xref: /openbmc/qemu/blockdev.c (revision 0a9516c2)
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 "sysemu/block-backend.h"
34 #include "sysemu/blockdev.h"
35 #include "hw/block/block.h"
36 #include "block/blockjob.h"
37 #include "block/throttle-groups.h"
38 #include "monitor/monitor.h"
39 #include "qemu/error-report.h"
40 #include "qemu/option.h"
41 #include "qemu/config-file.h"
42 #include "qapi/qmp/types.h"
43 #include "qapi-visit.h"
44 #include "qapi/qmp/qerror.h"
45 #include "qapi/qmp-output-visitor.h"
46 #include "qapi/util.h"
47 #include "sysemu/sysemu.h"
48 #include "block/block_int.h"
49 #include "qmp-commands.h"
50 #include "trace.h"
51 #include "sysemu/arch_init.h"
52 
53 static const char *const if_name[IF_COUNT] = {
54     [IF_NONE] = "none",
55     [IF_IDE] = "ide",
56     [IF_SCSI] = "scsi",
57     [IF_FLOPPY] = "floppy",
58     [IF_PFLASH] = "pflash",
59     [IF_MTD] = "mtd",
60     [IF_SD] = "sd",
61     [IF_VIRTIO] = "virtio",
62     [IF_XEN] = "xen",
63 };
64 
65 static int if_max_devs[IF_COUNT] = {
66     /*
67      * Do not change these numbers!  They govern how drive option
68      * index maps to unit and bus.  That mapping is ABI.
69      *
70      * All controllers used to imlement if=T drives need to support
71      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
72      * Otherwise, some index values map to "impossible" bus, unit
73      * values.
74      *
75      * For instance, if you change [IF_SCSI] to 255, -drive
76      * if=scsi,index=12 no longer means bus=1,unit=5, but
77      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
78      * the drive can't be set up.  Regression.
79      */
80     [IF_IDE] = 2,
81     [IF_SCSI] = 7,
82 };
83 
84 /**
85  * Boards may call this to offer board-by-board overrides
86  * of the default, global values.
87  */
88 void override_max_devs(BlockInterfaceType type, int max_devs)
89 {
90     BlockBackend *blk;
91     DriveInfo *dinfo;
92 
93     if (max_devs <= 0) {
94         return;
95     }
96 
97     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
98         dinfo = blk_legacy_dinfo(blk);
99         if (dinfo->type == type) {
100             fprintf(stderr, "Cannot override units-per-bus property of"
101                     " the %s interface, because a drive of that type has"
102                     " already been added.\n", if_name[type]);
103             g_assert_not_reached();
104         }
105     }
106 
107     if_max_devs[type] = max_devs;
108 }
109 
110 /*
111  * We automatically delete the drive when a device using it gets
112  * unplugged.  Questionable feature, but we can't just drop it.
113  * Device models call blockdev_mark_auto_del() to schedule the
114  * automatic deletion, and generic qdev code calls blockdev_auto_del()
115  * when deletion is actually safe.
116  */
117 void blockdev_mark_auto_del(BlockBackend *blk)
118 {
119     DriveInfo *dinfo = blk_legacy_dinfo(blk);
120     BlockDriverState *bs = blk_bs(blk);
121     AioContext *aio_context;
122 
123     if (!dinfo) {
124         return;
125     }
126 
127     if (bs) {
128         aio_context = bdrv_get_aio_context(bs);
129         aio_context_acquire(aio_context);
130 
131         if (bs->job) {
132             block_job_cancel(bs->job);
133         }
134 
135         aio_context_release(aio_context);
136     }
137 
138     dinfo->auto_del = 1;
139 }
140 
141 void blockdev_auto_del(BlockBackend *blk)
142 {
143     DriveInfo *dinfo = blk_legacy_dinfo(blk);
144 
145     if (dinfo && dinfo->auto_del) {
146         blk_unref(blk);
147     }
148 }
149 
150 /**
151  * Returns the current mapping of how many units per bus
152  * a particular interface can support.
153  *
154  *  A positive integer indicates n units per bus.
155  *  0 implies the mapping has not been established.
156  * -1 indicates an invalid BlockInterfaceType was given.
