?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
tcg/tci: Split out tcg_out_op_p
[qemu/kevin.git] / blockdev.c
blob5cc7c7effe9f411d5ea5cab7663d3a0ec2225f1d
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/runstate.h"
61 #include "sysemu/replay.h"
62 #include "qemu/cutils.h"
63 #include "qemu/help_option.h"
64 #include "qemu/main-loop.h"
65 #include "qemu/throttle-options.h"
66
67 QTAILQ_HEAD(, BlockDriverState) monitor_bdrv_states =
68 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states);
69
70 void bdrv_set_monitor_owned(BlockDriverState *bs)
71 {
72 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
73 }
74
75 static const char *const if_name[IF_COUNT] = {
76 [IF_NONE] = "none",
77 [IF_IDE] = "ide",
78 [IF_SCSI] = "scsi",
79 [IF_FLOPPY] = "floppy",
80 [IF_PFLASH] = "pflash",
81 [IF_MTD] = "mtd",
82 [IF_SD] = "sd",
83 [IF_VIRTIO] = "virtio",
84 [IF_XEN] = "xen",
85 };
86
87 static int if_max_devs[IF_COUNT] = {
88 /*
89 * Do not change these numbers! They govern how drive option
90 * index maps to unit and bus. That mapping is ABI.
91 *
92 * All controllers used to implement if=T drives need to support
93 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
94 * Otherwise, some index values map to "impossible" bus, unit
95 * values.
96 *
97 * For instance, if you change [IF_SCSI] to 255, -drive
98 * if=scsi,index=12 no longer means bus=1,unit=5, but
99 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
100 * the drive can't be set up. Regression.
101 */
102 [IF_IDE] = 2,
103 [IF_SCSI] = 7,
104 };
105
106 /**
107 * Boards may call this to offer board-by-board overrides
108 * of the default, global values.
109 */
110 void override_max_devs(BlockInterfaceType type, int max_devs)
111 {
112 BlockBackend *blk;
113 DriveInfo *dinfo;
114
115 if (max_devs <= 0) {
116 return;
117 }
118
119 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
120 dinfo = blk_legacy_dinfo(blk);
121 if (dinfo->type == type) {
122 fprintf(stderr, "Cannot override units-per-bus property of"
123 " the %s interface, because a drive of that type has"
124 " already been added.\n", if_name[type]);
125 g_assert_not_reached();
126 }
127 }
128
129 if_max_devs[type] = max_devs;
130 }
131
132 /*
133 * We automatically delete the drive when a device using it gets
134 * unplugged. Questionable feature, but we can't just drop it.
135 * Device models call blockdev_mark_auto_del() to schedule the
136 * automatic deletion, and generic qdev code calls blockdev_auto_del()
137 * when deletion is actually safe.
138 */
139 void blockdev_mark_auto_del(BlockBackend *blk)
140 {
141 DriveInfo *dinfo = blk_legacy_dinfo(blk);
142 BlockJob *job;
143
144 if (!dinfo) {
145 return;
146 }
147
148 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
149 if (block_job_has_bdrv(job, blk_bs(blk))) {
150 AioContext *aio_context = job->job.aio_context;
151 aio_context_acquire(aio_context);
152
153 job_cancel(&job->job, false);
154
155 aio_context_release(aio_context);
156 }
157 }
158
159 dinfo->auto_del = 1;
160 }
161
162 void blockdev_auto_del(BlockBackend *blk)
163 {
164 DriveInfo *dinfo = blk_legacy_dinfo(blk);
165
166 if (dinfo && dinfo->auto_del) {
167 monitor_remove_blk(blk);
168 blk_unref(blk);
169 }
170 }
171
172 /**
173 * Returns the current mapping of how many units per bus
174 * a particular interface can support.
175 *
176 * A positive integer indicates n units per bus.
177 * 0 implies the mapping has not been established.
178 * -1 indicates an invalid BlockInterfaceType was given.
179 */
180 int drive_get_max_devs(BlockInterfaceType type)
181 {
182 if (type >= IF_IDE && type < IF_COUNT) {
183 return if_max_devs[type];
184 }
185
186 return -1;
187 }
188
189 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
190 {
191 int max_devs = if_max_devs[type];
192 return max_devs ? index / max_devs : 0;
193 }
194
195 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
196 {
197 int max_devs = if_max_devs[type];
198 return max_devs ? index % max_devs : index;
199 }
200
201 QemuOpts *drive_def(const char *optstr)
202 {
203 return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
204 }
205
206 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
207 const char *optstr)
208 {
209 QemuOpts *opts;
210
211 opts = drive_def(optstr);
212 if (!opts) {
213 return NULL;
214 }
215 if (type != IF_DEFAULT) {
216 qemu_opt_set(opts, "if", if_name[type], &error_abort);
217 }
218 if (index >= 0) {
219 qemu_opt_set_number(opts, "index", index, &error_abort);
220 }
221 if (file)
222 qemu_opt_set(opts, "file", file, &error_abort);
223 return opts;
224 }
225
226 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
227 {
228 BlockBackend *blk;
229 DriveInfo *dinfo;
230
231 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
232 dinfo = blk_legacy_dinfo(blk);
233 if (dinfo && dinfo->type == type
234 && dinfo->bus == bus && dinfo->unit == unit) {
235 return dinfo;
236 }
237 }
238
239 return NULL;
240 }
241
242 void drive_mark_claimed_by_board(void)
243 {
244 BlockBackend *blk;
245 DriveInfo *dinfo;
246
247 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
248 dinfo = blk_legacy_dinfo(blk);
249 if (dinfo && blk_get_attached_dev(blk)) {
250 dinfo->claimed_by_board = true;
251 }
252 }
253 }
254
255 void drive_check_orphaned(void)
256 {
257 BlockBackend *blk;
258 DriveInfo *dinfo;
259 Location loc;
260 bool orphans = false;
261
262 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
263 dinfo = blk_legacy_dinfo(blk);
264 if (dinfo->is_default || dinfo->type == IF_NONE) {
265 continue;
266 }
267 if (!blk_get_attached_dev(blk)) {
268 loc_push_none(&loc);
269 qemu_opts_loc_restore(dinfo->opts);
270 error_report("machine type does not support"
271 " if=%s,bus=%d,unit=%d",
272 if_name[dinfo->type], dinfo->bus, dinfo->unit);
273 loc_pop(&loc);
274 orphans = true;
275 continue;
276 }
277 if (!dinfo->claimed_by_board && dinfo->type != IF_VIRTIO) {
278 loc_push_none(&loc);
279 qemu_opts_loc_restore(dinfo->opts);
280 warn_report("bogus if=%s is deprecated, use if=none",
281 if_name[dinfo->type]);
282 loc_pop(&loc);
283 }
284 }
285
286 if (orphans) {
287 exit(1);
288 }
289 }
290
291 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
292 {
293 return drive_get(type,
294 drive_index_to_bus_id(type, index),
295 drive_index_to_unit_id(type, index));
296 }
297
298 int drive_get_max_bus(BlockInterfaceType type)
299 {
300 int max_bus;
301 BlockBackend *blk;
302 DriveInfo *dinfo;
303
304 max_bus = -1;
305 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
306 dinfo = blk_legacy_dinfo(blk);
307 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
308 max_bus = dinfo->bus;
309 }
310 }
311 return max_bus;
312 }
313
314 /* Get a block device. This should only be used for single-drive devices
315 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
316 appropriate bus. */
317 DriveInfo *drive_get_next(BlockInterfaceType type)
318 {
319 static int next_block_unit[IF_COUNT];
320
321 return drive_get(type, 0, next_block_unit[type]++);
322 }
323
324 static void bdrv_format_print(void *opaque, const char *name)
325 {
326 qemu_printf(" %s", name);
327 }
328
329 typedef struct {
330 QEMUBH *bh;
331 BlockDriverState *bs;
332 } BDRVPutRefBH;
333
334 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
335 {
336 if (!strcmp(buf, "ignore")) {
337 return BLOCKDEV_ON_ERROR_IGNORE;
338 } else if (!is_read && !strcmp(buf, "enospc")) {
339 return BLOCKDEV_ON_ERROR_ENOSPC;
340 } else if (!strcmp(buf, "stop")) {
341 return BLOCKDEV_ON_ERROR_STOP;
342 } else if (!strcmp(buf, "report")) {
343 return BLOCKDEV_ON_ERROR_REPORT;
344 } else {
345 error_setg(errp, "'%s' invalid %s error action",
346 buf, is_read ? "read" : "write");
347 return -1;
348 }
349 }
350
351 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
352 Error **errp)
353 {
354 const QListEntry *entry;
355 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
356 switch (qobject_type(entry->value)) {
357
358 case QTYPE_QSTRING: {
359 unsigned long long length;
360 const char *str = qstring_get_str(qobject_to(QString,
361 entry->value));
362 if (parse_uint_full(str, &length, 10) == 0 &&
363 length > 0 && length <= UINT_MAX) {
364 block_acct_add_interval(stats, (unsigned) length);
365 } else {
366 error_setg(errp, "Invalid interval length: %s", str);
367 return false;
368 }
369 break;
370 }
371
372 case QTYPE_QNUM: {
373 int64_t length = qnum_get_int(qobject_to(QNum, entry->value));
374
375 if (length > 0 && length <= UINT_MAX) {
376 block_acct_add_interval(stats, (unsigned) length);
377 } else {
378 error_setg(errp, "Invalid interval length: %" PRId64, length);
379 return false;
380 }
381 break;
382 }
383
384 default:
385 error_setg(errp, "The specification of stats-intervals is invalid");
386 return false;
387 }
388 }
389 return true;
390 }
391
392 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
393
394 /* All parameters but @opts are optional and may be set to NULL. */
395 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
396 const char **throttling_group, ThrottleConfig *throttle_cfg,
397 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
398 {
399 Error *local_error = NULL;
400 const char *aio;
401
402 if (bdrv_flags) {
403 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
404 *bdrv_flags |= BDRV_O_COPY_ON_READ;
405 }
406
407 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
408 if (bdrv_parse_aio(aio, bdrv_flags) < 0) {
409 error_setg(errp, "invalid aio option");
410 return;
411 }
412 }
413 }
414
415 /* disk I/O throttling */
416 if (throttling_group) {
417 *throttling_group = qemu_opt_get(opts, "throttling.group");
418 }
419
420 if (throttle_cfg) {
421 throttle_config_init(throttle_cfg);
422 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
423 qemu_opt_get_number(opts, "throttling.bps-total", 0);
424 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
425 qemu_opt_get_number(opts, "throttling.bps-read", 0);
426 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
427 qemu_opt_get_number(opts, "throttling.bps-write", 0);
428 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
429 qemu_opt_get_number(opts, "throttling.iops-total", 0);
430 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
431 qemu_opt_get_number(opts, "throttling.iops-read", 0);
432 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
433 qemu_opt_get_number(opts, "throttling.iops-write", 0);
434
435 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
436 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
437 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
438 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
439 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
440 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
441 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
442 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
443 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
444 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
445 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
446 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
447
448 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
449 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
450 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
451 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
452 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
453 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
454 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
455 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
456 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
457 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
458 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
459 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
460
461 throttle_cfg->op_size =
462 qemu_opt_get_number(opts, "throttling.iops-size", 0);
463
464 if (!throttle_is_valid(throttle_cfg, errp)) {
465 return;
466 }
467 }
468
469 if (detect_zeroes) {
470 *detect_zeroes =
471 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
472 qemu_opt_get(opts, "detect-zeroes"),
473 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
474 &local_error);
475 if (local_error) {
476 error_propagate(errp, local_error);
477 return;
478 }
479 }
480 }
481
482 /* Takes the ownership of bs_opts */
483 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
484 Error **errp)
485 {
486 const char *buf;
487 int bdrv_flags = 0;
488 int on_read_error, on_write_error;
489 bool account_invalid, account_failed;
490 bool writethrough, read_only;
491 BlockBackend *blk;
492 BlockDriverState *bs;
493 ThrottleConfig cfg;
494 int snapshot = 0;
495 Error *error = NULL;
496 QemuOpts *opts;
497 QDict *interval_dict = NULL;
498 QList *interval_list = NULL;
499 const char *id;
500 BlockdevDetectZeroesOptions detect_zeroes =
501 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
502 const char *throttling_group = NULL;
503
504 /* Check common options by copying from bs_opts to opts, all other options
505 * stay in bs_opts for processing by bdrv_open(). */
506 id = qdict_get_try_str(bs_opts, "id");
507 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, errp);
508 if (!opts) {
509 goto err_no_opts;
510 }
511
512 if (!qemu_opts_absorb_qdict(opts, bs_opts, errp)) {
513 goto early_err;
514 }
515
516 if (id) {
517 qdict_del(bs_opts, "id");
518 }
519
520 /* extract parameters */
521 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
522
523 account_invalid = qemu_opt_get_bool(opts, "stats-account-invalid", true);
524 account_failed = qemu_opt_get_bool(opts, "stats-account-failed", true);
525
526 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
527
528 id = qemu_opts_id(opts);
529
530 qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
531 qdict_array_split(interval_dict, &interval_list);
532
533 if (qdict_size(interval_dict) != 0) {
534 error_setg(errp, "Invalid option stats-intervals.%s",
535 qdict_first(interval_dict)->key);
536 goto early_err;
537 }
538
539 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
540 &detect_zeroes, &error);
541 if (error) {
542 error_propagate(errp, error);
543 goto early_err;
544 }
545
546 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
547 if (is_help_option(buf)) {
548 qemu_printf("Supported formats:");
549 bdrv_iterate_format(bdrv_format_print, NULL, false);
550 qemu_printf("\nSupported formats (read-only):");
551 bdrv_iterate_format(bdrv_format_print, NULL, true);
552 qemu_printf("\n");
553 goto early_err;
554 }
555
556 if (qdict_haskey(bs_opts, "driver")) {
557 error_setg(errp, "Cannot specify both 'driver' and 'format'");
558 goto early_err;
559 }
560 qdict_put_str(bs_opts, "driver", buf);
561 }
562
563 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
564 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
565 on_write_error = parse_block_error_action(buf, 0, &error);
566 if (error) {
567 error_propagate(errp, error);
568 goto early_err;
569 }
570 }
571
572 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
573 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
574 on_read_error = parse_block_error_action(buf, 1, &error);
575 if (error) {
576 error_propagate(errp, error);
577 goto early_err;
578 }
579 }
580
581 if (snapshot) {
582 bdrv_flags |= BDRV_O_SNAPSHOT;
583 }
584
585 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
586
587 /* init */
588 if ((!file || !*file) && !qdict_size(bs_opts)) {
589 BlockBackendRootState *blk_rs;
590
591 blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
592 blk_rs = blk_get_root_state(blk);
593 blk_rs->open_flags = bdrv_flags;
594 blk_rs->read_only = read_only;
595 blk_rs->detect_zeroes = detect_zeroes;
596
597 qobject_unref(bs_opts);
598 } else {
599 if (file && !*file) {
600 file = NULL;
601 }
602
603 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
604 * with other callers) rather than what we want as the real defaults.
