDriverState *iter;
2649 BlockDriverState *base_bs, *top_bs;
2650 AioContext *aio_context;
2651 Error *local_err = NULL;
2652 int job_flags = JOB_DEFAULT;
2653 uint64_t top_perm, top_shared;
2654
2655 if (!has_speed) {
2656 speed = 0;
2657 }
2658 if (!has_on_error) {
2659 on_error = BLOCKDEV_ON_ERROR_REPORT;
2660 }
2661 if (!has_filter_node_name) {
2662 filter_node_name = NULL;
2663 }
2664 if (has_auto_finalize && !auto_finalize) {
2665 job_flags |= JOB_MANUAL_FINALIZE;
2666 }
2667 if (has_auto_dismiss && !auto_dismiss) {
2668 job_flags |= JOB_MANUAL_DISMISS;
2669 }
2670
2671 /* Important Note:
2672 * libvirt relies on the DeviceNotFound error class in order to probe for
2673 * live commit feature versions; for this to work, we must make sure to
2674 * perform the device lookup before any generic errors that may occur in a
2675 * scenario in which all optional arguments are omitted. */
2676 bs = qmp_get_root_bs(device, &local_err);
2677 if (!bs) {
2678 bs = bdrv_lookup_bs(device, device, NULL);
2679 if (!bs) {
2680 error_free(local_err);
2681 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2682 "Device '%s' not found", device);
2683 } else {
2684 error_propagate(errp, local_err);
2685 }
2686 return;
2687 }
2688
2689 aio_context = bdrv_get_aio_context(bs);
2690 aio_context_acquire(aio_context);
2691
2692 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2693 goto out;
2694 }
2695
2696 /* default top_bs is the active layer */
2697 top_bs = bs;
2698
2699 if (has_top_node && has_top) {
2700 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2701 goto out;
2702 } else if (has_top_node) {
2703 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2704 if (top_bs == NULL) {
2705 goto out;
2706 }
2707 if (!bdrv_chain_contains(bs, top_bs)) {
2708 error_setg(errp, "'%s' is not in this backing file chain",
2709 top_node);
2710 goto out;
2711 }
2712 } else if (has_top && top) {
2713 /* This strcmp() is just a shortcut, there is no need to
2714 * refresh @bs's filename. If it mismatches,
2715 * bdrv_find_backing_image() will do the refresh and may still
2716 * return @bs. */
2717 if (strcmp(bs->filename, top) != 0) {
2718 top_bs = bdrv_find_backing_image(bs, top);
2719 }
2720 }
2721
2722 if (top_bs == NULL) {
2723 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2724 goto out;
2725 }
2726
2727 assert(bdrv_get_aio_context(top_bs) == aio_context);
2728
2729 if (has_base_node && has_base) {
2730 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2731 goto out;
2732 } else if (has_base_node) {
2733 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2734 if (base_bs == NULL) {
2735 goto out;
2736 }
2737 if (!bdrv_chain_contains(top_bs, base_bs)) {
2738 error_setg(errp, "'%s' is not in this backing file chain",
2739 base_node);
2740 goto out;
2741 }
2742 } else if (has_base && base) {
2743 base_bs = bdrv_find_backing_image(top_bs, base);
2744 if (base_bs == NULL) {
2745 error_setg(errp, "Can't find '%s' in the backing chain", base);
2746 goto out;
2747 }
2748 } else {
2749 base_bs = bdrv_find_base(top_bs);
2750 if (base_bs == NULL) {
2751 error_setg(errp, "There is no backimg image");
2752 goto out;
2753 }
2754 }
2755
2756 assert(bdrv_get_aio_context(base_bs) == aio_context);
2757
2758 for (iter = top_bs; iter != bdrv_filter_or_cow_bs(base_bs);
2759 iter = bdrv_filter_or_cow_bs(iter))
2760 {
2761 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2762 goto out;
2763 }
2764 }
2765
2766 /* Do not allow attempts to commit an image into itself */
2767 if (top_bs == base_bs) {
2768 error_setg(errp, "cannot commit an image into itself");
2769 goto out;
2770 }
2771
2772 /*
2773 * Active commit is required if and only if someone has taken a
2774 * WRITE permission on the top node. Historically, we have always
2775 * used active commit for top nodes, so continue that practice
2776 * lest we possibly break clients that rely on this behavior, e.g.
2777 * to later attach this node to a writing parent.
2778 * (Active commit is never really wrong.)
