xref: /openbmc/qemu/blockdev.c (revision b55d7d34)
1 /*
2  * QEMU host block devices
3  *
4  * Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  * This work is licensed under the terms of the GNU GPL, version 2 or
7  * later.  See the COPYING file in the top-level directory.
8  *
9  * This file incorporates work covered by the following copyright and
10  * permission notice:
11  *
12  * Copyright (c) 2003-2008 Fabrice Bellard
13  *
14  * Permission is hereby granted, free of charge, to any person obtaining a copy
15  * of this software and associated documentation files (the "Software"), to deal
16  * in the Software without restriction, including without limitation the rights
17  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
18  * copies of the Software, and to permit persons to whom the Software is
19  * furnished to do so, subject to the following conditions:
20  *
21  * The above copyright notice and this permission notice shall be included in
22  * all copies or substantial portions of the Software.
23  *
24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
27  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
30  * THE SOFTWARE.
31  */
32 
33 #include "qemu/osdep.h"
34 #include "sysemu/block-backend.h"
35 #include "sysemu/blockdev.h"
36 #include "hw/block/block.h"
37 #include "block/blockjob.h"
38 #include "block/qdict.h"
39 #include "block/throttle-groups.h"
40 #include "monitor/monitor.h"
41 #include "qemu/error-report.h"
42 #include "qemu/option.h"
43 #include "qemu/qemu-print.h"
44 #include "qemu/config-file.h"
45 #include "qapi/qapi-commands-block.h"
46 #include "qapi/qapi-commands-transaction.h"
47 #include "qapi/qapi-visit-block-core.h"
48 #include "qapi/qmp/qdict.h"
49 #include "qapi/qmp/qnum.h"
50 #include "qapi/qmp/qstring.h"
51 #include "qapi/error.h"
52 #include "qapi/qmp/qerror.h"
53 #include "qapi/qmp/qlist.h"
54 #include "qapi/qobject-output-visitor.h"
55 #include "sysemu/sysemu.h"
56 #include "sysemu/iothread.h"
57 #include "block/block_int.h"
58 #include "block/trace.h"
59 #include "sysemu/arch_init.h"
60 #include "sysemu/qtest.h"
61 #include "qemu/cutils.h"
62 #include "qemu/help_option.h"
63 #include "qemu/throttle-options.h"
64 
65 static QTAILQ_HEAD(, BlockDriverState) monitor_bdrv_states =
66     QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states);
67 
68 static int do_open_tray(const char *blk_name, const char *qdev_id,
69                         bool force, Error **errp);
70 static void blockdev_remove_medium(bool has_device, const char *device,
71                                    bool has_id, const char *id, Error **errp);
72 static void blockdev_insert_medium(bool has_device, const char *device,
73                                    bool has_id, const char *id,
74                                    const char *node_name, Error **errp);
75 
76 static const char *const if_name[IF_COUNT] = {
77     [IF_NONE] = "none",
78     [IF_IDE] = "ide",
79     [IF_SCSI] = "scsi",
80     [IF_FLOPPY] = "floppy",
81     [IF_PFLASH] = "pflash",
82     [IF_MTD] = "mtd",
83     [IF_SD] = "sd",
84     [IF_VIRTIO] = "virtio",
85     [IF_XEN] = "xen",
86 };
87 
88 static int if_max_devs[IF_COUNT] = {
89     /*
90      * Do not change these numbers!  They govern how drive option
91      * index maps to unit and bus.  That mapping is ABI.
92      *
93      * All controllers used to implement if=T drives need to support
94      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
95      * Otherwise, some index values map to "impossible" bus, unit
96      * values.
97      *
98      * For instance, if you change [IF_SCSI] to 255, -drive
99      * if=scsi,index=12 no longer means bus=1,unit=5, but
100      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
101      * the drive can't be set up.  Regression.
102      */
103     [IF_IDE] = 2,
104     [IF_SCSI] = 7,
105 };
106 
107 /**
108  * Boards may call this to offer board-by-board overrides
109  * of the default, global values.
110  */
111 void override_max_devs(BlockInterfaceType type, int max_devs)
112 {
113     BlockBackend *blk;
114     DriveInfo *dinfo;
115 
116     if (max_devs <= 0) {
117         return;
118     }
119 
120     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
121         dinfo = blk_legacy_dinfo(blk);
122         if (dinfo->type == type) {
123             fprintf(stderr, "Cannot override units-per-bus property of"
124                     " the %s interface, because a drive of that type has"
125                     " already been added.\n", if_name[type]);
126             g_assert_not_reached();
127         }
128     }
129 
130     if_max_devs[type] = max_devs;
131 }
132 
133 /*
134  * We automatically delete the drive when a device using it gets
135  * unplugged.  Questionable feature, but we can't just drop it.
136  * Device models call blockdev_mark_auto_del() to schedule the
137  * automatic deletion, and generic qdev code calls blockdev_auto_del()
138  * when deletion is actually safe.
139  */
140 void blockdev_mark_auto_del(BlockBackend *blk)
141 {
142     DriveInfo *dinfo = blk_legacy_dinfo(blk);
143     BlockJob *job;
144 
145     if (!dinfo) {
146         return;
147     }
148 
149     for (job = block_job_next(NULL); job; job = block_job_next(job)) {
150         if (block_job_has_bdrv(job, blk_bs(blk))) {
151             AioContext *aio_context = job->job.aio_context;
152             aio_context_acquire(aio_context);
153 
154             job_cancel(&job->job, false);
155 
156             aio_context_release(aio_context);
157         }
158     }
159 
160     dinfo->auto_del = 1;
161 }
162 
163 void blockdev_auto_del(BlockBackend *blk)
164 {
165     DriveInfo *dinfo = blk_legacy_dinfo(blk);
166 
167     if (dinfo && dinfo->auto_del) {
168         monitor_remove_blk(blk);
169         blk_unref(blk);
170     }
171 }
172 
173 /**
174  * Returns the current mapping of how many units per bus
175  * a particular interface can support.
176  *
177  *  A positive integer indicates n units per bus.
178  *  0 implies the mapping has not been established.
179  * -1 indicates an invalid BlockInterfaceType was given.
180  */
181 int drive_get_max_devs(BlockInterfaceType type)
182 {
183     if (type >= IF_IDE && type < IF_COUNT) {
184         return if_max_devs[type];
185     }
186 
187     return -1;
188 }
189 
190 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
191 {
192     int max_devs = if_max_devs[type];
193     return max_devs ? index / max_devs : 0;
194 }
195 
196 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
197 {
198     int max_devs = if_max_devs[type];
199     return max_devs ? index % max_devs : index;
200 }
201 
202 QemuOpts *drive_def(const char *optstr)
203 {
204     return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
205 }
206 
207 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
208                     const char *optstr)
209 {
210     QemuOpts *opts;
211 
212     opts = drive_def(optstr);
213     if (!opts) {
214         return NULL;
215     }
216     if (type != IF_DEFAULT) {
217         qemu_opt_set(opts, "if", if_name[type], &error_abort);
218     }
219     if (index >= 0) {
220         qemu_opt_set_number(opts, "index", index, &error_abort);
221     }
222     if (file)
223         qemu_opt_set(opts, "file", file, &error_abort);
224     return opts;
225 }
226 
227 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
228 {
229     BlockBackend *blk;
230     DriveInfo *dinfo;
231 
232     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
233         dinfo = blk_legacy_dinfo(blk);
234         if (dinfo && dinfo->type == type
235             && dinfo->bus == bus && dinfo->unit == unit) {
236             return dinfo;
237         }
238     }
239 
240     return NULL;
241 }
242 
243 void drive_check_orphaned(void)
244 {
245     BlockBackend *blk;
246     DriveInfo *dinfo;
247     Location loc;
248     bool orphans = false;
249 
250     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
251         dinfo = blk_legacy_dinfo(blk);
252         if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
253             dinfo->type != IF_NONE) {
254             loc_push_none(&loc);
255             qemu_opts_loc_restore(dinfo->opts);
256             error_report("machine type does not support"
257                          " if=%s,bus=%d,unit=%d",
258                          if_name[dinfo->type], dinfo->bus, dinfo->unit);
259             loc_pop(&loc);
260             orphans = true;
261         }
262     }
263 
264     if (orphans) {
265         exit(1);
266     }
267 }
268 
269 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
270 {
271     return drive_get(type,
272                      drive_index_to_bus_id(type, index),
273                      drive_index_to_unit_id(type, index));
274 }
275 
276 int drive_get_max_bus(BlockInterfaceType type)
277 {
278     int max_bus;
279     BlockBackend *blk;
280     DriveInfo *dinfo;
281 
282     max_bus = -1;
283     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
284         dinfo = blk_legacy_dinfo(blk);
285         if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
286             max_bus = dinfo->bus;
287         }
288     }
289     return max_bus;
290 }
291 
292 /* Get a block device.  This should only be used for single-drive devices
293    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
294    appropriate bus.  */
295 DriveInfo *drive_get_next(BlockInterfaceType type)
296 {
297     static int next_block_unit[IF_COUNT];
298 
299     return drive_get(type, 0, next_block_unit[type]++);
300 }
301 
302 static void bdrv_format_print(void *opaque, const char *name)
303 {
304     qemu_printf(" %s", name);
305 }
306 
307 typedef struct {
308     QEMUBH *bh;
309     BlockDriverState *bs;
310 } BDRVPutRefBH;
311 
312 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
313 {
314     if (!strcmp(buf, "ignore")) {
315         return BLOCKDEV_ON_ERROR_IGNORE;
316     } else if (!is_read && !strcmp(buf, "enospc")) {
317         return BLOCKDEV_ON_ERROR_ENOSPC;
318     } else if (!strcmp(buf, "stop")) {
319         return BLOCKDEV_ON_ERROR_STOP;
320     } else if (!strcmp(buf, "report")) {
321         return BLOCKDEV_ON_ERROR_REPORT;
322     } else {
323         error_setg(errp, "'%s' invalid %s error action",
324                    buf, is_read ? "read" : "write");
325         return -1;
326     }
327 }
328 
329 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
330                                   Error **errp)
331 {
332     const QListEntry *entry;
333     for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
334         switch (qobject_type(entry->value)) {
335 
336         case QTYPE_QSTRING: {
337             unsigned long long length;
338             const char *str = qstring_get_str(qobject_to(QString,
339                                                          entry->value));
340             if (parse_uint_full(str, &length, 10) == 0 &&
341                 length > 0 && length <= UINT_MAX) {
342                 block_acct_add_interval(stats, (unsigned) length);
343             } else {
344                 error_setg(errp, "Invalid interval length: %s", str);
345                 return false;
346             }
347             break;
348         }
349 
350         case QTYPE_QNUM: {
351             int64_t length = qnum_get_int(qobject_to(QNum, entry->value));
352 
353             if (length > 0 && length <= UINT_MAX) {
354                 block_acct_add_interval(stats, (unsigned) length);
355             } else {
356                 error_setg(errp, "Invalid interval length: %" PRId64, length);
357                 return false;
358             }
359             break;
360         }
361 
362         default:
363             error_setg(errp, "The specification of stats-intervals is invalid");
364             return false;
365         }
366     }
367     return true;
368 }
369 
370 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
371 
372 /* All parameters but @opts are optional and may be set to NULL. */
373 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
374     const char **throttling_group, ThrottleConfig *throttle_cfg,
375     BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
376 {
377     Error *local_error = NULL;
378     const char *aio;
379 
380     if (bdrv_flags) {
381         if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
382             *bdrv_flags |= BDRV_O_COPY_ON_READ;
383         }
384 
385         if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
386             if (!strcmp(aio, "native")) {
387                 *bdrv_flags |= BDRV_O_NATIVE_AIO;
388             } else if (!strcmp(aio, "threads")) {
389                 /* this is the default */
390             } else {
391                error_setg(errp, "invalid aio option");
392                return;
393             }
394         }
395     }
396 
397     /* disk I/O throttling */
398     if (throttling_group) {
399         *throttling_group = qemu_opt_get(opts, "throttling.group");
400     }
401 
402     if (throttle_cfg) {
403         throttle_config_init(throttle_cfg);
404         throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
405             qemu_opt_get_number(opts, "throttling.bps-total", 0);
406         throttle_cfg->buckets[THROTTLE_BPS_READ].avg  =
407             qemu_opt_get_number(opts, "throttling.bps-read", 0);
408         throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
409             qemu_opt_get_number(opts, "throttling.bps-write", 0);
410         throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
411             qemu_opt_get_number(opts, "throttling.iops-total", 0);
412         throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
413             qemu_opt_get_number(opts, "throttling.iops-read", 0);
414         throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
415             qemu_opt_get_number(opts, "throttling.iops-write", 0);
416 
417         throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
418             qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
419         throttle_cfg->buckets[THROTTLE_BPS_READ].max  =
420             qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
421         throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
422             qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
423         throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
424             qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
425         throttle_cfg->buckets[THROTTLE_OPS_READ].max =
426             qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
427         throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
428             qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
429 
430         throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
431             qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
432         throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length  =
433             qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
434         throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
435             qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
436         throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
437             qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
438         throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
439             qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
440         throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
441             qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
442 
443         throttle_cfg->op_size =
444             qemu_opt_get_number(opts, "throttling.iops-size", 0);
445 
446         if (!throttle_is_valid(throttle_cfg, errp)) {
447             return;
448         }
449     }
450 
451     if (detect_zeroes) {
452         *detect_zeroes =
453             qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
454                             qemu_opt_get(opts, "detect-zeroes"),
455                             BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
456                             &local_error);
457         if (local_error) {
458             error_propagate(errp, local_error);
459             return;
460         }
461     }
462 }
463 
464 /* Takes the ownership of bs_opts */
465 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
466                                    Error **errp)
467 {
468     const char *buf;
469     int bdrv_flags = 0;
470     int on_read_error, on_write_error;
471     bool account_invalid, account_failed;
472     bool writethrough, read_only;
473     BlockBackend *blk;
474     BlockDriverState *bs;
475     ThrottleConfig cfg;
476     int snapshot = 0;
477     Error *error = NULL;
478     QemuOpts *opts;
479     QDict *interval_dict = NULL;
480     QList *interval_list = NULL;
481     const char *id;
482     BlockdevDetectZeroesOptions detect_zeroes =
483         BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
484     const char *throttling_group = NULL;
485 
486     /* Check common options by copying from bs_opts to opts, all other options
487      * stay in bs_opts for processing by bdrv_open(). */
488     id = qdict_get_try_str(bs_opts, "id");
489     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
490     if (error) {
491         error_propagate(errp, error);
492         goto err_no_opts;
493     }
494 
495     qemu_opts_absorb_qdict(opts, bs_opts, &error);
496     if (error) {
497         error_propagate(errp, error);
498         goto early_err;
499     }
500 
501     if (id) {
502         qdict_del(bs_opts, "id");
503     }
504 
505     /* extract parameters */
506     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
507 
508     account_invalid = qemu_opt_get_bool(opts, "stats-account-invalid", true);
509     account_failed = qemu_opt_get_bool(opts, "stats-account-failed", true);
510 
511     writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
512 
513     id = qemu_opts_id(opts);
514 
515     qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
516     qdict_array_split(interval_dict, &interval_list);
517 
518     if (qdict_size(interval_dict) != 0) {
519         error_setg(errp, "Invalid option stats-intervals.%s",
520                    qdict_first(interval_dict)->key);
521         goto early_err;
522     }
523 
524     extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
525                                     &detect_zeroes, &error);
526     if (error) {
527         error_propagate(errp, error);
528         goto early_err;
529     }
530 
531     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
532         if (is_help_option(buf)) {
533             qemu_printf("Supported formats:");
534             bdrv_iterate_format(bdrv_format_print, NULL, false);
535             qemu_printf("\nSupported formats (read-only):");
536             bdrv_iterate_format(bdrv_format_print, NULL, true);
537             qemu_printf("\n");
538             goto early_err;
539         }
540 
541         if (qdict_haskey(bs_opts, "driver")) {
542             error_setg(errp, "Cannot specify both 'driver' and 'format'");
543             goto early_err;
544         }
545         qdict_put_str(bs_opts, "driver", buf);
546     }
547 
548     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
549     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
550         on_write_error = parse_block_error_action(buf, 0, &error);
551         if (error) {
552             error_propagate(errp, error);
553             goto early_err;
554         }
555     }
556 
557     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
558     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
559         on_read_error = parse_block_error_action(buf, 1, &error);
560         if (error) {
561             error_propagate(errp, error);
562             goto early_err;
563         }
564     }
565 
566     if (snapshot) {
567         bdrv_flags |= BDRV_O_SNAPSHOT;
568     }
569 
570     read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
571 
572     /* init */
573     if ((!file || !*file) && !qdict_size(bs_opts)) {
574         BlockBackendRootState *blk_rs;
575 
576         blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
577         blk_rs = blk_get_root_state(blk);
578         blk_rs->open_flags    = bdrv_flags;
579         blk_rs->read_only     = read_only;
580         blk_rs->detect_zeroes = detect_zeroes;
581 
582         qobject_unref(bs_opts);
583     } else {
584         if (file && !*file) {
585             file = NULL;
586         }
587 
588         /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
589          * with other callers) rather than what we want as the real defaults.