605 * Apply the defaults here instead. */
606 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
607 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
608 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
609 read_only ? "on" : "off");
610 qdict_set_default_str(bs_opts, BDRV_OPT_AUTO_READ_ONLY, "on");
611 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
612
613 if (runstate_check(RUN_STATE_INMIGRATE)) {
614 bdrv_flags |= BDRV_O_INACTIVE;
615 }
616
617 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
618 if (!blk) {
619 goto err_no_bs_opts;
620 }
621 bs = blk_bs(blk);
622
623 bs->detect_zeroes = detect_zeroes;
624
625 block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
626
627 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
628 blk_unref(blk);
629 blk = NULL;
630 goto err_no_bs_opts;
631 }
632 }
633
634 /* disk I/O throttling */
635 if (throttle_enabled(&cfg)) {
636 if (!throttling_group) {
637 throttling_group = id;
638 }
639 blk_io_limits_enable(blk, throttling_group);
640 blk_set_io_limits(blk, &cfg);
641 }
642
643 blk_set_enable_write_cache(blk, !writethrough);
644 blk_set_on_error(blk, on_read_error, on_write_error);
645
646 if (!monitor_add_blk(blk, id, errp)) {
647 blk_unref(blk);
648 blk = NULL;
649 goto err_no_bs_opts;
650 }
651
652 err_no_bs_opts:
653 qemu_opts_del(opts);
654 qobject_unref(interval_dict);
655 qobject_unref(interval_list);
656 return blk;
657
658 early_err:
659 qemu_opts_del(opts);
660 qobject_unref(interval_dict);
661 qobject_unref(interval_list);
662 err_no_opts:
663 qobject_unref(bs_opts);
664 return NULL;
665 }
666
667 /* Takes the ownership of bs_opts */
668 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
669 {
670 int bdrv_flags = 0;
671
672 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
673 * with other callers) rather than what we want as the real defaults.
674 * Apply the defaults here instead. */
675 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
676 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
677 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
678
679 if (runstate_check(RUN_STATE_INMIGRATE)) {
680 bdrv_flags |= BDRV_O_INACTIVE;
681 }
682
683 return bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
684 }
685
686 void blockdev_close_all_bdrv_states(void)
687 {
688 BlockDriverState *bs, *next_bs;
689
690 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
691 AioContext *ctx = bdrv_get_aio_context(bs);
692
693 aio_context_acquire(ctx);
694 bdrv_unref(bs);
695 aio_context_release(ctx);
696 }
697 }
698
699 /* Iterates over the list of monitor-owned BlockDriverStates */
700 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
701 {
702 return bs ? QTAILQ_NEXT(bs, monitor_list)
703 : QTAILQ_FIRST(&monitor_bdrv_states);
704 }
705
706 static bool qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
707 Error **errp)
708 {
709 const char *value;
710
711 value = qemu_opt_get(opts, from);
712 if (value) {
713 if (qemu_opt_find(opts, to)) {
714 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
715 "same time", to, from);
716 return false;
717 }
718 }
719
720 /* rename all items in opts */
721 while ((value = qemu_opt_get(opts, from))) {
722 qemu_opt_set(opts, to, value, &error_abort);
723 qemu_opt_unset(opts, from);
724 }
725 return true;
726 }
727
728 QemuOptsList qemu_legacy_drive_opts = {
729 .name = "drive",
730 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
731 .desc = {
732 {
733 .name = "bus",
734 .type = QEMU_OPT_NUMBER,
735 .help = "bus number",
736 },{
737 .name = "unit",
738 .type = QEMU_OPT_NUMBER,
739 .help = "unit number (i.e. lun for scsi)",
740 },{
741 .name = "index",
742 .type = QEMU_OPT_NUMBER,
743 .help = "index number",
744 },{
745 .name = "media",
746 .type = QEMU_OPT_STRING,
747 .help = "media type (disk, cdrom)",
748 },{
749 .name = "if",
750 .type = QEMU_OPT_STRING,
751 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
752 },{
753 .name = "file",
754 .type = QEMU_OPT_STRING,
755 .help = "file name",
756 },
757
758 /* Options that are passed on, but have special semantics with -drive */
759 {
760 .name = BDRV_OPT_READ_ONLY,
761 .type = QEMU_OPT_BOOL,
762 .help = "open drive file as read-only",
763 },{
764 .name = "rerror",
765 .type = QEMU_OPT_STRING,
766 .help = "read error action",
767 },{
768 .name = "werror",
769 .type = QEMU_OPT_STRING,
770 .help = "write error action",
771 },{
772 .name = "copy-on-read",
773 .type = QEMU_OPT_BOOL,
774 .help = "copy read data from backing file into image file",
775 },
776
777 { /* end of list */ }
778 },
779 };
780
781 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
782 Error **errp)
783 {
784 const char *value;
785 BlockBackend *blk;
786 DriveInfo *dinfo = NULL;
787 QDict *bs_opts;
788 QemuOpts *legacy_opts;
789 DriveMediaType media = MEDIA_DISK;
790 BlockInterfaceType type;
791 int max_devs, bus_id, unit_id, index;
792 const char *werror, *rerror;
793 bool read_only = false;
794 bool copy_on_read;
795 const char *filename;
796 int i;
797
798 /* Change legacy command line options into QMP ones */
799 static const struct {
800 const char *from;
801 const char *to;
802 } opt_renames[] = {
803 { "iops", "throttling.iops-total" },
804 { "iops_rd", "throttling.iops-read" },
805 { "iops_wr", "throttling.iops-write" },
806
807 { "bps", "throttling.bps-total" },
808 { "bps_rd", "throttling.bps-read" },
809 { "bps_wr", "throttling.bps-write" },
810
811 { "iops_max", "throttling.iops-total-max" },
812 { "iops_rd_max", "throttling.iops-read-max" },
813 { "iops_wr_max", "throttling.iops-write-max" },
814
815 { "bps_max", "throttling.bps-total-max" },
816 { "bps_rd_max", "throttling.bps-read-max" },
817 { "bps_wr_max", "throttling.bps-write-max" },
818
819 { "iops_size", "throttling.iops-size" },
820
821 { "group", "throttling.group" },
822
823 { "readonly", BDRV_OPT_READ_ONLY },
824 };
825
826 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
827 if (!qemu_opt_rename(all_opts, opt_renames[i].from,
828 opt_renames[i].to, errp)) {
829 return NULL;
830 }
831 }
832
833 value = qemu_opt_get(all_opts, "cache");
834 if (value) {
835 int flags = 0;
836 bool writethrough;
837
838 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
839 error_setg(errp, "invalid cache option");
840 return NULL;
841 }
842
843 /* Specific options take precedence */
844 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
845 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
846 !writethrough, &error_abort);
847 }
848 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
849 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
850 !!(flags & BDRV_O_NOCACHE), &error_abort);
851 }
852 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
853 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
854 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
855 }
856 qemu_opt_unset(all_opts, "cache");
857 }
858
859 /* Get a QDict for processing the options */
860 bs_opts = qdict_new();
861 qemu_opts_to_qdict(all_opts, bs_opts);
862
863 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
864 &error_abort);
865 if (!qemu_opts_absorb_qdict(legacy_opts, bs_opts, errp)) {
866 goto fail;
867 }
868
869 /* Media type */
870 value = qemu_opt_get(legacy_opts, "media");
871 if (value) {
872 if (!strcmp(value, "disk")) {
873 media = MEDIA_DISK;
874 } else if (!strcmp(value, "cdrom")) {
875 media = MEDIA_CDROM;
876 read_only = true;
877 } else {
878 error_setg(errp, "'%s' invalid media", value);
879 goto fail;
880 }
881 }
882
883 /* copy-on-read is disabled with a warning for read-only devices */
884 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
885 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
886
887 if (read_only && copy_on_read) {
888 warn_report("disabling copy-on-read on read-only drive");
889 copy_on_read = false;
890 }
891
892 qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
893 qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
894
895 /* Controller type */
896 value = qemu_opt_get(legacy_opts, "if");
897 if (value) {
898 for (type = 0;
899 type < IF_COUNT && strcmp(value, if_name[type]);
900 type++) {
901 }
902 if (type == IF_COUNT) {
903 error_setg(errp, "unsupported bus type '%s'", value);
904 goto fail;
905 }
906 } else {
907 type = block_default_type;
908 }
909
910 /* Device address specified by bus/unit or index.
911 * If none was specified, try to find the first free one. */
912 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
913 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
914 index = qemu_opt_get_number(legacy_opts, "index", -1);
915
916 max_devs = if_max_devs[type];
917
918 if (index != -1) {
919 if (bus_id != 0 || unit_id != -1) {
920 error_setg(errp, "index cannot be used with bus and unit");
921 goto fail;
922 }
923 bus_id = drive_index_to_bus_id(type, index);
924 unit_id = drive_index_to_unit_id(type, index);
925 }
926
927 if (unit_id == -1) {
928 unit_id = 0;
929 while (drive_get(type, bus_id, unit_id) != NULL) {
930 unit_id++;
931 if (max_devs && unit_id >= max_devs) {
932 unit_id -= max_devs;
933 bus_id++;
934 }
935 }
936 }
937
938 if (max_devs && unit_id >= max_devs) {
939 error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
940 goto fail;
941 }
942
943 if (drive_get(type, bus_id, unit_id) != NULL) {
944 error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
945 bus_id, unit_id, index);
946 goto fail;
947 }
948
949 /* no id supplied -> create one */
950 if (qemu_opts_id(all_opts) == NULL) {
951 char *new_id;
952 const char *mediastr = "";
953 if (type == IF_IDE || type == IF_SCSI) {
954 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
955 }
956 if (max_devs) {
957 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
958 mediastr, unit_id);
959 } else {
960 new_id = g_strdup_printf("%s%s%i", if_name[type],
961 mediastr, unit_id);
962 }
963 qdict_put_str(bs_opts, "id", new_id);
964 g_free(new_id);
965 }
966
967 /* Add virtio block device */
968 if (type == IF_VIRTIO) {
969 QemuOpts *devopts;
970 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
971 &error_abort);
972 if (arch_type == QEMU_ARCH_S390X) {
973 qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
974 } else {
975 qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
976 }
977 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
978 &error_abort);
979 }
980
981 filename = qemu_opt_get(legacy_opts, "file");
982
983 /* Check werror/rerror compatibility with if=... */
984 werror = qemu_opt_get(legacy_opts, "werror");
985 if (werror != NULL) {
986 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
987 type != IF_NONE) {
988 error_setg(errp, "werror is not supported by this bus type");
989 goto fail;
990 }
991 qdict_put_str(bs_opts, "werror", werror);
992 }
993
994 rerror = qemu_opt_get(legacy_opts, "rerror");
995 if (rerror != NULL) {
996 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
997 type != IF_NONE) {
998 error_setg(errp, "rerror is not supported by this bus type");
999 goto fail;
1000 }
1001 qdict_put_str(bs_opts, "rerror", rerror);
1002 }
1003
1004 /* Actual block device init: Functionality shared with blockdev-add */
1005 blk = blockdev_init(filename, bs_opts, errp);
1006 bs_opts = NULL;
1007 if (!blk) {
1008 goto fail;
1009 }
1010
1011 /* Create legacy DriveInfo */
1012 dinfo = g_malloc0(sizeof(*dinfo));
1013 dinfo->opts = all_opts;
1014
1015 dinfo->type = type;
1016 dinfo->bus = bus_id;
1017 dinfo->unit = unit_id;
1018
1019 blk_set_legacy_dinfo(blk, dinfo);
1020
1021 switch(type) {
1022 case IF_IDE:
1023 case IF_SCSI:
1024 case IF_XEN:
1025 case IF_NONE:
1026 dinfo->media_cd = media == MEDIA_CDROM;
1027 break;
1028 default:
1029 break;
1030 }
1031
1032 fail:
1033 qemu_opts_del(legacy_opts);
1034 qobject_unref(bs_opts);
1035 return dinfo;
1036 }
1037
1038 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1039 {
1040 BlockDriverState *bs;
1041
1042 bs = bdrv_lookup_bs(name, name, errp);
1043 if (bs == NULL) {
1044 return NULL;
1045 }
1046
1047 if (!bdrv_is_root_node(bs)) {
1048 error_setg(errp, "Need a root block node");
1049 return NULL;
1050 }
1051
1052 if (!bdrv_is_inserted(bs)) {
1053 error_setg(errp, "Device has no medium");
1054 return NULL;
1055 }
1056
1057 return bs;
1058 }
1059
1060 static void blockdev_do_action(TransactionAction *action, Error **errp)
1061 {
1062 TransactionActionList list;
1063
1064 list.value = action;
1065 list.next = NULL;
1066 qmp_transaction(&list, false, NULL, errp);
1067 }
1068
1069 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1070 bool has_node_name, const char *node_name,
1071 const char *snapshot_file,
1072 bool has_snapshot_node_name,
1073 const char *snapshot_node_name,
1074 bool has_format, const char *format,
1075 bool has_mode, NewImageMode mode, Error **errp)
1076 {
1077 BlockdevSnapshotSync snapshot = {
1078 .has_device = has_device,
1079 .device = (char *) device,
1080 .has_node_name = has_node_name,
1081 .node_name = (char *) node_name,
1082 .snapshot_file = (char *) snapshot_file,
1083 .has_snapshot_node_name = has_snapshot_node_name,
1084 .snapshot_node_name = (char *) snapshot_node_name,
1085 .has_format = has_format,
1086 .format = (char *) format,
1087 .has_mode = has_mode,
1088 .mode = mode,
1089 };
1090 TransactionAction action = {
1091 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1092 .u.blockdev_snapshot_sync.data = &snapshot,
1093 };
1094 blockdev_do_action(&action, errp);
1095 }
1096
1097 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1098 Error **errp)
1099 {
1100 BlockdevSnapshot snapshot_data = {
1101 .node = (char *) node,
1102 .overlay = (char *) overlay
1103 };
1104 TransactionAction action = {
1105 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1106 .u.blockdev_snapshot.data = &snapshot_data,
1107 };
1108 blockdev_do_action(&action, errp);
1109 }
1110
1111 void qmp_blockdev_snapshot_internal_sync(const char *device,
1112 const char *name,
1113 Error **errp)
1114 {
1115 BlockdevSnapshotInternal snapshot = {
1116 .device = (char *) device,
1117 .name = (char *) name
1118 };
1119 TransactionAction action = {
1120 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1121 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1122 };
1123 blockdev_do_action(&action, errp);
1124 }
1125
1126 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1127 bool has_id,
1128 const char *id,
1129 bool has_name,
1130 const char *name,
1131 Error **errp)
1132 {
1133 BlockDriverState *bs;
1134 AioContext *aio_context;
1135 QEMUSnapshotInfo sn;
1136 Error *local_err = NULL;
1137 SnapshotInfo *info = NULL;
1138 int ret;
1139
1140 bs = qmp_get_root_bs(device, errp);
1141 if (!bs) {
1142 return NULL;
1143 }
1144 aio_context = bdrv_get_aio_context(bs);
1145 aio_context_acquire(aio_context);
1146
1147 if (!has_id) {
1148 id = NULL;
1149 }
1150
1151 if (!has_name) {
1152 name = NULL;
1153 }
1154
1155 if (!id && !name) {
1156 error_setg(errp, "Name or id must be provided");
1157 goto out_aio_context;
1158 }
1159
1160 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1161 goto out_aio_context;
1162 }
1163
1164 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1165 if (local_err) {
1166 error_propagate(errp, local_err);
1167 goto out_aio_context;
1168 }
1169 if (!ret) {
1170 error_setg(errp,
1171 "Snapshot with id '%s' and name '%s' does not exist on "
1172 "device '%s'",
1173 STR_OR_NULL(id), STR_OR_NULL(name), device);
1174 goto out_aio_context;
1175 }
1176
1177 bdrv_snapshot_delete(bs, id, name, &local_err);
1178 if (local_err) {
1179 error_propagate(errp, local_err);
1180 goto out_aio_context;
1181 }
1182
1183 aio_context_release(aio_context);
1184
1185 info = g_new0(SnapshotInfo, 1);
1186 info->id = g_strdup(sn.id_str);
1187 info->name = g_strdup(sn.name);
1188 info->date_nsec = sn.date_nsec;
1189 info->date_sec = sn.date_sec;
1190 info->vm_state_size = sn.vm_state_size;
1191 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1192 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1193 if (sn.icount != -1ULL) {
1194 info->icount = sn.icount;
1195 info->has_icount = true;
1196 }
1197
1198 return info;
1199
1200 out_aio_context:
1201 aio_context_release(aio_context);
1202 return NULL;
1203 }
1204
1205 /* New and old BlockDriverState structs for atomic group operations */
1206
1207 typedef struct BlkActionState BlkActionState;
1208
1209 /**
1210 * BlkActionOps:
1211 * Table of operations that define an Action.