157  */
158 int drive_get_max_devs(BlockInterfaceType type)
159 {
160     if (type >= IF_IDE && type < IF_COUNT) {
161         return if_max_devs[type];
162     }
163 
164     return -1;
165 }
166 
167 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
168 {
169     int max_devs = if_max_devs[type];
170     return max_devs ? index / max_devs : 0;
171 }
172 
173 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
174 {
175     int max_devs = if_max_devs[type];
176     return max_devs ? index % max_devs : index;
177 }
178 
179 QemuOpts *drive_def(const char *optstr)
180 {
181     return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
182 }
183 
184 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
185                     const char *optstr)
186 {
187     QemuOpts *opts;
188 
189     opts = drive_def(optstr);
190     if (!opts) {
191         return NULL;
192     }
193     if (type != IF_DEFAULT) {
194         qemu_opt_set(opts, "if", if_name[type], &error_abort);
195     }
196     if (index >= 0) {
197         qemu_opt_set_number(opts, "index", index, &error_abort);
198     }
199     if (file)
200         qemu_opt_set(opts, "file", file, &error_abort);
201     return opts;
202 }
203 
204 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
205 {
206     BlockBackend *blk;
207     DriveInfo *dinfo;
208 
209     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
210         dinfo = blk_legacy_dinfo(blk);
211         if (dinfo && dinfo->type == type
212             && dinfo->bus == bus && dinfo->unit == unit) {
213             return dinfo;
214         }
215     }
216 
217     return NULL;
218 }
219 
220 bool drive_check_orphaned(void)
221 {
222     BlockBackend *blk;
223     DriveInfo *dinfo;
224     bool rs = false;
225 
226     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
227         dinfo = blk_legacy_dinfo(blk);
228         /* If dinfo->bdrv->dev is NULL, it has no device attached. */
229         /* Unless this is a default drive, this may be an oversight. */
230         if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
231             dinfo->type != IF_NONE) {
232             fprintf(stderr, "Warning: Orphaned drive without device: "
233                     "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
234                     blk_name(blk), blk_bs(blk) ? blk_bs(blk)->filename : "",
235                     if_name[dinfo->type], dinfo->bus, dinfo->unit);
236             rs = true;
237         }
238     }
239 
240     return rs;
241 }
242 
243 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
244 {
245     return drive_get(type,
246                      drive_index_to_bus_id(type, index),
247                      drive_index_to_unit_id(type, index));
248 }
249 
250 int drive_get_max_bus(BlockInterfaceType type)
251 {
252     int max_bus;
253     BlockBackend *blk;
254     DriveInfo *dinfo;
255 
256     max_bus = -1;
257     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
258         dinfo = blk_legacy_dinfo(blk);
259         if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
260             max_bus = dinfo->bus;
261         }
262     }
263     return max_bus;
264 }
265 
266 /* Get a block device.  This should only be used for single-drive devices
267    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
268    appropriate bus.  */
269 DriveInfo *drive_get_next(BlockInterfaceType type)
270 {
271     static int next_block_unit[IF_COUNT];
272 
273     return drive_get(type, 0, next_block_unit[type]++);
274 }
275 
276 static void bdrv_format_print(void *opaque, const char *name)
277 {
278     error_printf(" %s", name);
279 }
280 
281 typedef struct {
282     QEMUBH *bh;
283     BlockDriverState *bs;
284 } BDRVPutRefBH;
285 
286 static void bdrv_put_ref_bh(void *opaque)
287 {
288     BDRVPutRefBH *s = opaque;
289 
290     bdrv_unref(s->bs);
291     qemu_bh_delete(s->bh);
292     g_free(s);
293 }
294 
295 /*
296  * Release a BDS reference in a BH
297  *
298  * It is not safe to use bdrv_unref() from a callback function when the callers
299  * still need the BlockDriverState.  In such cases we schedule a BH to release
300  * the reference.