590          * Apply the defaults here instead. */
591         qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
592         qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
593         qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
594                               read_only ? "on" : "off");
595         qdict_set_default_str(bs_opts, BDRV_OPT_AUTO_READ_ONLY, "on");
596         assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
597 
598         if (runstate_check(RUN_STATE_INMIGRATE)) {
599             bdrv_flags |= BDRV_O_INACTIVE;
600         }
601 
602         blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
603         if (!blk) {
604             goto err_no_bs_opts;
605         }
606         bs = blk_bs(blk);
607 
608         bs->detect_zeroes = detect_zeroes;
609 
610         block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
611 
612         if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
613             blk_unref(blk);
614             blk = NULL;
615             goto err_no_bs_opts;
616         }
617     }
618 
619     /* disk I/O throttling */
620     if (throttle_enabled(&cfg)) {
621         if (!throttling_group) {
622             throttling_group = id;
623         }
624         blk_io_limits_enable(blk, throttling_group);
625         blk_set_io_limits(blk, &cfg);
626     }
627 
628     blk_set_enable_write_cache(blk, !writethrough);
629     blk_set_on_error(blk, on_read_error, on_write_error);
630 
631     if (!monitor_add_blk(blk, id, errp)) {
632         blk_unref(blk);
633         blk = NULL;
634         goto err_no_bs_opts;
635     }
636 
637 err_no_bs_opts:
638     qemu_opts_del(opts);
639     qobject_unref(interval_dict);
640     qobject_unref(interval_list);
641     return blk;
642 
643 early_err:
644     qemu_opts_del(opts);
645     qobject_unref(interval_dict);
646     qobject_unref(interval_list);
647 err_no_opts:
648     qobject_unref(bs_opts);
649     return NULL;
650 }
651 
652 /* Takes the ownership of bs_opts */
653 static BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
654 {
655     int bdrv_flags = 0;
656 
657     /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
658      * with other callers) rather than what we want as the real defaults.
659      * Apply the defaults here instead. */
660     qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
661     qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
662     qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
663 
664     if (runstate_check(RUN_STATE_INMIGRATE)) {
665         bdrv_flags |= BDRV_O_INACTIVE;
666     }
667 
668     return bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
669 }
670 
671 void blockdev_close_all_bdrv_states(void)
672 {
673     BlockDriverState *bs, *next_bs;
674 
675     QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
676         AioContext *ctx = bdrv_get_aio_context(bs);
677 
678         aio_context_acquire(ctx);
679         bdrv_unref(bs);
680         aio_context_release(ctx);
681     }
682 }
683 
684 /* Iterates over the list of monitor-owned BlockDriverStates */
685 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
686 {
687     return bs ? QTAILQ_NEXT(bs, monitor_list)
688               : QTAILQ_FIRST(&monitor_bdrv_states);
689 }
690 
691 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
692                             Error **errp)
693 {
694     const char *value;
695 
696     value = qemu_opt_get(opts, from);
697     if (value) {
698         if (qemu_opt_find(opts, to)) {
699             error_setg(errp, "'%s' and its alias '%s' can't be used at the "
700                        "same time", to, from);
701             return;
702         }
703     }
704 
705     /* rename all items in opts */
706     while ((value = qemu_opt_get(opts, from))) {
707         qemu_opt_set(opts, to, value, &error_abort);
708         qemu_opt_unset(opts, from);
709     }
710 }
711 
712 QemuOptsList qemu_legacy_drive_opts = {
713     .name = "drive",
714     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
715     .desc = {
716         {
717             .name = "bus",
718             .type = QEMU_OPT_NUMBER,
719             .help = "bus number",
720         },{
721             .name = "unit",
722             .type = QEMU_OPT_NUMBER,
723             .help = "unit number (i.e. lun for scsi)",
724         },{
725             .name = "index",
726             .type = QEMU_OPT_NUMBER,
727             .help = "index number",
728         },{
729             .name = "media",
730             .type = QEMU_OPT_STRING,
731             .help = "media type (disk, cdrom)",
732         },{
733             .name = "if",
734             .type = QEMU_OPT_STRING,
735             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
736         },{
737             .name = "file",
738             .type = QEMU_OPT_STRING,
739             .help = "file name",
740         },
741 
742         /* Options that are passed on, but have special semantics with -drive */
743         {
744             .name = BDRV_OPT_READ_ONLY,
745             .type = QEMU_OPT_BOOL,
746             .help = "open drive file as read-only",
747         },{
748             .name = "rerror",
749             .type = QEMU_OPT_STRING,
750             .help = "read error action",
751         },{
752             .name = "werror",
753             .type = QEMU_OPT_STRING,
754             .help = "write error action",
755         },{
756             .name = "copy-on-read",
757             .type = QEMU_OPT_BOOL,
758             .help = "copy read data from backing file into image file",
759         },
760 
761         { /* end of list */ }
762     },
763 };
764 
765 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
766                      Error **errp)
767 {
768     const char *value;
769     BlockBackend *blk;
770     DriveInfo *dinfo = NULL;
771     QDict *bs_opts;
772     QemuOpts *legacy_opts;
773     DriveMediaType media = MEDIA_DISK;
774     BlockInterfaceType type;
775     int max_devs, bus_id, unit_id, index;
776     const char *werror, *rerror;
777     bool read_only = false;
778     bool copy_on_read;
779     const char *filename;
780     Error *local_err = NULL;
781     int i;
782 
783     /* Change legacy command line options into QMP ones */
784     static const struct {
785         const char *from;
786         const char *to;
787     } opt_renames[] = {
788         { "iops",           "throttling.iops-total" },
789         { "iops_rd",        "throttling.iops-read" },
790         { "iops_wr",        "throttling.iops-write" },
791 
792         { "bps",            "throttling.bps-total" },
793         { "bps_rd",         "throttling.bps-read" },
794         { "bps_wr",         "throttling.bps-write" },
795 
796         { "iops_max",       "throttling.iops-total-max" },
797         { "iops_rd_max",    "throttling.iops-read-max" },
798         { "iops_wr_max",    "throttling.iops-write-max" },
799 
800         { "bps_max",        "throttling.bps-total-max" },
801         { "bps_rd_max",     "throttling.bps-read-max" },
802         { "bps_wr_max",     "throttling.bps-write-max" },
803 
804         { "iops_size",      "throttling.iops-size" },
805 
806         { "group",          "throttling.group" },
807 
808         { "readonly",       BDRV_OPT_READ_ONLY },
809     };
810 
811     for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
812         qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
813                         &local_err);
814         if (local_err) {
815             error_propagate(errp, local_err);
816             return NULL;
817         }
818     }
819 
820     value = qemu_opt_get(all_opts, "cache");
821     if (value) {
822         int flags = 0;
823         bool writethrough;
824 
825         if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
826             error_setg(errp, "invalid cache option");
827             return NULL;
828         }
829 
830         /* Specific options take precedence */
831         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
832             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
833                               !writethrough, &error_abort);
834         }
835         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
836             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
837                               !!(flags & BDRV_O_NOCACHE), &error_abort);
838         }
839         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
840             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
841                               !!(flags & BDRV_O_NO_FLUSH), &error_abort);
842         }
843         qemu_opt_unset(all_opts, "cache");
844     }
845 
846     /* Get a QDict for processing the options */
847     bs_opts = qdict_new();
848     qemu_opts_to_qdict(all_opts, bs_opts);
849 
850     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
851                                    &error_abort);
852     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
853     if (local_err) {
854         error_propagate(errp, local_err);
855         goto fail;
856     }
857 
858     /* Media type */
859     value = qemu_opt_get(legacy_opts, "media");
860     if (value) {
861         if (!strcmp(value, "disk")) {
862             media = MEDIA_DISK;
863         } else if (!strcmp(value, "cdrom")) {
864             media = MEDIA_CDROM;
865             read_only = true;
866         } else {
867             error_setg(errp, "'%s' invalid media", value);
868             goto fail;
869         }
870     }
871 
872     /* copy-on-read is disabled with a warning for read-only devices */
873     read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
874     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
875 
876     if (read_only && copy_on_read) {
877         warn_report("disabling copy-on-read on read-only drive");
878         copy_on_read = false;
879     }
880 
881     qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
882     qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
883 
884     /* Controller type */
885     value = qemu_opt_get(legacy_opts, "if");
886     if (value) {
887         for (type = 0;
888              type < IF_COUNT && strcmp(value, if_name[type]);
889              type++) {
890         }
891         if (type == IF_COUNT) {
892             error_setg(errp, "unsupported bus type '%s'", value);
893             goto fail;
894         }
895     } else {
896         type = block_default_type;
897     }
898 
899     /* Device address specified by bus/unit or index.
900      * If none was specified, try to find the first free one. */
901     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
902     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
903     index   = qemu_opt_get_number(legacy_opts, "index", -1);
904 
905     max_devs = if_max_devs[type];
906 
907     if (index != -1) {
908         if (bus_id != 0 || unit_id != -1) {
909             error_setg(errp, "index cannot be used with bus and unit");
910             goto fail;
911         }
912         bus_id = drive_index_to_bus_id(type, index);
913         unit_id = drive_index_to_unit_id(type, index);
914     }
915 
916     if (unit_id == -1) {
917        unit_id = 0;
918        while (drive_get(type, bus_id, unit_id) != NULL) {
919            unit_id++;
920            if (max_devs && unit_id >= max_devs) {
921                unit_id -= max_devs;
922                bus_id++;
923            }
924        }
925     }
926 
927     if (max_devs && unit_id >= max_devs) {
928         error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
929         goto fail;
930     }
931 
932     if (drive_get(type, bus_id, unit_id) != NULL) {
933         error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
934                    bus_id, unit_id, index);
935         goto fail;
936     }
937 
938     /* no id supplied -> create one */
939     if (qemu_opts_id(all_opts) == NULL) {
940         char *new_id;
941         const char *mediastr = "";
942         if (type == IF_IDE || type == IF_SCSI) {
943             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
944         }
945         if (max_devs) {
946             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
947                                      mediastr, unit_id);
948         } else {
949             new_id = g_strdup_printf("%s%s%i", if_name[type],
950                                      mediastr, unit_id);
951         }
952         qdict_put_str(bs_opts, "id", new_id);
953         g_free(new_id);
954     }
955 
956     /* Add virtio block device */
957     if (type == IF_VIRTIO) {
958         QemuOpts *devopts;
959         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
960                                    &error_abort);
961         if (arch_type == QEMU_ARCH_S390X) {
962             qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
963         } else {
964             qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
965         }
966         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
967                      &error_abort);
968     }
969 
970     filename = qemu_opt_get(legacy_opts, "file");
971 
972     /* Check werror/rerror compatibility with if=... */
973     werror = qemu_opt_get(legacy_opts, "werror");
974     if (werror != NULL) {
975         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
976             type != IF_NONE) {
977             error_setg(errp, "werror is not supported by this bus type");
978             goto fail;
979         }
980         qdict_put_str(bs_opts, "werror", werror);
981     }
982 
983     rerror = qemu_opt_get(legacy_opts, "rerror");
984     if (rerror != NULL) {
985         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
986             type != IF_NONE) {
987             error_setg(errp, "rerror is not supported by this bus type");
988             goto fail;
989         }
990         qdict_put_str(bs_opts, "rerror", rerror);
991     }
992 
993     /* Actual block device init: Functionality shared with blockdev-add */
994     blk = blockdev_init(filename, bs_opts, &local_err);
995     bs_opts = NULL;
996     if (!blk) {
997         error_propagate(errp, local_err);
998         goto fail;
999     } else {
1000         assert(!local_err);
1001     }
1002 
1003     /* Create legacy DriveInfo */
1004     dinfo = g_malloc0(sizeof(*dinfo));
1005     dinfo->opts = all_opts;
1006 
1007     dinfo->type = type;
1008     dinfo->bus = bus_id;
1009     dinfo->unit = unit_id;
1010 
1011     blk_set_legacy_dinfo(blk, dinfo);
1012 
1013     switch(type) {
1014     case IF_IDE:
1015     case IF_SCSI:
1016     case IF_XEN:
1017     case IF_NONE:
1018         dinfo->media_cd = media == MEDIA_CDROM;
1019         break;
1020     default:
1021         break;
1022     }
1023 
1024 fail:
1025     qemu_opts_del(legacy_opts);
1026     qobject_unref(bs_opts);
1027     return dinfo;
1028 }
1029 
1030 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1031 {
1032     BlockDriverState *bs;
1033 
1034     bs = bdrv_lookup_bs(name, name, errp);
1035     if (bs == NULL) {
1036         return NULL;
1037     }
1038 
1039     if (!bdrv_is_root_node(bs)) {
1040         error_setg(errp, "Need a root block node");
1041         return NULL;
1042     }
1043 
1044     if (!bdrv_is_inserted(bs)) {
1045         error_setg(errp, "Device has no medium");
1046         return NULL;
1047     }
1048 
1049     return bs;
1050 }
1051 
1052 static BlockBackend *qmp_get_blk(const char *blk_name, const char *qdev_id,
1053                                  Error **errp)
1054 {
1055     BlockBackend *blk;
1056 
1057     if (!blk_name == !qdev_id) {
1058         error_setg(errp, "Need exactly one of 'device' and 'id'");
1059         return NULL;
1060     }
1061 
1062     if (qdev_id) {
1063         blk = blk_by_qdev_id(qdev_id, errp);
1064     } else {
1065         blk = blk_by_name(blk_name);
1066         if (blk == NULL) {
1067             error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1068                       "Device '%s' not found", blk_name);
1069         }
1070     }
1071 
1072     return blk;
1073 }
1074 
1075 void hmp_commit(Monitor *mon, const QDict *qdict)
1076 {
1077     const char *device = qdict_get_str(qdict, "device");
1078     BlockBackend *blk;
1079     int ret;
1080 
1081     if (!strcmp(device, "all")) {
1082         ret = blk_commit_all();
1083     } else {
1084         BlockDriverState *bs;
1085         AioContext *aio_context;
1086 
1087         blk = blk_by_name(device);
1088         if (!blk) {
1089             monitor_printf(mon, "Device '%s' not found\n", device);
1090             return;
1091         }
1092         if (!blk_is_available(blk)) {
1093             monitor_printf(mon, "Device '%s' has no medium\n", device);
1094             return;
1095         }
1096 
1097         bs = blk_bs(blk);
1098         aio_context = bdrv_get_aio_context(bs);
1099         aio_context_acquire(aio_context);
1100 
1101         ret = bdrv_commit(bs);
1102 
1103         aio_context_release(aio_context);
1104     }
1105     if (ret < 0) {
1106         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1107                        strerror(-ret));
1108     }
1109 }
1110 
1111 static void blockdev_do_action(TransactionAction *action, Error **errp)
1112 {
1113     TransactionActionList list;
1114 
1115     list.value = action;
1116     list.next = NULL;
1117     qmp_transaction(&list, false, NULL, errp);
1118 }
1119 
1120 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1121                                 bool has_node_name, const char *node_name,
1122                                 const char *snapshot_file,
1123                                 bool has_snapshot_node_name,
1124                                 const char *snapshot_node_name,
1125                                 bool has_format, const char *format,
1126                                 bool has_mode, NewImageMode mode, Error **errp)
1127 {
1128     BlockdevSnapshotSync snapshot = {
1129         .has_device = has_device,
1130         .device = (char *) device,
1131         .has_node_name = has_node_name,
1132         .node_name = (char *) node_name,
1133         .snapshot_file = (char *) snapshot_file,
1134         .has_snapshot_node_name = has_snapshot_node_name,
1135         .snapshot_node_name = (char *) snapshot_node_name,
1136         .has_format = has_format,
1137         .format = (char *) format,
1138         .has_mode = has_mode,
1139         .mode = mode,
1140     };
1141     TransactionAction action = {
1142         .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1143         .u.blockdev_snapshot_sync.data = &snapshot,
1144     };
1145     blockdev_do_action(&action, errp);
1146 }
1147 
1148 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1149                            Error **errp)
1150 {
1151     BlockdevSnapshot snapshot_data = {
1152         .node = (char *) node,
1153         .overlay = (char *) overlay
1154     };
1155     TransactionAction action = {
1156         .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1157         .u.blockdev_snapshot.data = &snapshot_data,
1158     };
1159     blockdev_do_action(&action, errp);
1160 }
1161 
1162 void qmp_blockdev_snapshot_internal_sync(const char *device,
1163                                          const char *name,
1164                                          Error **errp)
1165 {
1166     BlockdevSnapshotInternal snapshot = {
1167         .device = (char *) device,
1168         .name = (char *) name
1169     };
1170     TransactionAction action = {
1171         .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1172         .u.blockdev_snapshot_internal_sync.data = &snapshot,
1173     };
1174     blockdev_do_action(&action, errp);
1175 }
1176 
1177 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1178                                                          bool has_id,
1179                                                          const char *id,
1180                                                          bool has_name,
1181                                                          const char *name,
1182                                                          Error **errp)
1183 {
1184     BlockDriverState *bs;
1185     AioContext *aio_context;
1186     QEMUSnapshotInfo sn;
1187     Error *local_err = NULL;
1188     SnapshotInfo *info = NULL;
1189     int ret;
1190 
1191     bs = qmp_get_root_bs(device, errp);
1192     if (!bs) {
1193         return NULL;
1194     }
1195     aio_context = bdrv_get_aio_context(bs);
1196     aio_context_acquire(aio_context);
1197 
1198     if (!has_id) {
1199         id = NULL;
1200     }
1201 
1202     if (!has_name) {
1203         name = NULL;
1204     }
1205 
1206     if (!id && !name) {
1207         error_setg(errp, "Name or id must be provided");
1208         goto out_aio_context;
1209     }
1210 
1211     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1212         goto out_aio_context;
1213     }
1214 
1215     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1216     if (local_err) {
1217         error_propagate(errp, local_err);
1218         goto out_aio_context;
1219     }
1220     if (!ret) {
1221         error_setg(errp,
1222                    "Snapshot with id '%s' and name '%s' does not exist on "
1223                    "device '%s'",
1224                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1225         goto out_aio_context;
1226     }
1227 
1228     bdrv_snapshot_delete(bs, id, name, &local_err);
1229     if (local_err) {
1230         error_propagate(errp, local_err);
1231         goto out_aio_context;
1232     }
1233 
1234     aio_context_release(aio_context);
1235 
1236     info = g_new0(SnapshotInfo, 1);
1237     info->id = g_strdup(sn.id_str);
1238     info->name = g_strdup(sn.name);
1239     info->date_nsec = sn.date_nsec;
1240     info->date_sec = sn.date_sec;
1241     info->vm_state_size = sn.vm_state_size;
1242     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1243     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1244 
1245     return info;
1246 
1247 out_aio_context:
1248     aio_context_release(aio_context);
1249     return NULL;
1250 }
1251 
1252 /**
1253  * block_dirty_bitmap_lookup:
1254  * Return a dirty bitmap (if present), after validating
1255  * the node reference and bitmap names.