1212 *
1213 * @instance_size: Size of state struct, in bytes.
1214 * @prepare: Prepare the work, must NOT be NULL.
1215 * @commit: Commit the changes, can be NULL.
1216 * @abort: Abort the changes on fail, can be NULL.
1217 * @clean: Clean up resources after all transaction actions have called
1218 * commit() or abort(). Can be NULL.
1219 *
1220 * Only prepare() may fail. In a single transaction, only one of commit() or
1221 * abort() will be called. clean() will always be called if it is present.
1222 */
1223 typedef struct BlkActionOps {
1224 size_t instance_size;
1225 void (*prepare)(BlkActionState *common, Error **errp);
1226 void (*commit)(BlkActionState *common);
1227 void (*abort)(BlkActionState *common);
1228 void (*clean)(BlkActionState *common);
1229 } BlkActionOps;
1230
1231 /**
1232 * BlkActionState:
1233 * Describes one Action's state within a Transaction.
1234 *
1235 * @action: QAPI-defined enum identifying which Action to perform.
1236 * @ops: Table of ActionOps this Action can perform.
1237 * @block_job_txn: Transaction which this action belongs to.
1238 * @entry: List membership for all Actions in this Transaction.
1239 *
1240 * This structure must be arranged as first member in a subclassed type,
1241 * assuming that the compiler will also arrange it to the same offsets as the
1242 * base class.
1243 */
1244 struct BlkActionState {
1245 TransactionAction *action;
1246 const BlkActionOps *ops;
1247 JobTxn *block_job_txn;
1248 TransactionProperties *txn_props;
1249 QTAILQ_ENTRY(BlkActionState) entry;
1250 };
1251
1252 /* internal snapshot private data */
1253 typedef struct InternalSnapshotState {
1254 BlkActionState common;
1255 BlockDriverState *bs;
1256 QEMUSnapshotInfo sn;
1257 bool created;
1258 } InternalSnapshotState;
1259
1260
1261 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1262 {
1263 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1264 error_setg(errp,
1265 "Action '%s' does not support Transaction property "
1266 "completion-mode = %s",
1267 TransactionActionKind_str(s->action->type),
1268 ActionCompletionMode_str(s->txn_props->completion_mode));
1269 return -1;
1270 }
1271 return 0;
1272 }
1273
1274 static void internal_snapshot_prepare(BlkActionState *common,
1275 Error **errp)
1276 {
1277 Error *local_err = NULL;
1278 const char *device;
1279 const char *name;
1280 BlockDriverState *bs;
1281 QEMUSnapshotInfo old_sn, *sn;
1282 bool ret;
1283 qemu_timeval tv;
1284 BlockdevSnapshotInternal *internal;
1285 InternalSnapshotState *state;
1286 AioContext *aio_context;
1287 int ret1;
1288
1289 g_assert(common->action->type ==
1290 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1291 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1292 state = DO_UPCAST(InternalSnapshotState, common, common);
1293
1294 /* 1. parse input */
1295 device = internal->device;
1296 name = internal->name;
1297
1298 /* 2. check for validation */
1299 if (action_check_completion_mode(common, errp) < 0) {
1300 return;
1301 }
1302
1303 bs = qmp_get_root_bs(device, errp);
1304 if (!bs) {
1305 return;
1306 }
1307
1308 aio_context = bdrv_get_aio_context(bs);
1309 aio_context_acquire(aio_context);
1310
1311 state->bs = bs;
1312
1313 /* Paired with .clean() */
1314 bdrv_drained_begin(bs);
1315
1316 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1317 goto out;
1318 }
1319
1320 if (bdrv_is_read_only(bs)) {
1321 error_setg(errp, "Device '%s' is read only", device);
1322 goto out;
1323 }
1324
1325 if (!bdrv_can_snapshot(bs)) {
1326 error_setg(errp, "Block format '%s' used by device '%s' "
1327 "does not support internal snapshots",
1328 bs->drv->format_name, device);
1329 goto out;
1330 }
1331
1332 if (!strlen(name)) {
1333 error_setg(errp, "Name is empty");
1334 goto out;
1335 }
1336
1337 /* check whether a snapshot with name exist */
1338 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1339 &local_err);
1340 if (local_err) {
1341 error_propagate(errp, local_err);
1342 goto out;
1343 } else if (ret) {
1344 error_setg(errp,
1345 "Snapshot with name '%s' already exists on device '%s'",
1346 name, device);
1347 goto out;
1348 }
1349
1350 /* 3. take the snapshot */
1351 sn = &state->sn;
1352 pstrcpy(sn->name, sizeof(sn->name), name);
1353 qemu_gettimeofday(&tv);
1354 sn->date_sec = tv.tv_sec;
1355 sn->date_nsec = tv.tv_usec * 1000;
1356 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1357 if (replay_mode != REPLAY_MODE_NONE) {
1358 sn->icount = replay_get_current_icount();
1359 } else {
1360 sn->icount = -1ULL;
1361 }
1362
1363 ret1 = bdrv_snapshot_create(bs, sn);
1364 if (ret1 < 0) {
1365 error_setg_errno(errp, -ret1,
1366 "Failed to create snapshot '%s' on device '%s'",
1367 name, device);
1368 goto out;
1369 }
1370
1371 /* 4. succeed, mark a snapshot is created */
1372 state->created = true;
1373
1374 out:
1375 aio_context_release(aio_context);
1376 }
1377
1378 static void internal_snapshot_abort(BlkActionState *common)
1379 {
1380 InternalSnapshotState *state =
1381 DO_UPCAST(InternalSnapshotState, common, common);
1382 BlockDriverState *bs = state->bs;
1383 QEMUSnapshotInfo *sn = &state->sn;
1384 AioContext *aio_context;
1385 Error *local_error = NULL;
1386
1387 if (!state->created) {
1388 return;
1389 }
1390
1391 aio_context = bdrv_get_aio_context(state->bs);
1392 aio_context_acquire(aio_context);
1393
1394 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1395 error_reportf_err(local_error,
1396 "Failed to delete snapshot with id '%s' and "
1397 "name '%s' on device '%s' in abort: ",
1398 sn->id_str, sn->name,
1399 bdrv_get_device_name(bs));
1400 }
1401
1402 aio_context_release(aio_context);
1403 }
1404
1405 static void internal_snapshot_clean(BlkActionState *common)
1406 {
1407 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1408 common, common);
1409 AioContext *aio_context;
1410
1411 if (!state->bs) {
1412 return;
1413 }
1414
1415 aio_context = bdrv_get_aio_context(state->bs);
1416 aio_context_acquire(aio_context);
1417
1418 bdrv_drained_end(state->bs);
1419
1420 aio_context_release(aio_context);
1421 }
1422
1423 /* external snapshot private data */
1424 typedef struct ExternalSnapshotState {
1425 BlkActionState common;
1426 BlockDriverState *old_bs;
1427 BlockDriverState *new_bs;
1428 bool overlay_appended;
1429 } ExternalSnapshotState;
1430
1431 static void external_snapshot_prepare(BlkActionState *common,
1432 Error **errp)
1433 {
1434 int ret;
1435 int flags = 0;
1436 QDict *options = NULL;
1437 Error *local_err = NULL;
1438 /* Device and node name of the image to generate the snapshot from */
1439 const char *device;
1440 const char *node_name;
1441 /* Reference to the new image (for 'blockdev-snapshot') */
1442 const char *snapshot_ref;
1443 /* File name of the new image (for 'blockdev-snapshot-sync') */
1444 const char *new_image_file;
1445 ExternalSnapshotState *state =
1446 DO_UPCAST(ExternalSnapshotState, common, common);
1447 TransactionAction *action = common->action;
1448 AioContext *aio_context;
1449 uint64_t perm, shared;
1450
1451 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1452 * purpose but a different set of parameters */
1453 switch (action->type) {
1454 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1455 {
1456 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1457 device = s->node;
1458 node_name = s->node;
1459 new_image_file = NULL;
1460 snapshot_ref = s->overlay;
1461 }
1462 break;
1463 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1464 {
1465 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1466 device = s->has_device ? s->device : NULL;
1467 node_name = s->has_node_name ? s->node_name : NULL;
1468 new_image_file = s->snapshot_file;
1469 snapshot_ref = NULL;
1470 }
1471 break;
1472 default:
1473 g_assert_not_reached();
1474 }
1475
1476 /* start processing */
1477 if (action_check_completion_mode(common, errp) < 0) {
1478 return;
1479 }
1480
1481 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1482 if (!state->old_bs) {
1483 return;
1484 }
1485
1486 aio_context = bdrv_get_aio_context(state->old_bs);
1487 aio_context_acquire(aio_context);
1488
1489 /* Paired with .clean() */
1490 bdrv_drained_begin(state->old_bs);
1491
1492 if (!bdrv_is_inserted(state->old_bs)) {
1493 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1494 goto out;
1495 }
1496
1497 if (bdrv_op_is_blocked(state->old_bs,
1498 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1499 goto out;
1500 }
1501
1502 if (!bdrv_is_read_only(state->old_bs)) {
1503 if (bdrv_flush(state->old_bs)) {
1504 error_setg(errp, QERR_IO_ERROR);
1505 goto out;
1506 }
1507 }
1508
1509 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1510 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1511 const char *format = s->has_format ? s->format : "qcow2";
1512 enum NewImageMode mode;
1513 const char *snapshot_node_name =
1514 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1515
1516 if (node_name && !snapshot_node_name) {
1517 error_setg(errp, "New overlay node-name missing");
1518 goto out;
1519 }
1520
1521 if (snapshot_node_name &&
1522 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1523 error_setg(errp, "New overlay node-name already in use");
1524 goto out;
1525 }
1526
1527 flags = state->old_bs->open_flags;
1528 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1529 flags |= BDRV_O_NO_BACKING;
1530
1531 /* create new image w/backing file */
1532 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1533 if (mode != NEW_IMAGE_MODE_EXISTING) {
1534 int64_t size = bdrv_getlength(state->old_bs);
1535 if (size < 0) {
1536 error_setg_errno(errp, -size, "bdrv_getlength failed");
1537 goto out;
1538 }
1539 bdrv_refresh_filename(state->old_bs);
1540 bdrv_img_create(new_image_file, format,
1541 state->old_bs->filename,
1542 state->old_bs->drv->format_name,
1543 NULL, size, flags, false, &local_err);
1544 if (local_err) {
1545 error_propagate(errp, local_err);
1546 goto out;
1547 }
1548 }
1549
1550 options = qdict_new();
1551 if (snapshot_node_name) {
1552 qdict_put_str(options, "node-name", snapshot_node_name);
1553 }
1554 qdict_put_str(options, "driver", format);
1555 }
1556
1557 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1558 errp);
1559 /* We will manually add the backing_hd field to the bs later */
1560 if (!state->new_bs) {
1561 goto out;
1562 }
1563
1564 /*
1565 * Allow attaching a backing file to an overlay that's already in use only
1566 * if the parents don't assume that they are already seeing a valid image.