301  */
302 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
303 {
304     BDRVPutRefBH *s;
305 
306     s = g_new(BDRVPutRefBH, 1);
307     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
308     s->bs = bs;
309     qemu_bh_schedule(s->bh);
310 }
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 check_throttle_config(ThrottleConfig *cfg, Error **errp)
330 {
331     if (throttle_conflicting(cfg)) {
332         error_setg(errp, "bps/iops/max total values and read/write values"
333                          " cannot be used at the same time");
334         return false;
335     }
336 
337     if (!throttle_is_valid(cfg)) {
338         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
339         return false;
340     }
341 
342     if (throttle_max_is_missing_limit(cfg)) {
343         error_setg(errp, "bps_max/iops_max require corresponding"
344                          " bps/iops values");
345         return false;
346     }
347 
348     return true;
349 }
350 
351 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
352 
353 /* All parameters but @opts are optional and may be set to NULL. */
354 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
355     const char **throttling_group, ThrottleConfig *throttle_cfg,
356     BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
357 {
358     const char *discard;
359     Error *local_error = NULL;
360     const char *aio;
361 
362     if (bdrv_flags) {
363         if (!qemu_opt_get_bool(opts, "read-only", false)) {
364             *bdrv_flags |= BDRV_O_RDWR;
365         }
366         if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
367             *bdrv_flags |= BDRV_O_COPY_ON_READ;
368         }
369 
370         if ((discard = qemu_opt_get(opts, "discard")) != NULL) {
371             if (bdrv_parse_discard_flags(discard, bdrv_flags) != 0) {
372                 error_setg(errp, "Invalid discard option");
373                 return;
374             }
375         }
376 
377         if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true)) {
378             *bdrv_flags |= BDRV_O_CACHE_WB;
379         }
380         if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_DIRECT, false)) {
381             *bdrv_flags |= BDRV_O_NOCACHE;
382         }
383         if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) {
384             *bdrv_flags |= BDRV_O_NO_FLUSH;
385         }
386 
387         if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
388             if (!strcmp(aio, "native")) {
389                 *bdrv_flags |= BDRV_O_NATIVE_AIO;
390             } else if (!strcmp(aio, "threads")) {
391                 /* this is the default */
392             } else {
393                error_setg(errp, "invalid aio option");
394                return;
395             }
396         }
397     }
398 
399     /* disk I/O throttling */
400     if (throttling_group) {
401         *throttling_group = qemu_opt_get(opts, "throttling.group");
402     }
403 
404     if (throttle_cfg) {
405         memset(throttle_cfg, 0, sizeof(*throttle_cfg));
406         throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
407             qemu_opt_get_number(opts, "throttling.bps-total", 0);
408         throttle_cfg->buckets[THROTTLE_BPS_READ].avg  =
409             qemu_opt_get_number(opts, "throttling.bps-read", 0);
410         throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
411             qemu_opt_get_number(opts, "throttling.bps-write", 0);
412         throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
413             qemu_opt_get_number(opts, "throttling.iops-total", 0);
414         throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
415             qemu_opt_get_number(opts, "throttling.iops-read", 0);
416         throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
417             qemu_opt_get_number(opts, "throttling.iops-write", 0);
418 
419         throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
420             qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
421         throttle_cfg->buckets[THROTTLE_BPS_READ].max  =
422             qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
423         throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
424             qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
425         throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
426             qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
427         throttle_cfg->buckets[THROTTLE_OPS_READ].max =
428             qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
429         throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
430             qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
431 
432         throttle_cfg->op_size =
433             qemu_opt_get_number(opts, "throttling.iops-size", 0);
434 