1256  *
1257  * @node: The name of the BDS node to search for bitmaps
1258  * @name: The name of the bitmap to search for
1259  * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1260  * @errp: Output pointer for error information. Can be NULL.
1261  *
1262  * @return: A bitmap object on success, or NULL on failure.
1263  */
1264 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1265                                                   const char *name,
1266                                                   BlockDriverState **pbs,
1267                                                   Error **errp)
1268 {
1269     BlockDriverState *bs;
1270     BdrvDirtyBitmap *bitmap;
1271 
1272     if (!node) {
1273         error_setg(errp, "Node cannot be NULL");
1274         return NULL;
1275     }
1276     if (!name) {
1277         error_setg(errp, "Bitmap name cannot be NULL");
1278         return NULL;
1279     }
1280     bs = bdrv_lookup_bs(node, node, NULL);
1281     if (!bs) {
1282         error_setg(errp, "Node '%s' not found", node);
1283         return NULL;
1284     }
1285 
1286     bitmap = bdrv_find_dirty_bitmap(bs, name);
1287     if (!bitmap) {
1288         error_setg(errp, "Dirty bitmap '%s' not found", name);
1289         return NULL;
1290     }
1291 
1292     if (pbs) {
1293         *pbs = bs;
1294     }
1295 
1296     return bitmap;
1297 }
1298 
1299 /* New and old BlockDriverState structs for atomic group operations */
1300 
1301 typedef struct BlkActionState BlkActionState;
1302 
1303 /**
1304  * BlkActionOps:
1305  * Table of operations that define an Action.
1306  *
1307  * @instance_size: Size of state struct, in bytes.
1308  * @prepare: Prepare the work, must NOT be NULL.
1309  * @commit: Commit the changes, can be NULL.
1310  * @abort: Abort the changes on fail, can be NULL.
1311  * @clean: Clean up resources after all transaction actions have called
1312  *         commit() or abort(). Can be NULL.
1313  *
1314  * Only prepare() may fail. In a single transaction, only one of commit() or
1315  * abort() will be called. clean() will always be called if it is present.
1316  */
1317 typedef struct BlkActionOps {
1318     size_t instance_size;
1319     void (*prepare)(BlkActionState *common, Error **errp);
1320     void (*commit)(BlkActionState *common);
1321     void (*abort)(BlkActionState *common);
1322     void (*clean)(BlkActionState *common);
1323 } BlkActionOps;
1324 
1325 /**
1326  * BlkActionState:
1327  * Describes one Action's state within a Transaction.
1328  *
1329  * @action: QAPI-defined enum identifying which Action to perform.
1330  * @ops: Table of ActionOps this Action can perform.
1331  * @block_job_txn: Transaction which this action belongs to.
1332  * @entry: List membership for all Actions in this Transaction.
1333  *
1334  * This structure must be arranged as first member in a subclassed type,
1335  * assuming that the compiler will also arrange it to the same offsets as the
1336  * base class.
1337  */
1338 struct BlkActionState {
1339     TransactionAction *action;
1340     const BlkActionOps *ops;
1341     JobTxn *block_job_txn;
1342     TransactionProperties *txn_props;
1343     QTAILQ_ENTRY(BlkActionState) entry;
1344 };
1345 
1346 /* internal snapshot private data */
1347 typedef struct InternalSnapshotState {
1348     BlkActionState common;
1349     BlockDriverState *bs;
1350     QEMUSnapshotInfo sn;
1351     bool created;
1352 } InternalSnapshotState;
1353 
1354 
1355 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1356 {
1357     if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1358         error_setg(errp,
1359                    "Action '%s' does not support Transaction property "
1360                    "completion-mode = %s",
1361                    TransactionActionKind_str(s->action->type),
1362                    ActionCompletionMode_str(s->txn_props->completion_mode));
1363         return -1;
1364     }
1365     return 0;
1366 }
1367 
1368 static void internal_snapshot_prepare(BlkActionState *common,
1369                                       Error **errp)
1370 {
1371     Error *local_err = NULL;
1372     const char *device;
1373     const char *name;
1374     BlockDriverState *bs;
1375     QEMUSnapshotInfo old_sn, *sn;
1376     bool ret;
1377     qemu_timeval tv;
1378     BlockdevSnapshotInternal *internal;
1379     InternalSnapshotState *state;
1380     AioContext *aio_context;
1381     int ret1;
1382 
1383     g_assert(common->action->type ==
1384              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1385     internal = common->action->u.blockdev_snapshot_internal_sync.data;
1386     state = DO_UPCAST(InternalSnapshotState, common, common);
1387 
1388     /* 1. parse input */
1389     device = internal->device;
1390     name = internal->name;
1391 
1392     /* 2. check for validation */
1393     if (action_check_completion_mode(common, errp) < 0) {
1394         return;
1395     }
1396 
1397     bs = qmp_get_root_bs(device, errp);
1398     if (!bs) {
1399         return;
1400     }
1401 
1402     aio_context = bdrv_get_aio_context(bs);
1403     aio_context_acquire(aio_context);
1404 
1405     state->bs = bs;
1406 
1407     /* Paired with .clean() */
1408     bdrv_drained_begin(bs);
1409 
1410     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1411         goto out;
1412     }
1413 
1414     if (bdrv_is_read_only(bs)) {
1415         error_setg(errp, "Device '%s' is read only", device);
1416         goto out;
1417     }
1418 
1419     if (!bdrv_can_snapshot(bs)) {
1420         error_setg(errp, "Block format '%s' used by device '%s' "
1421                    "does not support internal snapshots",
1422                    bs->drv->format_name, device);
1423         goto out;
1424     }
1425 
1426     if (!strlen(name)) {
1427         error_setg(errp, "Name is empty");
1428         goto out;
1429     }
1430 
1431     /* check whether a snapshot with name exist */
1432     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1433                                             &local_err);
1434     if (local_err) {
1435         error_propagate(errp, local_err);
1436         goto out;
1437     } else if (ret) {
1438         error_setg(errp,
1439                    "Snapshot with name '%s' already exists on device '%s'",
1440                    name, device);
1441         goto out;
1442     }
1443 
1444     /* 3. take the snapshot */
1445     sn = &state->sn;
1446     pstrcpy(sn->name, sizeof(sn->name), name);
1447     qemu_gettimeofday(&tv);
1448     sn->date_sec = tv.tv_sec;
1449     sn->date_nsec = tv.tv_usec * 1000;
1450     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1451 
1452     ret1 = bdrv_snapshot_create(bs, sn);
1453     if (ret1 < 0) {
1454         error_setg_errno(errp, -ret1,
1455                          "Failed to create snapshot '%s' on device '%s'",
1456                          name, device);
1457         goto out;
1458     }
1459 
1460     /* 4. succeed, mark a snapshot is created */
1461     state->created = true;
1462 
1463 out:
1464     aio_context_release(aio_context);
1465 }
1466 
1467 static void internal_snapshot_abort(BlkActionState *common)
1468 {
1469     InternalSnapshotState *state =
1470                              DO_UPCAST(InternalSnapshotState, common, common);
1471     BlockDriverState *bs = state->bs;
1472     QEMUSnapshotInfo *sn = &state->sn;
1473     AioContext *aio_context;
1474     Error *local_error = NULL;
1475 
1476     if (!state->created) {
1477         return;
1478     }
1479 
1480     aio_context = bdrv_get_aio_context(state->bs);
1481     aio_context_acquire(aio_context);
1482 
1483     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1484         error_reportf_err(local_error,
1485                           "Failed to delete snapshot with id '%s' and "
1486                           "name '%s' on device '%s' in abort: ",
1487                           sn->id_str, sn->name,
1488                           bdrv_get_device_name(bs));
1489     }
1490 
1491     aio_context_release(aio_context);
1492 }
1493 
1494 static void internal_snapshot_clean(BlkActionState *common)
1495 {
1496     InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1497                                              common, common);
1498     AioContext *aio_context;
1499 
1500     if (!state->bs) {
1501         return;
1502     }
1503 
1504     aio_context = bdrv_get_aio_context(state->bs);
1505     aio_context_acquire(aio_context);
1506 
1507     bdrv_drained_end(state->bs);
1508 
1509     aio_context_release(aio_context);
1510 }
1511 
1512 /* external snapshot private data */
1513 typedef struct ExternalSnapshotState {
1514     BlkActionState common;
1515     BlockDriverState *old_bs;
1516     BlockDriverState *new_bs;
1517     bool overlay_appended;
1518 } ExternalSnapshotState;
1519 
1520 static void external_snapshot_prepare(BlkActionState *common,
1521                                       Error **errp)
1522 {
1523     int flags = 0;
1524     QDict *options = NULL;
1525     Error *local_err = NULL;
1526     /* Device and node name of the image to generate the snapshot from */
1527     const char *device;
1528     const char *node_name;
1529     /* Reference to the new image (for 'blockdev-snapshot') */
1530     const char *snapshot_ref;
1531     /* File name of the new image (for 'blockdev-snapshot-sync') */
1532     const char *new_image_file;
1533     ExternalSnapshotState *state =
1534                              DO_UPCAST(ExternalSnapshotState, common, common);
1535     TransactionAction *action = common->action;
1536     AioContext *aio_context;
1537     int ret;
1538 
1539     /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1540      * purpose but a different set of parameters */
1541     switch (action->type) {
1542     case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1543         {
1544             BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1545             device = s->node;
1546             node_name = s->node;
1547             new_image_file = NULL;
1548             snapshot_ref = s->overlay;
1549         }
1550         break;
1551     case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1552         {
1553             BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1554             device = s->has_device ? s->device : NULL;
1555             node_name = s->has_node_name ? s->node_name : NULL;
1556             new_image_file = s->snapshot_file;
1557             snapshot_ref = NULL;
1558         }
1559         break;
1560     default:
1561         g_assert_not_reached();
1562     }
1563 
1564     /* start processing */
1565     if (action_check_completion_mode(common, errp) < 0) {
1566         return;
1567     }
1568 
1569     state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1570     if (!state->old_bs) {
1571         return;
1572     }
1573 
1574     aio_context = bdrv_get_aio_context(state->old_bs);
1575     aio_context_acquire(aio_context);
1576 
1577     /* Paired with .clean() */
1578     bdrv_drained_begin(state->old_bs);
1579 
1580     if (!bdrv_is_inserted(state->old_bs)) {
1581         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1582         goto out;
1583     }
1584 
1585     if (bdrv_op_is_blocked(state->old_bs,
1586                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1587         goto out;
1588     }
1589 
1590     if (!bdrv_is_read_only(state->old_bs)) {
1591         if (bdrv_flush(state->old_bs)) {
1592             error_setg(errp, QERR_IO_ERROR);
1593             goto out;
1594         }
1595     }
1596 
1597     if (!bdrv_is_first_non_filter(state->old_bs)) {
1598         error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1599         goto out;
1600     }
1601 
1602     if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1603         BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1604         const char *format = s->has_format ? s->format : "qcow2";
1605         enum NewImageMode mode;
1606         const char *snapshot_node_name =
1607             s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1608 
1609         if (node_name && !snapshot_node_name) {
1610             error_setg(errp, "New overlay node name missing");
1611             goto out;
1612         }
1613 
1614         if (snapshot_node_name &&
1615             bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1616             error_setg(errp, "New overlay node name already in use");
1617             goto out;
1618         }
1619 
1620         flags = state->old_bs->open_flags;
1621         flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1622         flags |= BDRV_O_NO_BACKING;
1623 
1624         /* create new image w/backing file */
1625         mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1626         if (mode != NEW_IMAGE_MODE_EXISTING) {
1627             int64_t size = bdrv_getlength(state->old_bs);
1628             if (size < 0) {
1629                 error_setg_errno(errp, -size, "bdrv_getlength failed");
1630                 goto out;
1631             }
1632             bdrv_refresh_filename(state->old_bs);
1633             bdrv_img_create(new_image_file, format,
1634                             state->old_bs->filename,
1635                             state->old_bs->drv->format_name,
1636                             NULL, size, flags, false, &local_err);
1637             if (local_err) {
1638                 error_propagate(errp, local_err);
1639                 goto out;
1640             }
1641         }
1642 
1643         options = qdict_new();
1644         if (snapshot_node_name) {
1645             qdict_put_str(options, "node-name", snapshot_node_name);
1646         }
1647         qdict_put_str(options, "driver", format);
1648     }
1649 
1650     state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1651                               errp);
1652     /* We will manually add the backing_hd field to the bs later */
1653     if (!state->new_bs) {
1654         goto out;
1655     }
1656 
1657     if (bdrv_has_blk(state->new_bs)) {
1658         error_setg(errp, "The overlay is already in use");
1659         goto out;
1660     }
1661 
1662     if (bdrv_op_is_blocked(state->new_bs, BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
1663                            errp)) {
1664         goto out;
1665     }
1666 
1667     if (state->new_bs->backing != NULL) {
1668         error_setg(errp, "The overlay already has a backing image");
1669         goto out;
1670     }
1671 
1672     if (!state->new_bs->drv->supports_backing) {
1673         error_setg(errp, "The overlay does not support backing images");
1674         goto out;
1675     }
1676 
1677     ret = bdrv_try_set_aio_context(state->new_bs, aio_context, errp);
1678     if (ret < 0) {
1679         goto out;
1680     }
1681 
1682     /* This removes our old bs and adds the new bs. This is an operation that
1683      * can fail, so we need to do it in .prepare; undoing it for abort is
1684      * always possible. */
1685     bdrv_ref(state->new_bs);
1686     bdrv_append(state->new_bs, state->old_bs, &local_err);
1687     if (local_err) {
1688         error_propagate(errp, local_err);
1689         goto out;
1690     }
1691     state->overlay_appended = true;
1692 
1693 out:
1694     aio_context_release(aio_context);
1695 }
1696 
1697 static void external_snapshot_commit(BlkActionState *common)
1698 {
1699     ExternalSnapshotState *state =
1700                              DO_UPCAST(ExternalSnapshotState, common, common);
1701     AioContext *aio_context;
1702 
1703     aio_context = bdrv_get_aio_context(state->old_bs);
1704     aio_context_acquire(aio_context);
1705 
1706     /* We don't need (or want) to use the transactional
1707      * bdrv_reopen_multiple() across all the entries at once, because we
1708      * don't want to abort all of them if one of them fails the reopen */
1709     if (!atomic_read(&state->old_bs->copy_on_read)) {
1710         bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1711     }
1712 
1713     aio_context_release(aio_context);
1714 }
1715 
1716 static void external_snapshot_abort(BlkActionState *common)
1717 {
1718     ExternalSnapshotState *state =
1719                              DO_UPCAST(ExternalSnapshotState, common, common);
1720     if (state->new_bs) {
1721         if (state->overlay_appended) {
1722             AioContext *aio_context;
1723 
1724             aio_context = bdrv_get_aio_context(state->old_bs);
1725             aio_context_acquire(aio_context);
1726 
1727             bdrv_ref(state->old_bs);   /* we can't let bdrv_set_backind_hd()
1728                                           close state->old_bs; we need it */
1729             bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1730             bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1731             bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1732 
1733             aio_context_release(aio_context);
1734         }
1735     }
1736 }
1737 
1738 static void external_snapshot_clean(BlkActionState *common)
1739 {
1740     ExternalSnapshotState *state =
1741                              DO_UPCAST(ExternalSnapshotState, common, common);
1742     AioContext *aio_context;
1743 
1744     if (!state->old_bs) {
1745         return;
1746     }
1747 
1748     aio_context = bdrv_get_aio_context(state->old_bs);
1749     aio_context_acquire(aio_context);
1750 
1751     bdrv_drained_end(state->old_bs);
1752     bdrv_unref(state->new_bs);
1753 
1754     aio_context_release(aio_context);
1755 }
1756 
1757 typedef struct DriveBackupState {
1758     BlkActionState common;
1759     BlockDriverState *bs;
1760     BlockJob *job;
1761 } DriveBackupState;
1762 
1763 static BlockJob *do_drive_backup(DriveBackup *backup, JobTxn *txn,
1764                             Error **errp);
1765 
1766 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1767 {
1768     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1769     BlockDriverState *bs;
1770     DriveBackup *backup;
1771     AioContext *aio_context;
1772     Error *local_err = NULL;
1773 
1774     assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1775     backup = common->action->u.drive_backup.data;
1776 
1777     bs = qmp_get_root_bs(backup->device, errp);
1778     if (!bs) {
1779         return;
1780     }
1781 
1782     aio_context = bdrv_get_aio_context(bs);
1783     aio_context_acquire(aio_context);
1784 
1785     /* Paired with .clean() */
1786     bdrv_drained_begin(bs);
1787 
1788     state->bs = bs;
1789 
1790     state->job = do_drive_backup(backup, common->block_job_txn, &local_err);
1791     if (local_err) {
1792         error_propagate(errp, local_err);
1793         goto out;
1794     }
1795 
1796 out:
1797     aio_context_release(aio_context);
1798 }
1799 
1800 static void drive_backup_commit(BlkActionState *common)
1801 {
1802     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1803     AioContext *aio_context;
1804 
1805     aio_context = bdrv_get_aio_context(state->bs);
1806     aio_context_acquire(aio_context);
1807 
1808     assert(state->job);
1809     job_start(&state->job->job);
1810 
1811     aio_context_release(aio_context);
1812 }
1813 
1814 static void drive_backup_abort(BlkActionState *common)
1815 {
1816     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1817 
1818     if (state->job) {
1819         AioContext *aio_context;
1820 
1821         aio_context = bdrv_get_aio_context(state->bs);
1822         aio_context_acquire(aio_context);
1823 
1824         job_cancel_sync(&state->job->job);
1825 
1826         aio_context_release(aio_context);
1827     }
1828 }
1829 
1830 static void drive_backup_clean(BlkActionState *common)
1831 {
1832     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1833     AioContext *aio_context;
1834 
1835     if (!state->bs) {
1836         return;
1837     }
1838 
1839     aio_context = bdrv_get_aio_context(state->bs);
1840     aio_context_acquire(aio_context);
1841 
1842     bdrv_drained_end(state->bs);
1843 
1844     aio_context_release(aio_context);
1845 }
1846 
1847 typedef struct BlockdevBackupState {
1848     BlkActionState common;
1849     BlockDriverState *bs;
1850     BlockJob *job;
1851 } BlockdevBackupState;
1852 
1853 static BlockJob *do_blockdev_backup(BlockdevBackup *backup, JobTxn *txn,
1854                                     Error **errp);
1855 
1856 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1857 {
1858     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1859     BlockdevBackup *backup;
1860     BlockDriverState *bs, *target;
1861     AioContext *aio_context;
1862     Error *local_err = NULL;
1863 
1864     assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1865     backup = common->action->u.blockdev_backup.data;
1866 
1867     bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1868     if (!bs) {
1869         return;
1870     }
1871 
1872     target = bdrv_lookup_bs(backup->target, backup->target, errp);
1873     if (!target) {
1874         return;
1875     }
1876 
1877     aio_context = bdrv_get_aio_context(bs);
1878     aio_context_acquire(aio_context);
1879     state->bs = bs;
1880 
1881     /* Paired with .clean() */
1882     bdrv_drained_begin(state->bs);
1883 
1884     state->job = do_blockdev_backup(backup, common->block_job_txn, &local_err);
1885     if (local_err) {
1886         error_propagate(errp, local_err);
1887         goto out;
1888     }
1889 
1890 out:
1891     aio_context_release(aio_context);
1892 }
1893 
1894 static void blockdev_backup_commit(BlkActionState *common)
1895 {
1896     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1897     AioContext *aio_context;
1898 
1899     aio_context = bdrv_get_aio_context(state->bs);
1900     aio_context_acquire(aio_context);
1901 
1902     assert(state->job);
1903     job_start(&state->job->job);
1904 
1905     aio_context_release(aio_context);
1906 }
1907 
1908 static void blockdev_backup_abort(BlkActionState *common)
1909 {
1910     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1911 
1912     if (state->job) {
1913         AioContext *aio_context;
1914 
1915         aio_context = bdrv_get_aio_context(state->bs);
1916         aio_context_acquire(aio_context);
1917 
1918         job_cancel_sync(&state->job->job);
1919 
1920         aio_context_release(aio_context);
1921     }
1922 }
1923 
1924 static void blockdev_backup_clean(BlkActionState *common)
1925 {
1926     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1927     AioContext *aio_context;
1928 
1929     if (!state->bs) {
1930         return;
1931     }
1932 
1933     aio_context = bdrv_get_aio_context(state->bs);
1934     aio_context_acquire(aio_context);
1935 
1936     bdrv_drained_end(state->bs);
1937 
1938     aio_context_release(aio_context);
1939 }
1940 
1941 typedef struct BlockDirtyBitmapState {
1942     BlkActionState common;
1943     BdrvDirtyBitmap *bitmap;
1944     BlockDriverState *bs;
1945     HBitmap *backup;
1946     bool prepared;
1947     bool was_enabled;
1948 } BlockDirtyBitmapState;
1949 
1950 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
1951                                            Error **errp)
1952 {
1953     Error *local_err = NULL;
1954     BlockDirtyBitmapAdd *action;
1955     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1956                                              common, common);
1957 
1958     if (action_check_completion_mode(common, errp) < 0) {
1959         return;
1960     }
1961 
1962     action = common->action->u.block_dirty_bitmap_add.data;
1963     /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1964     qmp_block_dirty_bitmap_add(action->node, action->name,
1965                                action->has_granularity, action->granularity,
1966                                action->has_persistent, action->persistent,
1967                                action->has_autoload, action->autoload,
1968                                action->has_disabled, action->disabled,
1969                                &local_err);
1970 
1971     if (!local_err) {
1972         state->prepared = true;
1973     } else {
1974         error_propagate(errp, local_err);
1975     }
1976 }
1977 
1978 static void block_dirty_bitmap_add_abort(BlkActionState *common)
1979 {
1980     BlockDirtyBitmapAdd *action;
1981     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1982                                              common, common);
1983 
1984     action = common->action->u.block_dirty_bitmap_add.data;
1985     /* Should not be able to fail: IF the bitmap was added via .prepare(),
1986      * then the node reference and bitmap name must have been valid.
