?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
m68k: add the virtio devices aliases
[qemu/kevin.git] / blockdev.c
blob621cc3b7c4e5c7fbc4927305d53250a6335e13de
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 /*
243 * Check board claimed all -drive that are meant to be claimed.
244 * Fatal error if any remain unclaimed.
245 */
246 void drive_check_orphaned(void)
247 {
248 BlockBackend *blk;
249 DriveInfo *dinfo;
250 Location loc;
251 bool orphans = false;
252
253 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
254 dinfo = blk_legacy_dinfo(blk);
255 /*
256 * Ignore default drives, because we create certain default
257 * drives unconditionally, then leave them unclaimed. Not the
258 * users fault.
259 * Ignore IF_VIRTIO, because it gets desugared into -device,
260 * so we can leave failing to -device.
261 * Ignore IF_NONE, because leaving unclaimed IF_NONE remains
262 * available for device_add is a feature.
263 */
264 if (dinfo->is_default || dinfo->type == IF_VIRTIO
265 || dinfo->type == IF_NONE) {
266 continue;
267 }
268 if (!blk_get_attached_dev(blk)) {
269 loc_push_none(&loc);
270 qemu_opts_loc_restore(dinfo->opts);
271 error_report("machine type does not support"
272 " if=%s,bus=%d,unit=%d",
273 if_name[dinfo->type], dinfo->bus, dinfo->unit);
274 loc_pop(&loc);
275 orphans = true;
276 }
277 }
278
279 if (orphans) {
280 exit(1);
281 }
282 }
283
284 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
285 {
286 return drive_get(type,
287 drive_index_to_bus_id(type, index),
288 drive_index_to_unit_id(type, index));
289 }
290
291 int drive_get_max_bus(BlockInterfaceType type)
292 {
293 int max_bus;
294 BlockBackend *blk;
295 DriveInfo *dinfo;
296
297 max_bus = -1;
298 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
299 dinfo = blk_legacy_dinfo(blk);
300 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
301 max_bus = dinfo->bus;
302 }
303 }
304 return max_bus;
305 }
306
307 /* Get a block device. This should only be used for single-drive devices
308 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
309 appropriate bus. */
310 DriveInfo *drive_get_next(BlockInterfaceType type)
311 {
312 static int next_block_unit[IF_COUNT];
313
314 return drive_get(type, 0, next_block_unit[type]++);
315 }
316
317 static void bdrv_format_print(void *opaque, const char *name)
318 {
319 qemu_printf(" %s", name);
320 }
321
322 typedef struct {
323 QEMUBH *bh;
324 BlockDriverState *bs;
325 } BDRVPutRefBH;
326
327 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
328 {
329 if (!strcmp(buf, "ignore")) {
330 return BLOCKDEV_ON_ERROR_IGNORE;
331 } else if (!is_read && !strcmp(buf, "enospc")) {
332 return BLOCKDEV_ON_ERROR_ENOSPC;
333 } else if (!strcmp(buf, "stop")) {
334 return BLOCKDEV_ON_ERROR_STOP;
335 } else if (!strcmp(buf, "report")) {
336 return BLOCKDEV_ON_ERROR_REPORT;
337 } else {
338 error_setg(errp, "'%s' invalid %s error action",
339 buf, is_read ? "read" : "write");
340 return -1;
341 }
342 }
343
344 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
345 Error **errp)
346 {
347 const QListEntry *entry;
348 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
349 switch (qobject_type(entry->value)) {
350
351 case QTYPE_QSTRING: {
352 unsigned long long length;
353 const char *str = qstring_get_str(qobject_to(QString,
354 entry->value));
355 if (parse_uint_full(str, &length, 10) == 0 &&
356 length > 0 && length <= UINT_MAX) {
357 block_acct_add_interval(stats, (unsigned) length);
358 } else {
359 error_setg(errp, "Invalid interval length: %s", str);
360 return false;
361 }
362 break;
363 }
364
365 case QTYPE_QNUM: {
366 int64_t length = qnum_get_int(qobject_to(QNum, entry->value));
367
368 if (length > 0 && length <= UINT_MAX) {
369 block_acct_add_interval(stats, (unsigned) length);
370 } else {
371 error_setg(errp, "Invalid interval length: %" PRId64, length);
372 return false;
373 }
374 break;
375 }
376
377 default:
378 error_setg(errp, "The specification of stats-intervals is invalid");
379 return false;
380 }
381 }
382 return true;
383 }
384
385 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
386
387 /* All parameters but @opts are optional and may be set to NULL. */
388 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
389 const char **throttling_group, ThrottleConfig *throttle_cfg,
390 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
391 {
392 Error *local_error = NULL;
393 const char *aio;
394
395 if (bdrv_flags) {
396 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
397 *bdrv_flags |= BDRV_O_COPY_ON_READ;
398 }
399
400 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
401 if (bdrv_parse_aio(aio, bdrv_flags) < 0) {
402 error_setg(errp, "invalid aio option");
403 return;
404 }
405 }
406 }
407
408 /* disk I/O throttling */
409 if (throttling_group) {
410 *throttling_group = qemu_opt_get(opts, "throttling.group");
411 }
412
413 if (throttle_cfg) {
414 throttle_config_init(throttle_cfg);
415 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
416 qemu_opt_get_number(opts, "throttling.bps-total", 0);
417 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
418 qemu_opt_get_number(opts, "throttling.bps-read", 0);
419 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
420 qemu_opt_get_number(opts, "throttling.bps-write", 0);
421 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
422 qemu_opt_get_number(opts, "throttling.iops-total", 0);
423 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
424 qemu_opt_get_number(opts, "throttling.iops-read", 0);
425 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
426 qemu_opt_get_number(opts, "throttling.iops-write", 0);
427
428 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
429 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
430 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
431 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
432 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
433 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
434 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
435 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
436 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
437 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
438 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
439 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
440
441 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
442 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
443 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
444 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
445 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
446 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
447 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
448 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
449 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
450 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
451 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
452 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
453
454 throttle_cfg->op_size =
455 qemu_opt_get_number(opts, "throttling.iops-size", 0);
456
457 if (!throttle_is_valid(throttle_cfg, errp)) {
458 return;
459 }
460 }
461
462 if (detect_zeroes) {
463 *detect_zeroes =
464 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
465 qemu_opt_get(opts, "detect-zeroes"),
466 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
467 &local_error);
468 if (local_error) {
469 error_propagate(errp, local_error);
470 return;
471 }
472 }
473 }
474
475 /* Takes the ownership of bs_opts */
476 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
477 Error **errp)
478 {
479 const char *buf;
480 int bdrv_flags = 0;
481 int on_read_error, on_write_error;
482 bool account_invalid, account_failed;
483 bool writethrough, read_only;
484 BlockBackend *blk;
485 BlockDriverState *bs;
486 ThrottleConfig cfg;
487 int snapshot = 0;
488 Error *error = NULL;
489 QemuOpts *opts;
490 QDict *interval_dict = NULL;
491 QList *interval_list = NULL;
492 const char *id;
493 BlockdevDetectZeroesOptions detect_zeroes =
494 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
495 const char *throttling_group = NULL;
496
497 /* Check common options by copying from bs_opts to opts, all other options
498 * stay in bs_opts for processing by bdrv_open(). */
499 id = qdict_get_try_str(bs_opts, "id");
500 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, errp);
501 if (!opts) {
502 goto err_no_opts;
503 }
504
505 if (!qemu_opts_absorb_qdict(opts, bs_opts, errp)) {
506 goto early_err;
507 }
508
509 if (id) {
510 qdict_del(bs_opts, "id");
511 }
512
513 /* extract parameters */
514 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
515
516 account_invalid = qemu_opt_get_bool(opts, "stats-account-invalid", true);
517 account_failed = qemu_opt_get_bool(opts, "stats-account-failed", true);
518
519 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
520
521 id = qemu_opts_id(opts);
522
523 qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
524 qdict_array_split(interval_dict, &interval_list);
525
526 if (qdict_size(interval_dict) != 0) {
527 error_setg(errp, "Invalid option stats-intervals.%s",
528 qdict_first(interval_dict)->key);
529 goto early_err;
530 }
531
532 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
533 &detect_zeroes, &error);
534 if (error) {
535 error_propagate(errp, error);
536 goto early_err;
537 }
538
539 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
540 if (is_help_option(buf)) {
541 qemu_printf("Supported formats:");
542 bdrv_iterate_format(bdrv_format_print, NULL, false);
543 qemu_printf("\nSupported formats (read-only):");
544 bdrv_iterate_format(bdrv_format_print, NULL, true);
545 qemu_printf("\n");
546 goto early_err;
547 }
548
549 if (qdict_haskey(bs_opts, "driver")) {
550 error_setg(errp, "Cannot specify both 'driver' and 'format'");
551 goto early_err;
552 }
553 qdict_put_str(bs_opts, "driver", buf);
554 }
555
556 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
557 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
558 on_write_error = parse_block_error_action(buf, 0, &error);
559 if (error) {
560 error_propagate(errp, error);
561 goto early_err;
562 }
563 }
564
565 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
566 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
567 on_read_error = parse_block_error_action(buf, 1, &error);
568 if (error) {
569 error_propagate(errp, error);
570 goto early_err;
571 }
572 }
573
574 if (snapshot) {
575 bdrv_flags |= BDRV_O_SNAPSHOT;
576 }
577
578 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
579
580 /* init */
581 if ((!file || !*file) && !qdict_size(bs_opts)) {
582 BlockBackendRootState *blk_rs;
583
584 blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
585 blk_rs = blk_get_root_state(blk);
586 blk_rs->open_flags = bdrv_flags;
587 blk_rs->read_only = read_only;
588 blk_rs->detect_zeroes = detect_zeroes;
589
590 qobject_unref(bs_opts);
591 } else {
592 if (file && !*file) {
593 file = NULL;
594 }
595
596 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
597 * with other callers) rather than what we want as the real defaults.
598 * Apply the defaults here instead. */
599 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
600 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
601 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
602 read_only ? "on" : "off");
603 qdict_set_default_str(bs_opts, BDRV_OPT_AUTO_READ_ONLY, "on");
604 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
605
606 if (runstate_check(RUN_STATE_INMIGRATE)) {
607 bdrv_flags |= BDRV_O_INACTIVE;
608 }
609
610 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
611 if (!blk) {
612 goto err_no_bs_opts;
613 }
614 bs = blk_bs(blk);
615
616 bs->detect_zeroes = detect_zeroes;
617
618 block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
619
620 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
621 blk_unref(blk);
622 blk = NULL;
623 goto err_no_bs_opts;
624 }
625 }
626
627 /* disk I/O throttling */
628 if (throttle_enabled(&cfg)) {
629 if (!throttling_group) {
630 throttling_group = id;
631 }
632 blk_io_limits_enable(blk, throttling_group);
633 blk_set_io_limits(blk, &cfg);
634 }
635
636 blk_set_enable_write_cache(blk, !writethrough);
637 blk_set_on_error(blk, on_read_error, on_write_error);
638
639 if (!monitor_add_blk(blk, id, errp)) {
640 blk_unref(blk);
641 blk = NULL;
642 goto err_no_bs_opts;
643 }
644
645 err_no_bs_opts:
646 qemu_opts_del(opts);
647 qobject_unref(interval_dict);
648 qobject_unref(interval_list);
649 return blk;
650
651 early_err:
652 qemu_opts_del(opts);
653 qobject_unref(interval_dict);
654 qobject_unref(interval_list);
655 err_no_opts:
656 qobject_unref(bs_opts);
657 return NULL;
658 }
659
660 /* Takes the ownership of bs_opts */
661 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
662 {
663 int bdrv_flags = 0;
664
665 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
666 * with other callers) rather than what we want as the real defaults.