2779 */
2780 bdrv_get_cumulative_perm(top_bs, &top_perm, &top_shared);
2781 if (top_perm & BLK_PERM_WRITE ||
2782 bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs))
2783 {
2784 if (has_backing_file) {
2785 if (bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs)) {
2786 error_setg(errp, "'backing-file' specified,"
2787 " but 'top' is the active layer");
2788 } else {
2789 error_setg(errp, "'backing-file' specified, but 'top' has a "
2790 "writer on it");
2791 }
2792 goto out;
2793 }
2794 if (!has_job_id) {
2795 /*
2796 * Emulate here what block_job_create() does, because it
2797 * is possible that @bs != @top_bs (the block job should
2798 * be named after @bs, even if @top_bs is the actual
2799 * source)
2800 */
2801 job_id = bdrv_get_device_name(bs);
2802 }
2803 commit_active_start(job_id, top_bs, base_bs, job_flags, speed, on_error,
2804 filter_node_name, NULL, NULL, false, &local_err);
2805 } else {
2806 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2807 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2808 goto out;
2809 }
2810 commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, job_flags,
2811 speed, on_error, has_backing_file ? backing_file : NULL,
2812 filter_node_name, &local_err);
2813 }
2814 if (local_err != NULL) {
2815 error_propagate(errp, local_err);
2816 goto out;
2817 }
2818
2819 out:
2820 aio_context_release(aio_context);
2821 }
2822
2823 /* Common QMP interface for drive-backup and blockdev-backup */
2824 static BlockJob *do_backup_common(BackupCommon *backup,
2825 BlockDriverState *bs,
2826 BlockDriverState *target_bs,
2827 AioContext *aio_context,
2828 JobTxn *txn, Error **errp)
2829 {
2830 BlockJob *job = NULL;
2831 BdrvDirtyBitmap *bmap = NULL;
2832 BackupPerf perf = { .max_workers = 64 };
2833 int job_flags = JOB_DEFAULT;
2834
2835 if (!backup->has_speed) {
2836 backup->speed = 0;
2837 }
2838 if (!backup->has_on_source_error) {
2839 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2840 }
2841 if (!backup->has_on_target_error) {
2842 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2843 }
2844 if (!backup->has_job_id) {
2845 backup->job_id = NULL;
2846 }
2847 if (!backup->has_auto_finalize) {
2848 backup->auto_finalize = true;
2849 }
2850 if (!backup->has_auto_dismiss) {
2851 backup->auto_dismiss = true;
2852 }
2853 if (!backup->has_compress) {
2854 backup->compress = false;
2855 }
2856
2857 if (backup->x_perf) {
2858 if (backup->x_perf->has_use_copy_range) {
2859 perf.use_copy_range = backup->x_perf->use_copy_range;
2860 }
2861 if (backup->x_perf->has_max_workers) {
2862 perf.max_workers = backup->x_perf->max_workers;
2863 }
2864 if (backup->x_perf->has_max_chunk) {
2865 perf.max_chunk = backup->x_perf->max_chunk;
2866 }
2867 }
2868
2869 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2870 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2871 /* done before desugaring 'incremental' to print the right message */
2872 if (!backup->has_bitmap) {
2873 error_setg(errp, "must provide a valid bitmap name for "
2874 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2875 return NULL;
2876 }
2877 }
2878
2879 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2880 if (backup->has_bitmap_mode &&
2881 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2882 error_setg(errp, "Bitmap sync mode must be '%s' "
2883 "when using sync mode '%s'",
2884 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2885 MirrorSyncMode_str(backup->sync));
2886 return NULL;
2887 }
2888 backup->has_bitmap_mode = true;
2889 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2890 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2891 }
2892
2893 if (backup->has_bitmap) {
2894 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2895 if (!bmap) {
2896 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2897 return NULL;
2898 }
2899 if (!backup->has_bitmap_mode) {
2900 error_setg(errp, "Bitmap sync mode must be given "
2901 "when providing a bitmap");
2902 return NULL;
2903 }
2904 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2905 return NULL;
2906 }
2907
2908 /* This does not produce a useful bitmap artifact: */
2909 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2910 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2911 " outputs", MirrorSyncMode_str(backup->sync));
2912 return NULL;
2913 }
2914
2915 /* If the bitmap isn't used for input or output, this is useless: */
2916 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2917 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2918 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2919 " when combined with sync mode '%s'",
2920 BitmapSyncMode_str(backup->bitmap_mode),
2921 MirrorSyncMode_str(backup->sync));
2922 return NULL;
2923 }
2924 }
2925
2926 if (!backup->has_bitmap && backup->has_bitmap_mode) {
2927 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2928 return NULL;
2929 }
2930
2931 if (!backup->auto_finalize) {
2932 job_flags |= JOB_MANUAL_FINALIZE;
2933 }
2934 if (!backup->auto_dismiss) {
2935 job_flags |= JOB_MANUAL_DISMISS;
2936 }
2937
2938 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2939 backup->sync, bmap, backup->bitmap_mode,
2940 backup->compress,
2941 backup->filter_node_name,
2942 &perf,
2943 backup->on_source_error,
2944 backup->on_target_error,
2945 job_flags, NULL, NULL, txn, errp);
2946 return job;
2947 }
2948
2949 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2950 {
2951 TransactionAction action = {
2952 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2953 .u.drive_backup.data = backup,
2954 };
2955 blockdev_do_action(&action, errp);
2956 }
2957
2958 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2959 bool flat,
2960 Error **errp)
2961 {
2962 bool return_flat = has_flat && flat;
2963
2964 return bdrv_named_nodes_list(return_flat, errp);
2965 }
2966
2967 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2968 {
2969 return bdrv_get_xdbg_block_graph(errp);
2970 }
2971
2972 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2973 {
2974 TransactionAction action = {
2975 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2976 .u.blockdev_backup.data = backup,
2977 };
2978 blockdev_do_action(&action, errp);
2979 }
2980
2981 /* Parameter check and block job starting for drive mirroring.