1987      */
1988     if (state->prepared) {
1989         qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
1990     }
1991 }
1992 
1993 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
1994                                              Error **errp)
1995 {
1996     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1997                                              common, common);
1998     BlockDirtyBitmap *action;
1999 
2000     if (action_check_completion_mode(common, errp) < 0) {
2001         return;
2002     }
2003 
2004     action = common->action->u.block_dirty_bitmap_clear.data;
2005     state->bitmap = block_dirty_bitmap_lookup(action->node,
2006                                               action->name,
2007                                               &state->bs,
2008                                               errp);
2009     if (!state->bitmap) {
2010         return;
2011     }
2012 
2013     if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2014         return;
2015     }
2016 
2017     bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2018 }
2019 
2020 static void block_dirty_bitmap_restore(BlkActionState *common)
2021 {
2022     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2023                                              common, common);
2024 
2025     if (state->backup) {
2026         bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2027     }
2028 }
2029 
2030 static void block_dirty_bitmap_free_backup(BlkActionState *common)
2031 {
2032     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2033                                              common, common);
2034 
2035     hbitmap_free(state->backup);
2036 }
2037 
2038 static void block_dirty_bitmap_enable_prepare(BlkActionState *common,
2039                                               Error **errp)
2040 {
2041     BlockDirtyBitmap *action;
2042     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2043                                              common, common);
2044 
2045     if (action_check_completion_mode(common, errp) < 0) {
2046         return;
2047     }
2048 
2049     action = common->action->u.block_dirty_bitmap_enable.data;
2050     state->bitmap = block_dirty_bitmap_lookup(action->node,
2051                                               action->name,
2052                                               NULL,
2053                                               errp);
2054     if (!state->bitmap) {
2055         return;
2056     }
2057 
2058     if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2059         return;
2060     }
2061 
2062     state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2063     bdrv_enable_dirty_bitmap(state->bitmap);
2064 }
2065 
2066 static void block_dirty_bitmap_enable_abort(BlkActionState *common)
2067 {
2068     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2069                                              common, common);
2070 
2071     if (!state->was_enabled) {
2072         bdrv_disable_dirty_bitmap(state->bitmap);
2073     }
2074 }
2075 
2076 static void block_dirty_bitmap_disable_prepare(BlkActionState *common,
2077                                                Error **errp)
2078 {
2079     BlockDirtyBitmap *action;
2080     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2081                                              common, common);
2082 
2083     if (action_check_completion_mode(common, errp) < 0) {
2084         return;
2085     }
2086 
2087     action = common->action->u.block_dirty_bitmap_disable.data;
2088     state->bitmap = block_dirty_bitmap_lookup(action->node,
2089                                               action->name,
2090                                               NULL,
2091                                               errp);
2092     if (!state->bitmap) {
2093         return;
2094     }
2095 
2096     if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2097         return;
2098     }
2099 
2100     state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2101     bdrv_disable_dirty_bitmap(state->bitmap);
2102 }
2103 
2104 static void block_dirty_bitmap_disable_abort(BlkActionState *common)
2105 {
2106     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2107                                              common, common);
2108 
2109     if (state->was_enabled) {
2110         bdrv_enable_dirty_bitmap(state->bitmap);
2111     }
2112 }
2113 
2114 static BdrvDirtyBitmap *do_block_dirty_bitmap_merge(
2115         const char *node, const char *target,
2116         BlockDirtyBitmapMergeSourceList *bitmaps,
2117         HBitmap **backup, Error **errp);
2118 
2119 static void block_dirty_bitmap_merge_prepare(BlkActionState *common,
2120                                              Error **errp)
2121 {
2122     BlockDirtyBitmapMerge *action;
2123     BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2124                                              common, common);
2125 
2126     if (action_check_completion_mode(common, errp) < 0) {
2127         return;
2128     }
2129 
2130     action = common->action->u.block_dirty_bitmap_merge.data;
2131 
2132     state->bitmap = do_block_dirty_bitmap_merge(action->node, action->target,
2133                                                 action->bitmaps, &state->backup,
2134                                                 errp);
2135 }
2136 
2137 static void abort_prepare(BlkActionState *common, Error **errp)
2138 {
2139     error_setg(errp, "Transaction aborted using Abort action");
2140 }
2141 
2142 static void abort_commit(BlkActionState *common)
2143 {
2144     g_assert_not_reached(); /* this action never succeeds */
2145 }
2146 
2147 static const BlkActionOps actions[] = {
2148     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2149         .instance_size = sizeof(ExternalSnapshotState),
2150         .prepare  = external_snapshot_prepare,
2151         .commit   = external_snapshot_commit,
2152         .abort = external_snapshot_abort,
2153         .clean = external_snapshot_clean,
2154     },
2155     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2156         .instance_size = sizeof(ExternalSnapshotState),
2157         .prepare  = external_snapshot_prepare,
2158         .commit   = external_snapshot_commit,
2159         .abort = external_snapshot_abort,
2160         .clean = external_snapshot_clean,
2161     },
2162     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2163         .instance_size = sizeof(DriveBackupState),
2164         .prepare = drive_backup_prepare,
2165         .commit = drive_backup_commit,
2166         .abort = drive_backup_abort,
2167         .clean = drive_backup_clean,
2168     },
2169     [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2170         .instance_size = sizeof(BlockdevBackupState),
2171         .prepare = blockdev_backup_prepare,
2172         .commit = blockdev_backup_commit,
2173         .abort = blockdev_backup_abort,
2174         .clean = blockdev_backup_clean,
2175     },
2176     [TRANSACTION_ACTION_KIND_ABORT] = {
2177         .instance_size = sizeof(BlkActionState),
2178         .prepare = abort_prepare,
2179         .commit = abort_commit,
2180     },
2181     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2182         .instance_size = sizeof(InternalSnapshotState),
2183         .prepare  = internal_snapshot_prepare,
2184         .abort = internal_snapshot_abort,
2185         .clean = internal_snapshot_clean,
2186     },
2187     [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2188         .instance_size = sizeof(BlockDirtyBitmapState),
2189         .prepare = block_dirty_bitmap_add_prepare,
2190         .abort = block_dirty_bitmap_add_abort,
2191     },
2192     [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2193         .instance_size = sizeof(BlockDirtyBitmapState),
2194         .prepare = block_dirty_bitmap_clear_prepare,
2195         .commit = block_dirty_bitmap_free_backup,
2196         .abort = block_dirty_bitmap_restore,
2197     },
2198     [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2199         .instance_size = sizeof(BlockDirtyBitmapState),
2200         .prepare = block_dirty_bitmap_enable_prepare,
2201         .abort = block_dirty_bitmap_enable_abort,
2202     },
2203     [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2204         .instance_size = sizeof(BlockDirtyBitmapState),
2205         .prepare = block_dirty_bitmap_disable_prepare,
2206         .abort = block_dirty_bitmap_disable_abort,
2207     },
2208     [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2209         .instance_size = sizeof(BlockDirtyBitmapState),
2210         .prepare = block_dirty_bitmap_merge_prepare,
2211         .commit = block_dirty_bitmap_free_backup,
2212         .abort = block_dirty_bitmap_restore,
2213     },
2214     /* Where are transactions for MIRROR, COMMIT and STREAM?
2215      * Although these blockjobs use transaction callbacks like the backup job,
2216      * these jobs do not necessarily adhere to transaction semantics.
2217      * These jobs may not fully undo all of their actions on abort, nor do they
2218      * necessarily work in transactions with more than one job in them.
2219      */
2220 };
2221 
2222 /**
2223  * Allocate a TransactionProperties structure if necessary, and fill
2224  * that structure with desired defaults if they are unset.
2225  */
2226 static TransactionProperties *get_transaction_properties(
2227     TransactionProperties *props)
2228 {
2229     if (!props) {
2230         props = g_new0(TransactionProperties, 1);
2231     }
2232 
2233     if (!props->has_completion_mode) {
2234         props->has_completion_mode = true;
2235         props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2236     }
2237 
2238     return props;
2239 }
2240 
2241 /*
2242  * 'Atomic' group operations.  The operations are performed as a set, and if
2243  * any fail then we roll back all operations in the group.
2244  */
2245 void qmp_transaction(TransactionActionList *dev_list,
2246                      bool has_props,
2247                      struct TransactionProperties *props,
2248                      Error **errp)
2249 {
2250     TransactionActionList *dev_entry = dev_list;
2251     JobTxn *block_job_txn = NULL;
2252     BlkActionState *state, *next;
2253     Error *local_err = NULL;
2254 
2255     QTAILQ_HEAD(, BlkActionState) snap_bdrv_states;
2256     QTAILQ_INIT(&snap_bdrv_states);
2257 
2258     /* Does this transaction get canceled as a group on failure?
2259      * If not, we don't really need to make a JobTxn.
2260      */
2261     props = get_transaction_properties(props);
2262     if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2263         block_job_txn = job_txn_new();
2264     }
2265 
2266     /* drain all i/o before any operations */
2267     bdrv_drain_all();
2268 
2269     /* We don't do anything in this loop that commits us to the operations */
2270     while (NULL != dev_entry) {
2271         TransactionAction *dev_info = NULL;
2272         const BlkActionOps *ops;
2273 
2274         dev_info = dev_entry->value;
2275         dev_entry = dev_entry->next;
2276 
2277         assert(dev_info->type < ARRAY_SIZE(actions));
2278 
2279         ops = &actions[dev_info->type];
2280         assert(ops->instance_size > 0);
2281 
2282         state = g_malloc0(ops->instance_size);
2283         state->ops = ops;
2284         state->action = dev_info;
2285         state->block_job_txn = block_job_txn;
2286         state->txn_props = props;
2287         QTAILQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2288 
2289         state->ops->prepare(state, &local_err);
2290         if (local_err) {
2291             error_propagate(errp, local_err);
2292             goto delete_and_fail;
2293         }
2294     }
2295 
2296     QTAILQ_FOREACH(state, &snap_bdrv_states, entry) {
2297         if (state->ops->commit) {
2298             state->ops->commit(state);
2299         }
2300     }
2301 
2302     /* success */
2303     goto exit;
2304 
2305 delete_and_fail:
2306     /* failure, and it is all-or-none; roll back all operations */
2307     QTAILQ_FOREACH_REVERSE(state, &snap_bdrv_states, entry) {
2308         if (state->ops->abort) {
2309             state->ops->abort(state);
2310         }
2311     }
2312 exit:
2313     QTAILQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2314         if (state->ops->clean) {
2315             state->ops->clean(state);
2316         }
2317         g_free(state);
2318     }
2319     if (!has_props) {
2320         qapi_free_TransactionProperties(props);
2321     }
2322     job_txn_unref(block_job_txn);
2323 }
2324 
2325 void qmp_eject(bool has_device, const char *device,
2326                bool has_id, const char *id,
2327                bool has_force, bool force, Error **errp)
2328 {
2329     Error *local_err = NULL;
2330     int rc;
2331 
2332     if (!has_force) {
2333         force = false;
2334     }
2335 
2336     rc = do_open_tray(has_device ? device : NULL,
2337                       has_id ? id : NULL,
2338                       force, &local_err);
2339     if (rc && rc != -ENOSYS) {
2340         error_propagate(errp, local_err);
2341         return;
2342     }
2343     error_free(local_err);
2344 
2345     blockdev_remove_medium(has_device, device, has_id, id, errp);
2346 }
2347 
2348 void qmp_block_passwd(bool has_device, const char *device,
2349                       bool has_node_name, const char *node_name,
2350                       const char *password, Error **errp)
2351 {
2352     error_setg(errp,
2353                "Setting block passwords directly is no longer supported");
2354 }
2355 
2356 /*
2357  * Attempt to open the tray of @device.
2358  * If @force, ignore its tray lock.
2359  * Else, if the tray is locked, don't open it, but ask the guest to open it.
2360  * On error, store an error through @errp and return -errno.
2361  * If @device does not exist, return -ENODEV.
2362  * If it has no removable media, return -ENOTSUP.
2363  * If it has no tray, return -ENOSYS.
2364  * If the guest was asked to open the tray, return -EINPROGRESS.
2365  * Else, return 0.