667 * Apply the defaults here instead. */
668 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
669 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
670 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
671
672 if (runstate_check(RUN_STATE_INMIGRATE)) {
673 bdrv_flags |= BDRV_O_INACTIVE;
674 }
675
676 return bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
677 }
678
679 void blockdev_close_all_bdrv_states(void)
680 {
681 BlockDriverState *bs, *next_bs;
682
683 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
684 AioContext *ctx = bdrv_get_aio_context(bs);
685
686 aio_context_acquire(ctx);
687 bdrv_unref(bs);
688 aio_context_release(ctx);
689 }
690 }
691
692 /* Iterates over the list of monitor-owned BlockDriverStates */
693 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
694 {
695 return bs ? QTAILQ_NEXT(bs, monitor_list)
696 : QTAILQ_FIRST(&monitor_bdrv_states);
697 }
698
699 static bool qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
700 Error **errp)
701 {
702 const char *value;
703
704 value = qemu_opt_get(opts, from);
705 if (value) {
706 if (qemu_opt_find(opts, to)) {
707 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
708 "same time", to, from);
709 return false;
710 }
711 }
712
713 /* rename all items in opts */
714 while ((value = qemu_opt_get(opts, from))) {
715 qemu_opt_set(opts, to, value, &error_abort);
716 qemu_opt_unset(opts, from);
717 }
718 return true;
719 }
720
721 QemuOptsList qemu_legacy_drive_opts = {
722 .name = "drive",
723 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
724 .desc = {
725 {
726 .name = "bus",
727 .type = QEMU_OPT_NUMBER,
728 .help = "bus number",
729 },{
730 .name = "unit",
731 .type = QEMU_OPT_NUMBER,
732 .help = "unit number (i.e. lun for scsi)",
733 },{
734 .name = "index",
735 .type = QEMU_OPT_NUMBER,
736 .help = "index number",
737 },{
738 .name = "media",
739 .type = QEMU_OPT_STRING,
740 .help = "media type (disk, cdrom)",
741 },{
742 .name = "if",
743 .type = QEMU_OPT_STRING,
744 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
745 },{
746 .name = "file",
747 .type = QEMU_OPT_STRING,
748 .help = "file name",
749 },
750
751 /* Options that are passed on, but have special semantics with -drive */
752 {
753 .name = BDRV_OPT_READ_ONLY,
754 .type = QEMU_OPT_BOOL,
755 .help = "open drive file as read-only",
756 },{
757 .name = "rerror",
758 .type = QEMU_OPT_STRING,
759 .help = "read error action",
760 },{
761 .name = "werror",
762 .type = QEMU_OPT_STRING,
763 .help = "write error action",
764 },{
765 .name = "copy-on-read",
766 .type = QEMU_OPT_BOOL,
767 .help = "copy read data from backing file into image file",
768 },
769
770 { /* end of list */ }
771 },
772 };
773
774 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
775 Error **errp)
776 {
777 const char *value;
778 BlockBackend *blk;
779 DriveInfo *dinfo = NULL;
780 QDict *bs_opts;
781 QemuOpts *legacy_opts;
782 DriveMediaType media = MEDIA_DISK;
783 BlockInterfaceType type;
784 int max_devs, bus_id, unit_id, index;
785 const char *werror, *rerror;
786 bool read_only = false;
787 bool copy_on_read;
788 const char *filename;
789 int i;
790
791 /* Change legacy command line options into QMP ones */
792 static const struct {
793 const char *from;
794 const char *to;
795 } opt_renames[] = {
796 { "iops", "throttling.iops-total" },
797 { "iops_rd", "throttling.iops-read" },
798 { "iops_wr", "throttling.iops-write" },
799
800 { "bps", "throttling.bps-total" },
801 { "bps_rd", "throttling.bps-read" },
802 { "bps_wr", "throttling.bps-write" },
803
804 { "iops_max", "throttling.iops-total-max" },
805 { "iops_rd_max", "throttling.iops-read-max" },
806 { "iops_wr_max", "throttling.iops-write-max" },
807
808 { "bps_max", "throttling.bps-total-max" },
809 { "bps_rd_max", "throttling.bps-read-max" },
810 { "bps_wr_max", "throttling.bps-write-max" },
811
812 { "iops_size", "throttling.iops-size" },
813
814 { "group", "throttling.group" },
815
816 { "readonly", BDRV_OPT_READ_ONLY },
817 };
818
819 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
820 if (!qemu_opt_rename(all_opts, opt_renames[i].from,
821 opt_renames[i].to, errp)) {
822 return NULL;
823 }
824 }
825
826 value = qemu_opt_get(all_opts, "cache");
827 if (value) {
828 int flags = 0;
829 bool writethrough;
830
831 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
832 error_setg(errp, "invalid cache option");
833 return NULL;
834 }
835
836 /* Specific options take precedence */
837 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
838 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
839 !writethrough, &error_abort);
840 }
841 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
842 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
843 !!(flags & BDRV_O_NOCACHE), &error_abort);
844 }
845 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
846 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
847 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
848 }
849 qemu_opt_unset(all_opts, "cache");
850 }
851
852 /* Get a QDict for processing the options */
853 bs_opts = qdict_new();
854 qemu_opts_to_qdict(all_opts, bs_opts);
855
856 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
857 &error_abort);
858 if (!qemu_opts_absorb_qdict(legacy_opts, bs_opts, errp)) {
859 goto fail;
860 }
861
862 /* Media type */
863 value = qemu_opt_get(legacy_opts, "media");
864 if (value) {
865 if (!strcmp(value, "disk")) {
866 media = MEDIA_DISK;
867 } else if (!strcmp(value, "cdrom")) {
868 media = MEDIA_CDROM;
869 read_only = true;
870 } else {
871 error_setg(errp, "'%s' invalid media", value);
872 goto fail;
873 }
874 }
875
876 /* copy-on-read is disabled with a warning for read-only devices */
877 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
878 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
879
880 if (read_only && copy_on_read) {
881 warn_report("disabling copy-on-read on read-only drive");
882 copy_on_read = false;
883 }
884
885 qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
886 qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
887
888 /* Controller type */
889 value = qemu_opt_get(legacy_opts, "if");
890 if (value) {
891 for (type = 0;
892 type < IF_COUNT && strcmp(value, if_name[type]);
893 type++) {
894 }
895 if (type == IF_COUNT) {
896 error_setg(errp, "unsupported bus type '%s'", value);
897 goto fail;
898 }
899 } else {
900 type = block_default_type;
901 }
902
903 /* Device address specified by bus/unit or index.
904 * If none was specified, try to find the first free one. */
905 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
906 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
907 index = qemu_opt_get_number(legacy_opts, "index", -1);
908
909 max_devs = if_max_devs[type];
910
911 if (index != -1) {
912 if (bus_id != 0 || unit_id != -1) {
913 error_setg(errp, "index cannot be used with bus and unit");
914 goto fail;
915 }
916 bus_id = drive_index_to_bus_id(type, index);
917 unit_id = drive_index_to_unit_id(type, index);
918 }
919
920 if (unit_id == -1) {
921 unit_id = 0;
922 while (drive_get(type, bus_id, unit_id) != NULL) {
923 unit_id++;
924 if (max_devs && unit_id >= max_devs) {
925 unit_id -= max_devs;
926 bus_id++;
927 }
928 }
929 }
930
931 if (max_devs && unit_id >= max_devs) {
932 error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
933 goto fail;
934 }
935
936 if (drive_get(type, bus_id, unit_id) != NULL) {
937 error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
938 bus_id, unit_id, index);
939 goto fail;
940 }
941
942 /* no id supplied -> create one */
943 if (qemu_opts_id(all_opts) == NULL) {
944 char *new_id;
945 const char *mediastr = "";
946 if (type == IF_IDE || type == IF_SCSI) {
947 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
948 }
949 if (max_devs) {
950 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
951 mediastr, unit_id);
952 } else {
953 new_id = g_strdup_printf("%s%s%i", if_name[type],
954 mediastr, unit_id);
955 }
956 qdict_put_str(bs_opts, "id", new_id);
957 g_free(new_id);
958 }
959
960 /* Add virtio block device */
961 if (type == IF_VIRTIO) {
962 QemuOpts *devopts;
963 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
964 &error_abort);
965 if (arch_type == QEMU_ARCH_S390X) {
966 qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
967 } else {
968 qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
969 }
970 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
971 &error_abort);
972 }
973
974 filename = qemu_opt_get(legacy_opts, "file");
975
976 /* Check werror/rerror compatibility with if=... */
977 werror = qemu_opt_get(legacy_opts, "werror");
978 if (werror != NULL) {
979 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
980 type != IF_NONE) {
981 error_setg(errp, "werror is not supported by this bus type");
982 goto fail;
983 }
984 qdict_put_str(bs_opts, "werror", werror);
985 }
986
987 rerror = qemu_opt_get(legacy_opts, "rerror");
988 if (rerror != NULL) {
989 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
990 type != IF_NONE) {
991 error_setg(errp, "rerror is not supported by this bus type");
992 goto fail;
993 }
994 qdict_put_str(bs_opts, "rerror", rerror);
995 }
996
997 /* Actual block device init: Functionality shared with blockdev-add */
998 blk = blockdev_init(filename, bs_opts, errp);
999 bs_opts = NULL;
1000 if (!blk) {
1001 goto fail;
1002 }
1003
1004 /* Create legacy DriveInfo */
1005 dinfo = g_malloc0(sizeof(*dinfo));
1006 dinfo->opts = all_opts;
1007
1008 dinfo->type = type;
1009 dinfo->bus = bus_id;
1010 dinfo->unit = unit_id;
1011
1012 blk_set_legacy_dinfo(blk, dinfo);
1013
1014 switch(type) {
1015 case IF_IDE:
1016 case IF_SCSI:
1017 case IF_XEN:
1018 case IF_NONE:
1019 dinfo->media_cd = media == MEDIA_CDROM;
1020 break;
1021 default:
1022 break;
1023 }
1024
1025 fail:
1026 qemu_opts_del(legacy_opts);
1027 qobject_unref(bs_opts);
1028 return dinfo;
1029 }
1030
1031 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1032 {
1033 BlockDriverState *bs;
1034
1035 bs = bdrv_lookup_bs(name, name, errp);
1036 if (bs == NULL) {
1037 return NULL;
1038 }
1039
1040 if (!bdrv_is_root_node(bs)) {
1041 error_setg(errp, "Need a root block node");
1042 return NULL;
1043 }
1044
1045 if (!bdrv_is_inserted(bs)) {
1046 error_setg(errp, "Device has no medium");
1047 return NULL;
1048 }
1049
1050 return bs;
1051 }
1052
1053 static void blockdev_do_action(TransactionAction *action, Error **errp)
1054 {
1055 TransactionActionList list;
1056
1057 list.value = action;
1058 list.next = NULL;
1059 qmp_transaction(&list, false, NULL, errp);
1060 }
1061
1062 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1063 bool has_node_name, const char *node_name,
1064 const char *snapshot_file,
1065 bool has_snapshot_node_name,
1066 const char *snapshot_node_name,
1067 bool has_format, const char *format,
1068 bool has_mode, NewImageMode mode, Error **errp)
1069 {
1070 BlockdevSnapshotSync snapshot = {
1071 .has_device = has_device,
1072 .device = (char *) device,
1073 .has_node_name = has_node_name,
1074 .node_name = (char *) node_name,
1075 .snapshot_file = (char *) snapshot_file,
1076 .has_snapshot_node_name = has_snapshot_node_name,
1077 .snapshot_node_name = (char *) snapshot_node_name,
1078 .has_format = has_format,
1079 .format = (char *) format,
1080 .has_mode = has_mode,
1081 .mode = mode,
1082 };
1083 TransactionAction action = {
1084 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1085 .u.blockdev_snapshot_sync.data = &snapshot,
1086 };
1087 blockdev_do_action(&action, errp);
1088 }
1089
1090 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1091 Error **errp)
1092 {
1093 BlockdevSnapshot snapshot_data = {
1094 .node = (char *) node,
1095 .overlay = (char *) overlay
1096 };
1097 TransactionAction action = {
1098 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1099 .u.blockdev_snapshot.data = &snapshot_data,
1100 };
1101 blockdev_do_action(&action, errp);
1102 }
1103
1104 void qmp_blockdev_snapshot_internal_sync(const char *device,
1105 const char *name,
1106 Error **errp)
1107 {
1108 BlockdevSnapshotInternal snapshot = {
1109 .device = (char *) device,
1110 .name = (char *) name
1111 };
1112 TransactionAction action = {
1113 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1114 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1115 };
1116 blockdev_do_action(&action, errp);
1117 }
1118
1119 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1120 bool has_id,
1121 const char *id,
1122 bool has_name,
1123 const char *name,
1124 Error **errp)
1125 {
1126 BlockDriverState *bs;
1127 AioContext *aio_context;
1128 QEMUSnapshotInfo sn;
1129 Error *local_err = NULL;
1130 SnapshotInfo *info = NULL;
1131 int ret;
1132
1133 bs = qmp_get_root_bs(device, errp);
1134 if (!bs) {
1135 return NULL;
1136 }
1137 aio_context = bdrv_get_aio_context(bs);
1138 aio_context_acquire(aio_context);
1139
1140 if (!has_id) {
1141 id = NULL;
1142 }
1143
1144 if (!has_name) {
1145 name = NULL;
1146 }
1147
1148 if (!id && !name) {
1149 error_setg(errp, "Name or id must be provided");
1150 goto out_aio_context;
1151 }
1152
1153 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1154 goto out_aio_context;
1155 }
1156
1157 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1158 if (local_err) {
1159 error_propagate(errp, local_err);
1160 goto out_aio_context;
1161 }
1162 if (!ret) {
1163 error_setg(errp,
1164 "Snapshot with id '%s' and name '%s' does not exist on "
1165 "device '%s'",
1166 STR_OR_NULL(id), STR_OR_NULL(name), device);
1167 goto out_aio_context;
1168 }
1169
1170 bdrv_snapshot_delete(bs, id, name, &local_err);
1171 if (local_err) {
1172 error_propagate(errp, local_err);
1173 goto out_aio_context;
1174 }
1175
1176 aio_context_release(aio_context);
1177
1178 info = g_new0(SnapshotInfo, 1);
1179 info->id = g_strdup(sn.id_str);
1180 info->name = g_strdup(sn.name);
1181 info->date_nsec = sn.date_nsec;
1182 info->date_sec = sn.date_sec;
1183 info->vm_state_size = sn.vm_state_size;
1184 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1185 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1186 if (sn.icount != -1ULL) {
1187 info->icount = sn.icount;
1188 info->has_icount = true;
1189 }
1190
1191 return info;
1192
1193 out_aio_context:
1194 aio_context_release(aio_context);
1195 return NULL;
1196 }
1197
1198 /* New and old BlockDriverState structs for atomic group operations */
1199
1200 typedef struct BlkActionState BlkActionState;
1201
1202 /**
1203 * BlkActionOps:
1204 * Table of operations that define an Action.