2982 * Caller should hold @device and @target's aio context (must be the same).
2983 **/
2984 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2985 BlockDriverState *target,
2986 bool has_replaces, const char *replaces,
2987 enum MirrorSyncMode sync,
2988 BlockMirrorBackingMode backing_mode,
2989 bool zero_target,
2990 bool has_speed, int64_t speed,
2991 bool has_granularity, uint32_t granularity,
2992 bool has_buf_size, int64_t buf_size,
2993 bool has_on_source_error,
2994 BlockdevOnError on_source_error,
2995 bool has_on_target_error,
2996 BlockdevOnError on_target_error,
2997 bool has_unmap, bool unmap,
2998 bool has_filter_node_name,
2999 const char *filter_node_name,
3000 bool has_copy_mode, MirrorCopyMode copy_mode,
3001 bool has_auto_finalize, bool auto_finalize,
3002 bool has_auto_dismiss, bool auto_dismiss,
3003 Error **errp)
3004 {
3005 BlockDriverState *unfiltered_bs;
3006 int job_flags = JOB_DEFAULT;
3007
3008 if (!has_speed) {
3009 speed = 0;
3010 }
3011 if (!has_on_source_error) {
3012 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3013 }
3014 if (!has_on_target_error) {
3015 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3016 }
3017 if (!has_granularity) {
3018 granularity = 0;
3019 }
3020 if (!has_buf_size) {
3021 buf_size = 0;
3022 }
3023 if (!has_unmap) {
3024 unmap = true;
3025 }
3026 if (!has_filter_node_name) {
3027 filter_node_name = NULL;
3028 }
3029 if (!has_copy_mode) {
3030 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
3031 }
3032 if (has_auto_finalize && !auto_finalize) {
3033 job_flags |= JOB_MANUAL_FINALIZE;
3034 }
3035 if (has_auto_dismiss && !auto_dismiss) {
3036 job_flags |= JOB_MANUAL_DISMISS;
3037 }
3038
3039 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3040 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3041 "a value in range [512B, 64MB]");
3042 return;
3043 }
3044 if (granularity & (granularity - 1)) {
3045 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3046 "a power of 2");
3047 return;
3048 }
3049
3050 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3051 return;
3052 }
3053 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3054 return;
3055 }
3056
3057 if (!bdrv_backing_chain_next(bs) && sync == MIRROR_SYNC_MODE_TOP) {
3058 sync = MIRROR_SYNC_MODE_FULL;
3059 }
3060
3061 if (!has_replaces) {
3062 /* We want to mirror from @bs, but keep implicit filters on top */
3063 unfiltered_bs = bdrv_skip_implicit_filters(bs);
3064 if (unfiltered_bs != bs) {
3065 replaces = unfiltered_bs->node_name;
3066 has_replaces = true;
3067 }
3068 }
3069
3070 if (has_replaces) {
3071 BlockDriverState *to_replace_bs;
3072 AioContext *replace_aio_context;
3073 int64_t bs_size, replace_size;
3074
3075 bs_size = bdrv_getlength(bs);
3076 if (bs_size < 0) {
3077 error_setg_errno(errp, -bs_size, "Failed to query device's size");
3078 return;
3079 }
3080
3081 to_replace_bs = check_to_replace_node(bs, replaces, errp);
3082 if (!to_replace_bs) {
3083 return;
3084 }
3085
3086 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3087 aio_context_acquire(replace_aio_context);
3088 replace_size = bdrv_getlength(to_replace_bs);
3089 aio_context_release(replace_aio_context);
3090
3091 if (replace_size < 0) {
3092 error_setg_errno(errp, -replace_size,
3093 "Failed to query the replacement node's size");
3094 return;
3095 }
3096 if (bs_size != replace_size) {
3097 error_setg(errp, "cannot replace image with a mirror image of "
3098 "different size");
3099 return;
3100 }
3101 }
3102
3103 /* pass the node name to replace to mirror start since it's loose coupling