2366  */
2367 static int do_open_tray(const char *blk_name, const char *qdev_id,
2368                         bool force, Error **errp)
2369 {
2370     BlockBackend *blk;
2371     const char *device = qdev_id ?: blk_name;
2372     bool locked;
2373 
2374     blk = qmp_get_blk(blk_name, qdev_id, errp);
2375     if (!blk) {
2376         return -ENODEV;
2377     }
2378 
2379     if (!blk_dev_has_removable_media(blk)) {
2380         error_setg(errp, "Device '%s' is not removable", device);
2381         return -ENOTSUP;
2382     }
2383 
2384     if (!blk_dev_has_tray(blk)) {
2385         error_setg(errp, "Device '%s' does not have a tray", device);
2386         return -ENOSYS;
2387     }
2388 
2389     if (blk_dev_is_tray_open(blk)) {
2390         return 0;
2391     }
2392 
2393     locked = blk_dev_is_medium_locked(blk);
2394     if (locked) {
2395         blk_dev_eject_request(blk, force);
2396     }
2397 
2398     if (!locked || force) {
2399         blk_dev_change_media_cb(blk, false, &error_abort);
2400     }
2401 
2402     if (locked && !force) {
2403         error_setg(errp, "Device '%s' is locked and force was not specified, "
2404                    "wait for tray to open and try again", device);
2405         return -EINPROGRESS;
2406     }
2407 
2408     return 0;
2409 }
2410 
2411 void qmp_blockdev_open_tray(bool has_device, const char *device,
2412                             bool has_id, const char *id,
2413                             bool has_force, bool force,
2414                             Error **errp)
2415 {
2416     Error *local_err = NULL;
2417     int rc;
2418 
2419     if (!has_force) {
2420         force = false;
2421     }
2422     rc = do_open_tray(has_device ? device : NULL,
2423                       has_id ? id : NULL,
2424                       force, &local_err);
2425     if (rc && rc != -ENOSYS && rc != -EINPROGRESS) {
2426         error_propagate(errp, local_err);
2427         return;
2428     }
2429     error_free(local_err);
2430 }
2431 
2432 void qmp_blockdev_close_tray(bool has_device, const char *device,
2433                              bool has_id, const char *id,
2434                              Error **errp)
2435 {
2436     BlockBackend *blk;
2437     Error *local_err = NULL;
2438 
2439     device = has_device ? device : NULL;
2440     id = has_id ? id : NULL;
2441 
2442     blk = qmp_get_blk(device, id, errp);
2443     if (!blk) {
2444         return;
2445     }
2446 
2447     if (!blk_dev_has_removable_media(blk)) {
2448         error_setg(errp, "Device '%s' is not removable", device ?: id);
2449         return;
2450     }
2451 
2452     if (!blk_dev_has_tray(blk)) {
2453         /* Ignore this command on tray-less devices */
2454         return;
2455     }
2456 
2457     if (!blk_dev_is_tray_open(blk)) {
2458         return;
2459     }
2460 
2461     blk_dev_change_media_cb(blk, true, &local_err);
2462     if (local_err) {
2463         error_propagate(errp, local_err);
2464         return;
2465     }
2466 }
2467 
2468 static void blockdev_remove_medium(bool has_device, const char *device,
2469                                    bool has_id, const char *id, Error **errp)
2470 {
2471     BlockBackend *blk;
2472     BlockDriverState *bs;
2473     AioContext *aio_context;
2474     bool has_attached_device;
2475 
2476     device = has_device ? device : NULL;
2477     id = has_id ? id : NULL;
2478 
2479     blk = qmp_get_blk(device, id, errp);
2480     if (!blk) {
2481         return;
2482     }
2483 
2484     /* For BBs without a device, we can exchange the BDS tree at will */
2485     has_attached_device = blk_get_attached_dev(blk);
2486 
2487     if (has_attached_device && !blk_dev_has_removable_media(blk)) {
2488         error_setg(errp, "Device '%s' is not removable", device ?: id);
2489         return;
2490     }
2491 
2492     if (has_attached_device && blk_dev_has_tray(blk) &&
2493         !blk_dev_is_tray_open(blk))
2494     {
2495         error_setg(errp, "Tray of device '%s' is not open", device ?: id);
2496         return;
2497     }
2498 
2499     bs = blk_bs(blk);
2500     if (!bs) {
2501         return;
2502     }
2503 
2504     aio_context = bdrv_get_aio_context(bs);
2505     aio_context_acquire(aio_context);
2506 
2507     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
2508         goto out;
2509     }
2510 
2511     blk_remove_bs(blk);
2512 
2513     if (!blk_dev_has_tray(blk)) {
2514         /* For tray-less devices, blockdev-open-tray is a no-op (or may not be
2515          * called at all); therefore, the medium needs to be ejected here.
2516          * Do it after blk_remove_bs() so blk_is_inserted(blk) returns the @load
2517          * value passed here (i.e. false). */
2518         blk_dev_change_media_cb(blk, false, &error_abort);
2519     }
2520 
2521 out:
2522     aio_context_release(aio_context);
2523 }
2524 
2525 void qmp_blockdev_remove_medium(const char *id, Error **errp)
2526 {
2527     blockdev_remove_medium(false, NULL, true, id, errp);
2528 }
2529 
2530 static void qmp_blockdev_insert_anon_medium(BlockBackend *blk,
2531                                             BlockDriverState *bs, Error **errp)
2532 {
2533     Error *local_err = NULL;
2534     bool has_device;
2535     int ret;
2536 
2537     /* For BBs without a device, we can exchange the BDS tree at will */
2538     has_device = blk_get_attached_dev(blk);
2539 
2540     if (has_device && !blk_dev_has_removable_media(blk)) {
2541         error_setg(errp, "Device is not removable");
2542         return;
2543     }
2544 
2545     if (has_device && blk_dev_has_tray(blk) && !blk_dev_is_tray_open(blk)) {
2546         error_setg(errp, "Tray of the device is not open");
2547         return;
2548     }
2549 
2550     if (blk_bs(blk)) {
2551         error_setg(errp, "There already is a medium in the device");
2552         return;
2553     }
2554 
2555     ret = blk_insert_bs(blk, bs, errp);
2556     if (ret < 0) {
2557         return;
2558     }
2559 
2560     if (!blk_dev_has_tray(blk)) {
2561         /* For tray-less devices, blockdev-close-tray is a no-op (or may not be
2562          * called at all); therefore, the medium needs to be pushed into the
2563          * slot here.
2564          * Do it after blk_insert_bs() so blk_is_inserted(blk) returns the @load
2565          * value passed here (i.e. true). */
2566         blk_dev_change_media_cb(blk, true, &local_err);
2567         if (local_err) {
2568             error_propagate(errp, local_err);
2569             blk_remove_bs(blk);
2570             return;
2571         }
2572     }
2573 }
2574 
2575 static void blockdev_insert_medium(bool has_device, const char *device,
2576                                    bool has_id, const char *id,
2577                                    const char *node_name, Error **errp)
2578 {
2579     BlockBackend *blk;
2580     BlockDriverState *bs;
2581 
2582     blk = qmp_get_blk(has_device ? device : NULL,
2583                       has_id ? id : NULL,
2584                       errp);
2585     if (!blk) {
2586         return;
2587     }
2588 
2589     bs = bdrv_find_node(node_name);
2590     if (!bs) {
2591         error_setg(errp, "Node '%s' not found", node_name);
2592         return;
2593     }
2594 
2595     if (bdrv_has_blk(bs)) {
2596         error_setg(errp, "Node '%s' is already in use", node_name);
2597         return;
2598     }
2599 
2600     qmp_blockdev_insert_anon_medium(blk, bs, errp);
2601 }
2602 
2603 void qmp_blockdev_insert_medium(const char *id, const char *node_name,
2604                                 Error **errp)
2605 {
2606     blockdev_insert_medium(false, NULL, true, id, node_name, errp);
2607 }
2608 
2609 void qmp_blockdev_change_medium(bool has_device, const char *device,
2610                                 bool has_id, const char *id,
2611                                 const char *filename,
2612                                 bool has_format, const char *format,
2613                                 bool has_read_only,
2614                                 BlockdevChangeReadOnlyMode read_only,
2615                                 Error **errp)
2616 {
2617     BlockBackend *blk;
2618     BlockDriverState *medium_bs = NULL;
2619     int bdrv_flags;
2620     bool detect_zeroes;
2621     int rc;
2622     QDict *options = NULL;
2623     Error *err = NULL;
2624 
2625     blk = qmp_get_blk(has_device ? device : NULL,
2626                       has_id ? id : NULL,
2627                       errp);
2628     if (!blk) {
2629         goto fail;
2630     }
2631 
2632     if (blk_bs(blk)) {
2633         blk_update_root_state(blk);
2634     }
2635 
2636     bdrv_flags = blk_get_open_flags_from_root_state(blk);
2637     bdrv_flags &= ~(BDRV_O_TEMPORARY | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING |
2638         BDRV_O_PROTOCOL | BDRV_O_AUTO_RDONLY);
2639 
2640     if (!has_read_only) {
2641         read_only = BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN;
2642     }
2643 
2644     switch (read_only) {
2645     case BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN:
2646         break;
2647 
2648     case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_ONLY:
2649         bdrv_flags &= ~BDRV_O_RDWR;
2650         break;
2651 
2652     case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_WRITE:
2653         bdrv_flags |= BDRV_O_RDWR;
2654         break;
2655 
2656     default:
2657         abort();
2658     }
2659 
2660     options = qdict_new();
2661     detect_zeroes = blk_get_detect_zeroes_from_root_state(blk);
2662     qdict_put_str(options, "detect-zeroes", detect_zeroes ? "on" : "off");
2663 
2664     if (has_format) {
2665         qdict_put_str(options, "driver", format);
2666     }
2667 
2668     medium_bs = bdrv_open(filename, NULL, options, bdrv_flags, errp);
2669     if (!medium_bs) {
2670         goto fail;
2671     }
2672 
2673     rc = do_open_tray(has_device ? device : NULL,
2674                       has_id ? id : NULL,
2675                       false, &err);
2676     if (rc && rc != -ENOSYS) {
2677         error_propagate(errp, err);
2678         goto fail;
2679     }
2680     error_free(err);
2681     err = NULL;
2682 
2683     blockdev_remove_medium(has_device, device, has_id, id, &err);
2684     if (err) {
2685         error_propagate(errp, err);
2686         goto fail;
2687     }
2688 
2689     qmp_blockdev_insert_anon_medium(blk, medium_bs, &err);
2690     if (err) {
2691         error_propagate(errp, err);
2692         goto fail;
2693     }
2694 
2695     qmp_blockdev_close_tray(has_device, device, has_id, id, errp);
2696 
2697 fail:
2698     /* If the medium has been inserted, the device has its own reference, so
2699      * ours must be relinquished; and if it has not been inserted successfully,
2700      * the reference must be relinquished anyway */
2701     bdrv_unref(medium_bs);
2702 }
2703 
2704 /* throttling disk I/O limits */
2705 void qmp_block_set_io_throttle(BlockIOThrottle *arg, Error **errp)
2706 {
2707     ThrottleConfig cfg;
2708     BlockDriverState *bs;
2709     BlockBackend *blk;
2710     AioContext *aio_context;
2711 
2712     blk = qmp_get_blk(arg->has_device ? arg->device : NULL,
2713                       arg->has_id ? arg->id : NULL,
2714                       errp);
2715     if (!blk) {
2716         return;
2717     }
2718 
2719     aio_context = blk_get_aio_context(blk);
2720     aio_context_acquire(aio_context);
2721 
2722     bs = blk_bs(blk);
2723     if (!bs) {
2724         error_setg(errp, "Device has no medium");
2725         goto out;
2726     }
2727 
2728     throttle_config_init(&cfg);
2729     cfg.buckets[THROTTLE_BPS_TOTAL].avg = arg->bps;
2730     cfg.buckets[THROTTLE_BPS_READ].avg  = arg->bps_rd;
2731     cfg.buckets[THROTTLE_BPS_WRITE].avg = arg->bps_wr;
2732 
2733     cfg.buckets[THROTTLE_OPS_TOTAL].avg = arg->iops;
2734     cfg.buckets[THROTTLE_OPS_READ].avg  = arg->iops_rd;
2735     cfg.buckets[THROTTLE_OPS_WRITE].avg = arg->iops_wr;
2736 
2737     if (arg->has_bps_max) {
2738         cfg.buckets[THROTTLE_BPS_TOTAL].max = arg->bps_max;
2739     }
2740     if (arg->has_bps_rd_max) {
2741         cfg.buckets[THROTTLE_BPS_READ].max = arg->bps_rd_max;
2742     }
2743     if (arg->has_bps_wr_max) {
2744         cfg.buckets[THROTTLE_BPS_WRITE].max = arg->bps_wr_max;
2745     }
2746     if (arg->has_iops_max) {
2747         cfg.buckets[THROTTLE_OPS_TOTAL].max = arg->iops_max;
2748     }
2749     if (arg->has_iops_rd_max) {
2750         cfg.buckets[THROTTLE_OPS_READ].max = arg->iops_rd_max;
2751     }
2752     if (arg->has_iops_wr_max) {
2753         cfg.buckets[THROTTLE_OPS_WRITE].max = arg->iops_wr_max;
2754     }
2755 
2756     if (arg->has_bps_max_length) {
2757         cfg.buckets[THROTTLE_BPS_TOTAL].burst_length = arg->bps_max_length;
2758     }
2759     if (arg->has_bps_rd_max_length) {
2760         cfg.buckets[THROTTLE_BPS_READ].burst_length = arg->bps_rd_max_length;
2761     }
2762     if (arg->has_bps_wr_max_length) {
2763         cfg.buckets[THROTTLE_BPS_WRITE].burst_length = arg->bps_wr_max_length;
2764     }
2765     if (arg->has_iops_max_length) {
2766         cfg.buckets[THROTTLE_OPS_TOTAL].burst_length = arg->iops_max_length;
2767     }
2768     if (arg->has_iops_rd_max_length) {
2769         cfg.buckets[THROTTLE_OPS_READ].burst_length = arg->iops_rd_max_length;
2770     }
2771     if (arg->has_iops_wr_max_length) {
2772         cfg.buckets[THROTTLE_OPS_WRITE].burst_length = arg->iops_wr_max_length;
2773     }
2774 
2775     if (arg->has_iops_size) {
2776         cfg.op_size = arg->iops_size;
2777     }
2778 
2779     if (!throttle_is_valid(&cfg, errp)) {
2780         goto out;
2781     }
2782 
2783     if (throttle_enabled(&cfg)) {
2784         /* Enable I/O limits if they're not enabled yet, otherwise
2785          * just update the throttling group. */
2786         if (!blk_get_public(blk)->throttle_group_member.throttle_state) {
2787             blk_io_limits_enable(blk,
2788                                  arg->has_group ? arg->group :
2789                                  arg->has_device ? arg->device :
2790                                  arg->id);
2791         } else if (arg->has_group) {
2792             blk_io_limits_update_group(blk, arg->group);
2793         }
2794         /* Set the new throttling configuration */
2795         blk_set_io_limits(blk, &cfg);
2796     } else if (blk_get_public(blk)->throttle_group_member.throttle_state) {
2797         /* If all throttling settings are set to 0, disable I/O limits */
2798         blk_io_limits_disable(blk);
2799     }
2800 
2801 out:
2802     aio_context_release(aio_context);
2803 }
2804 
2805 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2806                                 bool has_granularity, uint32_t granularity,
2807                                 bool has_persistent, bool persistent,
2808                                 bool has_autoload, bool autoload,
2809                                 bool has_disabled, bool disabled,
2810                                 Error **errp)
2811 {
2812     BlockDriverState *bs;
2813     BdrvDirtyBitmap *bitmap;
2814     AioContext *aio_context = NULL;
2815 
2816     if (!name || name[0] == '\0') {
2817         error_setg(errp, "Bitmap name cannot be empty");
2818         return;
2819     }
2820 
2821     bs = bdrv_lookup_bs(node, node, errp);
2822     if (!bs) {
2823         return;
2824     }
2825 
2826     if (has_granularity) {
2827         if (granularity < 512 || !is_power_of_2(granularity)) {
2828             error_setg(errp, "Granularity must be power of 2 "
2829                              "and at least 512");
2830             return;
2831         }
2832     } else {
2833         /* Default to cluster size, if available: */
2834         granularity = bdrv_get_default_bitmap_granularity(bs);
2835     }
2836 
2837     if (!has_persistent) {
2838         persistent = false;
2839     }
2840 
2841     if (has_autoload) {
2842         warn_report("Autoload option is deprecated and its value is ignored");
2843     }
2844 
2845     if (!has_disabled) {
2846         disabled = false;
2847     }
2848 
2849     if (persistent) {
2850         aio_context = bdrv_get_aio_context(bs);
2851         aio_context_acquire(aio_context);
2852         if (!bdrv_can_store_new_dirty_bitmap(bs, name, granularity, errp)) {
2853             goto out;
2854         }
2855     }
2856 
2857     bitmap = bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2858     if (bitmap == NULL) {
2859         goto out;
2860     }
2861 
2862     if (disabled) {
2863         bdrv_disable_dirty_bitmap(bitmap);
2864     }
2865 
2866     bdrv_dirty_bitmap_set_persistence(bitmap, persistent);
2867  out:
2868     if (aio_context) {
2869         aio_context_release(aio_context);
2870     }
2871 }
2872 
2873 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2874                                    Error **errp)
2875 {
2876     BlockDriverState *bs;
2877     BdrvDirtyBitmap *bitmap;
2878     Error *local_err = NULL;
2879     AioContext *aio_context = NULL;
2880 
2881     bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2882     if (!bitmap || !bs) {
2883         return;
2884     }
2885 
2886     if (bdrv_dirty_bitmap_check(bitmap, BDRV_BITMAP_BUSY | BDRV_BITMAP_RO,
2887                                 errp)) {
2888         return;
2889     }
2890 
2891     if (bdrv_dirty_bitmap_get_persistence(bitmap)) {
2892         aio_context = bdrv_get_aio_context(bs);
2893         aio_context_acquire(aio_context);
2894         bdrv_remove_persistent_dirty_bitmap(bs, name, &local_err);
2895         if (local_err != NULL) {
2896             error_propagate(errp, local_err);
2897             goto out;
2898         }
2899     }
2900 
2901     bdrv_release_dirty_bitmap(bs, bitmap);
2902  out:
2903     if (aio_context) {
2904         aio_context_release(aio_context);
2905     }
2906 }
2907 
2908 /**
2909  * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2910  * immediately after a full backup operation.