435         if (!check_throttle_config(throttle_cfg, errp)) {
436             return;
437         }
438     }
439 
440     if (detect_zeroes) {
441         *detect_zeroes =
442             qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
443                             qemu_opt_get(opts, "detect-zeroes"),
444                             BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
445                             BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
446                             &local_error);
447         if (local_error) {
448             error_propagate(errp, local_error);
449             return;
450         }
451 
452         if (bdrv_flags &&
453             *detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
454             !(*bdrv_flags & BDRV_O_UNMAP))
455         {
456             error_setg(errp, "setting detect-zeroes to unmap is not allowed "
457                              "without setting discard operation to unmap");
458             return;
459         }
460     }
461 }
462 
463 /* Takes the ownership of bs_opts */
464 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
465                                    Error **errp)
466 {
467     const char *buf;
468     int bdrv_flags = 0;
469     int on_read_error, on_write_error;
470     BlockBackend *blk;
471     BlockDriverState *bs;
472     ThrottleConfig cfg;
473     int snapshot = 0;
474     Error *error = NULL;
475     QemuOpts *opts;
476     const char *id;
477     bool has_driver_specific_opts;
478     BlockdevDetectZeroesOptions detect_zeroes =
479         BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
480     const char *throttling_group = NULL;
481 
482     /* Check common options by copying from bs_opts to opts, all other options
483      * stay in bs_opts for processing by bdrv_open(). */
484     id = qdict_get_try_str(bs_opts, "id");
485     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
486     if (error) {
487         error_propagate(errp, error);
488         goto err_no_opts;
489     }
490 
491     qemu_opts_absorb_qdict(opts, bs_opts, &error);
492     if (error) {
493         error_propagate(errp, error);
494         goto early_err;
495     }
496 
497     if (id) {
498         qdict_del(bs_opts, "id");
499     }
500 
501     has_driver_specific_opts = !!qdict_size(bs_opts);
502 
503     /* extract parameters */
504     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
505 
506     extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
507                                     &detect_zeroes, &error);
508     if (error) {
509         error_propagate(errp, error);
510         goto early_err;
511     }
512 
513     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
514         if (is_help_option(buf)) {
515             error_printf("Supported formats:");
516             bdrv_iterate_format(bdrv_format_print, NULL);
517             error_printf("\n");
518             goto early_err;
519         }
520 
521         if (qdict_haskey(bs_opts, "driver")) {
522             error_setg(errp, "Cannot specify both 'driver' and 'format'");
523             goto early_err;
524         }
525         qdict_put(bs_opts, "driver", qstring_from_str(buf));
526     }
527 
528     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
529     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
530         on_write_error = parse_block_error_action(buf, 0, &error);
531         if (error) {
532             error_propagate(errp, error);
533             goto early_err;
534         }
535     }
536 
537     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
538     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
539         on_read_error = parse_block_error_action(buf, 1, &error);
540         if (error) {
541             error_propagate(errp, error);
542             goto early_err;
543         }
544     }
545 
546     if (snapshot) {
547         /* always use cache=unsafe with snapshot */
548         bdrv_flags &= ~BDRV_O_CACHE_MASK;
549         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
550     }
551 
552     /* init */
553     if ((!file || !*file) && !has_driver_specific_opts) {
554         BlockBackendRootState *blk_rs;
555 
556         blk = blk_new(qemu_opts_id(opts), errp);
557         if (!blk) {
558             goto early_err;
559         }
560 
561         blk_rs = blk_get_root_state(blk);
562         blk_rs->open_flags    = bdrv_flags;
563         blk_rs->read_only     = !(bdrv_flags & BDRV_O_RDWR);
564         blk_rs->detect_zeroes = detect_zeroes;
565 
566         if (throttle_enabled(&cfg)) {
567             if (!throttling_group) {
568                 throttling_group = blk_name(blk);
569             }
570             blk_rs->throttle_group = g_strdup(throttling_group);
571             blk_rs->throttle_state = throttle_group_incref(throttling_group);