1567 * (Specifically, allow it as a mirror target, which is write-only access.)
1568 */
1569 bdrv_get_cumulative_perm(state->new_bs, &perm, &shared);
1570 if (perm & BLK_PERM_CONSISTENT_READ) {
1571 error_setg(errp, "The overlay is already in use");
1572 goto out;
1573 }
1574
1575 if (state->new_bs->drv->is_filter) {
1576 error_setg(errp, "Filters cannot be used as overlays");
1577 goto out;
1578 }
1579
1580 if (bdrv_cow_child(state->new_bs)) {
1581 error_setg(errp, "The overlay already has a backing image");
1582 goto out;
1583 }
1584
1585 if (!state->new_bs->drv->supports_backing) {
1586 error_setg(errp, "The overlay does not support backing images");
1587 goto out;
1588 }
1589
1590 /* This removes our old bs and adds the new bs. This is an operation that
1591 * can fail, so we need to do it in .prepare; undoing it for abort is
1592 * always possible. */
1593 bdrv_ref(state->new_bs);
1594 ret = bdrv_append(state->new_bs, state->old_bs, errp);
1595 if (ret < 0) {
1596 goto out;
1597 }
1598 state->overlay_appended = true;
1599
1600 out:
1601 aio_context_release(aio_context);
1602 }
1603
1604 static void external_snapshot_commit(BlkActionState *common)
1605 {
1606 ExternalSnapshotState *state =
1607 DO_UPCAST(ExternalSnapshotState, common, common);
1608 AioContext *aio_context;
1609
1610 aio_context = bdrv_get_aio_context(state->old_bs);
1611 aio_context_acquire(aio_context);
1612
1613 /* We don't need (or want) to use the transactional
1614 * bdrv_reopen_multiple() across all the entries at once, because we
1615 * don't want to abort all of them if one of them fails the reopen */
1616 if (!qatomic_read(&state->old_bs->copy_on_read)) {
1617 bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1618 }
1619
1620 aio_context_release(aio_context);
1621 }
1622
1623 static void external_snapshot_abort(BlkActionState *common)
1624 {
1625 ExternalSnapshotState *state =
1626 DO_UPCAST(ExternalSnapshotState, common, common);
1627 if (state->new_bs) {
1628 if (state->overlay_appended) {
1629 AioContext *aio_context;
1630 AioContext *tmp_context;
1631 int ret;
1632
1633 aio_context = bdrv_get_aio_context(state->old_bs);
1634 aio_context_acquire(aio_context);
1635
1636 bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd()
1637 close state->old_bs; we need it */
1638 bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1639
1640 /*
1641 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1642 * the main AioContext. As we're still going to be using it, return
1643 * it to the AioContext it was before.
1644 */
1645 tmp_context = bdrv_get_aio_context(state->old_bs);
1646 if (aio_context != tmp_context) {
1647 aio_context_release(aio_context);
1648 aio_context_acquire(tmp_context);
1649
1650 ret = bdrv_try_set_aio_context(state->old_bs,
1651 aio_context, NULL);
1652 assert(ret == 0);
1653
1654 aio_context_release(tmp_context);
1655 aio_context_acquire(aio_context);
1656 }
1657
1658 bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1659 bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1660
1661 aio_context_release(aio_context);
1662 }
1663 }
1664 }
1665
1666 static void external_snapshot_clean(BlkActionState *common)
1667 {
1668 ExternalSnapshotState *state =
1669 DO_UPCAST(ExternalSnapshotState, common, common);
1670 AioContext *aio_context;
1671
1672 if (!state->old_bs) {
1673 return;
1674 }
1675
1676 aio_context = bdrv_get_aio_context(state->old_bs);
1677 aio_context_acquire(aio_context);
1678
1679 bdrv_drained_end(state->old_bs);
1680 bdrv_unref(state->new_bs);
1681
1682 aio_context_release(aio_context);
1683 }
1684
1685 typedef struct DriveBackupState {
1686 BlkActionState common;
1687 BlockDriverState *bs;
1688 BlockJob *job;
1689 } DriveBackupState;
1690
1691 static BlockJob *do_backup_common(BackupCommon *backup,
1692 BlockDriverState *bs,
1693 BlockDriverState *target_bs,
1694 AioContext *aio_context,
1695 JobTxn *txn, Error **errp);
1696
1697 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1698 {
1699 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1700 DriveBackup *backup;
1701 BlockDriverState *bs;
1702 BlockDriverState *target_bs;
1703 BlockDriverState *source = NULL;
1704 AioContext *aio_context;
1705 AioContext *old_context;
1706 QDict *options;
1707 Error *local_err = NULL;
1708 int flags;
1709 int64_t size;
1710 bool set_backing_hd = false;
1711 int ret;
1712
1713 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1714 backup = common->action->u.drive_backup.data;
1715
1716 if (!backup->has_mode) {
1717 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1718 }
1719
1720 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1721 if (!bs) {
1722 return;
1723 }
1724
1725 if (!bs->drv) {
1726 error_setg(errp, "Device has no medium");
1727 return;
1728 }
1729
1730 aio_context = bdrv_get_aio_context(bs);
1731 aio_context_acquire(aio_context);
1732
1733 /* Paired with .clean() */
1734 bdrv_drained_begin(bs);
1735
1736 if (!backup->has_format) {
1737 backup->format = backup->mode == NEW_IMAGE_MODE_EXISTING ?
1738 NULL : (char *) bs->drv->format_name;
1739 }
1740
1741 /* Early check to avoid creating target */
1742 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
1743 goto out;
1744 }
1745
1746 flags = bs->open_flags | BDRV_O_RDWR;
1747
1748 /*
1749 * See if we have a backing HD we can use to create our new image
1750 * on top of.
1751 */
1752 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
1753 /*
1754 * Backup will not replace the source by the target, so none
1755 * of the filters skipped here will be removed (in contrast to
1756 * mirror). Therefore, we can skip all of them when looking
1757 * for the first COW relationship.
1758 */
1759 source = bdrv_cow_bs(bdrv_skip_filters(bs));
1760 if (!source) {
1761 backup->sync = MIRROR_SYNC_MODE_FULL;
1762 }
1763 }
1764 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
1765 source = bs;
1766 flags |= BDRV_O_NO_BACKING;
1767 set_backing_hd = true;
1768 }
1769
1770 size = bdrv_getlength(bs);
1771 if (size < 0) {
1772 error_setg_errno(errp, -size, "bdrv_getlength failed");
1773 goto out;
1774 }
1775
1776 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
1777 assert(backup->format);
1778 if (source) {
1779 /* Implicit filters should not appear in the filename */
1780 BlockDriverState *explicit_backing =
1781 bdrv_skip_implicit_filters(source);
1782
1783 bdrv_refresh_filename(explicit_backing);
1784 bdrv_img_create(backup->target, backup->format,
1785 explicit_backing->filename,
1786 explicit_backing->drv->format_name, NULL,
1787 size, flags, false, &local_err);
1788 } else {
1789 bdrv_img_create(backup->target, backup->format, NULL, NULL, NULL,
1790 size, flags, false, &local_err);
1791 }
1792 }
1793
1794 if (local_err) {
1795 error_propagate(errp, local_err);
1796 goto out;
1797 }
1798
1799 options = qdict_new();
1800 qdict_put_str(options, "discard", "unmap");
1801 qdict_put_str(options, "detect-zeroes", "unmap");
1802 if (backup->format) {
1803 qdict_put_str(options, "driver", backup->format);
1804 }
1805
1806 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
1807 if (!target_bs) {
1808 goto out;
1809 }
1810
1811 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1812 old_context = bdrv_get_aio_context(target_bs);
1813 aio_context_release(aio_context);
1814 aio_context_acquire(old_context);
1815
1816 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1817 if (ret < 0) {
1818 bdrv_unref(target_bs);
1819 aio_context_release(old_context);
1820 return;
1821 }
1822
1823 aio_context_release(old_context);
1824 aio_context_acquire(aio_context);
1825
1826 if (set_backing_hd) {
1827 if (bdrv_set_backing_hd(target_bs, source, errp) < 0) {
1828 goto unref;
1829 }
1830 }
1831
1832 state->bs = bs;
1833
1834 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1835 bs, target_bs, aio_context,
1836 common->block_job_txn, errp);
1837
1838 unref:
1839 bdrv_unref(target_bs);
1840 out:
1841 aio_context_release(aio_context);
1842 }
1843
1844 static void drive_backup_commit(BlkActionState *common)
1845 {
1846 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1847 AioContext *aio_context;
1848
1849 aio_context = bdrv_get_aio_context(state->bs);
1850 aio_context_acquire(aio_context);
1851
1852 assert(state->job);
1853 job_start(&state->job->job);
1854
1855 aio_context_release(aio_context);
1856 }
1857
1858 static void drive_backup_abort(BlkActionState *common)
1859 {
1860 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1861
1862 if (state->job) {
1863 AioContext *aio_context;
1864
1865 aio_context = bdrv_get_aio_context(state->bs);
1866 aio_context_acquire(aio_context);
1867
1868 job_cancel_sync(&state->job->job);
1869
1870 aio_context_release(aio_context);
1871 }
1872 }
1873
1874 static void drive_backup_clean(BlkActionState *common)
1875 {
1876 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1877 AioContext *aio_context;
1878
1879 if (!state->bs) {
1880 return;
1881 }
1882
1883 aio_context = bdrv_get_aio_context(state->bs);
1884 aio_context_acquire(aio_context);
1885
1886 bdrv_drained_end(state->bs);
1887
1888 aio_context_release(aio_context);
1889 }
1890
1891 typedef struct BlockdevBackupState {
1892 BlkActionState common;
1893 BlockDriverState *bs;
1894 BlockJob *job;
1895 } BlockdevBackupState;
1896
1897 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1898 {
1899 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1900 BlockdevBackup *backup;
1901 BlockDriverState *bs;
1902 BlockDriverState *target_bs;
1903 AioContext *aio_context;
1904 AioContext *old_context;
1905 int ret;
1906
1907 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1908 backup = common->action->u.blockdev_backup.data;
1909
1910 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1911 if (!bs) {
1912 return;
1913 }
1914
1915 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1916 if (!target_bs) {
1917 return;
1918 }
1919
1920 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1921 aio_context = bdrv_get_aio_context(bs);
1922 old_context = bdrv_get_aio_context(target_bs);
1923 aio_context_acquire(old_context);
1924
1925 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1926 if (ret < 0) {
1927 aio_context_release(old_context);
1928 return;
1929 }
1930
1931 aio_context_release(old_context);
1932 aio_context_acquire(aio_context);
1933 state->bs = bs;
1934
1935 /* Paired with .clean() */
1936 bdrv_drained_begin(state->bs);
1937
1938 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1939 bs, target_bs, aio_context,
1940 common->block_job_txn, errp);
1941
1942 aio_context_release(aio_context);
1943 }
1944
1945 static void blockdev_backup_commit(BlkActionState *common)
1946 {
1947 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1948 AioContext *aio_context;
1949
1950 aio_context = bdrv_get_aio_context(state->bs);
1951 aio_context_acquire(aio_context);
1952
1953 assert(state->job);
1954 job_start(&state->job->job);
1955
1956 aio_context_release(aio_context);
1957 }
1958
1959 static void blockdev_backup_abort(BlkActionState *common)
1960 {
1961 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1962
1963 if (state->job) {
1964 AioContext *aio_context;
1965
1966 aio_context = bdrv_get_aio_context(state->bs);
1967 aio_context_acquire(aio_context);
1968
1969 job_cancel_sync(&state->job->job);
1970
1971 aio_context_release(aio_context);
1972 }
1973 }
1974
1975 static void blockdev_backup_clean(BlkActionState *common)
1976 {
1977 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1978 AioContext *aio_context;
1979
1980 if (!state->bs) {
1981 return;
1982 }
1983
1984 aio_context = bdrv_get_aio_context(state->bs);
1985 aio_context_acquire(aio_context);
1986
1987 bdrv_drained_end(state->bs);
1988
1989 aio_context_release(aio_context);
1990 }
1991
1992 typedef struct BlockDirtyBitmapState {
1993 BlkActionState common;
1994 BdrvDirtyBitmap *bitmap;
1995 BlockDriverState *bs;
1996 HBitmap *backup;
1997 bool prepared;
1998 bool was_enabled;
1999 } BlockDirtyBitmapState;
2000
2001 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
2002 Error **errp)
2003 {
2004 Error *local_err = NULL;
2005 BlockDirtyBitmapAdd *action;
2006 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2007 common, common);
2008
2009 if (action_check_completion_mode(common, errp) < 0) {
2010 return;
2011 }
2012
2013 action = common->action->u.block_dirty_bitmap_add.data;
2014 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2015 qmp_block_dirty_bitmap_add(action->node, action->name,
2016 action->has_granularity, action->granularity,
2017 action->has_persistent, action->persistent,
2018 action->has_disabled, action->disabled,
2019 &local_err);
2020
2021 if (!local_err) {
2022 state->prepared = true;
2023 } else {
2024 error_propagate(errp, local_err);
2025 }
2026 }
2027
2028 static void block_dirty_bitmap_add_abort(BlkActionState *common)
2029 {
2030 BlockDirtyBitmapAdd *action;
2031 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2032 common, common);
2033
2034 action = common->action->u.block_dirty_bitmap_add.data;
2035 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2036 * then the node reference and bitmap name must have been valid.