1205 *
1206 * @instance_size: Size of state struct, in bytes.
1207 * @prepare: Prepare the work, must NOT be NULL.
1208 * @commit: Commit the changes, can be NULL.
1209 * @abort: Abort the changes on fail, can be NULL.
1210 * @clean: Clean up resources after all transaction actions have called
1211 * commit() or abort(). Can be NULL.
1212 *
1213 * Only prepare() may fail. In a single transaction, only one of commit() or
1214 * abort() will be called. clean() will always be called if it is present.
1215 */
1216 typedef struct BlkActionOps {
1217 size_t instance_size;
1218 void (*prepare)(BlkActionState *common, Error **errp);
1219 void (*commit)(BlkActionState *common);
1220 void (*abort)(BlkActionState *common);
1221 void (*clean)(BlkActionState *common);
1222 } BlkActionOps;
1223
1224 /**
1225 * BlkActionState:
1226 * Describes one Action's state within a Transaction.
1227 *
1228 * @action: QAPI-defined enum identifying which Action to perform.
1229 * @ops: Table of ActionOps this Action can perform.
1230 * @block_job_txn: Transaction which this action belongs to.
1231 * @entry: List membership for all Actions in this Transaction.
1232 *
1233 * This structure must be arranged as first member in a subclassed type,
1234 * assuming that the compiler will also arrange it to the same offsets as the
1235 * base class.
1236 */
1237 struct BlkActionState {
1238 TransactionAction *action;
1239 const BlkActionOps *ops;
1240 JobTxn *block_job_txn;
1241 TransactionProperties *txn_props;
1242 QTAILQ_ENTRY(BlkActionState) entry;
1243 };
1244
1245 /* internal snapshot private data */
1246 typedef struct InternalSnapshotState {
1247 BlkActionState common;
1248 BlockDriverState *bs;
1249 QEMUSnapshotInfo sn;
1250 bool created;
1251 } InternalSnapshotState;
1252
1253
1254 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1255 {
1256 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1257 error_setg(errp,
1258 "Action '%s' does not support Transaction property "
1259 "completion-mode = %s",
1260 TransactionActionKind_str(s->action->type),
1261 ActionCompletionMode_str(s->txn_props->completion_mode));
1262 return -1;
1263 }
1264 return 0;
1265 }
1266
1267 static void internal_snapshot_prepare(BlkActionState *common,
1268 Error **errp)
1269 {
1270 Error *local_err = NULL;
1271 const char *device;
1272 const char *name;
1273 BlockDriverState *bs;
1274 QEMUSnapshotInfo old_sn, *sn;
1275 bool ret;
1276 qemu_timeval tv;
1277 BlockdevSnapshotInternal *internal;
1278 InternalSnapshotState *state;
1279 AioContext *aio_context;
1280 int ret1;
1281
1282 g_assert(common->action->type ==
1283 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1284 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1285 state = DO_UPCAST(InternalSnapshotState, common, common);
1286
1287 /* 1. parse input */
1288 device = internal->device;
1289 name = internal->name;
1290
1291 /* 2. check for validation */
1292 if (action_check_completion_mode(common, errp) < 0) {
1293 return;
1294 }
1295
1296 bs = qmp_get_root_bs(device, errp);
1297 if (!bs) {
1298 return;
1299 }
1300
1301 aio_context = bdrv_get_aio_context(bs);
1302 aio_context_acquire(aio_context);
1303
1304 state->bs = bs;
1305
1306 /* Paired with .clean() */
1307 bdrv_drained_begin(bs);
1308
1309 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1310 goto out;
1311 }
1312
1313 if (bdrv_is_read_only(bs)) {
1314 error_setg(errp, "Device '%s' is read only", device);
1315 goto out;
1316 }
1317
1318 if (!bdrv_can_snapshot(bs)) {
1319 error_setg(errp, "Block format '%s' used by device '%s' "
1320 "does not support internal snapshots",
1321 bs->drv->format_name, device);
1322 goto out;
1323 }
1324
1325 if (!strlen(name)) {
1326 error_setg(errp, "Name is empty");
1327 goto out;
1328 }
1329
1330 /* check whether a snapshot with name exist */
1331 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1332 &local_err);
1333 if (local_err) {
1334 error_propagate(errp, local_err);
1335 goto out;
1336 } else if (ret) {
1337 error_setg(errp,
1338 "Snapshot with name '%s' already exists on device '%s'",
1339 name, device);
1340 goto out;
1341 }
1342
1343 /* 3. take the snapshot */
1344 sn = &state->sn;
1345 pstrcpy(sn->name, sizeof(sn->name), name);
1346 qemu_gettimeofday(&tv);
1347 sn->date_sec = tv.tv_sec;
1348 sn->date_nsec = tv.tv_usec * 1000;
1349 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1350 if (replay_mode != REPLAY_MODE_NONE) {
1351 sn->icount = replay_get_current_icount();
1352 } else {
1353 sn->icount = -1ULL;
1354 }
1355
1356 ret1 = bdrv_snapshot_create(bs, sn);
1357 if (ret1 < 0) {
1358 error_setg_errno(errp, -ret1,
1359 "Failed to create snapshot '%s' on device '%s'",
1360 name, device);
1361 goto out;
1362 }
1363
1364 /* 4. succeed, mark a snapshot is created */
1365 state->created = true;
1366
1367 out:
1368 aio_context_release(aio_context);
1369 }
1370
1371 static void internal_snapshot_abort(BlkActionState *common)
1372 {
1373 InternalSnapshotState *state =
1374 DO_UPCAST(InternalSnapshotState, common, common);
1375 BlockDriverState *bs = state->bs;
1376 QEMUSnapshotInfo *sn = &state->sn;
1377 AioContext *aio_context;
1378 Error *local_error = NULL;
1379
1380 if (!state->created) {
1381 return;
1382 }
1383
1384 aio_context = bdrv_get_aio_context(state->bs);
1385 aio_context_acquire(aio_context);
1386
1387 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1388 error_reportf_err(local_error,
1389 "Failed to delete snapshot with id '%s' and "
1390 "name '%s' on device '%s' in abort: ",
1391 sn->id_str, sn->name,
1392 bdrv_get_device_name(bs));
1393 }
1394
1395 aio_context_release(aio_context);
1396 }
1397
1398 static void internal_snapshot_clean(BlkActionState *common)
1399 {
1400 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1401 common, common);
1402 AioContext *aio_context;
1403
1404 if (!state->bs) {
1405 return;
1406 }
1407
1408 aio_context = bdrv_get_aio_context(state->bs);
1409 aio_context_acquire(aio_context);
1410
1411 bdrv_drained_end(state->bs);
1412
1413 aio_context_release(aio_context);
1414 }
1415
1416 /* external snapshot private data */
1417 typedef struct ExternalSnapshotState {
1418 BlkActionState common;
1419 BlockDriverState *old_bs;
1420 BlockDriverState *new_bs;
1421 bool overlay_appended;
1422 } ExternalSnapshotState;
1423
1424 static void external_snapshot_prepare(BlkActionState *common,
1425 Error **errp)
1426 {
1427 int ret;
1428 int flags = 0;
1429 QDict *options = NULL;
1430 Error *local_err = NULL;
1431 /* Device and node name of the image to generate the snapshot from */
1432 const char *device;
1433 const char *node_name;
1434 /* Reference to the new image (for 'blockdev-snapshot') */
1435 const char *snapshot_ref;
1436 /* File name of the new image (for 'blockdev-snapshot-sync') */
1437 const char *new_image_file;
1438 ExternalSnapshotState *state =
1439 DO_UPCAST(ExternalSnapshotState, common, common);
1440 TransactionAction *action = common->action;
1441 AioContext *aio_context;
1442 uint64_t perm, shared;
1443
1444 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1445 * purpose but a different set of parameters */
1446 switch (action->type) {
1447 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1448 {
1449 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1450 device = s->node;
1451 node_name = s->node;
1452 new_image_file = NULL;
1453 snapshot_ref = s->overlay;
1454 }
1455 break;
1456 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1457 {
1458 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1459 device = s->has_device ? s->device : NULL;
1460 node_name = s->has_node_name ? s->node_name : NULL;
1461 new_image_file = s->snapshot_file;
1462 snapshot_ref = NULL;
1463 }
1464 break;
1465 default:
1466 g_assert_not_reached();
1467 }
1468
1469 /* start processing */
1470 if (action_check_completion_mode(common, errp) < 0) {
1471 return;
1472 }
1473
1474 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1475 if (!state->old_bs) {
1476 return;
1477 }
1478
1479 aio_context = bdrv_get_aio_context(state->old_bs);
1480 aio_context_acquire(aio_context);
1481
1482 /* Paired with .clean() */
1483 bdrv_drained_begin(state->old_bs);
1484
1485 if (!bdrv_is_inserted(state->old_bs)) {
1486 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1487 goto out;
1488 }
1489
1490 if (bdrv_op_is_blocked(state->old_bs,
1491 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1492 goto out;
1493 }
1494
1495 if (!bdrv_is_read_only(state->old_bs)) {
1496 if (bdrv_flush(state->old_bs)) {
1497 error_setg(errp, QERR_IO_ERROR);
1498 goto out;
1499 }
1500 }
1501
1502 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1503 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1504 const char *format = s->has_format ? s->format : "qcow2";
1505 enum NewImageMode mode;
1506 const char *snapshot_node_name =
1507 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1508
1509 if (node_name && !snapshot_node_name) {
1510 error_setg(errp, "New overlay node-name missing");
1511 goto out;
1512 }
1513
1514 if (snapshot_node_name &&
1515 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1516 error_setg(errp, "New overlay node-name already in use");
1517 goto out;
1518 }
1519
1520 flags = state->old_bs->open_flags;
1521 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1522 flags |= BDRV_O_NO_BACKING;
1523
1524 /* create new image w/backing file */
1525 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1526 if (mode != NEW_IMAGE_MODE_EXISTING) {
1527 int64_t size = bdrv_getlength(state->old_bs);
1528 if (size < 0) {
1529 error_setg_errno(errp, -size, "bdrv_getlength failed");
1530 goto out;
1531 }
1532 bdrv_refresh_filename(state->old_bs);
1533 bdrv_img_create(new_image_file, format,
1534 state->old_bs->filename,
1535 state->old_bs->drv->format_name,
1536 NULL, size, flags, false, &local_err);
1537 if (local_err) {
1538 error_propagate(errp, local_err);
1539 goto out;
1540 }
1541 }
1542
1543 options = qdict_new();
1544 if (snapshot_node_name) {
1545 qdict_put_str(options, "node-name", snapshot_node_name);
1546 }
1547 qdict_put_str(options, "driver", format);
1548 }
1549
1550 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1551 errp);
1552 /* We will manually add the backing_hd field to the bs later */
1553 if (!state->new_bs) {
1554 goto out;
1555 }
1556
1557 /*
1558 * Allow attaching a backing file to an overlay that's already in use only
1559 * if the parents don't assume that they are already seeing a valid image.
1560 * (Specifically, allow it as a mirror target, which is write-only access.)
1561 */
1562 bdrv_get_cumulative_perm(state->new_bs, &perm, &shared);
1563 if (perm & BLK_PERM_CONSISTENT_READ) {
1564 error_setg(errp, "The overlay is already in use");
1565 goto out;
1566 }
1567
1568 if (state->new_bs->drv->is_filter) {
1569 error_setg(errp, "Filters cannot be used as overlays");
1570 goto out;
1571 }
1572
1573 if (bdrv_cow_child(state->new_bs)) {
1574 error_setg(errp, "The overlay already has a backing image");
1575 goto out;
1576 }
1577
1578 if (!state->new_bs->drv->supports_backing) {
1579 error_setg(errp, "The overlay does not support backing images");
1580 goto out;
1581 }
1582
1583 /* This removes our old bs and adds the new bs. This is an operation that
1584 * can fail, so we need to do it in .prepare; undoing it for abort is
1585 * always possible. */
1586 bdrv_ref(state->new_bs);
1587 ret = bdrv_append(state->new_bs, state->old_bs, errp);
1588 if (ret < 0) {
1589 goto out;
1590 }
1591 state->overlay_appended = true;
1592
1593 out:
1594 aio_context_release(aio_context);
1595 }
1596
1597 static void external_snapshot_commit(BlkActionState *common)
1598 {
1599 ExternalSnapshotState *state =
1600 DO_UPCAST(ExternalSnapshotState, common, common);
1601 AioContext *aio_context;
1602
1603 aio_context = bdrv_get_aio_context(state->old_bs);
1604 aio_context_acquire(aio_context);
1605
1606 /* We don't need (or want) to use the transactional
1607 * bdrv_reopen_multiple() across all the entries at once, because we
1608 * don't want to abort all of them if one of them fails the reopen */
1609 if (!qatomic_read(&state->old_bs->copy_on_read)) {
1610 bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1611 }
1612
1613 aio_context_release(aio_context);
1614 }
1615
1616 static void external_snapshot_abort(BlkActionState *common)
1617 {
1618 ExternalSnapshotState *state =
1619 DO_UPCAST(ExternalSnapshotState, common, common);
1620 if (state->new_bs) {
1621 if (state->overlay_appended) {
1622 AioContext *aio_context;
1623 AioContext *tmp_context;
1624 int ret;
1625
1626 aio_context = bdrv_get_aio_context(state->old_bs);
1627 aio_context_acquire(aio_context);
1628
1629 bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd()
1630 close state->old_bs; we need it */
1631 bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1632
1633 /*
1634 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1635 * the main AioContext. As we're still going to be using it, return
1636 * it to the AioContext it was before.