572             blk_rs->throttle_state->cfg = cfg;
573         }
574 
575         QDECREF(bs_opts);
576     } else {
577         if (file && !*file) {
578             file = NULL;
579         }
580 
581         blk = blk_new_open(qemu_opts_id(opts), file, NULL, bs_opts, bdrv_flags,
582                            errp);
583         if (!blk) {
584             goto err_no_bs_opts;
585         }
586         bs = blk_bs(blk);
587 
588         bs->detect_zeroes = detect_zeroes;
589 
590         /* disk I/O throttling */
591         if (throttle_enabled(&cfg)) {
592             if (!throttling_group) {
593                 throttling_group = blk_name(blk);
594             }
595             bdrv_io_limits_enable(bs, throttling_group);
596             bdrv_set_io_limits(bs, &cfg);
597         }
598 
599         if (bdrv_key_required(bs)) {
600             autostart = 0;
601         }
602     }
603 
604     blk_set_on_error(blk, on_read_error, on_write_error);
605 
606 err_no_bs_opts:
607     qemu_opts_del(opts);
608     return blk;
609 
610 early_err:
611     qemu_opts_del(opts);
612 err_no_opts:
613     QDECREF(bs_opts);
614     return NULL;
615 }
616 
617 static QemuOptsList qemu_root_bds_opts;
618 
619 /* Takes the ownership of bs_opts */
620 static BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
621 {
622     BlockDriverState *bs;
623     QemuOpts *opts;
624     Error *local_error = NULL;
625     BlockdevDetectZeroesOptions detect_zeroes;
626     int ret;
627     int bdrv_flags = 0;
628 
629     opts = qemu_opts_create(&qemu_root_bds_opts, NULL, 1, errp);
630     if (!opts) {
631         goto fail;
632     }
633 
634     qemu_opts_absorb_qdict(opts, bs_opts, &local_error);
635     if (local_error) {
636         error_propagate(errp, local_error);
637         goto fail;
638     }
639 
640     extract_common_blockdev_options(opts, &bdrv_flags, NULL, NULL,
641                                     &detect_zeroes, &local_error);
642     if (local_error) {
643         error_propagate(errp, local_error);
644         goto fail;
645     }
646 
647     bs = NULL;
648     ret = bdrv_open(&bs, NULL, NULL, bs_opts, bdrv_flags, errp);
649     if (ret < 0) {
650         goto fail_no_bs_opts;
651     }
652 
653     bs->detect_zeroes = detect_zeroes;
654 
655 fail_no_bs_opts:
656     qemu_opts_del(opts);
657     return bs;
658 
659 fail:
660     qemu_opts_del(opts);
661     QDECREF(bs_opts);
662     return NULL;
663 }
664 
665 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
666                             Error **errp)
667 {
668     const char *value;
669 
670     value = qemu_opt_get(opts, from);
671     if (value) {
672         if (qemu_opt_find(opts, to)) {
673             error_setg(errp, "'%s' and its alias '%s' can't be used at the "
674                        "same time", to, from);
675             return;
676         }
677     }
678 
679     /* rename all items in opts */
680     while ((value = qemu_opt_get(opts, from))) {
681         qemu_opt_set(opts, to, value, &error_abort);
682         qemu_opt_unset(opts, from);
683     }
684 }
685 
686 QemuOptsList qemu_legacy_drive_opts = {
687     .name = "drive",
688     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
689     .desc = {
690         {
691             .name = "bus",
692             .type = QEMU_OPT_NUMBER,
693             .help = "bus number",
694         },{
695             .name = "unit",
696             .type = QEMU_OPT_NUMBER,
697             .help = "unit number (i.e. lun for scsi)",
698         },{
699             .name = "index",
700             .type = QEMU_OPT_NUMBER,
701             .help = "index number",
702         },{
703             .name = "media",
704             .type = QEMU_OPT_STRING,
705             .help = "media type (disk, cdrom)",
706         },{
707             .name = "if",
708             .type = QEMU_OPT_STRING,
709             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
710         },{
711             .name = "cyls",
712             .type = QEMU_OPT_NUMBER,
713             .help = "number of cylinders (ide disk geometry)",
714         },{
715             .name = "heads",
716             .type = QEMU_OPT_NUMBER,
717             .help = "number of heads (ide disk geometry)",
718         },{
719             .name = "secs",
720             .type = QEMU_OPT_NUMBER,
721             .help = "number of sectors (ide disk geometry)",
722         },{
723             .name = "trans",
724             .type = QEMU_OPT_STRING,
725             .help = "chs translation (auto, lba, none)",
726         },{
727             .name = "boot",
728             .type = QEMU_OPT_BOOL,
729             .help = "(deprecated, ignored)",
730         },{
731             .name = "addr",
732             .type = QEMU_OPT_STRING,
733             .help = "pci address (virtio only)",