3104 * and will allow to check whether the node still exist at mirror completion
3105 */
3106 mirror_start(job_id, bs, target,
3107 has_replaces ? replaces : NULL, job_flags,
3108 speed, granularity, buf_size, sync, backing_mode, zero_target,
3109 on_source_error, on_target_error, unmap, filter_node_name,
3110 copy_mode, errp);
3111 }
3112
3113 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3114 {
3115 BlockDriverState *bs;
3116 BlockDriverState *target_backing_bs, *target_bs;
3117 AioContext *aio_context;
3118 AioContext *old_context;
3119 BlockMirrorBackingMode backing_mode;
3120 Error *local_err = NULL;
3121 QDict *options = NULL;
3122 int flags;
3123 int64_t size;
3124 const char *format = arg->format;
3125 bool zero_target;
3126 int ret;
3127
3128 bs = qmp_get_root_bs(arg->device, errp);
3129 if (!bs) {
3130 return;
3131 }
3132
3133 /* Early check to avoid creating target */
3134 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3135 return;
3136 }
3137
3138 aio_context = bdrv_get_aio_context(bs);
3139 aio_context_acquire(aio_context);
3140
3141 if (!arg->has_mode) {
3142 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3143 }
3144
3145 if (!arg->has_format) {
3146 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3147 ? NULL : bs->drv->format_name);
3148 }
3149
3150 flags = bs->open_flags | BDRV_O_RDWR;
3151 target_backing_bs = bdrv_cow_bs(bdrv_skip_filters(bs));
3152 if (!target_backing_bs && arg->sync == MIRROR_SYNC_MODE_TOP) {
3153 arg->sync = MIRROR_SYNC_MODE_FULL;
3154 }
3155 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3156 target_backing_bs = bs;
3157 }
3158
3159 size = bdrv_getlength(bs);
3160 if (size < 0) {
3161 error_setg_errno(errp, -size, "bdrv_getlength failed");
3162 goto out;
3163 }
3164
3165 if (arg->has_replaces) {
3166 if (!arg->has_node_name) {
3167 error_setg(errp, "a node-name must be provided when replacing a"
3168 " named node of the graph");
3169 goto out;
3170 }
3171 }
3172
3173 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3174 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3175 } else {
3176 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3177 }
3178
3179 /* Don't open backing image in create() */
3180 flags |= BDRV_O_NO_BACKING;
3181
3182 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !target_backing_bs)
3183 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3184 {
3185 /* create new image w/o backing file */
3186 assert(format);
3187 bdrv_img_create(arg->target, format,
3188 NULL, NULL, NULL, size, flags, false, &local_err);
3189 } else {
3190 /* Implicit filters should not appear in the filename */
3191 BlockDriverState *explicit_backing =
3192 bdrv_skip_implicit_filters(target_backing_bs);
3193
3194 switch (arg->mode) {
3195 case NEW_IMAGE_MODE_EXISTING:
3196 break;
3197 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3198 /* create new image with backing file */
3199 bdrv_refresh_filename(explicit_backing);
3200 bdrv_img_create(arg->target, format,
3201 explicit_backing->filename,
3202 explicit_backing->drv->format_name,
3203 NULL, size, flags, false, &local_err);
3204 break;
3205 default:
3206 abort();
3207 }
3208 }
3209
3210 if (local_err) {
3211 error_propagate(errp, local_err);
3212 goto out;
3213 }
3214
3215 options = qdict_new();
3216 if (arg->has_node_name) {
3217 qdict_put_str(options, "node-name", arg->node_name);
3218 }
3219 if (format) {
3220 qdict_put_str(options, "driver", format);
3221 }
3222
3223 /* Mirroring takes care of copy-on-write using the source's backing
3224 * file.