1637 */
1638 tmp_context = bdrv_get_aio_context(state->old_bs);
1639 if (aio_context != tmp_context) {
1640 aio_context_release(aio_context);
1641 aio_context_acquire(tmp_context);
1642
1643 ret = bdrv_try_set_aio_context(state->old_bs,
1644 aio_context, NULL);
1645 assert(ret == 0);
1646
1647 aio_context_release(tmp_context);
1648 aio_context_acquire(aio_context);
1649 }
1650
1651 bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1652 bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1653
1654 aio_context_release(aio_context);
1655 }
1656 }
1657 }
1658
1659 static void external_snapshot_clean(BlkActionState *common)
1660 {
1661 ExternalSnapshotState *state =
1662 DO_UPCAST(ExternalSnapshotState, common, common);
1663 AioContext *aio_context;
1664
1665 if (!state->old_bs) {
1666 return;
1667 }
1668
1669 aio_context = bdrv_get_aio_context(state->old_bs);
1670 aio_context_acquire(aio_context);
1671
1672 bdrv_drained_end(state->old_bs);
1673 bdrv_unref(state->new_bs);
1674
1675 aio_context_release(aio_context);
1676 }
1677
1678 typedef struct DriveBackupState {
1679 BlkActionState common;
1680 BlockDriverState *bs;
1681 BlockJob *job;
1682 } DriveBackupState;
1683
1684 static BlockJob *do_backup_common(BackupCommon *backup,
1685 BlockDriverState *bs,
1686 BlockDriverState *target_bs,
1687 AioContext *aio_context,
1688 JobTxn *txn, Error **errp);
1689
1690 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1691 {
1692 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1693 DriveBackup *backup;
1694 BlockDriverState *bs;
1695 BlockDriverState *target_bs;
1696 BlockDriverState *source = NULL;
1697 AioContext *aio_context;
1698 AioContext *old_context;
1699 QDict *options;
1700 Error *local_err = NULL;
1701 int flags;
1702 int64_t size;
1703 bool set_backing_hd = false;
1704 int ret;
1705
1706 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1707 backup = common->action->u.drive_backup.data;
1708
1709 if (!backup->has_mode) {
1710 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1711 }
1712
1713 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1714 if (!bs) {
1715 return;
1716 }
1717
1718 if (!bs->drv) {
1719 error_setg(errp, "Device has no medium");
1720 return;
1721 }
1722
1723 aio_context = bdrv_get_aio_context(bs);
1724 aio_context_acquire(aio_context);
1725
1726 /* Paired with .clean() */
1727 bdrv_drained_begin(bs);
1728
1729 if (!backup->has_format) {
1730 backup->format = backup->mode == NEW_IMAGE_MODE_EXISTING ?
1731 NULL : (char *) bs->drv->format_name;
1732 }
1733
1734 /* Early check to avoid creating target */
1735 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
1736 goto out;
1737 }
1738
1739 flags = bs->open_flags | BDRV_O_RDWR;
1740
1741 /*
1742 * See if we have a backing HD we can use to create our new image
1743 * on top of.
1744 */
1745 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
1746 /*
1747 * Backup will not replace the source by the target, so none
1748 * of the filters skipped here will be removed (in contrast to
1749 * mirror). Therefore, we can skip all of them when looking
1750 * for the first COW relationship.
1751 */
1752 source = bdrv_cow_bs(bdrv_skip_filters(bs));
1753 if (!source) {
1754 backup->sync = MIRROR_SYNC_MODE_FULL;
1755 }
1756 }
1757 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
1758 source = bs;
1759 flags |= BDRV_O_NO_BACKING;
1760 set_backing_hd = true;
1761 }
1762
1763 size = bdrv_getlength(bs);
1764 if (size < 0) {
1765 error_setg_errno(errp, -size, "bdrv_getlength failed");
1766 goto out;
1767 }
1768
1769 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
1770 assert(backup->format);
1771 if (source) {
1772 /* Implicit filters should not appear in the filename */
1773 BlockDriverState *explicit_backing =
1774 bdrv_skip_implicit_filters(source);
1775
1776 bdrv_refresh_filename(explicit_backing);
1777 bdrv_img_create(backup->target, backup->format,
1778 explicit_backing->filename,
1779 explicit_backing->drv->format_name, NULL,
1780 size, flags, false, &local_err);
1781 } else {
1782 bdrv_img_create(backup->target, backup->format, NULL, NULL, NULL,
1783 size, flags, false, &local_err);
1784 }
1785 }
1786
1787 if (local_err) {
1788 error_propagate(errp, local_err);
1789 goto out;
1790 }
1791
1792 options = qdict_new();
1793 qdict_put_str(options, "discard", "unmap");
1794 qdict_put_str(options, "detect-zeroes", "unmap");
1795 if (backup->format) {
1796 qdict_put_str(options, "driver", backup->format);
1797 }
1798
1799 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
1800 if (!target_bs) {
1801 goto out;
1802 }
1803
1804 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1805 old_context = bdrv_get_aio_context(target_bs);
1806 aio_context_release(aio_context);
1807 aio_context_acquire(old_context);
1808
1809 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1810 if (ret < 0) {
1811 bdrv_unref(target_bs);
1812 aio_context_release(old_context);
1813 return;
1814 }
1815
1816 aio_context_release(old_context);
1817 aio_context_acquire(aio_context);
1818
1819 if (set_backing_hd) {
1820 if (bdrv_set_backing_hd(target_bs, source, errp) < 0) {
1821 goto unref;
1822 }
1823 }
1824
1825 state->bs = bs;
1826
1827 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1828 bs, target_bs, aio_context,
1829 common->block_job_txn, errp);
1830
1831 unref:
1832 bdrv_unref(target_bs);
1833 out:
1834 aio_context_release(aio_context);
1835 }
1836
1837 static void drive_backup_commit(BlkActionState *common)
1838 {
1839 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1840 AioContext *aio_context;
1841
1842 aio_context = bdrv_get_aio_context(state->bs);
1843 aio_context_acquire(aio_context);
1844
1845 assert(state->job);
1846 job_start(&state->job->job);
1847
1848 aio_context_release(aio_context);
1849 }
1850
1851 static void drive_backup_abort(BlkActionState *common)
1852 {
1853 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1854
1855 if (state->job) {
1856 AioContext *aio_context;
1857
1858 aio_context = bdrv_get_aio_context(state->bs);
1859 aio_context_acquire(aio_context);
1860
1861 job_cancel_sync(&state->job->job);
1862
1863 aio_context_release(aio_context);
1864 }
1865 }
1866
1867 static void drive_backup_clean(BlkActionState *common)
1868 {
1869 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1870 AioContext *aio_context;
1871
1872 if (!state->bs) {
1873 return;
1874 }
1875
1876 aio_context = bdrv_get_aio_context(state->bs);
1877 aio_context_acquire(aio_context);
1878
1879 bdrv_drained_end(state->bs);
1880
1881 aio_context_release(aio_context);
1882 }
1883
1884 typedef struct BlockdevBackupState {
1885 BlkActionState common;
1886 BlockDriverState *bs;
1887 BlockJob *job;
1888 } BlockdevBackupState;
1889
1890 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1891 {
1892 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1893 BlockdevBackup *backup;
1894 BlockDriverState *bs;
1895 BlockDriverState *target_bs;
1896 AioContext *aio_context;
1897 AioContext *old_context;
1898 int ret;
1899
1900 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1901 backup = common->action->u.blockdev_backup.data;
1902
1903 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1904 if (!bs) {
1905 return;
1906 }
1907
1908 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1909 if (!target_bs) {
1910 return;
1911 }
1912
1913 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1914 aio_context = bdrv_get_aio_context(bs);
1915 old_context = bdrv_get_aio_context(target_bs);
1916 aio_context_acquire(old_context);
1917
1918 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1919 if (ret < 0) {
1920 aio_context_release(old_context);
1921 return;
1922 }
1923
1924 aio_context_release(old_context);
1925 aio_context_acquire(aio_context);
1926 state->bs = bs;
1927
1928 /* Paired with .clean() */
1929 bdrv_drained_begin(state->bs);
1930
1931 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1932 bs, target_bs, aio_context,
1933 common->block_job_txn, errp);
1934
1935 aio_context_release(aio_context);
1936 }
1937
1938 static void blockdev_backup_commit(BlkActionState *common)
1939 {
1940 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1941 AioContext *aio_context;
1942
1943 aio_context = bdrv_get_aio_context(state->bs);
1944 aio_context_acquire(aio_context);
1945
1946 assert(state->job);
1947 job_start(&state->job->job);
1948
1949 aio_context_release(aio_context);
1950 }
1951
1952 static void blockdev_backup_abort(BlkActionState *common)
1953 {
1954 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1955
1956 if (state->job) {
1957 AioContext *aio_context;
1958
1959 aio_context = bdrv_get_aio_context(state->bs);
1960 aio_context_acquire(aio_context);
1961
1962 job_cancel_sync(&state->job->job);
1963
1964 aio_context_release(aio_context);
1965 }
1966 }
1967
1968 static void blockdev_backup_clean(BlkActionState *common)
1969 {
1970 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1971 AioContext *aio_context;
1972
1973 if (!state->bs) {
1974 return;
1975 }
1976
1977 aio_context = bdrv_get_aio_context(state->bs);
1978 aio_context_acquire(aio_context);
1979
1980 bdrv_drained_end(state->bs);
1981
1982 aio_context_release(aio_context);
1983 }
1984
1985 typedef struct BlockDirtyBitmapState {
1986 BlkActionState common;
1987 BdrvDirtyBitmap *bitmap;
1988 BlockDriverState *bs;
1989 HBitmap *backup;
1990 bool prepared;
1991 bool was_enabled;
1992 } BlockDirtyBitmapState;
1993
1994 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
1995 Error **errp)
1996 {
1997 Error *local_err = NULL;
1998 BlockDirtyBitmapAdd *action;
1999 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2000 common, common);
2001
2002 if (action_check_completion_mode(common, errp) < 0) {
2003 return;
2004 }
2005
2006 action = common->action->u.block_dirty_bitmap_add.data;
2007 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2008 qmp_block_dirty_bitmap_add(action->node, action->name,
2009 action->has_granularity, action->granularity,
2010 action->has_persistent, action->persistent,
2011 action->has_disabled, action->disabled,
2012 &local_err);
2013
2014 if (!local_err) {
2015 state->prepared = true;
2016 } else {
2017 error_propagate(errp, local_err);
2018 }
2019 }
2020
2021 static void block_dirty_bitmap_add_abort(BlkActionState *common)
2022 {
2023 BlockDirtyBitmapAdd *action;
2024 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2025 common, common);
2026
2027 action = common->action->u.block_dirty_bitmap_add.data;
2028 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2029 * then the node reference and bitmap name must have been valid.