734         },{
735             .name = "serial",
736             .type = QEMU_OPT_STRING,
737             .help = "disk serial number",
738         },{
739             .name = "file",
740             .type = QEMU_OPT_STRING,
741             .help = "file name",
742         },
743 
744         /* Options that are passed on, but have special semantics with -drive */
745         {
746             .name = "read-only",
747             .type = QEMU_OPT_BOOL,
748             .help = "open drive file as read-only",
749         },{
750             .name = "rerror",
751             .type = QEMU_OPT_STRING,
752             .help = "read error action",
753         },{
754             .name = "werror",
755             .type = QEMU_OPT_STRING,
756             .help = "write error action",
757         },{
758             .name = "copy-on-read",
759             .type = QEMU_OPT_BOOL,
760             .help = "copy read data from backing file into image file",
761         },
762 
763         { /* end of list */ }
764     },
765 };
766 
767 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
768 {
769     const char *value;
770     BlockBackend *blk;
771     DriveInfo *dinfo = NULL;
772     QDict *bs_opts;
773     QemuOpts *legacy_opts;
774     DriveMediaType media = MEDIA_DISK;
775     BlockInterfaceType type;
776     int cyls, heads, secs, translation;
777     int max_devs, bus_id, unit_id, index;
778     const char *devaddr;
779     const char *werror, *rerror;
780     bool read_only = false;
781     bool copy_on_read;
782     const char *serial;
783     const char *filename;
784     Error *local_err = NULL;
785     int i;
786 
787     /* Change legacy command line options into QMP ones */
788     static const struct {
789         const char *from;
790         const char *to;
791     } opt_renames[] = {
792         { "iops",           "throttling.iops-total" },
793         { "iops_rd",        "throttling.iops-read" },
794         { "iops_wr",        "throttling.iops-write" },
795 
796         { "bps",            "throttling.bps-total" },
797         { "bps_rd",         "throttling.bps-read" },
798         { "bps_wr",         "throttling.bps-write" },
799 
800         { "iops_max",       "throttling.iops-total-max" },
801         { "iops_rd_max",    "throttling.iops-read-max" },
802         { "iops_wr_max",    "throttling.iops-write-max" },
803 
804         { "bps_max",        "throttling.bps-total-max" },
805         { "bps_rd_max",     "throttling.bps-read-max" },
806         { "bps_wr_max",     "throttling.bps-write-max" },
807 
808         { "iops_size",      "throttling.iops-size" },
809 
810         { "group",          "throttling.group" },
811 
812         { "readonly",       "read-only" },
813     };
814 
815     for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
816         qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
817                         &local_err);
818         if (local_err) {
819             error_report_err(local_err);
820             return NULL;
821         }
822     }
823 
824     value = qemu_opt_get(all_opts, "cache");
825     if (value) {
826         int flags = 0;
827 
828         if (bdrv_parse_cache_flags(value, &flags) != 0) {
829             error_report("invalid cache option");
830             return NULL;
831         }
832 
833         /* Specific options take precedence */
834         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
835             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
836                               !!(flags & BDRV_O_CACHE_WB), &error_abort);
837         }
838         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
839             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
840                               !!(flags & BDRV_O_NOCACHE), &error_abort);
841         }
842         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
843             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
844                               !!(flags & BDRV_O_NO_FLUSH), &error_abort);
845         }
846         qemu_opt_unset(all_opts, "cache");
847     }
848 
849     /* Get a QDict for processing the options */
850     bs_opts = qdict_new();
851     qemu_opts_to_qdict(all_opts, bs_opts);
852 
853     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
854                                    &error_abort);
855     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
856     if (local_err) {
857         error_report_err(local_err);
858         goto fail;
859     }
860 
861     /* Deprecated option boot=[on|off] */
862     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
863         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
864                 "ignored. Future versions will reject this parameter. Please "
865                 "update your scripts.\n");