3225 */
3226 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3227 if (!target_bs) {
3228 goto out;
3229 }
3230
3231 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3232 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3233 !bdrv_has_zero_init(target_bs)));
3234
3235
3236 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3237 old_context = bdrv_get_aio_context(target_bs);
3238 aio_context_release(aio_context);
3239 aio_context_acquire(old_context);
3240
3241 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3242 if (ret < 0) {
3243 bdrv_unref(target_bs);
3244 aio_context_release(old_context);
3245 return;
3246 }
3247
3248 aio_context_release(old_context);
3249 aio_context_acquire(aio_context);
3250
3251 blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3252 arg->has_replaces, arg->replaces, arg->sync,
3253 backing_mode, zero_target,
3254 arg->has_speed, arg->speed,
3255 arg->has_granularity, arg->granularity,
3256 arg->has_buf_size, arg->buf_size,
3257 arg->has_on_source_error, arg->on_source_error,
3258 arg->has_on_target_error, arg->on_target_error,
3259 arg->has_unmap, arg->unmap,
3260 false, NULL,
3261 arg->has_copy_mode, arg->copy_mode,
3262 arg->has_auto_finalize, arg->auto_finalize,
3263 arg->has_auto_dismiss, arg->auto_dismiss,
3264 errp);
3265 bdrv_unref(target_bs);
3266 out:
3267 aio_context_release(aio_context);
3268 }
3269
3270 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3271 const char *device, const char *target,
3272 bool has_replaces, const char *replaces,
3273 MirrorSyncMode sync,
3274 bool has_speed, int64_t speed,
3275 bool has_granularity, uint32_t granularity,
3276 bool has_buf_size, int64_t buf_size,
3277 bool has_on_source_error,
3278 BlockdevOnError on_source_error,
3279 bool has_on_target_error,
3280 BlockdevOnError on_target_error,
3281 bool has_filter_node_name,
3282 const char *filter_node_name,
3283 bool has_copy_mode, MirrorCopyMode copy_mode,
3284 bool has_auto_finalize, bool auto_finalize,
3285 bool has_auto_dismiss, bool auto_dismiss,
3286 Error **errp)
3287 {
3288 BlockDriverState *bs;
3289 BlockDriverState *target_bs;
3290 AioContext *aio_context;
3291 AioContext *old_context;
3292 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3293 bool zero_target;
3294 int ret;
3295
3296 bs = qmp_get_root_bs(device, errp);
3297 if (!bs) {
3298 return;
3299 }
3300
3301 target_bs = bdrv_lookup_bs(target, target, errp);
3302 if (!target_bs) {
3303 return;
3304 }
3305
3306 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3307
3308 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3309 old_context = bdrv_get_aio_context(target_bs);
3310 aio_context = bdrv_get_aio_context(bs);
3311 aio_context_acquire(old_context);
3312
3313 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3314
3315 aio_context_release(old_context);
3316 aio_context_acquire(aio_context);
3317
3318 if (ret < 0) {
3319 goto out;
3320 }
3321
3322 blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
3323 has_replaces, replaces, sync, backing_mode,
3324 zero_target, has_speed, speed,
3325 has_granularity, granularity,
3326 has_buf_size, buf_size,
3327 has_on_source_error, on_source_error,
3328 has_on_target_error, on_target_error,
3329 true, true,
3330 has_filter_node_name, filter_node_name,
3331 has_copy_mode, copy_mode,
3332 has_auto_finalize, auto_finalize,
3333 has_auto_dismiss, auto_dismiss,
3334 errp);
3335 out:
3336 aio_context_release(aio_context);
3337 }
3338
3339 /* Get a block job using its ID and acquire its AioContext */
3340 static BlockJob *find_block_job(const char *id, AioContext **aio_context,
3341 Error **errp)
3342 {
3343 BlockJob *job;
3344
3345 assert(id != NULL);
3346
3347 *aio_context = NULL;
3348
3349 job = block_job_get(id);
3350
3351 if (!job) {
3352 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3353 "Block job '%s' not found", id);
3354 return NULL;
3355 }
3356
3357 *aio_context = blk_get_aio_context(job->blk);
3358 aio_context_acquire(*aio_context);
3359
3360 return job;
3361 }
3362
3363 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3364 {
3365 AioContext *aio_context;
3366 BlockJob *job = find_block_job(device, &aio_context, errp);
3367
3368 if (!job) {