2037 */
2038 if (state->prepared) {
2039 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2040 }
2041 }
2042
2043 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2044 Error **errp)
2045 {
2046 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2047 common, common);
2048 BlockDirtyBitmap *action;
2049
2050 if (action_check_completion_mode(common, errp) < 0) {
2051 return;
2052 }
2053
2054 action = common->action->u.block_dirty_bitmap_clear.data;
2055 state->bitmap = block_dirty_bitmap_lookup(action->node,
2056 action->name,
2057 &state->bs,
2058 errp);
2059 if (!state->bitmap) {
2060 return;
2061 }
2062
2063 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2064 return;
2065 }
2066
2067 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2068 }
2069
2070 static void block_dirty_bitmap_restore(BlkActionState *common)
2071 {
2072 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2073 common, common);
2074
2075 if (state->backup) {
2076 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2077 }
2078 }
2079
2080 static void block_dirty_bitmap_free_backup(BlkActionState *common)
2081 {
2082 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2083 common, common);
2084
2085 hbitmap_free(state->backup);
2086 }
2087
2088 static void block_dirty_bitmap_enable_prepare(BlkActionState *common,
2089 Error **errp)
2090 {
2091 BlockDirtyBitmap *action;
2092 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2093 common, common);
2094
2095 if (action_check_completion_mode(common, errp) < 0) {
2096 return;
2097 }
2098
2099 action = common->action->u.block_dirty_bitmap_enable.data;
2100 state->bitmap = block_dirty_bitmap_lookup(action->node,
2101 action->name,
2102 NULL,
2103 errp);
2104 if (!state->bitmap) {
2105 return;
2106 }
2107
2108 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2109 return;
2110 }
2111
2112 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2113 bdrv_enable_dirty_bitmap(state->bitmap);
2114 }
2115
2116 static void block_dirty_bitmap_enable_abort(BlkActionState *common)
2117 {
2118 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2119 common, common);
2120
2121 if (!state->was_enabled) {
2122 bdrv_disable_dirty_bitmap(state->bitmap);
2123 }
2124 }
2125
2126 static void block_dirty_bitmap_disable_prepare(BlkActionState *common,
2127 Error **errp)
2128 {
2129 BlockDirtyBitmap *action;
2130 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2131 common, common);
2132
2133 if (action_check_completion_mode(common, errp) < 0) {
2134 return;
2135 }
2136
2137 action = common->action->u.block_dirty_bitmap_disable.data;
2138 state->bitmap = block_dirty_bitmap_lookup(action->node,
2139 action->name,
2140 NULL,
2141 errp);
2142 if (!state->bitmap) {
2143 return;
2144 }
2145
2146 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2147 return;
2148 }
2149
2150 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2151 bdrv_disable_dirty_bitmap(state->bitmap);
2152 }
2153
2154 static void block_dirty_bitmap_disable_abort(BlkActionState *common)
2155 {
2156 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2157 common, common);
2158
2159 if (state->was_enabled) {
2160 bdrv_enable_dirty_bitmap(state->bitmap);
2161 }
2162 }
2163
2164 static void block_dirty_bitmap_merge_prepare(BlkActionState *common,
2165 Error **errp)
2166 {
2167 BlockDirtyBitmapMerge *action;
2168 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2169 common, common);
2170
2171 if (action_check_completion_mode(common, errp) < 0) {
2172 return;
2173 }
2174
2175 action = common->action->u.block_dirty_bitmap_merge.data;
2176
2177 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2178 action->bitmaps, &state->backup,
2179 errp);
2180 }
2181
2182 static void block_dirty_bitmap_remove_prepare(BlkActionState *common,
2183 Error **errp)
2184 {
2185 BlockDirtyBitmap *action;
2186 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2187 common, common);
2188
2189 if (action_check_completion_mode(common, errp) < 0) {
2190 return;
2191 }
2192
2193 action = common->action->u.block_dirty_bitmap_remove.data;
2194
2195 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2196 false, &state->bs, errp);
2197 if (state->bitmap) {
2198 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2199 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2200 }
2201 }
2202
2203 static void block_dirty_bitmap_remove_abort(BlkActionState *common)
2204 {
2205 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2206 common, common);
2207
2208 if (state->bitmap) {
2209 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2210 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2211 }
2212 }
2213
2214 static void block_dirty_bitmap_remove_commit(BlkActionState *common)
2215 {
2216 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2217 common, common);
2218
2219 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2220 bdrv_release_dirty_bitmap(state->bitmap);
2221 }
2222
2223 static void abort_prepare(BlkActionState *common, Error **errp)
2224 {
2225 error_setg(errp, "Transaction aborted using Abort action");
2226 }
2227
2228 static void abort_commit(BlkActionState *common)
2229 {
2230 g_assert_not_reached(); /* this action never succeeds */
2231 }
2232
2233 static const BlkActionOps actions[] = {
2234 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2235 .instance_size = sizeof(ExternalSnapshotState),
2236 .prepare = external_snapshot_prepare,
2237 .commit = external_snapshot_commit,
2238 .abort = external_snapshot_abort,
2239 .clean = external_snapshot_clean,
2240 },
2241 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2242 .instance_size = sizeof(ExternalSnapshotState),
2243 .prepare = external_snapshot_prepare,
2244 .commit = external_snapshot_commit,
2245 .abort = external_snapshot_abort,
2246 .clean = external_snapshot_clean,
2247 },
2248 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2249 .instance_size = sizeof(DriveBackupState),
2250 .prepare = drive_backup_prepare,
2251 .commit = drive_backup_commit,
2252 .abort = drive_backup_abort,
2253 .clean = drive_backup_clean,
2254 },
2255 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2256 .instance_size = sizeof(BlockdevBackupState),
2257 .prepare = blockdev_backup_prepare,
2258 .commit = blockdev_backup_commit,
2259 .abort = blockdev_backup_abort,
2260 .clean = blockdev_backup_clean,
2261 },
2262 [TRANSACTION_ACTION_KIND_ABORT] = {
2263 .instance_size = sizeof(BlkActionState),
2264 .prepare = abort_prepare,
2265 .commit = abort_commit,
2266 },
2267 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2268 .instance_size = sizeof(InternalSnapshotState),
2269 .prepare = internal_snapshot_prepare,
2270 .abort = internal_snapshot_abort,
2271 .clean = internal_snapshot_clean,
2272 },
2273 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2274 .instance_size = sizeof(BlockDirtyBitmapState),
2275 .prepare = block_dirty_bitmap_add_prepare,
2276 .abort = block_dirty_bitmap_add_abort,
2277 },
2278 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2279 .instance_size = sizeof(BlockDirtyBitmapState),
2280 .prepare = block_dirty_bitmap_clear_prepare,
2281 .commit = block_dirty_bitmap_free_backup,
2282 .abort = block_dirty_bitmap_restore,
2283 },
2284 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2285 .instance_size = sizeof(BlockDirtyBitmapState),
2286 .prepare = block_dirty_bitmap_enable_prepare,
2287 .abort = block_dirty_bitmap_enable_abort,
2288 },
2289 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2290 .instance_size = sizeof(BlockDirtyBitmapState),
2291 .prepare = block_dirty_bitmap_disable_prepare,
2292 .abort = block_dirty_bitmap_disable_abort,
2293 },
2294 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2295 .instance_size = sizeof(BlockDirtyBitmapState),
2296 .prepare = block_dirty_bitmap_merge_prepare,
2297 .commit = block_dirty_bitmap_free_backup,
2298 .abort = block_dirty_bitmap_restore,
2299 },
2300 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE] = {
2301 .instance_size = sizeof(BlockDirtyBitmapState),
2302 .prepare = block_dirty_bitmap_remove_prepare,
2303 .commit = block_dirty_bitmap_remove_commit,
2304 .abort = block_dirty_bitmap_remove_abort,
2305 },
2306 /* Where are transactions for MIRROR, COMMIT and STREAM?
2307 * Although these blockjobs use transaction callbacks like the backup job,
2308 * these jobs do not necessarily adhere to transaction semantics.
2309 * These jobs may not fully undo all of their actions on abort, nor do they
2310 * necessarily work in transactions with more than one job in them.
2311 */
2312 };
2313
2314 /**
2315 * Allocate a TransactionProperties structure if necessary, and fill
2316 * that structure with desired defaults if they are unset.
2317 */
2318 static TransactionProperties *get_transaction_properties(
2319 TransactionProperties *props)
2320 {
2321 if (!props) {
2322 props = g_new0(TransactionProperties, 1);
2323 }
2324
2325 if (!props->has_completion_mode) {
2326 props->has_completion_mode = true;
2327 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2328 }
2329
2330 return props;
2331 }
2332
2333 /*
2334 * 'Atomic' group operations. The operations are performed as a set, and if
2335 * any fail then we roll back all operations in the group.
2336 */
2337 void qmp_transaction(TransactionActionList *dev_list,
2338 bool has_props,
2339 struct TransactionProperties *props,
2340 Error **errp)
2341 {
2342 TransactionActionList *dev_entry = dev_list;
2343 JobTxn *block_job_txn = NULL;
2344 BlkActionState *state, *next;
2345 Error *local_err = NULL;
2346
2347 QTAILQ_HEAD(, BlkActionState) snap_bdrv_states;
2348 QTAILQ_INIT(&snap_bdrv_states);
2349
2350 /* Does this transaction get canceled as a group on failure?
2351 * If not, we don't really need to make a JobTxn.