866     }
867 
868     /* Media type */
869     value = qemu_opt_get(legacy_opts, "media");
870     if (value) {
871         if (!strcmp(value, "disk")) {
872             media = MEDIA_DISK;
873         } else if (!strcmp(value, "cdrom")) {
874             media = MEDIA_CDROM;
875             read_only = true;
876         } else {
877             error_report("'%s' invalid media", value);
878             goto fail;
879         }
880     }
881 
882     /* copy-on-read is disabled with a warning for read-only devices */
883     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
884     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
885 
886     if (read_only && copy_on_read) {
887         error_report("warning: disabling copy-on-read on read-only drive");
888         copy_on_read = false;
889     }
890 
891     qdict_put(bs_opts, "read-only",
892               qstring_from_str(read_only ? "on" : "off"));
893     qdict_put(bs_opts, "copy-on-read",
894               qstring_from_str(copy_on_read ? "on" :"off"));
895 
896     /* Controller type */
897     value = qemu_opt_get(legacy_opts, "if");
898     if (value) {
899         for (type = 0;
900              type < IF_COUNT && strcmp(value, if_name[type]);
901              type++) {
902         }
903         if (type == IF_COUNT) {
904             error_report("unsupported bus type '%s'", value);
905             goto fail;
906         }
907     } else {
908         type = block_default_type;
909     }
910 
911     /* Geometry */
912     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
913     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
914     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
915 
916     if (cyls || heads || secs) {
917         if (cyls < 1) {
918             error_report("invalid physical cyls number");
919             goto fail;
920         }
921         if (heads < 1) {
922             error_report("invalid physical heads number");
923             goto fail;
924         }
925         if (secs < 1) {
926             error_report("invalid physical secs number");
927             goto fail;
928         }
929     }
930 
931     translation = BIOS_ATA_TRANSLATION_AUTO;
932     value = qemu_opt_get(legacy_opts, "trans");
933     if (value != NULL) {
934         if (!cyls) {
935             error_report("'%s' trans must be used with cyls, heads and secs",
936                          value);
937             goto fail;
938         }
939         if (!strcmp(value, "none")) {
940             translation = BIOS_ATA_TRANSLATION_NONE;
941         } else if (!strcmp(value, "lba")) {
942             translation = BIOS_ATA_TRANSLATION_LBA;
943         } else if (!strcmp(value, "large")) {
944             translation = BIOS_ATA_TRANSLATION_LARGE;
945         } else if (!strcmp(value, "rechs")) {
946             translation = BIOS_ATA_TRANSLATION_RECHS;
947         } else if (!strcmp(value, "auto")) {
948             translation = BIOS_ATA_TRANSLATION_AUTO;
949         } else {
950             error_report("'%s' invalid translation type", value);
951             goto fail;
952         }
953     }
954 
955     if (media == MEDIA_CDROM) {
956         if (cyls || secs || heads) {
957             error_report("CHS can't be set with media=cdrom");
958             goto fail;
959         }
960     }
961 
962     /* Device address specified by bus/unit or index.
963      * If none was specified, try to find the first free one. */
964     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
965     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
966     index   = qemu_opt_get_number(legacy_opts, "index", -1);
967 
968     max_devs = if_max_devs[type];
969 
970     if (index != -1) {
971         if (bus_id != 0 || unit_id != -1) {
972             error_report("index cannot be used with bus and unit");
973             goto fail;
974         }
975         bus_id = drive_index_to_bus_id(type, index);
976         unit_id = drive_index_to_unit_id(type, index);
977     }
978 
979     if (unit_id == -1) {
980        unit_id = 0;
981        while (drive_get(type, bus_id, unit_id) != NULL) {
982            unit_id++;
983            if (max_devs && unit_id >= max_devs) {
984                unit_id -= max_devs;
985                bus_id++;
986            }
987        }
988     }
989 
990     if (max_devs && unit_id >= max_devs) {
991         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
992         goto fail;
993     }
994 
995     if (drive_get(type, bus_id, unit_id) != NULL) {
996         error_report("drive with bus=%d, unit=%d (index=%d) exists",
997                      bus_id, unit_id, index);
998         goto fail;
999     }
1000 
1001     /* Serial number */
1002     serial = qemu_opt_get(legacy_opts, "serial");
1003 
1004     /* no id supplied -> create one */