3369 return;
3370 }
3371
3372 block_job_set_speed(job, speed, errp);
3373 aio_context_release(aio_context);
3374 }
3375
3376 void qmp_block_job_cancel(const char *device,
3377 bool has_force, bool force, Error **errp)
3378 {
3379 AioContext *aio_context;
3380 BlockJob *job = find_block_job(device, &aio_context, errp);
3381
3382 if (!job) {
3383 return;
3384 }
3385
3386 if (!has_force) {
3387 force = false;
3388 }
3389
3390 if (job_user_paused(&job->job) && !force) {
3391 error_setg(errp, "The block job for device '%s' is currently paused",
3392 device);
3393 goto out;
3394 }
3395
3396 trace_qmp_block_job_cancel(job);
3397 job_user_cancel(&job->job, force, errp);
3398 out:
3399 aio_context_release(aio_context);
3400 }
3401
3402 void qmp_block_job_pause(const char *device, Error **errp)
3403 {
3404 AioContext *aio_context;
3405 BlockJob *job = find_block_job(device, &aio_context, errp);
3406
3407 if (!job) {
3408 return;
3409 }
3410
3411 trace_qmp_block_job_pause(job);
3412 job_user_pause(&job->job, errp);
3413 aio_context_release(aio_context);
3414 }
3415
3416 void qmp_block_job_resume(const char *device, Error **errp)
3417 {
3418 AioContext *aio_context;
3419 BlockJob *job = find_block_job(device, &aio_context, errp);
3420
3421 if (!job) {
3422 return;
3423 }
3424
3425 trace_qmp_block_job_resume(job);
3426 job_user_resume(&job->job, errp);
3427 aio_context_release(aio_context);
3428 }
3429
3430 void qmp_block_job_complete(const char *device, Error **errp)
3431 {
3432 AioContext *aio_context;
3433 BlockJob *job = find_block_job(device, &aio_context, errp);
3434
3435 if (!job) {
3436 return;
3437 }
3438
3439 trace_qmp_block_job_complete(job);
3440 job_complete(&job->job, errp);
3441 aio_context_release(aio_context);
3442 }
3443
3444 void qmp_block_job_finalize(const char *id, Error **errp)
3445 {
3446 AioContext *aio_context;
3447 BlockJob *job = find_block_job(id, &aio_context, errp);
3448
3449 if (!job) {
3450 return;
3451 }
3452
3453 trace_qmp_block_job_finalize(job);
3454 job_ref(&job->job);
3455 job_finalize(&job->job, errp);
3456
3457 /*
3458 * Job's context might have changed via job_finalize (and job_txn_apply
3459 * automatically acquires the new one), so make sure we release the correct
3460 * one.
3461 */
3462 aio_context = blk_get_aio_context(job->blk);
3463 job_unref(&job->job);
3464 aio_context_release(aio_context);
3465 }
3466
3467 void qmp_block_job_dismiss(const char *id, Error **errp)
3468 {
3469 AioContext *aio_context;
3470 BlockJob *bjob = find_block_job(id, &aio_context, errp);
3471 Job *job;
3472
3473 if (!bjob) {
3474 return;
3475 }
3476
3477 trace_qmp_block_job_dismiss(bjob);
3478 job = &bjob->job;
3479 job_dismiss(&job, errp);
3480 aio_context_release(aio_context);
3481 }
3482
3483 void qmp_change_backing_file(const char *device,
3484 const char *image_node_name,
3485 const char *backing_file,
3486 Error **errp)
3487 {
3488 BlockDriverState *bs = NULL;
3489 AioContext *aio_context;
3490 BlockDriverState *image_bs = NULL;
3491 Error *local_err = NULL;
3492 bool ro;
3493 int ret;
3494
3495 bs = qmp_get_root_bs(device, errp);
3496 if (!bs) {
3497 return;
3498 }
3499
3500 aio_context = bdrv_get_aio_context(bs);
3501 aio_context_acquire(aio_context);
3502
3503 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3504 if (local_err) {
3505 error_propagate(errp, local_err);
3506 goto out;
3507 }
3508
3509 if (!image_bs) {
3510 error_setg(errp, "image file not found");
3511 goto out;
3512 }
3513
3514 if (bdrv_find_base(image_bs) == image_bs) {
3515 error_setg(errp, "not allowing backing file change on an image "
3516 "without a backing file");
3517 goto out;
3518 }
3519
3520 /* even though we are not necessarily operating on bs, we need it to
3521 * determine if block ops are currently prohibited on the chain */
3522 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3523 goto out;
3524 }
3525
3526 /* final sanity check */
3527 if (!bdrv_chain_contains(bs, image_bs)) {
3528 error_setg(errp, "'%s' and image file are not in the same chain",
3529 device);
3530 goto out;
3531 }
3532
3533 /* if not r/w, reopen to make r/w */
3534 ro = bdrv_is_read_only(image_bs);
3535
3536 if (ro) {