2911  */
2912 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2913                                   Error **errp)
2914 {
2915     BdrvDirtyBitmap *bitmap;
2916     BlockDriverState *bs;
2917 
2918     bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2919     if (!bitmap || !bs) {
2920         return;
2921     }
2922 
2923     if (bdrv_dirty_bitmap_check(bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2924         return;
2925     }
2926 
2927     bdrv_clear_dirty_bitmap(bitmap, NULL);
2928 }
2929 
2930 void qmp_block_dirty_bitmap_enable(const char *node, const char *name,
2931                                    Error **errp)
2932 {
2933     BlockDriverState *bs;
2934     BdrvDirtyBitmap *bitmap;
2935 
2936     bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2937     if (!bitmap) {
2938         return;
2939     }
2940 
2941     if (bdrv_dirty_bitmap_check(bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2942         return;
2943     }
2944 
2945     bdrv_enable_dirty_bitmap(bitmap);
2946 }
2947 
2948 void qmp_block_dirty_bitmap_disable(const char *node, const char *name,
2949                                     Error **errp)
2950 {
2951     BlockDriverState *bs;
2952     BdrvDirtyBitmap *bitmap;
2953 
2954     bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2955     if (!bitmap) {
2956         return;
2957     }
2958 
2959     if (bdrv_dirty_bitmap_check(bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2960         return;
2961     }
2962 
2963     bdrv_disable_dirty_bitmap(bitmap);
2964 }
2965 
2966 static BdrvDirtyBitmap *do_block_dirty_bitmap_merge(
2967         const char *node, const char *target,
2968         BlockDirtyBitmapMergeSourceList *bitmaps,
2969         HBitmap **backup, Error **errp)
2970 {
2971     BlockDriverState *bs;
2972     BdrvDirtyBitmap *dst, *src, *anon;
2973     BlockDirtyBitmapMergeSourceList *lst;
2974     Error *local_err = NULL;
2975 
2976     dst = block_dirty_bitmap_lookup(node, target, &bs, errp);
2977     if (!dst) {
2978         return NULL;
2979     }
2980 
2981     anon = bdrv_create_dirty_bitmap(bs, bdrv_dirty_bitmap_granularity(dst),
2982                                     NULL, errp);
2983     if (!anon) {
2984         return NULL;
2985     }
2986 
2987     for (lst = bitmaps; lst; lst = lst->next) {
2988         switch (lst->value->type) {
2989             const char *name, *node;
2990         case QTYPE_QSTRING:
2991             name = lst->value->u.local;
2992             src = bdrv_find_dirty_bitmap(bs, name);
2993             if (!src) {
2994                 error_setg(errp, "Dirty bitmap '%s' not found", name);
2995                 dst = NULL;
2996                 goto out;
2997             }
2998             break;
2999         case QTYPE_QDICT:
3000             node = lst->value->u.external.node;
3001             name = lst->value->u.external.name;
3002             src = block_dirty_bitmap_lookup(node, name, NULL, errp);
3003             if (!src) {
3004                 dst = NULL;
3005                 goto out;
3006             }
3007             break;
3008         default:
3009             abort();
3010         }
3011 
3012         bdrv_merge_dirty_bitmap(anon, src, NULL, &local_err);
3013         if (local_err) {
3014             error_propagate(errp, local_err);
3015             dst = NULL;
3016             goto out;
3017         }
3018     }
3019 
3020     /* Merge into dst; dst is unchanged on failure. */
3021     bdrv_merge_dirty_bitmap(dst, anon, backup, errp);
3022 
3023  out:
3024     bdrv_release_dirty_bitmap(bs, anon);
3025     return dst;
3026 }
3027 
3028 void qmp_block_dirty_bitmap_merge(const char *node, const char *target,
3029                                   BlockDirtyBitmapMergeSourceList *bitmaps,
3030                                   Error **errp)
3031 {
3032     do_block_dirty_bitmap_merge(node, target, bitmaps, NULL, errp);
3033 }
3034 
3035 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
3036                                                               const char *name,
3037                                                               Error **errp)
3038 {
3039     BdrvDirtyBitmap *bitmap;
3040     BlockDriverState *bs;
3041     BlockDirtyBitmapSha256 *ret = NULL;
3042     char *sha256;
3043 
3044     bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
3045     if (!bitmap || !bs) {
3046         return NULL;
3047     }
3048 
3049     sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
3050     if (sha256 == NULL) {
3051         return NULL;
3052     }
3053 
3054     ret = g_new(BlockDirtyBitmapSha256, 1);
3055     ret->sha256 = sha256;
3056 
3057     return ret;
3058 }
3059 
3060 void hmp_drive_del(Monitor *mon, const QDict *qdict)
3061 {
3062     const char *id = qdict_get_str(qdict, "id");
3063     BlockBackend *blk;
3064     BlockDriverState *bs;
3065     AioContext *aio_context;
3066     Error *local_err = NULL;
3067 
3068     bs = bdrv_find_node(id);
3069     if (bs) {
3070         qmp_blockdev_del(id, &local_err);
3071         if (local_err) {
3072             error_report_err(local_err);
3073         }
3074         return;
3075     }
3076 
3077     blk = blk_by_name(id);
3078     if (!blk) {
3079         error_report("Device '%s' not found", id);
3080         return;
3081     }
3082 
3083     if (!blk_legacy_dinfo(blk)) {
3084         error_report("Deleting device added with blockdev-add"
3085                      " is not supported");
3086         return;
3087     }
3088 
3089     aio_context = blk_get_aio_context(blk);
3090     aio_context_acquire(aio_context);
3091 
3092     bs = blk_bs(blk);
3093     if (bs) {
3094         if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
3095             error_report_err(local_err);
3096             aio_context_release(aio_context);
3097             return;
3098         }
3099 
3100         blk_remove_bs(blk);
3101     }
3102 
3103     /* Make the BlockBackend and the attached BlockDriverState anonymous */
3104     monitor_remove_blk(blk);
3105 
3106     /* If this BlockBackend has a device attached to it, its refcount will be
3107      * decremented when the device is removed; otherwise we have to do so here.
3108      */
3109     if (blk_get_attached_dev(blk)) {
3110         /* Further I/O must not pause the guest */
3111         blk_set_on_error(blk, BLOCKDEV_ON_ERROR_REPORT,
3112                          BLOCKDEV_ON_ERROR_REPORT);
3113     } else {
3114         blk_unref(blk);
3115     }
3116 
3117     aio_context_release(aio_context);
3118 }
3119 
3120 void qmp_block_resize(bool has_device, const char *device,
3121                       bool has_node_name, const char *node_name,
3122                       int64_t size, Error **errp)
3123 {
3124     Error *local_err = NULL;
3125     BlockBackend *blk = NULL;
3126     BlockDriverState *bs;
3127     AioContext *aio_context;
3128     int ret;
3129 
3130     bs = bdrv_lookup_bs(has_device ? device : NULL,
3131                         has_node_name ? node_name : NULL,
3132                         &local_err);
3133     if (local_err) {
3134         error_propagate(errp, local_err);
3135         return;
3136     }
3137 
3138     aio_context = bdrv_get_aio_context(bs);
3139     aio_context_acquire(aio_context);
3140 
3141     if (!bdrv_is_first_non_filter(bs)) {
3142         error_setg(errp, QERR_FEATURE_DISABLED, "resize");
3143         goto out;
3144     }
3145 
3146     if (size < 0) {
3147         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
3148         goto out;
3149     }
3150 
3151     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
3152         error_setg(errp, QERR_DEVICE_IN_USE, device);
3153         goto out;
3154     }
3155 
3156     blk = blk_new(bdrv_get_aio_context(bs), BLK_PERM_RESIZE, BLK_PERM_ALL);
3157     ret = blk_insert_bs(blk, bs, errp);
3158     if (ret < 0) {
3159         goto out;
3160     }
3161 
3162     bdrv_drained_begin(bs);
3163     ret = blk_truncate(blk, size, PREALLOC_MODE_OFF, errp);
3164     bdrv_drained_end(bs);
3165 
3166 out:
3167     blk_unref(blk);
3168     aio_context_release(aio_context);
3169 }
3170 
3171 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
3172                       bool has_base, const char *base,
3173                       bool has_base_node, const char *base_node,
3174                       bool has_backing_file, const char *backing_file,
3175                       bool has_speed, int64_t speed,
3176                       bool has_on_error, BlockdevOnError on_error,
3177                       bool has_auto_finalize, bool auto_finalize,
3178                       bool has_auto_dismiss, bool auto_dismiss,
3179                       Error **errp)
3180 {
3181     BlockDriverState *bs, *iter;
3182     BlockDriverState *base_bs = NULL;
3183     AioContext *aio_context;
3184     Error *local_err = NULL;
3185     const char *base_name = NULL;
3186     int job_flags = JOB_DEFAULT;
3187 
3188     if (!has_on_error) {
3189         on_error = BLOCKDEV_ON_ERROR_REPORT;
3190     }
3191 
3192     bs = bdrv_lookup_bs(device, device, errp);
3193     if (!bs) {
3194         return;
3195     }
3196 
3197     aio_context = bdrv_get_aio_context(bs);
3198     aio_context_acquire(aio_context);
3199 
3200     if (has_base && has_base_node) {
3201         error_setg(errp, "'base' and 'base-node' cannot be specified "
3202                    "at the same time");
3203         goto out;
3204     }
3205 
3206     if (has_base) {
3207         base_bs = bdrv_find_backing_image(bs, base);
3208         if (base_bs == NULL) {
3209             error_setg(errp, QERR_BASE_NOT_FOUND, base);
3210             goto out;
3211         }
3212         assert(bdrv_get_aio_context(base_bs) == aio_context);
3213         base_name = base;
3214     }
3215 
3216     if (has_base_node) {
3217         base_bs = bdrv_lookup_bs(NULL, base_node, errp);
3218         if (!base_bs) {
3219             goto out;
3220         }
3221         if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
3222             error_setg(errp, "Node '%s' is not a backing image of '%s'",
3223                        base_node, device);
3224             goto out;
3225         }
3226         assert(bdrv_get_aio_context(base_bs) == aio_context);
3227         bdrv_refresh_filename(base_bs);
3228         base_name = base_bs->filename;
3229     }
3230 
3231     /* Check for op blockers in the whole chain between bs and base */
3232     for (iter = bs; iter && iter != base_bs; iter = backing_bs(iter)) {
3233         if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
3234             goto out;
3235         }
3236     }
3237 
3238     /* if we are streaming the entire chain, the result will have no backing
3239      * file, and specifying one is therefore an error */
3240     if (base_bs == NULL && has_backing_file) {
3241         error_setg(errp, "backing file specified, but streaming the "
3242                          "entire chain");
3243         goto out;
3244     }
3245 
3246     /* backing_file string overrides base bs filename */
3247     base_name = has_backing_file ? backing_file : base_name;
3248 
3249     if (has_auto_finalize && !auto_finalize) {
3250         job_flags |= JOB_MANUAL_FINALIZE;
3251     }
3252     if (has_auto_dismiss && !auto_dismiss) {
3253         job_flags |= JOB_MANUAL_DISMISS;
3254     }
3255 
3256     stream_start(has_job_id ? job_id : NULL, bs, base_bs, base_name,
3257                  job_flags, has_speed ? speed : 0, on_error, &local_err);
3258     if (local_err) {
3259         error_propagate(errp, local_err);
3260         goto out;
3261     }
3262 
3263     trace_qmp_block_stream(bs);
3264 
3265 out:
3266     aio_context_release(aio_context);
3267 }
3268 
3269 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
3270                       bool has_base_node, const char *base_node,
3271                       bool has_base, const char *base,
3272                       bool has_top_node, const char *top_node,
3273                       bool has_top, const char *top,
3274                       bool has_backing_file, const char *backing_file,
3275                       bool has_speed, int64_t speed,
3276                       bool has_filter_node_name, const char *filter_node_name,
3277                       bool has_auto_finalize, bool auto_finalize,
3278                       bool has_auto_dismiss, bool auto_dismiss,
3279                       Error **errp)
3280 {
3281     BlockDriverState *bs;
3282     BlockDriverState *iter;
3283     BlockDriverState *base_bs, *top_bs;
3284     AioContext *aio_context;
3285     Error *local_err = NULL;
3286     /* This will be part of the QMP command, if/when the
3287      * BlockdevOnError change for blkmirror makes it in
3288      */
3289     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
3290     int job_flags = JOB_DEFAULT;
3291 
3292     if (!has_speed) {
3293         speed = 0;
3294     }
3295     if (!has_filter_node_name) {
3296         filter_node_name = NULL;
3297     }
3298     if (has_auto_finalize && !auto_finalize) {
3299         job_flags |= JOB_MANUAL_FINALIZE;
3300     }
3301     if (has_auto_dismiss && !auto_dismiss) {
3302         job_flags |= JOB_MANUAL_DISMISS;
3303     }
3304 
3305     /* Important Note:
3306      *  libvirt relies on the DeviceNotFound error class in order to probe for
3307      *  live commit feature versions; for this to work, we must make sure to
3308      *  perform the device lookup before any generic errors that may occur in a
3309      *  scenario in which all optional arguments are omitted. */
3310     bs = qmp_get_root_bs(device, &local_err);
3311     if (!bs) {
3312         bs = bdrv_lookup_bs(device, device, NULL);
3313         if (!bs) {
3314             error_free(local_err);
3315             error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3316                       "Device '%s' not found", device);
3317         } else {
3318             error_propagate(errp, local_err);
3319         }
3320         return;
3321     }
3322 
3323     aio_context = bdrv_get_aio_context(bs);
3324     aio_context_acquire(aio_context);
3325 
3326     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
3327         goto out;
3328     }
3329 
3330     /* default top_bs is the active layer */
3331     top_bs = bs;
3332 
3333     if (has_top_node && has_top) {
3334         error_setg(errp, "'top-node' and 'top' are mutually exclusive");
3335         goto out;
3336     } else if (has_top_node) {
3337         top_bs = bdrv_lookup_bs(NULL, top_node, errp);
3338         if (top_bs == NULL) {
3339             goto out;
3340         }
3341         if (!bdrv_chain_contains(bs, top_bs)) {
3342             error_setg(errp, "'%s' is not in this backing file chain",
3343                        top_node);
3344             goto out;
3345         }
3346     } else if (has_top && top) {
3347         /* This strcmp() is just a shortcut, there is no need to
3348          * refresh @bs's filename.  If it mismatches,
3349          * bdrv_find_backing_image() will do the refresh and may still
3350          * return @bs. */
3351         if (strcmp(bs->filename, top) != 0) {
3352             top_bs = bdrv_find_backing_image(bs, top);
3353         }
3354     }
3355 
3356     if (top_bs == NULL) {
3357         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
3358         goto out;
3359     }
3360 
3361     assert(bdrv_get_aio_context(top_bs) == aio_context);
3362 
3363     if (has_base_node && has_base) {
3364         error_setg(errp, "'base-node' and 'base' are mutually exclusive");
3365         goto out;
3366     } else if (has_base_node) {
3367         base_bs = bdrv_lookup_bs(NULL, base_node, errp);
3368         if (base_bs == NULL) {
3369             goto out;
3370         }
3371         if (!bdrv_chain_contains(top_bs, base_bs)) {
3372             error_setg(errp, "'%s' is not in this backing file chain",
3373                        base_node);
3374             goto out;
3375         }
3376     } else if (has_base && base) {
3377         base_bs = bdrv_find_backing_image(top_bs, base);
3378     } else {
3379         base_bs = bdrv_find_base(top_bs);
3380     }
3381 
3382     if (base_bs == NULL) {
3383         error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
3384         goto out;
3385     }
3386 
3387     assert(bdrv_get_aio_context(base_bs) == aio_context);
3388 
3389     for (iter = top_bs; iter != backing_bs(base_bs); iter = backing_bs(iter)) {
3390         if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
3391             goto out;
3392         }
3393     }
3394 
3395     /* Do not allow attempts to commit an image into itself */
3396     if (top_bs == base_bs) {
3397         error_setg(errp, "cannot commit an image into itself");
3398         goto out;
3399     }
3400 
3401     if (top_bs == bs) {
3402         if (has_backing_file) {
3403             error_setg(errp, "'backing-file' specified,"
3404                              " but 'top' is the active layer");
3405             goto out;
3406         }
3407         commit_active_start(has_job_id ? job_id : NULL, bs, base_bs,
3408                             job_flags, speed, on_error,
3409                             filter_node_name, NULL, NULL, false, &local_err);
3410     } else {
3411         BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
3412         if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
3413             goto out;
3414         }
3415         commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, job_flags,
3416                      speed, on_error, has_backing_file ? backing_file : NULL,
3417                      filter_node_name, &local_err);
3418     }
3419     if (local_err != NULL) {
3420         error_propagate(errp, local_err);
3421         goto out;
3422     }
3423 
3424 out:
3425     aio_context_release(aio_context);
3426 }
3427 
3428 static BlockJob *do_drive_backup(DriveBackup *backup, JobTxn *txn,
3429                                  Error **errp)
3430 {
3431     BlockDriverState *bs;
3432     BlockDriverState *target_bs;
3433     BlockDriverState *source = NULL;
3434     BlockJob *job = NULL;
3435     BdrvDirtyBitmap *bmap = NULL;
3436     AioContext *aio_context;
3437     QDict *options = NULL;
3438     Error *local_err = NULL;
3439     int flags, job_flags = JOB_DEFAULT;
3440     int64_t size;
3441     bool set_backing_hd = false;
3442     int ret;
3443 
3444     if (!backup->has_speed) {
3445         backup->speed = 0;
3446     }
3447     if (!backup->has_on_source_error) {
3448         backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3449     }
3450     if (!backup->has_on_target_error) {
3451         backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3452     }
3453     if (!backup->has_mode) {
3454         backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3455     }
3456     if (!backup->has_job_id) {
3457         backup->job_id = NULL;
3458     }
3459     if (!backup->has_auto_finalize) {
3460         backup->auto_finalize = true;
3461     }
3462     if (!backup->has_auto_dismiss) {
3463         backup->auto_dismiss = true;
3464     }
3465     if (!backup->has_compress) {
3466         backup->compress = false;
3467     }
3468 
3469     bs = bdrv_lookup_bs(backup->device, backup->device, errp);
3470     if (!bs) {
3471         return NULL;
3472     }
3473 
3474     if (!bs->drv) {
3475         error_setg(errp, "Device has no medium");
3476         return NULL;
3477     }
3478 
3479     aio_context = bdrv_get_aio_context(bs);
3480     aio_context_acquire(aio_context);
3481 
3482     if (!backup->has_format) {
3483         backup->format = backup->mode == NEW_IMAGE_MODE_EXISTING ?