2030 */
2031 if (state->prepared) {
2032 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2033 }
2034 }
2035
2036 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2037 Error **errp)
2038 {
2039 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2040 common, common);
2041 BlockDirtyBitmap *action;
2042
2043 if (action_check_completion_mode(common, errp) < 0) {
2044 return;
2045 }
2046
2047 action = common->action->u.block_dirty_bitmap_clear.data;
2048 state->bitmap = block_dirty_bitmap_lookup(action->node,
2049 action->name,
2050 &state->bs,
2051 errp);
2052 if (!state->bitmap) {
2053 return;
2054 }
2055
2056 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2057 return;
2058 }
2059
2060 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2061 }
2062
2063 static void block_dirty_bitmap_restore(BlkActionState *common)
2064 {
2065 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2066 common, common);
2067
2068 if (state->backup) {
2069 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2070 }
2071 }
2072
2073 static void block_dirty_bitmap_free_backup(BlkActionState *common)
2074 {
2075 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2076 common, common);
2077
2078 hbitmap_free(state->backup);
2079 }
2080
2081 static void block_dirty_bitmap_enable_prepare(BlkActionState *common,
2082 Error **errp)
2083 {
2084 BlockDirtyBitmap *action;
2085 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2086 common, common);
2087
2088 if (action_check_completion_mode(common, errp) < 0) {
2089 return;
2090 }
2091
2092 action = common->action->u.block_dirty_bitmap_enable.data;
2093 state->bitmap = block_dirty_bitmap_lookup(action->node,
2094 action->name,
2095 NULL,
2096 errp);
2097 if (!state->bitmap) {
2098 return;
2099 }
2100
2101 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2102 return;
2103 }
2104
2105 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2106 bdrv_enable_dirty_bitmap(state->bitmap);
2107 }
2108
2109 static void block_dirty_bitmap_enable_abort(BlkActionState *common)
2110 {
2111 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2112 common, common);
2113
2114 if (!state->was_enabled) {
2115 bdrv_disable_dirty_bitmap(state->bitmap);
2116 }
2117 }
2118
2119 static void block_dirty_bitmap_disable_prepare(BlkActionState *common,
2120 Error **errp)
2121 {
2122 BlockDirtyBitmap *action;
2123 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2124 common, common);
2125
2126 if (action_check_completion_mode(common, errp) < 0) {
2127 return;
2128 }
2129
2130 action = common->action->u.block_dirty_bitmap_disable.data;
2131 state->bitmap = block_dirty_bitmap_lookup(action->node,
2132 action->name,
2133 NULL,
2134 errp);
2135 if (!state->bitmap) {
2136 return;
2137 }
2138
2139 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2140 return;
2141 }
2142
2143 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2144 bdrv_disable_dirty_bitmap(state->bitmap);
2145 }
2146
2147 static void block_dirty_bitmap_disable_abort(BlkActionState *common)
2148 {
2149 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2150 common, common);
2151
2152 if (state->was_enabled) {
2153 bdrv_enable_dirty_bitmap(state->bitmap);
2154 }
2155 }
2156
2157 static void block_dirty_bitmap_merge_prepare(BlkActionState *common,
2158 Error **errp)
2159 {
2160 BlockDirtyBitmapMerge *action;
2161 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2162 common, common);
2163
2164 if (action_check_completion_mode(common, errp) < 0) {
2165 return;
2166 }
2167
2168 action = common->action->u.block_dirty_bitmap_merge.data;
2169
2170 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2171 action->bitmaps, &state->backup,
2172 errp);
2173 }
2174
2175 static void block_dirty_bitmap_remove_prepare(BlkActionState *common,
2176 Error **errp)
2177 {
2178 BlockDirtyBitmap *action;
2179 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2180 common, common);
2181
2182 if (action_check_completion_mode(common, errp) < 0) {
2183 return;
2184 }
2185
2186 action = common->action->u.block_dirty_bitmap_remove.data;
2187
2188 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2189 false, &state->bs, errp);
2190 if (state->bitmap) {
2191 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2192 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2193 }
2194 }
2195
2196 static void block_dirty_bitmap_remove_abort(BlkActionState *common)
2197 {
2198 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2199 common, common);
2200
2201 if (state->bitmap) {
2202 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2203 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2204 }
2205 }
2206
2207 static void block_dirty_bitmap_remove_commit(BlkActionState *common)
2208 {
2209 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2210 common, common);
2211
2212 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2213 bdrv_release_dirty_bitmap(state->bitmap);
2214 }
2215
2216 static void abort_prepare(BlkActionState *common, Error **errp)
2217 {
2218 error_setg(errp, "Transaction aborted using Abort action");
2219 }
2220
2221 static void abort_commit(BlkActionState *common)
2222 {
2223 g_assert_not_reached(); /* this action never succeeds */
2224 }
2225
2226 static const BlkActionOps actions[] = {
2227 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2228 .instance_size = sizeof(ExternalSnapshotState),
2229 .prepare = external_snapshot_prepare,
2230 .commit = external_snapshot_commit,
2231 .abort = external_snapshot_abort,
2232 .clean = external_snapshot_clean,
2233 },
2234 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
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_DRIVE_BACKUP] = {
2242 .instance_size = sizeof(DriveBackupState),
2243 .prepare = drive_backup_prepare,
2244 .commit = drive_backup_commit,
2245 .abort = drive_backup_abort,
2246 .clean = drive_backup_clean,
2247 },
2248 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2249 .instance_size = sizeof(BlockdevBackupState),
2250 .prepare = blockdev_backup_prepare,
2251 .commit = blockdev_backup_commit,
2252 .abort = blockdev_backup_abort,
2253 .clean = blockdev_backup_clean,
2254 },
2255 [TRANSACTION_ACTION_KIND_ABORT] = {
2256 .instance_size = sizeof(BlkActionState),
2257 .prepare = abort_prepare,
2258 .commit = abort_commit,
2259 },
2260 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2261 .instance_size = sizeof(InternalSnapshotState),
2262 .prepare = internal_snapshot_prepare,
2263 .abort = internal_snapshot_abort,
2264 .clean = internal_snapshot_clean,
2265 },
2266 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2267 .instance_size = sizeof(BlockDirtyBitmapState),
2268 .prepare = block_dirty_bitmap_add_prepare,
2269 .abort = block_dirty_bitmap_add_abort,
2270 },
2271 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2272 .instance_size = sizeof(BlockDirtyBitmapState),
2273 .prepare = block_dirty_bitmap_clear_prepare,
2274 .commit = block_dirty_bitmap_free_backup,
2275 .abort = block_dirty_bitmap_restore,
2276 },
2277 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2278 .instance_size = sizeof(BlockDirtyBitmapState),
2279 .prepare = block_dirty_bitmap_enable_prepare,
2280 .abort = block_dirty_bitmap_enable_abort,
2281 },
2282 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2283 .instance_size = sizeof(BlockDirtyBitmapState),
2284 .prepare = block_dirty_bitmap_disable_prepare,
2285 .abort = block_dirty_bitmap_disable_abort,
2286 },
2287 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2288 .instance_size = sizeof(BlockDirtyBitmapState),
2289 .prepare = block_dirty_bitmap_merge_prepare,
2290 .commit = block_dirty_bitmap_free_backup,
2291 .abort = block_dirty_bitmap_restore,
2292 },
2293 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE] = {
2294 .instance_size = sizeof(BlockDirtyBitmapState),
2295 .prepare = block_dirty_bitmap_remove_prepare,
2296 .commit = block_dirty_bitmap_remove_commit,
2297 .abort = block_dirty_bitmap_remove_abort,
2298 },
2299 /* Where are transactions for MIRROR, COMMIT and STREAM?
2300 * Although these blockjobs use transaction callbacks like the backup job,
2301 * these jobs do not necessarily adhere to transaction semantics.
2302 * These jobs may not fully undo all of their actions on abort, nor do they
2303 * necessarily work in transactions with more than one job in them.
2304 */
2305 };
2306
2307 /**
2308 * Allocate a TransactionProperties structure if necessary, and fill
2309 * that structure with desired defaults if they are unset.
2310 */
2311 static TransactionProperties *get_transaction_properties(
2312 TransactionProperties *props)
2313 {
2314 if (!props) {
2315 props = g_new0(TransactionProperties, 1);
2316 }
2317
2318 if (!props->has_completion_mode) {
2319 props->has_completion_mode = true;
2320 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2321 }
2322
2323 return props;
2324 }
2325
2326 /*
2327 * 'Atomic' group operations. The operations are performed as a set, and if
2328 * any fail then we roll back all operations in the group.
2329 */
2330 void qmp_transaction(TransactionActionList *dev_list,
2331 bool has_props,
2332 struct TransactionProperties *props,
2333 Error **errp)
2334 {
2335 TransactionActionList *dev_entry = dev_list;
2336 JobTxn *block_job_txn = NULL;
2337 BlkActionState *state, *next;
2338 Error *local_err = NULL;
2339
2340 QTAILQ_HEAD(, BlkActionState) snap_bdrv_states;
2341 QTAILQ_INIT(&snap_bdrv_states);
2342
2343 /* Does this transaction get canceled as a group on failure?
2344 * If not, we don't really need to make a JobTxn.
2345 */
2346 props = get_transaction_properties(props);
2347 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2348 block_job_txn = job_txn_new();
2349 }
2350
2351 /* drain all i/o before any operations */
2352 bdrv_drain_all();
2353
2354 /* We don't do anything in this loop that commits us to the operations */
2355 while (NULL != dev_entry) {
2356 TransactionAction *dev_info = NULL;
2357 const BlkActionOps *ops;
2358
2359 dev_info = dev_entry->value;
2360 dev_entry = dev_entry->next;
2361
2362 assert(dev_info->type < ARRAY_SIZE(actions));
2363
2364 ops = &actions[dev_info->type];
2365 assert(ops->instance_size > 0);
2366
2367 state = g_malloc0(ops->instance_size);
2368 state->ops = ops;
2369 state->action = dev_info;
2370 state->block_job_txn = block_job_txn;
2371 state->txn_props = props;
2372 QTAILQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2373
2374 state->ops->prepare(state, &local_err);
2375 if (local_err) {
2376 error_propagate(errp, local_err);
2377 goto delete_and_fail;
2378 }
2379 }
2380
2381 QTAILQ_FOREACH(state, &snap_bdrv_states, entry) {
2382 if (state->ops->commit) {
2383 state->ops->commit(state);
2384 }
2385 }
2386
2387 /* success */
2388 goto exit;
2389
2390 delete_and_fail:
2391 /* failure, and it is all-or-none; roll back all operations */
2392 QTAILQ_FOREACH_REVERSE(state, &snap_bdrv_states, entry) {
2393 if (state->ops->abort) {
2394 state->ops->abort(state);
2395 }
2396 }
2397 exit:
2398 QTAILQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2399 if (state->ops->clean) {
2400 state->ops->clean(state);
2401 }
2402 g_free(state);
2403 }
2404 if (!has_props) {
2405 qapi_free_TransactionProperties(props);
2406 }
2407 job_txn_unref(block_job_txn);
2408 }
2409
2410 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2411 const char *name,
2412 Error **errp)
2413 {
2414 BdrvDirtyBitmap *bitmap;
2415 BlockDriverState *bs;
2416 BlockDirtyBitmapSha256 *ret = NULL;
2417 char *sha256;
2418
2419 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2420 if (!bitmap || !bs) {
2421 return NULL;
2422 }
2423
2424 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2425 if (sha256 == NULL) {
2426 return NULL;
2427 }
2428
2429 ret = g_new(BlockDirtyBitmapSha256, 1);
2430 ret->sha256 = sha256;
2431
2432 return ret;
2433 }
2434
2435 void coroutine_fn qmp_block_resize(bool has_device, const char *device,
2436 bool has_node_name, const char *node_name,
2437 int64_t size, Error **errp)
2438 {
2439 Error *local_err = NULL;
2440 BlockBackend *blk;
2441 BlockDriverState *bs;
2442 AioContext *old_ctx;
2443
2444 bs = bdrv_lookup_bs(has_device ? device : NULL,
2445 has_node_name ? node_name : NULL,
2446 &local_err);
2447 if (local_err) {
2448 error_propagate(errp, local_err);
2449 return;
2450 }
2451
2452 if (size < 0) {
2453 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2454 return;
2455 }
2456
2457 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2458 error_setg(errp, QERR_DEVICE_IN_USE, device);
2459 return;
2460 }
2461
2462 blk = blk_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2463 if (!blk) {
2464 return;
2465 }
2466
2467 bdrv_co_lock(bs);
2468 bdrv_drained_begin(bs);
2469 bdrv_co_unlock(bs);
2470
2471 old_ctx = bdrv_co_enter(bs);
2472 blk_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2473 bdrv_co_leave(bs, old_ctx);
2474
2475 bdrv_co_lock(bs);
2476 bdrv_drained_end(bs);
2477 blk_unref(blk);
2478 bdrv_co_unlock(bs);
2479 }
2480
2481 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
2482 bool has_base, const char *base,
2483 bool has_base_node, const char *base_node,
2484 bool has_backing_file, const char *backing_file,
2485 bool has_bottom, const char *bottom,
2486 bool has_speed, int64_t speed,
2487 bool has_on_error, BlockdevOnError on_error,
2488 bool has_filter_node_name, const char *filter_node_name,
2489 bool has_auto_finalize, bool auto_finalize,
2490 bool has_auto_dismiss, bool auto_dismiss,
2491 Error **errp)
2492 {
2493 BlockDriverState *bs, *iter, *iter_end;
2494 BlockDriverState *base_bs = NULL;
2495 BlockDriverState *bottom_bs = NULL;
2496 AioContext *aio_context;
2497 Error *local_err = NULL;
2498 int job_flags = JOB_DEFAULT;
2499
2500 if (has_base && has_base_node) {
2501 error_setg(errp, "'base' and 'base-node' cannot be specified "
2502 "at the same time");
2503 return;
2504 }
2505
2506 if (has_base && has_bottom) {