3537 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3538 goto out;
3539 }
3540 }
3541
3542 ret = bdrv_change_backing_file(image_bs, backing_file,
3543 image_bs->drv ? image_bs->drv->format_name : "",
3544 false);
3545
3546 if (ret < 0) {
3547 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3548 backing_file);
3549 /* don't exit here, so we can try to restore open flags if
3550 * appropriate */
3551 }
3552
3553 if (ro) {
3554 bdrv_reopen_set_read_only(image_bs, true, errp);
3555 }
3556
3557 out:
3558 aio_context_release(aio_context);
3559 }
3560
3561 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3562 {
3563 BlockDriverState *bs;
3564 QObject *obj;
3565 Visitor *v = qobject_output_visitor_new(&obj);
3566 QDict *qdict;
3567
3568 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3569 visit_complete(v, &obj);
3570 qdict = qobject_to(QDict, obj);
3571
3572 qdict_flatten(qdict);
3573
3574 if (!qdict_get_try_str(qdict, "node-name")) {
3575 error_setg(errp, "'node-name' must be specified for the root node");
3576 goto fail;
3577 }
3578
3579 bs = bds_tree_init(qdict, errp);
3580 if (!bs) {
3581 goto fail;
3582 }
3583
3584 bdrv_set_monitor_owned(bs);
3585
3586 fail:
3587 visit_free(v);
3588 }
3589
3590 void qmp_x_blockdev_reopen(BlockdevOptions *options, Error **errp)
3591 {
3592 BlockDriverState *bs;
3593 AioContext *ctx;
3594 QObject *obj;
3595 Visitor *v = qobject_output_visitor_new(&obj);
3596 BlockReopenQueue *queue;
3597 QDict *qdict;
3598
3599 /* Check for the selected node name */
3600 if (!options->has_node_name) {
3601 error_setg(errp, "Node name not specified");
3602 goto fail;
3603 }
3604
3605 bs = bdrv_find_node(options->node_name);
3606 if (!bs) {
3607 error_setg(errp, "Cannot find node named '%s'", options->node_name);
3608 goto fail;
3609 }
3610
3611 /* Put all options in a QDict and flatten it */
3612 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3613 visit_complete(v, &obj);
3614 qdict = qobject_to(QDict, obj);
3615
3616 qdict_flatten(qdict);
3617
3618 /* Perform the reopen operation */
3619 ctx = bdrv_get_aio_context(bs);
3620 aio_context_acquire(ctx);
3621 bdrv_subtree_drained_begin(bs);
3622 queue = bdrv_reopen_queue(NULL, bs, qdict, false);
3623 bdrv_reopen_multiple(queue, errp);
3624 bdrv_subtree_drained_end(bs);
3625 aio_context_release(ctx);
3626
3627 fail:
3628 visit_free(v);
3629 }
3630
3631 void qmp_blockdev_del(const char *node_name, Error **errp)
3632 {
3633 AioContext *aio_context;
3634 BlockDriverState *bs;
3635
3636 bs = bdrv_find_node(node_name);
3637 if (!bs) {
3638 error_setg(errp, "Cannot find node %s", node_name);
3639 return;
3640 }
3641 if (bdrv_has_blk(bs)) {
3642 error_setg(errp, "Node %s is in use", node_name);
3643 return;
3644 }
3645 aio_context = bdrv_get_aio_context(bs);
3646 aio_context_acquire(aio_context);
3647
3648 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3649 goto out;
3650 }
3651
3652 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3653 error_setg(errp, "Node %s is not owned by the monitor",
3654 bs->node_name);
3655 goto out;
3656 }
3657
3658 if (bs->refcnt > 1) {
3659 error_setg(errp, "Block device %s is in use",
3660 bdrv_get_device_or_node_name(bs));
3661 goto out;
3662 }
3663
3664 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3665 bdrv_unref(bs);
3666
3667 out:
3668 aio_context_release(aio_context);
3669 }
3670
3671 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3672 const char *child_name)
3673 {
3674 BdrvChild *child;
3675
3676 QLIST_FOREACH(child, &parent_bs->children, next) {
3677 if (strcmp(child->name, child_name) == 0) {
3678 return child;
3679 }
3680 }
3681
3682 return NULL;
3683 }
3684
3685 void qmp_x_blockdev_change(const char *parent, bool has_child,
3686 const char *child, bool has_node,
3687 const char *node, Error **errp)
3688 {
3689 BlockDriverState *parent_bs, *new_bs = NULL;
3690 BdrvChild *p_child;
3691
3692 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3693 if (!parent_bs) {
3694 return;
3695 }
3696
3697 if (has_child == has_node) {
3698 if (has_child) {
3699 error_setg(errp, "The parameters child and node are in conflict");
3700 } else {
3701 error_setg(errp, "Either child or node must be specified");
3702 }
3703 return;
3704 }