2352 */
2353 props = get_transaction_properties(props);
2354 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2355 block_job_txn = job_txn_new();
2356 }
2357
2358 /* drain all i/o before any operations */
2359 bdrv_drain_all();
2360
2361 /* We don't do anything in this loop that commits us to the operations */
2362 while (NULL != dev_entry) {
2363 TransactionAction *dev_info = NULL;
2364 const BlkActionOps *ops;
2365
2366 dev_info = dev_entry->value;
2367 dev_entry = dev_entry->next;
2368
2369 assert(dev_info->type < ARRAY_SIZE(actions));
2370
2371 ops = &actions[dev_info->type];
2372 assert(ops->instance_size > 0);
2373
2374 state = g_malloc0(ops->instance_size);
2375 state->ops = ops;
2376 state->action = dev_info;
2377 state->block_job_txn = block_job_txn;
2378 state->txn_props = props;
2379 QTAILQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2380
2381 state->ops->prepare(state, &local_err);
2382 if (local_err) {
2383 error_propagate(errp, local_err);
2384 goto delete_and_fail;
2385 }
2386 }
2387
2388 QTAILQ_FOREACH(state, &snap_bdrv_states, entry) {
2389 if (state->ops->commit) {
2390 state->ops->commit(state);
2391 }
2392 }
2393
2394 /* success */
2395 goto exit;
2396
2397 delete_and_fail:
2398 /* failure, and it is all-or-none; roll back all operations */
2399 QTAILQ_FOREACH_REVERSE(state, &snap_bdrv_states, entry) {
2400 if (state->ops->abort) {
2401 state->ops->abort(state);
2402 }
2403 }
2404 exit:
2405 QTAILQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2406 if (state->ops->clean) {
2407 state->ops->clean(state);
2408 }
2409 g_free(state);
2410 }
2411 if (!has_props) {
2412 qapi_free_TransactionProperties(props);
2413 }
2414 job_txn_unref(block_job_txn);
2415 }
2416
2417 void qmp_block_passwd(bool has_device, const char *device,
2418 bool has_node_name, const char *node_name,
2419 const char *password, Error **errp)
2420 {
2421 error_setg(errp,
2422 "Setting block passwords directly is no longer supported");
2423 }
2424
2425 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2426 const char *name,
2427 Error **errp)
2428 {
2429 BdrvDirtyBitmap *bitmap;
2430 BlockDriverState *bs;
2431 BlockDirtyBitmapSha256 *ret = NULL;
2432 char *sha256;
2433
2434 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2435 if (!bitmap || !bs) {
2436 return NULL;
2437 }
2438
2439 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2440 if (sha256 == NULL) {
2441 return NULL;
2442 }
2443
2444 ret = g_new(BlockDirtyBitmapSha256, 1);
2445 ret->sha256 = sha256;
2446
2447 return ret;
2448 }
2449
2450 void coroutine_fn qmp_block_resize(bool has_device, const char *device,
2451 bool has_node_name, const char *node_name,
2452 int64_t size, Error **errp)
2453 {
2454 Error *local_err = NULL;
2455 BlockBackend *blk;
2456 BlockDriverState *bs;
2457 AioContext *old_ctx;
2458
2459 bs = bdrv_lookup_bs(has_device ? device : NULL,
2460 has_node_name ? node_name : NULL,
2461 &local_err);
2462 if (local_err) {
2463 error_propagate(errp, local_err);
2464 return;
2465 }
2466
2467 if (size < 0) {
2468 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2469 return;
2470 }
2471
2472 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2473 error_setg(errp, QERR_DEVICE_IN_USE, device);
2474 return;
2475 }
2476
2477 blk = blk_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2478 if (!blk) {
2479 return;
2480 }
2481
2482 bdrv_co_lock(bs);
2483 bdrv_drained_begin(bs);
2484 bdrv_co_unlock(bs);
2485
2486 old_ctx = bdrv_co_enter(bs);
2487 blk_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2488 bdrv_co_leave(bs, old_ctx);
2489
2490 bdrv_co_lock(bs);
2491 bdrv_drained_end(bs);
2492 blk_unref(blk);
2493 bdrv_co_unlock(bs);
2494 }
2495
2496 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
2497 bool has_base, const char *base,
2498 bool has_base_node, const char *base_node,
2499 bool has_backing_file, const char *backing_file,
2500 bool has_bottom, const char *bottom,
2501 bool has_speed, int64_t speed,
2502 bool has_on_error, BlockdevOnError on_error,
2503 bool has_filter_node_name, const char *filter_node_name,
2504 bool has_auto_finalize, bool auto_finalize,
2505 bool has_auto_dismiss, bool auto_dismiss,
2506 Error **errp)
2507 {
2508 BlockDriverState *bs, *iter, *iter_end;
2509 BlockDriverState *base_bs = NULL;
2510 BlockDriverState *bottom_bs = NULL;
2511 AioContext *aio_context;
2512 Error *local_err = NULL;
2513 int job_flags = JOB_DEFAULT;
2514
2515 if (has_base && has_base_node) {
2516 error_setg(errp, "'base' and 'base-node' cannot be specified "
2517 "at the same time");
2518 return;
2519 }
2520
2521 if (has_base && has_bottom) {
2522 error_setg(errp, "'base' and 'bottom' cannot be specified "
2523 "at the same time");
2524 return;
2525 }
2526
2527 if (has_bottom && has_base_node) {
2528 error_setg(errp, "'bottom' and 'base-node' cannot be specified "
2529 "at the same time");
2530 return;
2531 }
2532
2533 if (!has_on_error) {
2534 on_error = BLOCKDEV_ON_ERROR_REPORT;
2535 }
2536
2537 bs = bdrv_lookup_bs(device, device, errp);
2538 if (!bs) {
2539 return;
2540 }
2541
2542 aio_context = bdrv_get_aio_context(bs);
2543 aio_context_acquire(aio_context);
2544
2545 if (has_base) {
2546 base_bs = bdrv_find_backing_image(bs, base);
2547 if (base_bs == NULL) {
2548 error_setg(errp, "Can't find '%s' in the backing chain", base);
2549 goto out;
2550 }
2551 assert(bdrv_get_aio_context(base_bs) == aio_context);
2552 }
2553
2554 if (has_base_node) {
2555 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2556 if (!base_bs) {
2557 goto out;
2558 }
2559 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2560 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2561 base_node, device);
2562 goto out;
2563 }
2564 assert(bdrv_get_aio_context(base_bs) == aio_context);
2565 bdrv_refresh_filename(base_bs);
2566 }
2567
2568 if (has_bottom) {
2569 bottom_bs = bdrv_lookup_bs(NULL, bottom, errp);
2570 if (!bottom_bs) {
2571 goto out;
2572 }
2573 if (!bottom_bs->drv) {
2574 error_setg(errp, "Node '%s' is not open", bottom);
2575 goto out;
2576 }
2577 if (bottom_bs->drv->is_filter) {
2578 error_setg(errp, "Node '%s' is a filter, use a non-filter node "
2579 "as 'bottom'", bottom);
2580 goto out;
2581 }
2582 if (!bdrv_chain_contains(bs, bottom_bs)) {
2583 error_setg(errp, "Node '%s' is not in a chain starting from '%s'",
2584 bottom, device);
2585 goto out;
2586 }
2587 assert(bdrv_get_aio_context(bottom_bs) == aio_context);
2588 }
2589
2590 /*
2591 * Check for op blockers in the whole chain between bs and base (or bottom)
2592 */
2593 iter_end = has_bottom ? bdrv_filter_or_cow_bs(bottom_bs) : base_bs;
2594 for (iter = bs; iter && iter != iter_end;
2595 iter = bdrv_filter_or_cow_bs(iter))
2596 {
2597 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2598 goto out;
2599 }
2600 }
2601
2602 /* if we are streaming the entire chain, the result will have no backing
2603 * file, and specifying one is therefore an error */
2604 if (base_bs == NULL && has_backing_file) {
2605 error_setg(errp, "backing file specified, but streaming the "
2606 "entire chain");
2607 goto out;
2608 }
2609
2610 if (has_auto_finalize && !auto_finalize) {
2611 job_flags |= JOB_MANUAL_FINALIZE;
2612 }
2613 if (has_auto_dismiss && !auto_dismiss) {
2614 job_flags |= JOB_MANUAL_DISMISS;
2615 }
2616
2617 stream_start(has_job_id ? job_id : NULL, bs, base_bs, backing_file,
2618 bottom_bs, job_flags, has_speed ? speed : 0, on_error,
2619 filter_node_name, &local_err);
2620 if (local_err) {
2621 error_propagate(errp, local_err);
2622 goto out;
2623 }
2624
2625 trace_qmp_block_stream(bs);
2626
2627 out:
2628 aio_context_release(aio_context);
2629 }
2630
2631 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
2632 bool has_base_node, const char *base_node,
2633 bool has_base, const char *base,
2634 bool has_top_node, const char *top_node,
2635 bool has_top, const char *top,
2636 bool has_backing_file, const char *backing_file,
2637 bool has_speed, int64_t speed,
2638 bool has_on_error, BlockdevOnError on_error,
2639 bool has_filter_node_name, const char *filter_node_name,
2640 bool has_auto_finalize, bool auto_finalize,
2641 bool has_auto_dismiss, bool auto_dismiss,
2642 Error **errp)
2643 {
2644 BlockDriverState *bs;
2645 BlockDriverState *iter;
2646 BlockDriverState *base_bs, *top_bs;
2647 AioContext *aio_context;
2648 Error *local_err = NULL;
2649 int job_flags = JOB_DEFAULT;
2650 uint64_t top_perm, top_shared;
2651
2652 if (!has_speed) {
2653 speed = 0;
2654 }
2655 if (!has_on_error) {
2656 on_error = BLOCKDEV_ON_ERROR_REPORT;
2657 }
2658 if (!has_filter_node_name) {
2659 filter_node_name = NULL;
2660 }
2661 if (has_auto_finalize && !auto_finalize) {
2662 job_flags |= JOB_MANUAL_FINALIZE;
2663 }
2664 if (has_auto_dismiss && !auto_dismiss) {
2665 job_flags |= JOB_MANUAL_DISMISS;
2666 }
2667
2668 /* Important Note:
2669 * libvirt relies on the DeviceNotFound error class in order to probe for
2670 * live commit feature versions; for this to work, we must make sure to
2671 * perform the device lookup before any generic errors that may occur in a
2672 * scenario in which all optional arguments are omitted. */
2673 bs = qmp_get_root_bs(device, &local_err);
2674 if (!bs) {
2675 bs = bdrv_lookup_bs(device, device, NULL);
2676 if (!bs) {
2677 error_free(local_err);
2678 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2679 "Device '%s' not found", device);
2680 } else {
2681 error_propagate(errp, local_err);
2682 }
2683 return;
2684 }
2685
2686 aio_context = bdrv_get_aio_context(bs);
2687 aio_context_acquire(aio_context);
2688
2689 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2690 goto out;
2691 }
2692
2693 /* default top_bs is the active layer */
2694 top_bs = bs;
2695
2696 if (has_top_node && has_top) {
2697 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2698 goto out;
2699 } else if (has_top_node) {
2700 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2701 if (top_bs == NULL) {
2702 goto out;
2703 }
2704 if (!bdrv_chain_contains(bs, top_bs)) {
2705 error_setg(errp, "'%s' is not in this backing file chain",
2706 top_node);
2707 goto out;
2708 }
2709 } else if (has_top && top) {
2710 /* This strcmp() is just a shortcut, there is no need to
2711 * refresh @bs's filename. If it mismatches,
2712 * bdrv_find_backing_image() will do the refresh and may still
2713 * return @bs. */
2714 if (strcmp(bs->filename, top) != 0) {
2715 top_bs = bdrv_find_backing_image(bs, top);
2716 }
2717 }
2718
2719 if (top_bs == NULL) {
2720 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2721 goto out;
2722 }
2723
2724 assert(bdrv_get_aio_context(top_bs) == aio_context);
2725
2726 if (has_base_node && has_base) {
2727 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2728 goto out;
2729 } else if (has_base_node) {
2730 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2731 if (base_bs == NULL) {
2732 goto out;
2733 }
2734 if (!bdrv_chain_contains(top_bs, base_bs)) {
2735 error_setg(errp, "'%s' is not in this backing file chain",
2736 base_node);
2737 goto out;
2738 }
2739 } else if (has_base && base) {
2740 base_bs = bdrv_find_backing_image(top_bs, base);
2741 if (base_bs == NULL) {
2742 error_setg(errp, "Can't find '%s' in the backing chain", base);
2743 goto out;
2744 }
2745 } else {
2746 base_bs = bdrv_find_base(top_bs);
2747 if (base_bs == NULL) {
2748 error_setg(errp, "There is no backimg image");
2749 goto out;
2750 }
2751 }
2752
2753 assert(bdrv_get_aio_context(base_bs) == aio_context);
2754
2755 for (iter = top_bs; iter != bdrv_filter_or_cow_bs(base_bs);
2756 iter = bdrv_filter_or_cow_bs(iter))
2757 {
2758 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2759 goto out;
2760 }
2761 }
2762
2763 /* Do not allow attempts to commit an image into itself */
2764 if (top_bs == base_bs) {
2765 error_setg(errp, "cannot commit an image into itself");
2766 goto out;
2767 }
2768
2769 /*
2770 * Active commit is required if and only if someone has taken a
2771 * WRITE permission on the top node. Historically, we have always
2772 * used active commit for top nodes, so continue that practice
2773 * lest we possibly break clients that rely on this behavior, e.g.
2774 * to later attach this node to a writing parent.
2775 * (Active commit is never really wrong.)
2776 */
2777 bdrv_get_cumulative_perm(top_bs, &top_perm, &top_shared);
2778 if (top_perm & BLK_PERM_WRITE ||
2779 bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs))
2780 {
2781 if (has_backing_file) {
2782 if (bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs)) {
2783 error_setg(errp, "'backing-file' specified,"
2784 " but 'top' is the active layer");
2785 } else {
2786 error_setg(errp, "'backing-file' specified, but 'top' has a "
2787 "writer on it");
2788 }
2789 goto out;
2790 }
2791 if (!has_job_id) {
2792 /*
2793 * Emulate here what block_job_create() does, because it
2794 * is possible that @bs != @top_bs (the block job should
2795 * be named after @bs, even if @top_bs is the actual
2796 * source)
2797 */
2798 job_id = bdrv_get_device_name(bs);
2799 }
2800 commit_active_start(job_id, top_bs, base_bs, job_flags, speed, on_error,
2801 filter_node_name, NULL, NULL, false, &local_err);
2802 } else {
2803 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2804 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2805 goto out;
2806 }
2807 commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, job_flags,
2808 speed, on_error, has_backing_file ? backing_file : NULL,
2809 filter_node_name, &local_err);
2810 }
2811 if (local_err != NULL) {
2812 error_propagate(errp, local_err);
2813 goto out;
2814 }
2815
2816 out:
2817 aio_context_release(aio_context);
2818 }
2819
2820 /* Common QMP interface for drive-backup and blockdev-backup */
2821 static BlockJob *do_backup_common(BackupCommon *backup,
2822 BlockDriverState *bs,
2823 BlockDriverState *target_bs,
2824 AioContext *aio_context,
2825 JobTxn *txn, Error **errp)
2826 {
2827 BlockJob *job = NULL;
2828 BdrvDirtyBitmap *bmap = NULL;
2829 BackupPerf perf = { .max_workers = 64 };
2830 int job_flags = JOB_DEFAULT;
2831
2832 if (!backup->has_speed) {
2833 backup->speed = 0;
2834 }
2835 if (!backup->has_on_source_error) {
2836 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2837 }
2838 if (!backup->has_on_target_error) {
2839 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2840 }
2841 if (!backup->has_job_id) {
2842 backup->job_id = NULL;
2843 }
2844 if (!backup->has_auto_finalize) {
2845 backup->auto_finalize = true;
2846 }
2847 if (!backup->has_auto_dismiss) {
2848 backup->auto_dismiss = true;
2849 }
2850 if (!backup->has_compress) {
2851 backup->compress = false;
2852 }
2853
2854 if (backup->x_perf) {
2855 if (backup->x_perf->has_use_copy_range) {
2856 perf.use_copy_range = backup->x_perf->use_copy_range;
2857 }
2858 if (backup->x_perf->has_max_workers) {
2859 perf.max_workers = backup->x_perf->max_workers;
2860 }
2861 if (backup->x_perf->has_max_chunk) {
2862 perf.max_chunk = backup->x_perf->max_chunk;
2863 }
2864 }
2865
2866 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2867 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2868 /* done before desugaring 'incremental' to print the right message */
2869 if (!backup->has_bitmap) {
2870 error_setg(errp, "must provide a valid bitmap name for "
2871 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2872 return NULL;
2873 }
2874 }
2875
2876 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2877 if (backup->has_bitmap_mode &&
2878 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2879 error_setg(errp, "Bitmap sync mode must be '%s' "
2880 "when using sync mode '%s'",
2881 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2882 MirrorSyncMode_str(backup->sync));
2883 return NULL;
2884 }
2885 backup->has_bitmap_mode = true;
2886 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2887 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2888 }
2889
2890 if (backup->has_bitmap) {
2891 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2892 if (!bmap) {
2893 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2894 return NULL;
2895 }
2896 if (!backup->has_bitmap_mode) {
2897 error_setg(errp, "Bitmap sync mode must be given "
2898 "when providing a bitmap");
2899 return NULL;
2900 }
2901 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2902 return NULL;
2903 }
2904
2905 /* This does not produce a useful bitmap artifact: */
2906 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2907 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2908 " outputs", MirrorSyncMode_str(backup->sync));
2909 return NULL;
2910 }
2911
2912 /* If the bitmap isn't used for input or output, this is useless: */
2913 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2914 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2915 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2916 " when combined with sync mode '%s'",
2917 BitmapSyncMode_str(backup->bitmap_mode),
2918 MirrorSyncMode_str(backup->sync));
2919 return NULL;
2920 }
2921 }
2922
2923 if (!backup->has_bitmap && backup->has_bitmap_mode) {
2924 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2925 return NULL;
2926 }
2927
2928 if (!backup->auto_finalize) {
2929 job_flags |= JOB_MANUAL_FINALIZE;
2930 }
2931 if (!backup->auto_dismiss) {
2932 job_flags |= JOB_MANUAL_DISMISS;
2933 }
2934
2935 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2936 backup->sync, bmap, backup->bitmap_mode,
2937 backup->compress,
2938 backup->filter_node_name,
2939 &perf,
2940 backup->on_source_error,
2941 backup->on_target_error,
2942 job_flags, NULL, NULL, txn, errp);
2943 return job;
2944 }
2945
2946 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2947 {
2948 TransactionAction action = {
2949 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2950 .u.drive_backup.data = backup,
2951 };
2952 blockdev_do_action(&action, errp);
2953 }
2954
2955 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2956 bool flat,
2957 Error **errp)
2958 {
2959 bool return_flat = has_flat && flat;
2960
2961 return bdrv_named_nodes_list(return_flat, errp);
2962 }
2963
2964 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2965 {
2966 return bdrv_get_xdbg_block_graph(errp);
2967 }
2968
2969 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2970 {
2971 TransactionAction action = {
2972 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2973 .u.blockdev_backup.data = backup,
2974 };
2975 blockdev_do_action(&action, errp);
2976 }
2977
2978 /* Parameter check and block job starting for drive mirroring.