2507 error_setg(errp, "'base' and 'bottom' cannot be specified "
2508 "at the same time");
2509 return;
2510 }
2511
2512 if (has_bottom && has_base_node) {
2513 error_setg(errp, "'bottom' and 'base-node' cannot be specified "
2514 "at the same time");
2515 return;
2516 }
2517
2518 if (!has_on_error) {
2519 on_error = BLOCKDEV_ON_ERROR_REPORT;
2520 }
2521
2522 bs = bdrv_lookup_bs(device, device, errp);
2523 if (!bs) {
2524 return;
2525 }
2526
2527 aio_context = bdrv_get_aio_context(bs);
2528 aio_context_acquire(aio_context);
2529
2530 if (has_base) {
2531 base_bs = bdrv_find_backing_image(bs, base);
2532 if (base_bs == NULL) {
2533 error_setg(errp, "Can't find '%s' in the backing chain", base);
2534 goto out;
2535 }
2536 assert(bdrv_get_aio_context(base_bs) == aio_context);
2537 }
2538
2539 if (has_base_node) {
2540 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2541 if (!base_bs) {
2542 goto out;
2543 }
2544 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2545 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2546 base_node, device);
2547 goto out;
2548 }
2549 assert(bdrv_get_aio_context(base_bs) == aio_context);
2550 bdrv_refresh_filename(base_bs);
2551 }
2552
2553 if (has_bottom) {
2554 bottom_bs = bdrv_lookup_bs(NULL, bottom, errp);
2555 if (!bottom_bs) {
2556 goto out;
2557 }
2558 if (!bottom_bs->drv) {
2559 error_setg(errp, "Node '%s' is not open", bottom);
2560 goto out;
2561 }
2562 if (bottom_bs->drv->is_filter) {
2563 error_setg(errp, "Node '%s' is a filter, use a non-filter node "
2564 "as 'bottom'", bottom);
2565 goto out;
2566 }
2567 if (!bdrv_chain_contains(bs, bottom_bs)) {
2568 error_setg(errp, "Node '%s' is not in a chain starting from '%s'",
2569 bottom, device);
2570 goto out;
2571 }
2572 assert(bdrv_get_aio_context(bottom_bs) == aio_context);
2573 }
2574
2575 /*
2576 * Check for op blockers in the whole chain between bs and base (or bottom)
2577 */
2578 iter_end = has_bottom ? bdrv_filter_or_cow_bs(bottom_bs) : base_bs;
2579 for (iter = bs; iter && iter != iter_end;
2580 iter = bdrv_filter_or_cow_bs(iter))
2581 {
2582 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2583 goto out;
2584 }
2585 }
2586
2587 /* if we are streaming the entire chain, the result will have no backing
2588 * file, and specifying one is therefore an error */
2589 if (base_bs == NULL && has_backing_file) {
2590 error_setg(errp, "backing file specified, but streaming the "
2591 "entire chain");
2592 goto out;
2593 }
2594
2595 if (has_auto_finalize && !auto_finalize) {
2596 job_flags |= JOB_MANUAL_FINALIZE;
2597 }
2598 if (has_auto_dismiss && !auto_dismiss) {
2599 job_flags |= JOB_MANUAL_DISMISS;
2600 }
2601
2602 stream_start(has_job_id ? job_id : NULL, bs, base_bs, backing_file,
2603 bottom_bs, job_flags, has_speed ? speed : 0, on_error,
2604 filter_node_name, &local_err);
2605 if (local_err) {
2606 error_propagate(errp, local_err);
2607 goto out;
2608 }
2609
2610 trace_qmp_block_stream(bs);
2611
2612 out:
2613 aio_context_release(aio_context);
2614 }
2615
2616 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
2617 bool has_base_node, const char *base_node,
2618 bool has_base, const char *base,
2619 bool has_top_node, const char *top_node,
2620 bool has_top, const char *top,
2621 bool has_backing_file, const char *backing_file,
2622 bool has_speed, int64_t speed,
2623 bool has_on_error, BlockdevOnError on_error,
2624 bool has_filter_node_name, const char *filter_node_name,
2625 bool has_auto_finalize, bool auto_finalize,
2626 bool has_auto_dismiss, bool auto_dismiss,
2627 Error **errp)
2628 {
2629 BlockDriverState *bs;
2630 BlockDriverState *iter;
2631 BlockDriverState *base_bs, *top_bs;
2632 AioContext *aio_context;
2633 Error *local_err = NULL;
2634 int job_flags = JOB_DEFAULT;
2635 uint64_t top_perm, top_shared;
2636
2637 if (!has_speed) {
2638 speed = 0;
2639 }
2640 if (!has_on_error) {
2641 on_error = BLOCKDEV_ON_ERROR_REPORT;
2642 }
2643 if (!has_filter_node_name) {
2644 filter_node_name = NULL;
2645 }
2646 if (has_auto_finalize && !auto_finalize) {
2647 job_flags |= JOB_MANUAL_FINALIZE;
2648 }
2649 if (has_auto_dismiss && !auto_dismiss) {
2650 job_flags |= JOB_MANUAL_DISMISS;
2651 }
2652
2653 /* Important Note:
2654 * libvirt relies on the DeviceNotFound error class in order to probe for
2655 * live commit feature versions; for this to work, we must make sure to
2656 * perform the device lookup before any generic errors that may occur in a
2657 * scenario in which all optional arguments are omitted. */
2658 bs = qmp_get_root_bs(device, &local_err);
2659 if (!bs) {
2660 bs = bdrv_lookup_bs(device, device, NULL);
2661 if (!bs) {
2662 error_free(local_err);
2663 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2664 "Device '%s' not found", device);
2665 } else {
2666 error_propagate(errp, local_err);
2667 }
2668 return;
2669 }
2670
2671 aio_context = bdrv_get_aio_context(bs);
2672 aio_context_acquire(aio_context);
2673
2674 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2675 goto out;
2676 }
2677
2678 /* default top_bs is the active layer */
2679 top_bs = bs;
2680
2681 if (has_top_node && has_top) {
2682 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2683 goto out;
2684 } else if (has_top_node) {
2685 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2686 if (top_bs == NULL) {
2687 goto out;
2688 }
2689 if (!bdrv_chain_contains(bs, top_bs)) {
2690 error_setg(errp, "'%s' is not in this backing file chain",
2691 top_node);
2692 goto out;
2693 }
2694 } else if (has_top && top) {
2695 /* This strcmp() is just a shortcut, there is no need to
2696 * refresh @bs's filename. If it mismatches,
2697 * bdrv_find_backing_image() will do the refresh and may still
2698 * return @bs. */
2699 if (strcmp(bs->filename, top) != 0) {
2700 top_bs = bdrv_find_backing_image(bs, top);
2701 }
2702 }
2703
2704 if (top_bs == NULL) {
2705 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2706 goto out;
2707 }
2708
2709 assert(bdrv_get_aio_context(top_bs) == aio_context);
2710
2711 if (has_base_node && has_base) {
2712 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2713 goto out;
2714 } else if (has_base_node) {
2715 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2716 if (base_bs == NULL) {
2717 goto out;
2718 }
2719 if (!bdrv_chain_contains(top_bs, base_bs)) {
2720 error_setg(errp, "'%s' is not in this backing file chain",
2721 base_node);
2722 goto out;
2723 }
2724 } else if (has_base && base) {
2725 base_bs = bdrv_find_backing_image(top_bs, base);
2726 if (base_bs == NULL) {
2727 error_setg(errp, "Can't find '%s' in the backing chain", base);
2728 goto out;
2729 }
2730 } else {
2731 base_bs = bdrv_find_base(top_bs);
2732 if (base_bs == NULL) {
2733 error_setg(errp, "There is no backimg image");
2734 goto out;
2735 }
2736 }
2737
2738 assert(bdrv_get_aio_context(base_bs) == aio_context);
2739
2740 for (iter = top_bs; iter != bdrv_filter_or_cow_bs(base_bs);
2741 iter = bdrv_filter_or_cow_bs(iter))
2742 {
2743 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2744 goto out;
2745 }
2746 }
2747
2748 /* Do not allow attempts to commit an image into itself */
2749 if (top_bs == base_bs) {
2750 error_setg(errp, "cannot commit an image into itself");
2751 goto out;
2752 }
2753
2754 /*
2755 * Active commit is required if and only if someone has taken a
2756 * WRITE permission on the top node. Historically, we have always
2757 * used active commit for top nodes, so continue that practice
2758 * lest we possibly break clients that rely on this behavior, e.g.
2759 * to later attach this node to a writing parent.
2760 * (Active commit is never really wrong.)
2761 */
2762 bdrv_get_cumulative_perm(top_bs, &top_perm, &top_shared);
2763 if (top_perm & BLK_PERM_WRITE ||
2764 bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs))
2765 {
2766 if (has_backing_file) {
2767 if (bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs)) {
2768 error_setg(errp, "'backing-file' specified,"
2769 " but 'top' is the active layer");
2770 } else {
2771 error_setg(errp, "'backing-file' specified, but 'top' has a "
2772 "writer on it");
2773 }
2774 goto out;
2775 }
2776 if (!has_job_id) {
2777 /*
2778 * Emulate here what block_job_create() does, because it
2779 * is possible that @bs != @top_bs (the block job should
2780 * be named after @bs, even if @top_bs is the actual
2781 * source)
2782 */
2783 job_id = bdrv_get_device_name(bs);
2784 }
2785 commit_active_start(job_id, top_bs, base_bs, job_flags, speed, on_error,
2786 filter_node_name, NULL, NULL, false, &local_err);
2787 } else {
2788 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2789 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2790 goto out;
2791 }
2792 commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, job_flags,
2793 speed, on_error, has_backing_file ? backing_file : NULL,
2794 filter_node_name, &local_err);
2795 }
2796 if (local_err != NULL) {
2797 error_propagate(errp, local_err);
2798 goto out;
2799 }
2800
2801 out:
2802 aio_context_release(aio_context);
2803 }
2804
2805 /* Common QMP interface for drive-backup and blockdev-backup */
2806 static BlockJob *do_backup_common(BackupCommon *backup,
2807 BlockDriverState *bs,
2808 BlockDriverState *target_bs,
2809 AioContext *aio_context,
2810 JobTxn *txn, Error **errp)
2811 {
2812 BlockJob *job = NULL;
2813 BdrvDirtyBitmap *bmap = NULL;
2814 BackupPerf perf = { .max_workers = 64 };
2815 int job_flags = JOB_DEFAULT;
2816
2817 if (!backup->has_speed) {
2818 backup->speed = 0;
2819 }
2820 if (!backup->has_on_source_error) {
2821 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2822 }
2823 if (!backup->has_on_target_error) {
2824 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2825 }
2826 if (!backup->has_job_id) {
2827 backup->job_id = NULL;
2828 }
2829 if (!backup->has_auto_finalize) {
2830 backup->auto_finalize = true;
2831 }
2832 if (!backup->has_auto_dismiss) {
2833 backup->auto_dismiss = true;
2834 }
2835 if (!backup->has_compress) {
2836 backup->compress = false;
2837 }
2838
2839 if (backup->x_perf) {
2840 if (backup->x_perf->has_use_copy_range) {
2841 perf.use_copy_range = backup->x_perf->use_copy_range;
2842 }
2843 if (backup->x_perf->has_max_workers) {
2844 perf.max_workers = backup->x_perf->max_workers;
2845 }
2846 if (backup->x_perf->has_max_chunk) {
2847 perf.max_chunk = backup->x_perf->max_chunk;
2848 }
2849 }
2850
2851 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2852 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2853 /* done before desugaring 'incremental' to print the right message */
2854 if (!backup->has_bitmap) {
2855 error_setg(errp, "must provide a valid bitmap name for "
2856 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2857 return NULL;
2858 }
2859 }
2860
2861 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2862 if (backup->has_bitmap_mode &&
2863 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2864 error_setg(errp, "Bitmap sync mode must be '%s' "
2865 "when using sync mode '%s'",
2866 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2867 MirrorSyncMode_str(backup->sync));
2868 return NULL;
2869 }
2870 backup->has_bitmap_mode = true;
2871 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2872 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2873 }
2874
2875 if (backup->has_bitmap) {
2876 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2877 if (!bmap) {
2878 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2879 return NULL;
2880 }
2881 if (!backup->has_bitmap_mode) {
2882 error_setg(errp, "Bitmap sync mode must be given "
2883 "when providing a bitmap");
2884 return NULL;
2885 }
2886 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2887 return NULL;
2888 }
2889
2890 /* This does not produce a useful bitmap artifact: */
2891 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2892 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2893 " outputs", MirrorSyncMode_str(backup->sync));
2894 return NULL;
2895 }
2896
2897 /* If the bitmap isn't used for input or output, this is useless: */
2898 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2899 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2900 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2901 " when combined with sync mode '%s'",
2902 BitmapSyncMode_str(backup->bitmap_mode),
2903 MirrorSyncMode_str(backup->sync));
2904 return NULL;
2905 }
2906 }
2907
2908 if (!backup->has_bitmap && backup->has_bitmap_mode) {
2909 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2910 return NULL;
2911 }
2912
2913 if (!backup->auto_finalize) {
2914 job_flags |= JOB_MANUAL_FINALIZE;
2915 }
2916 if (!backup->auto_dismiss) {
2917 job_flags |= JOB_MANUAL_DISMISS;
2918 }
2919
2920 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2921 backup->sync, bmap, backup->bitmap_mode,
2922 backup->compress,
2923 backup->filter_node_name,
2924 &perf,
2925 backup->on_source_error,
2926 backup->on_target_error,
2927 job_flags, NULL, NULL, txn, errp);
2928 return job;
2929 }
2930
2931 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2932 {
2933 TransactionAction action = {
2934 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2935 .u.drive_backup.data = backup,
2936 };
2937 blockdev_do_action(&action, errp);
2938 }
2939
2940 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2941 bool flat,
2942 Error **errp)
2943 {
2944 bool return_flat = has_flat && flat;
2945
2946 return bdrv_named_nodes_list(return_flat, errp);
2947 }
2948
2949 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2950 {
2951 return bdrv_get_xdbg_block_graph(errp);
2952 }
2953
2954 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2955 {
2956 TransactionAction action = {
2957 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2958 .u.blockdev_backup.data = backup,
2959 };
2960 blockdev_do_action(&action, errp);
2961 }
2962
2963 /* Parameter check and block job starting for drive mirroring.
