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