3484                          NULL : (char*) bs->drv->format_name;
3485     }
3486 
3487     /* Early check to avoid creating target */
3488     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
3489         goto out;
3490     }
3491 
3492     flags = bs->open_flags | BDRV_O_RDWR;
3493 
3494     /* See if we have a backing HD we can use to create our new image
3495      * on top of. */
3496     if (backup->sync == MIRROR_SYNC_MODE_TOP) {
3497         source = backing_bs(bs);
3498         if (!source) {
3499             backup->sync = MIRROR_SYNC_MODE_FULL;
3500         }
3501     }
3502     if (backup->sync == MIRROR_SYNC_MODE_NONE) {
3503         source = bs;
3504         flags |= BDRV_O_NO_BACKING;
3505         set_backing_hd = true;
3506     }
3507 
3508     size = bdrv_getlength(bs);
3509     if (size < 0) {
3510         error_setg_errno(errp, -size, "bdrv_getlength failed");
3511         goto out;
3512     }
3513 
3514     if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
3515         assert(backup->format);
3516         if (source) {
3517             bdrv_refresh_filename(source);
3518             bdrv_img_create(backup->target, backup->format, source->filename,
3519                             source->drv->format_name, NULL,
3520                             size, flags, false, &local_err);
3521         } else {
3522             bdrv_img_create(backup->target, backup->format, NULL, NULL, NULL,
3523                             size, flags, false, &local_err);
3524         }
3525     }
3526 
3527     if (local_err) {
3528         error_propagate(errp, local_err);
3529         goto out;
3530     }
3531 
3532     if (backup->format) {
3533         if (!options) {
3534             options = qdict_new();
3535         }
3536         qdict_put_str(options, "driver", backup->format);
3537     }
3538 
3539     target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
3540     if (!target_bs) {
3541         goto out;
3542     }
3543 
3544     ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3545     if (ret < 0) {
3546         bdrv_unref(target_bs);
3547         goto out;
3548     }
3549 
3550     if (set_backing_hd) {
3551         bdrv_set_backing_hd(target_bs, source, &local_err);
3552         if (local_err) {
3553             goto unref;
3554         }
3555     }
3556 
3557     if (backup->has_bitmap) {
3558         bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
3559         if (!bmap) {
3560             error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
3561             goto unref;
3562         }
3563         if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_DEFAULT, errp)) {
3564             goto unref;
3565         }
3566     }
3567     if (!backup->auto_finalize) {
3568         job_flags |= JOB_MANUAL_FINALIZE;
3569     }
3570     if (!backup->auto_dismiss) {
3571         job_flags |= JOB_MANUAL_DISMISS;
3572     }
3573 
3574     job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
3575                             backup->sync, bmap, backup->compress,
3576                             backup->on_source_error, backup->on_target_error,
3577                             job_flags, NULL, NULL, txn, &local_err);
3578     if (local_err != NULL) {
3579         error_propagate(errp, local_err);
3580         goto unref;
3581     }
3582 
3583 unref:
3584     bdrv_unref(target_bs);
3585 out:
3586     aio_context_release(aio_context);
3587     return job;
3588 }
3589 
3590 void qmp_drive_backup(DriveBackup *arg, Error **errp)
3591 {
3592 
3593     BlockJob *job;
3594     job = do_drive_backup(arg, NULL, errp);
3595     if (job) {
3596         job_start(&job->job);
3597     }
3598 }
3599 
3600 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
3601 {
3602     return bdrv_named_nodes_list(errp);
3603 }
3604 
3605 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
3606 {
3607     return bdrv_get_xdbg_block_graph(errp);
3608 }
3609 
3610 BlockJob *do_blockdev_backup(BlockdevBackup *backup, JobTxn *txn,
3611                              Error **errp)
3612 {
3613     BlockDriverState *bs;
3614     BlockDriverState *target_bs;
3615     Error *local_err = NULL;
3616     BdrvDirtyBitmap *bmap = NULL;
3617     AioContext *aio_context;
3618     BlockJob *job = NULL;
3619     int job_flags = JOB_DEFAULT;
3620     int ret;
3621 
3622     if (!backup->has_speed) {
3623         backup->speed = 0;
3624     }
3625     if (!backup->has_on_source_error) {
3626         backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3627     }
3628     if (!backup->has_on_target_error) {
3629         backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3630     }
3631     if (!backup->has_job_id) {
3632         backup->job_id = NULL;
3633     }
3634     if (!backup->has_auto_finalize) {
3635         backup->auto_finalize = true;
3636     }
3637     if (!backup->has_auto_dismiss) {
3638         backup->auto_dismiss = true;
3639     }
3640     if (!backup->has_compress) {
3641         backup->compress = false;
3642     }
3643 
3644     bs = bdrv_lookup_bs(backup->device, backup->device, errp);
3645     if (!bs) {
3646         return NULL;
3647     }
3648 
3649     aio_context = bdrv_get_aio_context(bs);
3650     aio_context_acquire(aio_context);
3651 
3652     target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
3653     if (!target_bs) {
3654         goto out;
3655     }
3656 
3657     ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3658     if (ret < 0) {
3659         goto out;
3660     }
3661 
3662     if (backup->has_bitmap) {
3663         bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
3664         if (!bmap) {
3665             error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
3666             goto out;
3667         }
3668         if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_DEFAULT, errp)) {
3669             goto out;
3670         }
3671     }
3672 
3673     if (!backup->auto_finalize) {
3674         job_flags |= JOB_MANUAL_FINALIZE;
3675     }
3676     if (!backup->auto_dismiss) {
3677         job_flags |= JOB_MANUAL_DISMISS;
3678     }
3679     job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
3680                             backup->sync, bmap, backup->compress,
3681                             backup->on_source_error, backup->on_target_error,
3682                             job_flags, NULL, NULL, txn, &local_err);
3683     if (local_err != NULL) {
3684         error_propagate(errp, local_err);
3685     }
3686 out:
3687     aio_context_release(aio_context);
3688     return job;
3689 }
3690 
3691 void qmp_blockdev_backup(BlockdevBackup *arg, Error **errp)
3692 {
3693     BlockJob *job;
3694     job = do_blockdev_backup(arg, NULL, errp);
3695     if (job) {
3696         job_start(&job->job);
3697     }
3698 }
3699 
3700 /* Parameter check and block job starting for drive mirroring.
3701  * Caller should hold @device and @target's aio context (must be the same).
3702  **/
3703 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
3704                                    BlockDriverState *target,
3705                                    bool has_replaces, const char *replaces,
3706                                    enum MirrorSyncMode sync,
3707                                    BlockMirrorBackingMode backing_mode,
3708                                    bool has_speed, int64_t speed,
3709                                    bool has_granularity, uint32_t granularity,
3710                                    bool has_buf_size, int64_t buf_size,
3711                                    bool has_on_source_error,
3712                                    BlockdevOnError on_source_error,
3713                                    bool has_on_target_error,
3714                                    BlockdevOnError on_target_error,
3715                                    bool has_unmap, bool unmap,
3716                                    bool has_filter_node_name,
3717                                    const char *filter_node_name,
3718                                    bool has_copy_mode, MirrorCopyMode copy_mode,
3719                                    bool has_auto_finalize, bool auto_finalize,
3720                                    bool has_auto_dismiss, bool auto_dismiss,
3721                                    Error **errp)
3722 {
3723     int job_flags = JOB_DEFAULT;
3724 
3725     if (!has_speed) {
3726         speed = 0;
3727     }
3728     if (!has_on_source_error) {
3729         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3730     }
3731     if (!has_on_target_error) {
3732         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3733     }
3734     if (!has_granularity) {
3735         granularity = 0;
3736     }
3737     if (!has_buf_size) {
3738         buf_size = 0;
3739     }
3740     if (!has_unmap) {
3741         unmap = true;
3742     }
3743     if (!has_filter_node_name) {
3744         filter_node_name = NULL;
3745     }
3746     if (!has_copy_mode) {
3747         copy_mode = MIRROR_COPY_MODE_BACKGROUND;
3748     }
3749     if (has_auto_finalize && !auto_finalize) {
3750         job_flags |= JOB_MANUAL_FINALIZE;
3751     }
3752     if (has_auto_dismiss && !auto_dismiss) {
3753         job_flags |= JOB_MANUAL_DISMISS;
3754     }
3755 
3756     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3757         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3758                    "a value in range [512B, 64MB]");
3759         return;
3760     }
3761     if (granularity & (granularity - 1)) {
3762         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3763                    "power of 2");
3764         return;
3765     }
3766 
3767     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3768         return;
3769     }
3770     if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3771         return;
3772     }
3773 
3774     if (!bs->backing && sync == MIRROR_SYNC_MODE_TOP) {
3775         sync = MIRROR_SYNC_MODE_FULL;
3776     }
3777 
3778     if (has_replaces) {
3779         BlockDriverState *to_replace_bs;
3780         AioContext *replace_aio_context;
3781         int64_t bs_size, replace_size;
3782 
3783         bs_size = bdrv_getlength(bs);
3784         if (bs_size < 0) {
3785             error_setg_errno(errp, -bs_size, "Failed to query device's size");
3786             return;
3787         }
3788 
3789         to_replace_bs = check_to_replace_node(bs, replaces, errp);
3790         if (!to_replace_bs) {
3791             return;
3792         }
3793 
3794         replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3795         aio_context_acquire(replace_aio_context);
3796         replace_size = bdrv_getlength(to_replace_bs);
3797         aio_context_release(replace_aio_context);
3798 
3799         if (replace_size < 0) {
3800             error_setg_errno(errp, -replace_size,
3801                              "Failed to query the replacement node's size");
3802             return;
3803         }
3804         if (bs_size != replace_size) {
3805             error_setg(errp, "cannot replace image with a mirror image of "
3806                              "different size");
3807             return;
3808         }
3809     }
3810 
3811     /* pass the node name to replace to mirror start since it's loose coupling
3812      * and will allow to check whether the node still exist at mirror completion
3813      */
3814     mirror_start(job_id, bs, target,
3815                  has_replaces ? replaces : NULL, job_flags,
3816                  speed, granularity, buf_size, sync, backing_mode,
3817                  on_source_error, on_target_error, unmap, filter_node_name,
3818                  copy_mode, errp);
3819 }
3820 
3821 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3822 {
3823     BlockDriverState *bs;
3824     BlockDriverState *source, *target_bs;
3825     AioContext *aio_context;
3826     BlockMirrorBackingMode backing_mode;
3827     Error *local_err = NULL;
3828     QDict *options = NULL;
3829     int flags;
3830     int64_t size;
3831     const char *format = arg->format;
3832     int ret;
3833 
3834     bs = qmp_get_root_bs(arg->device, errp);
3835     if (!bs) {
3836         return;
3837     }
3838 
3839     /* Early check to avoid creating target */
3840     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3841         return;
3842     }
3843 
3844     aio_context = bdrv_get_aio_context(bs);
3845     aio_context_acquire(aio_context);
3846 
3847     if (!arg->has_mode) {
3848         arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3849     }
3850 
3851     if (!arg->has_format) {
3852         format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3853                   ? NULL : bs->drv->format_name);
3854     }
3855 
3856     flags = bs->open_flags | BDRV_O_RDWR;
3857     source = backing_bs(bs);
3858     if (!source && arg->sync == MIRROR_SYNC_MODE_TOP) {
3859         arg->sync = MIRROR_SYNC_MODE_FULL;
3860     }
3861     if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3862         source = bs;
3863     }
3864 
3865     size = bdrv_getlength(bs);
3866     if (size < 0) {
3867         error_setg_errno(errp, -size, "bdrv_getlength failed");
3868         goto out;
3869     }
3870 
3871     if (arg->has_replaces) {
3872         if (!arg->has_node_name) {
3873             error_setg(errp, "a node-name must be provided when replacing a"
3874                              " named node of the graph");
3875             goto out;
3876         }
3877     }
3878 
3879     if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3880         backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3881     } else {
3882         backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3883     }
3884 
3885     /* Don't open backing image in create() */
3886     flags |= BDRV_O_NO_BACKING;
3887 
3888     if ((arg->sync == MIRROR_SYNC_MODE_FULL || !source)
3889         && arg->mode != NEW_IMAGE_MODE_EXISTING)
3890     {
3891         /* create new image w/o backing file */
3892         assert(format);
3893         bdrv_img_create(arg->target, format,
3894                         NULL, NULL, NULL, size, flags, false, &local_err);
3895     } else {
3896         switch (arg->mode) {
3897         case NEW_IMAGE_MODE_EXISTING:
3898             break;
3899         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3900             /* create new image with backing file */
3901             bdrv_refresh_filename(source);
3902             bdrv_img_create(arg->target, format,
3903                             source->filename,
3904                             source->drv->format_name,
3905                             NULL, size, flags, false, &local_err);
3906             break;
3907         default:
3908             abort();
3909         }
3910     }
3911 
3912     if (local_err) {
3913         error_propagate(errp, local_err);
3914         goto out;
3915     }
3916 
3917     options = qdict_new();
3918     if (arg->has_node_name) {
3919         qdict_put_str(options, "node-name", arg->node_name);
3920     }
3921     if (format) {
3922         qdict_put_str(options, "driver", format);
3923     }
3924 
3925     /* Mirroring takes care of copy-on-write using the source's backing
3926      * file.