2979 * Caller should hold @device and @target's aio context (must be the same).
2980 **/
2981 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2982 BlockDriverState *target,
2983 bool has_replaces, const char *replaces,
2984 enum MirrorSyncMode sync,
2985 BlockMirrorBackingMode backing_mode,
2986 bool zero_target,
2987 bool has_speed, int64_t speed,
2988 bool has_granularity, uint32_t granularity,
2989 bool has_buf_size, int64_t buf_size,
2990 bool has_on_source_error,
2991 BlockdevOnError on_source_error,
2992 bool has_on_target_error,
2993 BlockdevOnError on_target_error,
2994 bool has_unmap, bool unmap,
2995 bool has_filter_node_name,
2996 const char *filter_node_name,
2997 bool has_copy_mode, MirrorCopyMode copy_mode,
2998 bool has_auto_finalize, bool auto_finalize,
2999 bool has_auto_dismiss, bool auto_dismiss,
3000 Error **errp)
3001 {
3002 BlockDriverState *unfiltered_bs;
3003 int job_flags = JOB_DEFAULT;
3004
3005 if (!has_speed) {
3006 speed = 0;
3007 }
3008 if (!has_on_source_error) {
3009 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3010 }
3011 if (!has_on_target_error) {
3012 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3013 }
3014 if (!has_granularity) {
3015 granularity = 0;
3016 }
3017 if (!has_buf_size) {
3018 buf_size = 0;
3019 }
3020 if (!has_unmap) {
3021 unmap = true;
3022 }
3023 if (!has_filter_node_name) {
3024 filter_node_name = NULL;
3025 }
3026 if (!has_copy_mode) {
3027 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
3028 }
3029 if (has_auto_finalize && !auto_finalize) {
3030 job_flags |= JOB_MANUAL_FINALIZE;
3031 }
3032 if (has_auto_dismiss && !auto_dismiss) {
3033 job_flags |= JOB_MANUAL_DISMISS;
3034 }
3035
3036 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3037 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3038 "a value in range [512B, 64MB]");
3039 return;
3040 }
3041 if (granularity & (granularity - 1)) {
3042 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3043 "a power of 2");
3044 return;
3045 }
3046
3047 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3048 return;
3049 }
3050 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3051 return;
3052 }
3053
3054 if (!bdrv_backing_chain_next(bs) && sync == MIRROR_SYNC_MODE_TOP) {
3055 sync = MIRROR_SYNC_MODE_FULL;
3056 }
3057
3058 if (!has_replaces) {
3059 /* We want to mirror from @bs, but keep implicit filters on top */
3060 unfiltered_bs = bdrv_skip_implicit_filters(bs);
3061 if (unfiltered_bs != bs) {
3062 replaces = unfiltered_bs->node_name;
3063 has_replaces = true;
3064 }
3065 }
3066
3067 if (has_replaces) {
3068 BlockDriverState *to_replace_bs;
3069 AioContext *replace_aio_context;
3070 int64_t bs_size, replace_size;
3071
3072 bs_size = bdrv_getlength(bs);
3073 if (bs_size < 0) {
3074 error_setg_errno(errp, -bs_size, "Failed to query device's size");
3075 return;
3076 }
3077
3078 to_replace_bs = check_to_replace_node(bs, replaces, errp);
3079 if (!to_replace_bs) {
3080 return;
3081 }
3082
3083 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3084 aio_context_acquire(replace_aio_context);
3085 replace_size = bdrv_getlength(to_replace_bs);
3086 aio_context_release(replace_aio_context);
3087
3088 if (replace_size < 0) {
3089 error_setg_errno(errp, -replace_size,
3090 "Failed to query the replacement node's size");
3091 return;
3092 }
3093 if (bs_size != replace_size) {
3094 error_setg(errp, "cannot replace image with a mirror image of "
3095 "different size");
3096 return;
3097 }
3098 }
3099
3100 /* pass the node name to replace to mirror start since it's loose coupling
3101 * and will allow to check whether the node still exist at mirror completion
3102 */
3103 mirror_start(job_id, bs, target,
3104 has_replaces ? replaces : NULL, job_flags,
3105 speed, granularity, buf_size, sync, backing_mode, zero_target,
3106 on_source_error, on_target_error, unmap, filter_node_name,
3107 copy_mode, errp);
3108 }
3109
3110 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3111 {
3112 BlockDriverState *bs;
3113 BlockDriverState *target_backing_bs, *target_bs;
3114 AioContext *aio_context;
3115 AioContext *old_context;
3116 BlockMirrorBackingMode backing_mode;
3117 Error *local_err = NULL;
3118 QDict *options = NULL;
3119 int flags;
3120 int64_t size;
3121 const char *format = arg->format;
3122 bool zero_target;
3123 int ret;
3124
3125 bs = qmp_get_root_bs(arg->device, errp);
3126 if (!bs) {
3127 return;
3128 }
3129
3130 /* Early check to avoid creating target */
3131 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3132 return;
3133 }
3134
3135 aio_context = bdrv_get_aio_context(bs);
3136 aio_context_acquire(aio_context);
3137
3138 if (!arg->has_mode) {
3139 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3140 }
3141
3142 if (!arg->has_format) {
3143 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3144 ? NULL : bs->drv->format_name);
3145 }
3146
3147 flags = bs->open_flags | BDRV_O_RDWR;
3148 target_backing_bs = bdrv_cow_bs(bdrv_skip_filters(bs));
3149 if (!target_backing_bs && arg->sync == MIRROR_SYNC_MODE_TOP) {
3150 arg->sync = MIRROR_SYNC_MODE_FULL;
3151 }
3152 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3153 target_backing_bs = bs;
3154 }
3155
3156 size = bdrv_getlength(bs);
3157 if (size < 0) {
3158 error_setg_errno(errp, -size, "bdrv_getlength failed");
3159 goto out;
3160 }
3161
3162 if (arg->has_replaces) {
3163 if (!arg->has_node_name) {
3164 error_setg(errp, "a node-name must be provided when replacing a"
3165 " named node of the graph");
3166 goto out;
3167 }
3168 }
3169
3170 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3171 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3172 } else {
3173 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3174 }
3175
3176 /* Don't open backing image in create() */
3177 flags |= BDRV_O_NO_BACKING;
3178
3179 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !target_backing_bs)
3180 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3181 {
3182 /* create new image w/o backing file */
3183 assert(format);
3184 bdrv_img_create(arg->target, format,
3185 NULL, NULL, NULL, size, flags, false, &local_err);
3186 } else {
3187 /* Implicit filters should not appear in the filename */
3188 BlockDriverState *explicit_backing =
3189 bdrv_skip_implicit_filters(target_backing_bs);
3190
3191 switch (arg->mode) {
3192 case NEW_IMAGE_MODE_EXISTING:
3193 break;
3194 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3195 /* create new image with backing file */
3196 bdrv_refresh_filename(explicit_backing);
3197 bdrv_img_create(arg->target, format,
3198 explicit_backing->filename,
3199 explicit_backing->drv->format_name,
3200 NULL, size, flags, false, &local_err);
3201 break;
3202 default:
3203 abort();
3204 }
3205 }
3206
3207 if (local_err) {
3208 error_propagate(errp, local_err);
3209 goto out;
3210 }
3211
3212 options = qdict_new();
3213 if (arg->has_node_name) {
3214 qdict_put_str(options, "node-name", arg->node_name);
3215 }
3216 if (format) {
3217 qdict_put_str(options, "driver", format);
3218 }
3219
3220 /* Mirroring takes care of copy-on-write using the source's backing
3221 * file.
3222 */
3223 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3224 if (!target_bs) {
3225 goto out;
3226 }
3227
3228 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3229 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3230 !bdrv_has_zero_init(target_bs)));
3231
3232
3233 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3234 old_context = bdrv_get_aio_context(target_bs);
3235 aio_context_release(aio_context);
3236 aio_context_acquire(old_context);
3237
3238 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3239 if (ret < 0) {
3240 bdrv_unref(target_bs);
3241 aio_context_release(old_context);
3242 return;
3243 }
3244
3245 aio_context_release(old_context);
3246 aio_context_acquire(aio_context);
3247
3248 blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3249 arg->has_replaces, arg->replaces, arg->sync,
3250 backing_mode, zero_target,
3251 arg->has_speed, arg->speed,
3252 arg->has_granularity, arg->granularity,
3253 arg->has_buf_size, arg->buf_size,
3254 arg->has_on_source_error, arg->on_source_error,
3255 arg->has_on_target_error, arg->on_target_error,
3256 arg->has_unmap, arg->unmap,
3257 false, NULL,
3258 arg->has_copy_mode, arg->copy_mode,
3259 arg->has_auto_finalize, arg->auto_finalize,
3260 arg->has_auto_dismiss, arg->auto_dismiss,
3261 errp);
3262 bdrv_unref(target_bs);
3263 out:
3264 aio_context_release(aio_context);
3265 }
3266
3267 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3268 const char *device, const char *target,
3269 bool has_replaces, const char *replaces,
3270 MirrorSyncMode sync,
3271 bool has_speed, int64_t speed,
3272 bool has_granularity, uint32_t granularity,
3273 bool has_buf_size, int64_t buf_size,
3274 bool has_on_source_error,
3275 BlockdevOnError on_source_error,
3276 bool has_on_target_error,
3277 BlockdevOnError on_target_error,
3278 bool has_filter_node_name,
3279 const char *filter_node_name,
3280 bool has_copy_mode, MirrorCopyMode copy_mode,
3281 bool has_auto_finalize, bool auto_finalize,
3282 bool has_auto_dismiss, bool auto_dismiss,
3283 Error **errp)
3284 {
3285 BlockDriverState *bs;
3286 BlockDriverState *target_bs;
3287 AioContext *aio_context;
3288 AioContext *old_context;
3289 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3290 bool zero_target;
3291 int ret;
3292
3293 bs = qmp_get_root_bs(device, errp);
3294 if (!bs) {
3295 return;
3296 }
3297
3298 target_bs = bdrv_lookup_bs(target, target, errp);
3299 if (!target_bs) {
3300 return;
3301 }
3302
3303 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3304
3305 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3306 old_context = bdrv_get_aio_context(target_bs);
3307 aio_context = bdrv_get_aio_context(bs);
3308 aio_context_acquire(old_context);
3309
3310 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3311
3312 aio_context_release(old_context);
3313 aio_context_acquire(aio_context);
3314
3315 if (ret < 0) {
3316 goto out;
3317 }
3318
3319 blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
3320 has_replaces, replaces, sync, backing_mode,
3321 zero_target, has_speed, speed,
3322 has_granularity, granularity,
3323 has_buf_size, buf_size,
3324 has_on_source_error, on_source_error,
3325 has_on_target_error, on_target_error,
3326 true, true,
3327 has_filter_node_name, filter_node_name,
3328 has_copy_mode, copy_mode,
3329 has_auto_finalize, auto_finalize,
3330 has_auto_dismiss, auto_dismiss,
3331 errp);
3332 out:
3333 aio_context_release(aio_context);
3334 }
3335
3336 /* Get a block job using its ID and acquire its AioContext */
3337 static BlockJob *find_block_job(const char *id, AioContext **aio_context,
3338 Error **errp)
3339 {
3340 BlockJob *job;
3341
3342 assert(id != NULL);
3343
3344 *aio_context = NULL;
3345
3346 job = block_job_get(id);
3347
3348 if (!job) {
3349 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3350 "Block job '%s' not found", id);
3351 return NULL;
3352 }
3353
3354 *aio_context = blk_get_aio_context(job->blk);
3355 aio_context_acquire(*aio_context);
3356
3357 return job;
3358 }
3359
3360 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3361 {
3362 AioContext *aio_context;
3363 BlockJob *job = find_block_job(device, &aio_context, errp);
3364
3365 if (!job) {
3366 return;
3367 }
3368
3369 block_job_set_speed(job, speed, errp);
3370 aio_context_release(aio_context);
3371 }
3372
3373 void qmp_block_job_cancel(const char *device,
3374 bool has_force, bool force, Error **errp)
3375 {
3376 AioContext *aio_context;
3377 BlockJob *job = find_block_job(device, &aio_context, errp);
3378
3379 if (!job) {
3380 return;
3381 }
3382
3383 if (!has_force) {
3384 force = false;
3385 }
3386
3387 if (job_user_paused(&job->job) && !force) {
3388 error_setg(errp, "The block job for device '%s' is currently paused",
3389 device);
3390 goto out;
3391 }
3392
3393 trace_qmp_block_job_cancel(job);
3394 job_user_cancel(&job->job, force, errp);
3395 out:
3396 aio_context_release(aio_context);
3397 }
3398
3399 void qmp_block_job_pause(const char *device, Error **errp)
3400 {
3401 AioContext *aio_context;
3402 BlockJob *job = find_block_job(device, &aio_context, errp);
3403
3404 if (!job) {
3405 return;
3406 }
3407
3408 trace_qmp_block_job_pause(job);
3409 job_user_pause(&job->job, errp);
3410 aio_context_release(aio_context);
3411 }
3412
3413 void qmp_block_job_resume(const char *device, Error **errp)
3414 {
3415 AioContext *aio_context;
3416 BlockJob *job = find_block_job(device, &aio_context, errp);
3417
3418 if (!job) {
3419 return;
3420 }
3421
3422 trace_qmp_block_job_resume(job);
3423 job_user_resume(&job->job, errp);
3424 aio_context_release(aio_context);
3425 }
3426
3427 void qmp_block_job_complete(const char *device, Error **errp)
3428 {
3429 AioContext *aio_context;
3430 BlockJob *job = find_block_job(device, &aio_context, errp);
3431
3432 if (!job) {
3433 return;
3434 }
3435
3436 trace_qmp_block_job_complete(job);
3437 job_complete(&job->job, errp);
3438 aio_context_release(aio_context);
3439 }
3440
3441 void qmp_block_job_finalize(const char *id, Error **errp)
3442 {
3443 AioContext *aio_context;
3444 BlockJob *job = find_block_job(id, &aio_context, errp);
3445
3446 if (!job) {
3447 return;
3448 }
3449
3450 trace_qmp_block_job_finalize(job);
3451 job_ref(&job->job);
3452 job_finalize(&job->job, errp);
3453
3454 /*
3455 * Job's context might have changed via job_finalize (and job_txn_apply
3456 * automatically acquires the new one), so make sure we release the correct
3457 * one.