2964 * Caller should hold @device and @target's aio context (must be the same).
2965 **/
2966 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2967 BlockDriverState *target,
2968 bool has_replaces, const char *replaces,
2969 enum MirrorSyncMode sync,
2970 BlockMirrorBackingMode backing_mode,
2971 bool zero_target,
2972 bool has_speed, int64_t speed,
2973 bool has_granularity, uint32_t granularity,
2974 bool has_buf_size, int64_t buf_size,
2975 bool has_on_source_error,
2976 BlockdevOnError on_source_error,
2977 bool has_on_target_error,
2978 BlockdevOnError on_target_error,
2979 bool has_unmap, bool unmap,
2980 bool has_filter_node_name,
2981 const char *filter_node_name,
2982 bool has_copy_mode, MirrorCopyMode copy_mode,
2983 bool has_auto_finalize, bool auto_finalize,
2984 bool has_auto_dismiss, bool auto_dismiss,
2985 Error **errp)
2986 {
2987 BlockDriverState *unfiltered_bs;
2988 int job_flags = JOB_DEFAULT;
2989
2990 if (!has_speed) {
2991 speed = 0;
2992 }
2993 if (!has_on_source_error) {
2994 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2995 }
2996 if (!has_on_target_error) {
2997 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2998 }
2999 if (!has_granularity) {
3000 granularity = 0;
3001 }
3002 if (!has_buf_size) {
3003 buf_size = 0;
3004 }
3005 if (!has_unmap) {
3006 unmap = true;
3007 }
3008 if (!has_filter_node_name) {
3009 filter_node_name = NULL;
3010 }
3011 if (!has_copy_mode) {
3012 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
3013 }
3014 if (has_auto_finalize && !auto_finalize) {
3015 job_flags |= JOB_MANUAL_FINALIZE;
3016 }
3017 if (has_auto_dismiss && !auto_dismiss) {
3018 job_flags |= JOB_MANUAL_DISMISS;
3019 }
3020
3021 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3022 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3023 "a value in range [512B, 64MB]");
3024 return;
3025 }
3026 if (granularity & (granularity - 1)) {
3027 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3028 "a power of 2");
3029 return;
3030 }
3031
3032 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3033 return;
3034 }
3035 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3036 return;
3037 }
3038
3039 if (!bdrv_backing_chain_next(bs) && sync == MIRROR_SYNC_MODE_TOP) {
3040 sync = MIRROR_SYNC_MODE_FULL;
3041 }
3042
3043 if (!has_replaces) {
3044 /* We want to mirror from @bs, but keep implicit filters on top */
3045 unfiltered_bs = bdrv_skip_implicit_filters(bs);
3046 if (unfiltered_bs != bs) {
3047 replaces = unfiltered_bs->node_name;
3048 has_replaces = true;
3049 }
3050 }
3051
3052 if (has_replaces) {
3053 BlockDriverState *to_replace_bs;
3054 AioContext *replace_aio_context;
3055 int64_t bs_size, replace_size;
3056
3057 bs_size = bdrv_getlength(bs);
3058 if (bs_size < 0) {
3059 error_setg_errno(errp, -bs_size, "Failed to query device's size");
3060 return;
3061 }
3062
3063 to_replace_bs = check_to_replace_node(bs, replaces, errp);
3064 if (!to_replace_bs) {
3065 return;
3066 }
3067
3068 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3069 aio_context_acquire(replace_aio_context);
3070 replace_size = bdrv_getlength(to_replace_bs);
3071 aio_context_release(replace_aio_context);
3072
3073 if (replace_size < 0) {
3074 error_setg_errno(errp, -replace_size,
3075 "Failed to query the replacement node's size");
3076 return;
3077 }
3078 if (bs_size != replace_size) {
3079 error_setg(errp, "cannot replace image with a mirror image of "
3080 "different size");
3081 return;
3082 }
3083 }
3084
3085 /* pass the node name to replace to mirror start since it's loose coupling
3086 * and will allow to check whether the node still exist at mirror completion
3087 */
3088 mirror_start(job_id, bs, target,
3089 has_replaces ? replaces : NULL, job_flags,
3090 speed, granularity, buf_size, sync, backing_mode, zero_target,
3091 on_source_error, on_target_error, unmap, filter_node_name,
3092 copy_mode, errp);
3093 }
3094
3095 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3096 {
3097 BlockDriverState *bs;
3098 BlockDriverState *target_backing_bs, *target_bs;
3099 AioContext *aio_context;
3100 AioContext *old_context;
3101 BlockMirrorBackingMode backing_mode;
3102 Error *local_err = NULL;
3103 QDict *options = NULL;
3104 int flags;
3105 int64_t size;
3106 const char *format = arg->format;
3107 bool zero_target;
3108 int ret;
3109
3110 bs = qmp_get_root_bs(arg->device, errp);
3111 if (!bs) {
3112 return;
3113 }
3114
3115 /* Early check to avoid creating target */
3116 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3117 return;
3118 }
3119
3120 aio_context = bdrv_get_aio_context(bs);
3121 aio_context_acquire(aio_context);
3122
3123 if (!arg->has_mode) {
3124 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3125 }
3126
3127 if (!arg->has_format) {
3128 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3129 ? NULL : bs->drv->format_name);
3130 }
3131
3132 flags = bs->open_flags | BDRV_O_RDWR;
3133 target_backing_bs = bdrv_cow_bs(bdrv_skip_filters(bs));
3134 if (!target_backing_bs && arg->sync == MIRROR_SYNC_MODE_TOP) {
3135 arg->sync = MIRROR_SYNC_MODE_FULL;
3136 }
3137 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3138 target_backing_bs = bs;
3139 }
3140
3141 size = bdrv_getlength(bs);
3142 if (size < 0) {
3143 error_setg_errno(errp, -size, "bdrv_getlength failed");
3144 goto out;
3145 }
3146
3147 if (arg->has_replaces) {
3148 if (!arg->has_node_name) {
3149 error_setg(errp, "a node-name must be provided when replacing a"
3150 " named node of the graph");
3151 goto out;
3152 }
3153 }
3154
3155 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3156 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3157 } else {
3158 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3159 }
3160
3161 /* Don't open backing image in create() */
3162 flags |= BDRV_O_NO_BACKING;
3163
3164 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !target_backing_bs)
3165 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3166 {
3167 /* create new image w/o backing file */
3168 assert(format);
3169 bdrv_img_create(arg->target, format,
3170 NULL, NULL, NULL, size, flags, false, &local_err);
3171 } else {
3172 /* Implicit filters should not appear in the filename */
3173 BlockDriverState *explicit_backing =
3174 bdrv_skip_implicit_filters(target_backing_bs);
3175
3176 switch (arg->mode) {
3177 case NEW_IMAGE_MODE_EXISTING:
3178 break;
3179 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3180 /* create new image with backing file */
3181 bdrv_refresh_filename(explicit_backing);
3182 bdrv_img_create(arg->target, format,
3183 explicit_backing->filename,
3184 explicit_backing->drv->format_name,
3185 NULL, size, flags, false, &local_err);
3186 break;
3187 default:
3188 abort();
3189 }
3190 }
3191
3192 if (local_err) {
3193 error_propagate(errp, local_err);
3194 goto out;
3195 }
3196
3197 options = qdict_new();
3198 if (arg->has_node_name) {
3199 qdict_put_str(options, "node-name", arg->node_name);
3200 }
3201 if (format) {
3202 qdict_put_str(options, "driver", format);
3203 }
3204
3205 /* Mirroring takes care of copy-on-write using the source's backing
3206 * file.
3207 */
3208 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3209 if (!target_bs) {
3210 goto out;
3211 }
3212
3213 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3214 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3215 !bdrv_has_zero_init(target_bs)));
3216
3217
3218 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3219 old_context = bdrv_get_aio_context(target_bs);
3220 aio_context_release(aio_context);
3221 aio_context_acquire(old_context);
3222
3223 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3224 if (ret < 0) {
3225 bdrv_unref(target_bs);
3226 aio_context_release(old_context);
3227 return;
3228 }
3229
3230 aio_context_release(old_context);
3231 aio_context_acquire(aio_context);
3232
3233 blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3234 arg->has_replaces, arg->replaces, arg->sync,
3235 backing_mode, zero_target,
3236 arg->has_speed, arg->speed,
3237 arg->has_granularity, arg->granularity,
3238 arg->has_buf_size, arg->buf_size,
3239 arg->has_on_source_error, arg->on_source_error,
3240 arg->has_on_target_error, arg->on_target_error,
3241 arg->has_unmap, arg->unmap,
3242 false, NULL,
3243 arg->has_copy_mode, arg->copy_mode,
3244 arg->has_auto_finalize, arg->auto_finalize,
3245 arg->has_auto_dismiss, arg->auto_dismiss,
3246 errp);
3247 bdrv_unref(target_bs);
3248 out:
3249 aio_context_release(aio_context);
3250 }
3251
3252 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3253 const char *device, const char *target,
3254 bool has_replaces, const char *replaces,
3255 MirrorSyncMode sync,
3256 bool has_speed, int64_t speed,
3257 bool has_granularity, uint32_t granularity,
3258 bool has_buf_size, int64_t buf_size,
3259 bool has_on_source_error,
3260 BlockdevOnError on_source_error,
3261 bool has_on_target_error,
3262 BlockdevOnError on_target_error,
3263 bool has_filter_node_name,
3264 const char *filter_node_name,
3265 bool has_copy_mode, MirrorCopyMode copy_mode,
3266 bool has_auto_finalize, bool auto_finalize,
3267 bool has_auto_dismiss, bool auto_dismiss,
3268 Error **errp)
3269 {
3270 BlockDriverState *bs;
3271 BlockDriverState *target_bs;
3272 AioContext *aio_context;
3273 AioContext *old_context;
3274 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3275 bool zero_target;
3276 int ret;
3277
3278 bs = qmp_get_root_bs(device, errp);
3279 if (!bs) {
3280 return;
3281 }
3282
3283 target_bs = bdrv_lookup_bs(target, target, errp);
3284 if (!target_bs) {
3285 return;
3286 }
3287
3288 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3289
3290 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3291 old_context = bdrv_get_aio_context(target_bs);
3292 aio_context = bdrv_get_aio_context(bs);
3293 aio_context_acquire(old_context);
3294
3295 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3296
3297 aio_context_release(old_context);
3298 aio_context_acquire(aio_context);
3299
3300 if (ret < 0) {
3301 goto out;
3302 }
3303
3304 blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
3305 has_replaces, replaces, sync, backing_mode,
3306 zero_target, has_speed, speed,
3307 has_granularity, granularity,
3308 has_buf_size, buf_size,
3309 has_on_source_error, on_source_error,
3310 has_on_target_error, on_target_error,
3311 true, true,
3312 has_filter_node_name, filter_node_name,
3313 has_copy_mode, copy_mode,
3314 has_auto_finalize, auto_finalize,
3315 has_auto_dismiss, auto_dismiss,
3316 errp);
3317 out:
3318 aio_context_release(aio_context);
3319 }
3320
3321 /* Get a block job using its ID and acquire its AioContext */
3322 static BlockJob *find_block_job(const char *id, AioContext **aio_context,
3323 Error **errp)
3324 {
3325 BlockJob *job;
3326
3327 assert(id != NULL);
3328
3329 *aio_context = NULL;
3330
3331 job = block_job_get(id);
3332
3333 if (!job) {
3334 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3335 "Block job '%s' not found", id);
3336 return NULL;
3337 }
3338
3339 *aio_context = blk_get_aio_context(job->blk);
3340 aio_context_acquire(*aio_context);
3341
3342 return job;
3343 }
3344
3345 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3346 {
3347 AioContext *aio_context;
3348 BlockJob *job = find_block_job(device, &aio_context, errp);
3349
3350 if (!job) {
3351 return;
3352 }
3353
3354 block_job_set_speed(job, speed, errp);
3355 aio_context_release(aio_context);
3356 }
3357
3358 void qmp_block_job_cancel(const char *device,
3359 bool has_force, bool force, Error **errp)
3360 {
3361 AioContext *aio_context;
3362 BlockJob *job = find_block_job(device, &aio_context, errp);
3363
3364 if (!job) {
3365 return;
3366 }
3367
3368 if (!has_force) {
3369 force = false;
3370 }
3371
3372 if (job_user_paused(&job->job) && !force) {
3373 error_setg(errp, "The block job for device '%s' is currently paused",
3374 device);
3375 goto out;
3376 }
3377
3378 trace_qmp_block_job_cancel(job);
3379 job_user_cancel(&job->job, force, errp);
3380 out:
3381 aio_context_release(aio_context);
3382 }
3383
3384 void qmp_block_job_pause(const char *device, Error **errp)
3385 {
3386 AioContext *aio_context;
3387 BlockJob *job = find_block_job(device, &aio_context, errp);
3388
3389 if (!job) {
3390 return;
3391 }
3392
3393 trace_qmp_block_job_pause(job);
3394 job_user_pause(&job->job, errp);
3395 aio_context_release(aio_context);
3396 }
3397
3398 void qmp_block_job_resume(const char *device, Error **errp)
3399 {
3400 AioContext *aio_context;
3401 BlockJob *job = find_block_job(device, &aio_context, errp);
3402
3403 if (!job) {
3404 return;
3405 }
3406
3407 trace_qmp_block_job_resume(job);
3408 job_user_resume(&job->job, errp);
3409 aio_context_release(aio_context);
3410 }
3411
3412 void qmp_block_job_complete(const char *device, Error **errp)
3413 {
3414 AioContext *aio_context;
3415 BlockJob *job = find_block_job(device, &aio_context, errp);
3416
3417 if (!job) {
3418 return;
3419 }
3420
3421 trace_qmp_block_job_complete(job);
3422 job_complete(&job->job, errp);
3423 aio_context_release(aio_context);
3424 }
3425
3426 void qmp_block_job_finalize(const char *id, Error **errp)
3427 {
3428 AioContext *aio_context;
3429 BlockJob *job = find_block_job(id, &aio_context, errp);
3430
3431 if (!job) {
3432 return;
3433 }
3434
3435 trace_qmp_block_job_finalize(job);
3436 job_ref(&job->job);
3437 job_finalize(&job->job, errp);
3438
3439 /*
3440 * Job's context might have changed via job_finalize (and job_txn_apply
3441 * automatically acquires the new one), so make sure we release the correct
3442 * one.