3705
3706 if (has_child) {
3707 p_child = bdrv_find_child(parent_bs, child);
3708 if (!p_child) {
3709 error_setg(errp, "Node '%s' does not have child '%s'",
3710 parent, child);
3711 return;
3712 }
3713 bdrv_del_child(parent_bs, p_child, errp);
3714 }
3715
3716 if (has_node) {
3717 new_bs = bdrv_find_node(node);
3718 if (!new_bs) {
3719 error_setg(errp, "Node '%s' not found", node);
3720 return;
3721 }
3722 bdrv_add_child(parent_bs, new_bs, errp);
3723 }
3724 }
3725
3726 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3727 {
3728 BlockJobInfoList *head = NULL, **tail = &head;
3729 BlockJob *job;
3730
3731 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
3732 BlockJobInfo *value;
3733 AioContext *aio_context;
3734
3735 if (block_job_is_internal(job)) {
3736 continue;
3737 }
3738 aio_context = blk_get_aio_context(job->blk);
3739 aio_context_acquire(aio_context);
3740 value = block_job_query(job, errp);
3741 aio_context_release(aio_context);
3742 if (!value) {
3743 qapi_free_BlockJobInfoList(head);
3744 return NULL;
3745 }
3746 QAPI_LIST_APPEND(tail, value);
3747 }
3748
3749 return head;
3750 }
3751
3752 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3753 bool has_force, bool force, Error **errp)
3754 {
3755 AioContext *old_context;
3756 AioContext *new_context;
3757 BlockDriverState *bs;
3758
3759 bs = bdrv_find_node(node_name);
3760 if (!bs) {
3761 error_setg(errp, "Cannot find node %s", node_name);
3762 return;
3763 }
3764
3765 /* Protects against accidents. */
3766 if (!(has_force && force) && bdrv_has_blk(bs)) {
3767 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3768 "be in use (use force=true to override this check)",
3769 node_name);
3770 return;
3771 }
3772
3773 if (iothread->type == QTYPE_QSTRING) {
3774 IOThread *obj = iothread_by_id(iothread->u.s);
3775 if (!obj) {
3776 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3777 return;
3778 }
3779
3780 new_context = iothread_get_aio_context(obj);
3781 } else {
3782 new_context = qemu_get_aio_context();
3783 }
3784
3785 old_context = bdrv_get_aio_context(bs);
3786 aio_context_acquire(old_context);
3787
3788 bdrv_try_set_aio_context(bs, new_context, errp);
3789
3790 aio_context_release(old_context);
3791 }
3792
3793 QemuOptsList qemu_common_drive_opts = {
3794 .name = "drive",
3795 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3796 .desc = {
3797 {
3798 .name = "snapshot",
3799 .type = QEMU_OPT_BOOL,
3800 .help = "enable/disable snapshot mode",
3801 },{
3802 .name = "aio",
3803 .type = QEMU_OPT_STRING,
3804 .help = "host AIO implementation (threads, native, io_uring)",
3805 },{
3806 .name = BDRV_OPT_CACHE_WB,
3807 .type = QEMU_OPT_BOOL,
3808 .help = "Enable writeback mode",
3809 },{
3810 .name = "format",
3811 .type = QEMU_OPT_STRING,
3812 .help = "disk format (raw, qcow2, ...)",
3813 },{
3814 .name = "rerror",
3815 .type = QEMU_OPT_STRING,
3816 .help = "read error action",
3817 },{
3818 .name = "werror",
3819 .type = QEMU_OPT_STRING,
3820 .help = "write error action",
3821 },{
3822 .name = BDRV_OPT_READ_ONLY,
3823 .type = QEMU_OPT_BOOL,
3824 .help = "open drive file as read-only",
3825 },
3826
3827 THROTTLE_OPTS,
3828
3829 {
3830 .name = "throttling.group",
3831 .type = QEMU_OPT_STRING,
3832 .help = "name of the block throttling group",
3833 },{
3834 .name = "copy-on-read",
3835 .type = QEMU_OPT_BOOL,
3836 .help = "copy read data from backing file into image file",
3837 },{
3838 .name = "detect-zeroes",
3839 .type = QEMU_OPT_STRING,
3840 .help = "try to optimize zero writes (off, on, unmap)",
3841 },{
3842 .name = "stats-account-invalid",
3843 .type = QEMU_OPT_BOOL,
3844 .help = "whether to account for invalid I/O operations "
3845 "in the statistics",
3846 },{
3847 .name = "stats-account-failed",
3848 .type = QEMU_OPT_BOOL,
3849 .help = "whether to account for failed I/O operations "
3850 "in the statistics",
3851 },
3852 { /* end of list */ }
3853 },
3854 };
3855
3856 QemuOptsList qemu_drive_opts = {
3857 .name = "drive",
3858 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3859 .desc = {
3860 /*
3861 * no elements => accept any params
3862 * validation will happen later
3863 */
3864 { /* end of list */ }
3865 },
3866 };
Random patches