3927      */
3928     target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3929     if (!target_bs) {
3930         goto out;
3931     }
3932 
3933     ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3934     if (ret < 0) {
3935         bdrv_unref(target_bs);
3936         goto out;
3937     }
3938 
3939     blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3940                            arg->has_replaces, arg->replaces, arg->sync,
3941                            backing_mode, arg->has_speed, arg->speed,
3942                            arg->has_granularity, arg->granularity,
3943                            arg->has_buf_size, arg->buf_size,
3944                            arg->has_on_source_error, arg->on_source_error,
3945                            arg->has_on_target_error, arg->on_target_error,
3946                            arg->has_unmap, arg->unmap,
3947                            false, NULL,
3948                            arg->has_copy_mode, arg->copy_mode,
3949                            arg->has_auto_finalize, arg->auto_finalize,
3950                            arg->has_auto_dismiss, arg->auto_dismiss,
3951                            &local_err);
3952     bdrv_unref(target_bs);
3953     error_propagate(errp, local_err);
3954 out:
3955     aio_context_release(aio_context);
3956 }
3957 
3958 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3959                          const char *device, const char *target,
3960                          bool has_replaces, const char *replaces,
3961                          MirrorSyncMode sync,
3962                          bool has_speed, int64_t speed,
3963                          bool has_granularity, uint32_t granularity,
3964                          bool has_buf_size, int64_t buf_size,
3965                          bool has_on_source_error,
3966                          BlockdevOnError on_source_error,
3967                          bool has_on_target_error,
3968                          BlockdevOnError on_target_error,
3969                          bool has_filter_node_name,
3970                          const char *filter_node_name,
3971                          bool has_copy_mode, MirrorCopyMode copy_mode,
3972                          bool has_auto_finalize, bool auto_finalize,
3973                          bool has_auto_dismiss, bool auto_dismiss,
3974                          Error **errp)
3975 {
3976     BlockDriverState *bs;
3977     BlockDriverState *target_bs;
3978     AioContext *aio_context;
3979     BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3980     Error *local_err = NULL;
3981     int ret;
3982 
3983     bs = qmp_get_root_bs(device, errp);
3984     if (!bs) {
3985         return;
3986     }
3987 
3988     target_bs = bdrv_lookup_bs(target, target, errp);
3989     if (!target_bs) {
3990         return;
3991     }
3992 
3993     aio_context = bdrv_get_aio_context(bs);
3994     aio_context_acquire(aio_context);
3995 
3996     ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3997     if (ret < 0) {
3998         goto out;
3999     }
4000 
4001     blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
4002                            has_replaces, replaces, sync, backing_mode,
4003                            has_speed, speed,
4004                            has_granularity, granularity,
4005                            has_buf_size, buf_size,
4006                            has_on_source_error, on_source_error,
4007                            has_on_target_error, on_target_error,
4008                            true, true,
4009                            has_filter_node_name, filter_node_name,
4010                            has_copy_mode, copy_mode,
4011                            has_auto_finalize, auto_finalize,
4012                            has_auto_dismiss, auto_dismiss,
4013                            &local_err);
4014     error_propagate(errp, local_err);
4015 out:
4016     aio_context_release(aio_context);
4017 }
4018 
4019 /* Get a block job using its ID and acquire its AioContext */
4020 static BlockJob *find_block_job(const char *id, AioContext **aio_context,
4021                                 Error **errp)
4022 {
4023     BlockJob *job;
4024 
4025     assert(id != NULL);
4026 
4027     *aio_context = NULL;
4028 
4029     job = block_job_get(id);
4030 
4031     if (!job) {
4032         error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
4033                   "Block job '%s' not found", id);
4034         return NULL;
4035     }
4036 
4037     *aio_context = blk_get_aio_context(job->blk);
4038     aio_context_acquire(*aio_context);
4039 
4040     return job;
4041 }
4042 
4043 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
4044 {
4045     AioContext *aio_context;
4046     BlockJob *job = find_block_job(device, &aio_context, errp);
4047 
4048     if (!job) {
4049         return;
4050     }
4051 
4052     block_job_set_speed(job, speed, errp);
4053     aio_context_release(aio_context);
4054 }
4055 
4056 void qmp_block_job_cancel(const char *device,
4057                           bool has_force, bool force, Error **errp)
4058 {
4059     AioContext *aio_context;
4060     BlockJob *job = find_block_job(device, &aio_context, errp);
4061 
4062     if (!job) {
4063         return;
4064     }
4065 
4066     if (!has_force) {
4067         force = false;
4068     }
4069 
4070     if (job_user_paused(&job->job) && !force) {
4071         error_setg(errp, "The block job for device '%s' is currently paused",
4072                    device);
4073         goto out;
4074     }
4075 
4076     trace_qmp_block_job_cancel(job);
4077     job_user_cancel(&job->job, force, errp);
4078 out:
4079     aio_context_release(aio_context);
4080 }
4081 
4082 void qmp_block_job_pause(const char *device, Error **errp)
4083 {
4084     AioContext *aio_context;
4085     BlockJob *job = find_block_job(device, &aio_context, errp);
4086 
4087     if (!job) {
4088         return;
4089     }
4090 
4091     trace_qmp_block_job_pause(job);
4092     job_user_pause(&job->job, errp);
4093     aio_context_release(aio_context);
4094 }
4095 
4096 void qmp_block_job_resume(const char *device, Error **errp)
4097 {
4098     AioContext *aio_context;
4099     BlockJob *job = find_block_job(device, &aio_context, errp);
4100 
4101     if (!job) {
4102         return;
4103     }
4104 
4105     trace_qmp_block_job_resume(job);
4106     job_user_resume(&job->job, errp);
4107     aio_context_release(aio_context);
4108 }
4109 
4110 void qmp_block_job_complete(const char *device, Error **errp)
4111 {
4112     AioContext *aio_context;
4113     BlockJob *job = find_block_job(device, &aio_context, errp);
4114 
4115     if (!job) {
4116         return;
4117     }
4118 
4119     trace_qmp_block_job_complete(job);
4120     job_complete(&job->job, errp);
4121     aio_context_release(aio_context);
4122 }
4123 
4124 void qmp_block_job_finalize(const char *id, Error **errp)
4125 {
4126     AioContext *aio_context;
4127     BlockJob *job = find_block_job(id, &aio_context, errp);
4128 
4129     if (!job) {
4130         return;
4131     }
4132 
4133     trace_qmp_block_job_finalize(job);
4134     job_finalize(&job->job, errp);
4135     aio_context_release(aio_context);
4136 }
4137 
4138 void qmp_block_job_dismiss(const char *id, Error **errp)
4139 {
4140     AioContext *aio_context;
4141     BlockJob *bjob = find_block_job(id, &aio_context, errp);
4142     Job *job;
4143 
4144     if (!bjob) {
4145         return;
4146     }
4147 
4148     trace_qmp_block_job_dismiss(bjob);
4149     job = &bjob->job;
4150     job_dismiss(&job, errp);
4151     aio_context_release(aio_context);
4152 }
4153 
4154 void qmp_change_backing_file(const char *device,
4155                              const char *image_node_name,
4156                              const char *backing_file,
4157                              Error **errp)
4158 {
4159     BlockDriverState *bs = NULL;
4160     AioContext *aio_context;
4161     BlockDriverState *image_bs = NULL;
4162     Error *local_err = NULL;
4163     bool ro;
4164     int ret;
4165 
4166     bs = qmp_get_root_bs(device, errp);
4167     if (!bs) {
4168         return;
4169     }
4170 
4171     aio_context = bdrv_get_aio_context(bs);
4172     aio_context_acquire(aio_context);
4173 
4174     image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
4175     if (local_err) {
4176         error_propagate(errp, local_err);
4177         goto out;
4178     }
4179 
4180     if (!image_bs) {
4181         error_setg(errp, "image file not found");
4182         goto out;
4183     }
4184 
4185     if (bdrv_find_base(image_bs) == image_bs) {
4186         error_setg(errp, "not allowing backing file change on an image "
4187                          "without a backing file");
4188         goto out;
4189     }
4190 
4191     /* even though we are not necessarily operating on bs, we need it to
4192      * determine if block ops are currently prohibited on the chain */
4193     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
4194         goto out;
4195     }
4196 
4197     /* final sanity check */
4198     if (!bdrv_chain_contains(bs, image_bs)) {
4199         error_setg(errp, "'%s' and image file are not in the same chain",
4200                    device);
4201         goto out;
4202     }
4203 
4204     /* if not r/w, reopen to make r/w */
4205     ro = bdrv_is_read_only(image_bs);
4206 
4207     if (ro) {
4208         if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
4209             goto out;
4210         }
4211     }
4212 
4213     ret = bdrv_change_backing_file(image_bs, backing_file,
4214                                image_bs->drv ? image_bs->drv->format_name : "");
4215 
4216     if (ret < 0) {
4217         error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
4218                          backing_file);
4219         /* don't exit here, so we can try to restore open flags if
4220          * appropriate */
4221     }
4222 
4223     if (ro) {
4224         bdrv_reopen_set_read_only(image_bs, true, &local_err);
4225         error_propagate(errp, local_err);
4226     }
4227 
4228 out:
4229     aio_context_release(aio_context);
4230 }
4231 
4232 void hmp_drive_add_node(Monitor *mon, const char *optstr)
4233 {
4234     QemuOpts *opts;
4235     QDict *qdict;
4236     Error *local_err = NULL;
4237 
4238     opts = qemu_opts_parse_noisily(&qemu_drive_opts, optstr, false);
4239     if (!opts) {
4240         return;
4241     }
4242 
4243     qdict = qemu_opts_to_qdict(opts, NULL);
4244 
4245     if (!qdict_get_try_str(qdict, "node-name")) {
4246         qobject_unref(qdict);
4247         error_report("'node-name' needs to be specified");
4248         goto out;
4249     }
4250 
4251     BlockDriverState *bs = bds_tree_init(qdict, &local_err);
4252     if (!bs) {
4253         error_report_err(local_err);
4254         goto out;
4255     }
4256 
4257     QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
4258 
4259 out:
4260     qemu_opts_del(opts);
4261 }
4262 
4263 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
4264 {
4265     BlockDriverState *bs;
4266     QObject *obj;
4267     Visitor *v = qobject_output_visitor_new(&obj);
4268     QDict *qdict;
4269     Error *local_err = NULL;
4270 
4271     visit_type_BlockdevOptions(v, NULL, &options, &local_err);
4272     if (local_err) {
4273         error_propagate(errp, local_err);
4274         goto fail;
4275     }
4276 
4277     visit_complete(v, &obj);
4278     qdict = qobject_to(QDict, obj);
4279 
4280     qdict_flatten(qdict);
4281 
4282     if (!qdict_get_try_str(qdict, "node-name")) {
4283         error_setg(errp, "'node-name' must be specified for the root node");
4284         goto fail;
4285     }
4286 
4287     bs = bds_tree_init(qdict, errp);
4288     if (!bs) {
4289         goto fail;
4290     }
4291 
4292     QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
4293 
4294 fail:
4295     visit_free(v);
4296 }
4297 
4298 void qmp_x_blockdev_reopen(BlockdevOptions *options, Error **errp)
4299 {
4300     BlockDriverState *bs;
4301     AioContext *ctx;
4302     QObject *obj;
4303     Visitor *v = qobject_output_visitor_new(&obj);
4304     Error *local_err = NULL;
4305     BlockReopenQueue *queue;
4306     QDict *qdict;
4307 
4308     /* Check for the selected node name */
4309     if (!options->has_node_name) {
4310         error_setg(errp, "Node name not specified");
4311         goto fail;
4312     }
4313 
4314     bs = bdrv_find_node(options->node_name);
4315     if (!bs) {
4316         error_setg(errp, "Cannot find node named '%s'", options->node_name);
4317         goto fail;
4318     }
4319 
4320     /* Put all options in a QDict and flatten it */
4321     visit_type_BlockdevOptions(v, NULL, &options, &local_err);
4322     if (local_err) {
4323         error_propagate(errp, local_err);
4324         goto fail;
4325     }
4326 
4327     visit_complete(v, &obj);
4328     qdict = qobject_to(QDict, obj);
4329 
4330     qdict_flatten(qdict);
4331 
4332     /* Perform the reopen operation */
4333     ctx = bdrv_get_aio_context(bs);
4334     aio_context_acquire(ctx);
4335     bdrv_subtree_drained_begin(bs);
4336     queue = bdrv_reopen_queue(NULL, bs, qdict, false);
4337     bdrv_reopen_multiple(queue, errp);
4338     bdrv_subtree_drained_end(bs);
4339     aio_context_release(ctx);
4340 
4341 fail:
4342     visit_free(v);
4343 }
4344 
4345 void qmp_blockdev_del(const char *node_name, Error **errp)
4346 {
4347     AioContext *aio_context;
4348     BlockDriverState *bs;
4349 
4350     bs = bdrv_find_node(node_name);
4351     if (!bs) {
4352         error_setg(errp, "Cannot find node %s", node_name);
4353         return;
4354     }
4355     if (bdrv_has_blk(bs)) {
4356         error_setg(errp, "Node %s is in use", node_name);
4357         return;
4358     }
4359     aio_context = bdrv_get_aio_context(bs);
4360     aio_context_acquire(aio_context);
4361 
4362     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
4363         goto out;
4364     }
4365 
4366     if (!QTAILQ_IN_USE(bs, monitor_list)) {
4367         error_setg(errp, "Node %s is not owned by the monitor",
4368                    bs->node_name);
4369         goto out;
4370     }
4371 
4372     if (bs->refcnt > 1) {
4373         error_setg(errp, "Block device %s is in use",
4374                    bdrv_get_device_or_node_name(bs));
4375         goto out;
4376     }
4377 
4378     QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
4379     bdrv_unref(bs);
4380 
4381 out:
4382     aio_context_release(aio_context);
4383 }
4384 
4385 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
4386                                   const char *child_name)
4387 {
4388     BdrvChild *child;
4389 
4390     QLIST_FOREACH(child, &parent_bs->children, next) {
4391         if (strcmp(child->name, child_name) == 0) {
4392             return child;
4393         }
4394     }
4395 
4396     return NULL;
4397 }
4398 
4399 void qmp_x_blockdev_change(const char *parent, bool has_child,
4400                            const char *child, bool has_node,
4401                            const char *node, Error **errp)
4402 {
4403     BlockDriverState *parent_bs, *new_bs = NULL;
4404     BdrvChild *p_child;
4405 
4406     parent_bs = bdrv_lookup_bs(parent, parent, errp);
4407     if (!parent_bs) {
4408         return;
4409     }
4410 
4411     if (has_child == has_node) {
4412         if (has_child) {
4413             error_setg(errp, "The parameters child and node are in conflict");
4414         } else {
4415             error_setg(errp, "Either child or node must be specified");
4416         }
4417         return;
4418     }
4419 
4420     if (has_child) {
4421         p_child = bdrv_find_child(parent_bs, child);
4422         if (!p_child) {
4423             error_setg(errp, "Node '%s' does not have child '%s'",
4424                        parent, child);
4425             return;
4426         }
4427         bdrv_del_child(parent_bs, p_child, errp);
4428     }
4429 
4430     if (has_node) {
4431         new_bs = bdrv_find_node(node);
4432         if (!new_bs) {
4433             error_setg(errp, "Node '%s' not found", node);
4434             return;
4435         }
4436         bdrv_add_child(parent_bs, new_bs, errp);
4437     }
4438 }
4439 
4440 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
4441 {
4442     BlockJobInfoList *head = NULL, **p_next = &head;
4443     BlockJob *job;
4444 
4445     for (job = block_job_next(NULL); job; job = block_job_next(job)) {
4446         BlockJobInfoList *elem;
4447         AioContext *aio_context;
4448 
4449         if (block_job_is_internal(job)) {
4450             continue;
4451         }
4452         elem = g_new0(BlockJobInfoList, 1);
4453         aio_context = blk_get_aio_context(job->blk);
4454         aio_context_acquire(aio_context);
4455         elem->value = block_job_query(job, errp);
4456         aio_context_release(aio_context);
4457         if (!elem->value) {
4458             g_free(elem);
4459             qapi_free_BlockJobInfoList(head);
4460             return NULL;
4461         }
4462         *p_next = elem;
4463         p_next = &elem->next;
4464     }
4465 
4466     return head;
4467 }
4468 
4469 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
4470                                  bool has_force, bool force, Error **errp)
4471 {
4472     AioContext *old_context;
4473     AioContext *new_context;
4474     BlockDriverState *bs;
4475 
4476     bs = bdrv_find_node(node_name);
4477     if (!bs) {
4478         error_setg(errp, "Cannot find node %s", node_name);
4479         return;
4480     }
4481 
4482     /* Protects against accidents. */
4483     if (!(has_force && force) && bdrv_has_blk(bs)) {
4484         error_setg(errp, "Node %s is associated with a BlockBackend and could "
4485                          "be in use (use force=true to override this check)",
4486                          node_name);
4487         return;
4488     }
4489 
4490     if (iothread->type == QTYPE_QSTRING) {
4491         IOThread *obj = iothread_by_id(iothread->u.s);
4492         if (!obj) {
4493             error_setg(errp, "Cannot find iothread %s", iothread->u.s);
4494             return;
4495         }
4496 
4497         new_context = iothread_get_aio_context(obj);
4498     } else {
4499         new_context = qemu_get_aio_context();
4500     }
4501 
4502     old_context = bdrv_get_aio_context(bs);
4503     aio_context_acquire(old_context);
4504 
4505     bdrv_try_set_aio_context(bs, new_context, errp);
4506 
4507     aio_context_release(old_context);
4508 }
4509 
4510 void qmp_block_latency_histogram_set(
4511     const char *id,
4512     bool has_boundaries, uint64List *boundaries,
4513     bool has_boundaries_read, uint64List *boundaries_read,
4514     bool has_boundaries_write, uint64List *boundaries_write,
4515     bool has_boundaries_flush, uint64List *boundaries_flush,
4516     Error **errp)
4517 {
4518     BlockBackend *blk = qmp_get_blk(NULL, id, errp);
4519     BlockAcctStats *stats;
4520     int ret;
4521 
4522     if (!blk) {
4523         return;
4524     }
4525 
4526     stats = blk_get_stats(blk);
4527 
4528     if (!has_boundaries && !has_boundaries_read && !has_boundaries_write &&
4529         !has_boundaries_flush)
4530     {
4531         block_latency_histograms_clear(stats);
4532         return;
4533     }
4534 
4535     if (has_boundaries || has_boundaries_read) {
4536         ret = block_latency_histogram_set(
4537             stats, BLOCK_ACCT_READ,
4538             has_boundaries_read ? boundaries_read : boundaries);
4539         if (ret) {
4540             error_setg(errp, "Device '%s' set read boundaries fail", id);
4541             return;
4542         }
4543     }
4544 
4545     if (has_boundaries || has_boundaries_write) {
4546         ret = block_latency_histogram_set(
4547             stats, BLOCK_ACCT_WRITE,
4548             has_boundaries_write ? boundaries_write : boundaries);
4549         if (ret) {
4550             error_setg(errp, "Device '%s' set write boundaries fail", id);
4551             return;
4552         }
4553     }
4554 
4555     if (has_boundaries || has_boundaries_flush) {
4556         ret = block_latency_histogram_set(
4557             stats, BLOCK_ACCT_FLUSH,
4558             has_boundaries_flush ? boundaries_flush : boundaries);
4559         if (ret) {
4560             error_setg(errp, "Device '%s' set flush boundaries fail", id);
4561             return;
4562         }
4563     }
4564 }
4565 
4566 QemuOptsList qemu_common_drive_opts = {
4567     .name = "drive",
4568     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
4569     .desc = {
4570         {
4571             .name = "snapshot",
4572             .type = QEMU_OPT_BOOL,
4573             .help = "enable/disable snapshot mode",
4574         },{
4575             .name = "aio",
4576             .type = QEMU_OPT_STRING,
4577             .help = "host AIO implementation (threads, native)",
4578         },{
4579             .name = BDRV_OPT_CACHE_WB,
4580             .type = QEMU_OPT_BOOL,
4581             .help = "Enable writeback mode",
4582         },{
4583             .name = "format",
4584             .type = QEMU_OPT_STRING,
4585             .help = "disk format (raw, qcow2, ...)",
4586         },{
4587             .name = "rerror",
4588             .type = QEMU_OPT_STRING,
4589             .help = "read error action",
4590         },{
4591             .name = "werror",
4592             .type = QEMU_OPT_STRING,
4593             .help = "write error action",
4594         },{
4595             .name = BDRV_OPT_READ_ONLY,
4596             .type = QEMU_OPT_BOOL,
4597             .help = "open drive file as read-only",
4598         },
4599 
4600         THROTTLE_OPTS,
4601 
4602         {
4603             .name = "throttling.group",
4604             .type = QEMU_OPT_STRING,
4605             .help = "name of the block throttling group",
4606         },{
4607             .name = "copy-on-read",
4608             .type = QEMU_OPT_BOOL,
4609             .help = "copy read data from backing file into image file",
4610         },{
4611             .name = "detect-zeroes",
4612             .type = QEMU_OPT_STRING,
4613             .help = "try to optimize zero writes (off, on, unmap)",
4614         },{
4615             .name = "stats-account-invalid",
4616             .type = QEMU_OPT_BOOL,
4617             .help = "whether to account for invalid I/O operations "
4618                     "in the statistics",
4619         },{
4620             .name = "stats-account-failed",
4621             .type = QEMU_OPT_BOOL,
4622             .help = "whether to account for failed I/O operations "
4623                     "in the statistics",
4624         },
4625         { /* end of list */ }
4626     },
4627 };
4628 
4629 QemuOptsList qemu_drive_opts = {
4630     .name = "drive",
4631     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
4632     .desc = {
4633         /*
4634          * no elements => accept any params
4635          * validation will happen later
4636          */
4637         { /* end of list */ }
4638     },
4639 };
4640