3443 */
3444 aio_context = blk_get_aio_context(job->blk);
3445 job_unref(&job->job);
3446 aio_context_release(aio_context);
3447 }
3448
3449 void qmp_block_job_dismiss(const char *id, Error **errp)
3450 {
3451 AioContext *aio_context;
3452 BlockJob *bjob = find_block_job(id, &aio_context, errp);
3453 Job *job;
3454
3455 if (!bjob) {
3456 return;
3457 }
3458
3459 trace_qmp_block_job_dismiss(bjob);
3460 job = &bjob->job;
3461 job_dismiss(&job, errp);
3462 aio_context_release(aio_context);
3463 }
3464
3465 void qmp_change_backing_file(const char *device,
3466 const char *image_node_name,
3467 const char *backing_file,
3468 Error **errp)
3469 {
3470 BlockDriverState *bs = NULL;
3471 AioContext *aio_context;
3472 BlockDriverState *image_bs = NULL;
3473 Error *local_err = NULL;
3474 bool ro;
3475 int ret;
3476
3477 bs = qmp_get_root_bs(device, errp);
3478 if (!bs) {
3479 return;
3480 }
3481
3482 aio_context = bdrv_get_aio_context(bs);
3483 aio_context_acquire(aio_context);
3484
3485 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3486 if (local_err) {
3487 error_propagate(errp, local_err);
3488 goto out;
3489 }
3490
3491 if (!image_bs) {
3492 error_setg(errp, "image file not found");
3493 goto out;
3494 }
3495
3496 if (bdrv_find_base(image_bs) == image_bs) {
3497 error_setg(errp, "not allowing backing file change on an image "
3498 "without a backing file");
3499 goto out;
3500 }
3501
3502 /* even though we are not necessarily operating on bs, we need it to
3503 * determine if block ops are currently prohibited on the chain */
3504 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3505 goto out;
3506 }
3507
3508 /* final sanity check */
3509 if (!bdrv_chain_contains(bs, image_bs)) {
3510 error_setg(errp, "'%s' and image file are not in the same chain",
3511 device);
3512 goto out;
3513 }
3514
3515 /* if not r/w, reopen to make r/w */
3516 ro = bdrv_is_read_only(image_bs);
3517
3518 if (ro) {
3519 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3520 goto out;
3521 }
3522 }
3523
3524 ret = bdrv_change_backing_file(image_bs, backing_file,
3525 image_bs->drv ? image_bs->drv->format_name : "",
3526 false);
3527
3528 if (ret < 0) {
3529 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3530 backing_file);
3531 /* don't exit here, so we can try to restore open flags if
3532 * appropriate */
3533 }
3534
3535 if (ro) {
3536 bdrv_reopen_set_read_only(image_bs, true, errp);
3537 }
3538
3539 out:
3540 aio_context_release(aio_context);
3541 }
3542
3543 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3544 {
3545 BlockDriverState *bs;
3546 QObject *obj;
3547 Visitor *v = qobject_output_visitor_new(&obj);
3548 QDict *qdict;
3549
3550 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3551 visit_complete(v, &obj);
3552 qdict = qobject_to(QDict, obj);
3553
3554 qdict_flatten(qdict);
3555
3556 if (!qdict_get_try_str(qdict, "node-name")) {
3557 error_setg(errp, "'node-name' must be specified for the root node");
3558 goto fail;
3559 }
3560
3561 bs = bds_tree_init(qdict, errp);
3562 if (!bs) {
3563 goto fail;
3564 }
3565
3566 bdrv_set_monitor_owned(bs);
3567
3568 fail:
3569 visit_free(v);
3570 }
3571
3572 void qmp_x_blockdev_reopen(BlockdevOptions *options, Error **errp)
3573 {
3574 BlockDriverState *bs;
3575 AioContext *ctx;
3576 QObject *obj;
3577 Visitor *v = qobject_output_visitor_new(&obj);
3578 BlockReopenQueue *queue;
3579 QDict *qdict;
3580
3581 /* Check for the selected node name */
3582 if (!options->has_node_name) {
3583 error_setg(errp, "node-name not specified");
3584 goto fail;
3585 }
3586
3587 bs = bdrv_find_node(options->node_name);
3588 if (!bs) {
3589 error_setg(errp, "Failed to find node with node-name='%s'",
3590 options->node_name);
3591 goto fail;
3592 }
3593
3594 /* Put all options in a QDict and flatten it */
3595 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3596 visit_complete(v, &obj);
3597 qdict = qobject_to(QDict, obj);
3598
3599 qdict_flatten(qdict);
3600
3601 /* Perform the reopen operation */
3602 ctx = bdrv_get_aio_context(bs);
3603 aio_context_acquire(ctx);
3604 bdrv_subtree_drained_begin(bs);
3605 queue = bdrv_reopen_queue(NULL, bs, qdict, false);
3606 bdrv_reopen_multiple(queue, errp);
3607 bdrv_subtree_drained_end(bs);
3608 aio_context_release(ctx);
3609
3610 fail:
3611 visit_free(v);
3612 }
3613
3614 void qmp_blockdev_del(const char *node_name, Error **errp)
3615 {
3616 AioContext *aio_context;
3617 BlockDriverState *bs;
3618
3619 bs = bdrv_find_node(node_name);
3620 if (!bs) {
3621 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3622 return;
3623 }
3624 if (bdrv_has_blk(bs)) {
3625 error_setg(errp, "Node %s is in use", node_name);
3626 return;
3627 }
3628 aio_context = bdrv_get_aio_context(bs);
3629 aio_context_acquire(aio_context);
3630
3631 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3632 goto out;
3633 }
3634
3635 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3636 error_setg(errp, "Node %s is not owned by the monitor",
3637 bs->node_name);
3638 goto out;
3639 }
3640
3641 if (bs->refcnt > 1) {
3642 error_setg(errp, "Block device %s is in use",
3643 bdrv_get_device_or_node_name(bs));
3644 goto out;
3645 }
3646
3647 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3648 bdrv_unref(bs);
3649
3650 out:
3651 aio_context_release(aio_context);
3652 }
3653
3654 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3655 const char *child_name)
3656 {
3657 BdrvChild *child;
3658
3659 QLIST_FOREACH(child, &parent_bs->children, next) {
3660 if (strcmp(child->name, child_name) == 0) {
3661 return child;
3662 }
3663 }
3664
3665 return NULL;
3666 }
3667
3668 void qmp_x_blockdev_change(const char *parent, bool has_child,
3669 const char *child, bool has_node,
3670 const char *node, Error **errp)
3671 {
3672 BlockDriverState *parent_bs, *new_bs = NULL;
3673 BdrvChild *p_child;
3674
3675 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3676 if (!parent_bs) {
3677 return;
3678 }
3679
3680 if (has_child == has_node) {
3681 if (has_child) {
3682 error_setg(errp, "The parameters child and node are in conflict");
3683 } else {
3684 error_setg(errp, "Either child or node must be specified");
3685 }
3686 return;
3687 }
3688
3689 if (has_child) {
3690 p_child = bdrv_find_child(parent_bs, child);
3691 if (!p_child) {
3692 error_setg(errp, "Node '%s' does not have child '%s'",
3693 parent, child);
3694 return;
3695 }
3696 bdrv_del_child(parent_bs, p_child, errp);
3697 }
3698
3699 if (has_node) {
3700 new_bs = bdrv_find_node(node);
3701 if (!new_bs) {
3702 error_setg(errp, "Node '%s' not found", node);
3703 return;
3704 }
3705 bdrv_add_child(parent_bs, new_bs, errp);
3706 }
3707 }
3708
3709 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3710 {
3711 BlockJobInfoList *head = NULL, **tail = &head;
3712 BlockJob *job;
3713
3714 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
3715 BlockJobInfo *value;
3716 AioContext *aio_context;
3717
3718 if (block_job_is_internal(job)) {
3719 continue;
3720 }
3721 aio_context = blk_get_aio_context(job->blk);
3722 aio_context_acquire(aio_context);
3723 value = block_job_query(job, errp);
3724 aio_context_release(aio_context);
3725 if (!value) {
3726 qapi_free_BlockJobInfoList(head);
3727 return NULL;
3728 }
3729 QAPI_LIST_APPEND(tail, value);
3730 }
3731
3732 return head;
3733 }
3734
3735 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3736 bool has_force, bool force, Error **errp)
3737 {
3738 AioContext *old_context;
3739 AioContext *new_context;
3740 BlockDriverState *bs;
3741
3742 bs = bdrv_find_node(node_name);
3743 if (!bs) {
3744 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3745 return;
3746 }
3747
3748 /* Protects against accidents. */
3749 if (!(has_force && force) && bdrv_has_blk(bs)) {
3750 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3751 "be in use (use force=true to override this check)",
3752 node_name);
3753 return;
3754 }
3755
3756 if (iothread->type == QTYPE_QSTRING) {
3757 IOThread *obj = iothread_by_id(iothread->u.s);
3758 if (!obj) {
3759 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3760 return;
3761 }
3762
3763 new_context = iothread_get_aio_context(obj);
3764 } else {
3765 new_context = qemu_get_aio_context();
3766 }
3767
3768 old_context = bdrv_get_aio_context(bs);
3769 aio_context_acquire(old_context);
3770
3771 bdrv_try_set_aio_context(bs, new_context, errp);
3772
3773 aio_context_release(old_context);
3774 }
3775
3776 QemuOptsList qemu_common_drive_opts = {
3777 .name = "drive",
3778 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3779 .desc = {
3780 {
3781 .name = "snapshot",
3782 .type = QEMU_OPT_BOOL,
3783 .help = "enable/disable snapshot mode",
3784 },{
3785 .name = "aio",
3786 .type = QEMU_OPT_STRING,
3787 .help = "host AIO implementation (threads, native, io_uring)",
3788 },{
3789 .name = BDRV_OPT_CACHE_WB,
3790 .type = QEMU_OPT_BOOL,
3791 .help = "Enable writeback mode",
3792 },{
3793 .name = "format",
3794 .type = QEMU_OPT_STRING,
3795 .help = "disk format (raw, qcow2, ...)",
3796 },{
3797 .name = "rerror",
3798 .type = QEMU_OPT_STRING,
3799 .help = "read error action",
3800 },{
3801 .name = "werror",
3802 .type = QEMU_OPT_STRING,
3803 .help = "write error action",
3804 },{
3805 .name = BDRV_OPT_READ_ONLY,
3806 .type = QEMU_OPT_BOOL,
3807 .help = "open drive file as read-only",
3808 },
3809
3810 THROTTLE_OPTS,
3811
3812 {
3813 .name = "throttling.group",
3814 .type = QEMU_OPT_STRING,
3815 .help = "name of the block throttling group",
3816 },{
3817 .name = "copy-on-read",
3818 .type = QEMU_OPT_BOOL,
3819 .help = "copy read data from backing file into image file",
3820 },{
3821 .name = "detect-zeroes",
3822 .type = QEMU_OPT_STRING,
3823 .help = "try to optimize zero writes (off, on, unmap)",
3824 },{
3825 .name = "stats-account-invalid",
3826 .type = QEMU_OPT_BOOL,
3827 .help = "whether to account for invalid I/O operations "
3828 "in the statistics",
3829 },{
3830 .name = "stats-account-failed",
3831 .type = QEMU_OPT_BOOL,
3832 .help = "whether to account for failed I/O operations "
3833 "in the statistics",
3834 },
3835 { /* end of list */ }
3836 },
3837 };
3838
3839 QemuOptsList qemu_drive_opts = {
3840 .name = "drive",
3841 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3842 .desc = {
3843 /*
3844 * no elements => accept any params
3845 * validation will happen later
3846 */
3847 { /* end of list */ }
3848 },
3849 };
Random patches