3458 */
3459 aio_context = blk_get_aio_context(job->blk);
3460 job_unref(&job->job);
3461 aio_context_release(aio_context);
3462 }
3463
3464 void qmp_block_job_dismiss(const char *id, Error **errp)
3465 {
3466 AioContext *aio_context;
3467 BlockJob *bjob = find_block_job(id, &aio_context, errp);
3468 Job *job;
3469
3470 if (!bjob) {
3471 return;
3472 }
3473
3474 trace_qmp_block_job_dismiss(bjob);
3475 job = &bjob->job;
3476 job_dismiss(&job, errp);
3477 aio_context_release(aio_context);
3478 }
3479
3480 void qmp_change_backing_file(const char *device,
3481 const char *image_node_name,
3482 const char *backing_file,
3483 Error **errp)
3484 {
3485 BlockDriverState *bs = NULL;
3486 AioContext *aio_context;
3487 BlockDriverState *image_bs = NULL;
3488 Error *local_err = NULL;
3489 bool ro;
3490 int ret;
3491
3492 bs = qmp_get_root_bs(device, errp);
3493 if (!bs) {
3494 return;
3495 }
3496
3497 aio_context = bdrv_get_aio_context(bs);
3498 aio_context_acquire(aio_context);
3499
3500 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3501 if (local_err) {
3502 error_propagate(errp, local_err);
3503 goto out;
3504 }
3505
3506 if (!image_bs) {
3507 error_setg(errp, "image file not found");
3508 goto out;
3509 }
3510
3511 if (bdrv_find_base(image_bs) == image_bs) {
3512 error_setg(errp, "not allowing backing file change on an image "
3513 "without a backing file");
3514 goto out;
3515 }
3516
3517 /* even though we are not necessarily operating on bs, we need it to
3518 * determine if block ops are currently prohibited on the chain */
3519 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3520 goto out;
3521 }
3522
3523 /* final sanity check */
3524 if (!bdrv_chain_contains(bs, image_bs)) {
3525 error_setg(errp, "'%s' and image file are not in the same chain",
3526 device);
3527 goto out;
3528 }
3529
3530 /* if not r/w, reopen to make r/w */
3531 ro = bdrv_is_read_only(image_bs);
3532
3533 if (ro) {
3534 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3535 goto out;
3536 }
3537 }
3538
3539 ret = bdrv_change_backing_file(image_bs, backing_file,
3540 image_bs->drv ? image_bs->drv->format_name : "",
3541 false);
3542
3543 if (ret < 0) {
3544 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3545 backing_file);
3546 /* don't exit here, so we can try to restore open flags if
3547 * appropriate */
3548 }
3549
3550 if (ro) {
3551 bdrv_reopen_set_read_only(image_bs, true, errp);
3552 }
3553
3554 out:
3555 aio_context_release(aio_context);
3556 }
3557
3558 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3559 {
3560 BlockDriverState *bs;
3561 QObject *obj;
3562 Visitor *v = qobject_output_visitor_new(&obj);
3563 QDict *qdict;
3564
3565 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3566 visit_complete(v, &obj);
3567 qdict = qobject_to(QDict, obj);
3568
3569 qdict_flatten(qdict);
3570
3571 if (!qdict_get_try_str(qdict, "node-name")) {
3572 error_setg(errp, "'node-name' must be specified for the root node");
3573 goto fail;
3574 }
3575
3576 bs = bds_tree_init(qdict, errp);
3577 if (!bs) {
3578 goto fail;
3579 }
3580
3581 bdrv_set_monitor_owned(bs);
3582
3583 fail:
3584 visit_free(v);
3585 }
3586
3587 void qmp_x_blockdev_reopen(BlockdevOptions *options, Error **errp)
3588 {
3589 BlockDriverState *bs;
3590 AioContext *ctx;
3591 QObject *obj;
3592 Visitor *v = qobject_output_visitor_new(&obj);
3593 BlockReopenQueue *queue;
3594 QDict *qdict;
3595
3596 /* Check for the selected node name */
3597 if (!options->has_node_name) {
3598 error_setg(errp, "node-name not specified");
3599 goto fail;
3600 }
3601
3602 bs = bdrv_find_node(options->node_name);
3603 if (!bs) {
3604 error_setg(errp, "Failed to find node with node-name='%s'",
3605 options->node_name);
3606 goto fail;
3607 }
3608
3609 /* Put all options in a QDict and flatten it */
3610 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3611 visit_complete(v, &obj);
3612 qdict = qobject_to(QDict, obj);
3613
3614 qdict_flatten(qdict);
3615
3616 /* Perform the reopen operation */
3617 ctx = bdrv_get_aio_context(bs);
3618 aio_context_acquire(ctx);
3619 bdrv_subtree_drained_begin(bs);
3620 queue = bdrv_reopen_queue(NULL, bs, qdict, false);
3621 bdrv_reopen_multiple(queue, errp);
3622 bdrv_subtree_drained_end(bs);
3623 aio_context_release(ctx);
3624
3625 fail:
3626 visit_free(v);
3627 }
3628
3629 void qmp_blockdev_del(const char *node_name, Error **errp)
3630 {
3631 AioContext *aio_context;
3632 BlockDriverState *bs;
3633
3634 bs = bdrv_find_node(node_name);
3635 if (!bs) {
3636 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3637 return;
3638 }
3639 if (bdrv_has_blk(bs)) {
3640 error_setg(errp, "Node %s is in use", node_name);
3641 return;
3642 }
3643 aio_context = bdrv_get_aio_context(bs);
3644 aio_context_acquire(aio_context);
3645
3646 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3647 goto out;
3648 }
3649
3650 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3651 error_setg(errp, "Node %s is not owned by the monitor",
3652 bs->node_name);
3653 goto out;
3654 }
3655
3656 if (bs->refcnt > 1) {
3657 error_setg(errp, "Block device %s is in use",
3658 bdrv_get_device_or_node_name(bs));
3659 goto out;
3660 }
3661
3662 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3663 bdrv_unref(bs);
3664
3665 out:
3666 aio_context_release(aio_context);
3667 }
3668
3669 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3670 const char *child_name)
3671 {
3672 BdrvChild *child;
3673
3674 QLIST_FOREACH(child, &parent_bs->children, next) {
3675 if (strcmp(child->name, child_name) == 0) {
3676 return child;
3677 }
3678 }
3679
3680 return NULL;
3681 }
3682
3683 void qmp_x_blockdev_change(const char *parent, bool has_child,
3684 const char *child, bool has_node,
3685 const char *node, Error **errp)
3686 {
3687 BlockDriverState *parent_bs, *new_bs = NULL;
3688 BdrvChild *p_child;
3689
3690 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3691 if (!parent_bs) {
3692 return;
3693 }
3694
3695 if (has_child == has_node) {
3696 if (has_child) {
3697 error_setg(errp, "The parameters child and node are in conflict");
3698 } else {
3699 error_setg(errp, "Either child or node must be specified");
3700 }
3701 return;
3702 }
3703
3704 if (has_child) {
3705 p_child = bdrv_find_child(parent_bs, child);
3706 if (!p_child) {
3707 error_setg(errp, "Node '%s' does not have child '%s'",
3708 parent, child);
3709 return;
3710 }
3711 bdrv_del_child(parent_bs, p_child, errp);
3712 }
3713
3714 if (has_node) {
3715 new_bs = bdrv_find_node(node);
3716 if (!new_bs) {
3717 error_setg(errp, "Node '%s' not found", node);
3718 return;
3719 }
3720 bdrv_add_child(parent_bs, new_bs, errp);
3721 }
3722 }
3723
3724 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3725 {
3726 BlockJobInfoList *head = NULL, **tail = &head;
3727 BlockJob *job;
3728
3729 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
3730 BlockJobInfo *value;
3731 AioContext *aio_context;
3732
3733 if (block_job_is_internal(job)) {
3734 continue;
3735 }
3736 aio_context = blk_get_aio_context(job->blk);
3737 aio_context_acquire(aio_context);
3738 value = block_job_query(job, errp);
3739 aio_context_release(aio_context);
3740 if (!value) {
3741 qapi_free_BlockJobInfoList(head);
3742 return NULL;
3743 }
3744 QAPI_LIST_APPEND(tail, value);
3745 }
3746
3747 return head;
3748 }
3749
3750 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3751 bool has_force, bool force, Error **errp)
3752 {
3753 AioContext *old_context;
3754 AioContext *new_context;
3755 BlockDriverState *bs;
3756
3757 bs = bdrv_find_node(node_name);
3758 if (!bs) {
3759 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3760 return;
3761 }
3762
3763 /* Protects against accidents. */
3764 if (!(has_force && force) && bdrv_has_blk(bs)) {
3765 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3766 "be in use (use force=true to override this check)",
3767 node_name);
3768 return;
3769 }
3770
3771 if (iothread->type == QTYPE_QSTRING) {
3772 IOThread *obj = iothread_by_id(iothread->u.s);
3773 if (!obj) {
3774 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3775 return;
3776 }
3777
3778 new_context = iothread_get_aio_context(obj);
3779 } else {
3780 new_context = qemu_get_aio_context();
3781 }
3782
3783 old_context = bdrv_get_aio_context(bs);
3784 aio_context_acquire(old_context);
3785
3786 bdrv_try_set_aio_context(bs, new_context, errp);
3787
3788 aio_context_release(old_context);
3789 }
3790
3791 QemuOptsList qemu_common_drive_opts = {
3792 .name = "drive",
3793 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3794 .desc = {
3795 {
3796 .name = "snapshot",
3797 .type = QEMU_OPT_BOOL,
3798 .help = "enable/disable snapshot mode",
3799 },{
3800 .name = "aio",
3801 .type = QEMU_OPT_STRING,
3802 .help = "host AIO implementation (threads, native, io_uring)",
3803 },{
3804 .name = BDRV_OPT_CACHE_WB,
3805 .type = QEMU_OPT_BOOL,
3806 .help = "Enable writeback mode",
3807 },{
3808 .name = "format",
3809 .type = QEMU_OPT_STRING,
3810 .help = "disk format (raw, qcow2, ...)",
3811 },{
3812 .name = "rerror",
3813 .type = QEMU_OPT_STRING,
3814 .help = "read error action",
3815 },{
3816 .name = "werror",
3817 .type = QEMU_OPT_STRING,
3818 .help = "write error action",
3819 },{
3820 .name = BDRV_OPT_READ_ONLY,
3821 .type = QEMU_OPT_BOOL,
3822 .help = "open drive file as read-only",
3823 },
3824
3825 THROTTLE_OPTS,
3826
3827 {
3828 .name = "throttling.group",
3829 .type = QEMU_OPT_STRING,
3830 .help = "name of the block throttling group",
3831 },{
3832 .name = "copy-on-read",
3833 .type = QEMU_OPT_BOOL,
3834 .help = "copy read data from backing file into image file",
3835 },{
3836 .name = "detect-zeroes",
3837 .type = QEMU_OPT_STRING,
3838 .help = "try to optimize zero writes (off, on, unmap)",
3839 },{
3840 .name = "stats-account-invalid",
3841 .type = QEMU_OPT_BOOL,
3842 .help = "whether to account for invalid I/O operations "
3843 "in the statistics",
3844 },{
3845 .name = "stats-account-failed",
3846 .type = QEMU_OPT_BOOL,
3847 .help = "whether to account for failed I/O operations "
3848 "in the statistics",
3849 },
3850 { /* end of list */ }
3851 },
3852 };
3853
3854 QemuOptsList qemu_drive_opts = {
3855 .name = "drive",
3856 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3857 .desc = {
3858 /*
3859 * no elements => accept any params
3860 * validation will happen later
3861 */
3862 { /* end of list */ }
3863 },
3864 };
Random patches