1005     if (qemu_opts_id(all_opts) == NULL) {
1006         char *new_id;
1007         const char *mediastr = "";
1008         if (type == IF_IDE || type == IF_SCSI) {
1009             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
1010         }
1011         if (max_devs) {
1012             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
1013                                      mediastr, unit_id);
1014         } else {
1015             new_id = g_strdup_printf("%s%s%i", if_name[type],
1016                                      mediastr, unit_id);
1017         }
1018         qdict_put(bs_opts, "id", qstring_from_str(new_id));
1019         g_free(new_id);
1020     }
1021 
1022     /* Add virtio block device */
1023     devaddr = qemu_opt_get(legacy_opts, "addr");
1024     if (devaddr && type != IF_VIRTIO) {
1025         error_report("addr is not supported by this bus type");
1026         goto fail;
1027     }
1028 
1029     if (type == IF_VIRTIO) {
1030         QemuOpts *devopts;
1031         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
1032                                    &error_abort);
1033         if (arch_type == QEMU_ARCH_S390X) {
1034             qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
1035         } else {
1036             qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
1037         }
1038         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
1039                      &error_abort);
1040         if (devaddr) {
1041             qemu_opt_set(devopts, "addr", devaddr, &error_abort);
1042         }
1043     }
1044 
1045     filename = qemu_opt_get(legacy_opts, "file");
1046 
1047     /* Check werror/rerror compatibility with if=... */
1048     werror = qemu_opt_get(legacy_opts, "werror");
1049     if (werror != NULL) {
1050         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
1051             type != IF_NONE) {
1052             error_report("werror is not supported by this bus type");
1053             goto fail;
1054         }
1055         qdict_put(bs_opts, "werror", qstring_from_str(werror));
1056     }
1057 
1058     rerror = qemu_opt_get(legacy_opts, "rerror");
1059     if (rerror != NULL) {
1060         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
1061             type != IF_NONE) {
1062             error_report("rerror is not supported by this bus type");
1063             goto fail;
1064         }
1065         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
1066     }
1067 
1068     /* Actual block device init: Functionality shared with blockdev-add */
1069     blk = blockdev_init(filename, bs_opts, &local_err);
1070     bs_opts = NULL;
1071     if (!blk) {
1072         if (local_err) {
1073             error_report_err(local_err);
1074         }
1075         goto fail;
1076     } else {
1077         assert(!local_err);
1078     }
1079 
1080     /* Create legacy DriveInfo */
1081     dinfo = g_malloc0(sizeof(*dinfo));
1082     dinfo->opts = all_opts;
1083 
1084     dinfo->cyls = cyls;
1085     dinfo->heads = heads;
1086     dinfo->secs = secs;
1087     dinfo->trans = translation;
1088 
1089     dinfo->type = type;
1090     dinfo->bus = bus_id;
1091     dinfo->unit = unit_id;
1092     dinfo->devaddr = devaddr;
1093     dinfo->serial = g_strdup(serial);
1094 
1095     blk_set_legacy_dinfo(blk, dinfo);
1096 
1097     switch(type) {
1098     case IF_IDE:
1099     case IF_SCSI:
1100     case IF_XEN:
1101     case IF_NONE:
1102         dinfo->media_cd = media == MEDIA_CDROM;
1103         break;
1104     default:
1105         break;
1106     }
1107 
1108 fail:
1109     qemu_opts_del(legacy_opts);
1110     QDECREF(bs_opts);
1111     return dinfo;
1112 }
1113 
1114 void hmp_commit(Monitor *mon, const QDict *qdict)
1115 {
1116     const char *device = qdict_get_str(qdict, "device");
1117     BlockBackend *blk;
1118     int ret;
1119 
1120     if (!strcmp(device, "all")) {
1121         ret = bdrv_commit_all();
1122     } else {
1123         BlockDriverState *bs;
1124         AioContext *aio_context;
1125 
1126         blk = blk_by_name(device);
1127         if (!blk) {
1128             monitor_printf(mon, "Device '%s' not found\n", device);
1129             return;
1130         }
1131         if (!blk_is_available(blk)) {
1132             monitor_printf(mon, "Device '%s' has no medium\n", device);
1133             return;
1134         }
1135 
1136         bs = blk_bs(blk);
1137         aio_context = bdrv_get_aio_context(bs);
1138         aio_context_acquire(aio_context);
1139 
1140         ret = bdrv_commit(bs);
1141 
1142         aio_context_release(aio_context);
1143     }
1144     if (ret < 0) {
1145         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1146                        strerror(-ret));
1147     }
1148 }
1149 
1150 static void blockdev_do_action(TransactionActionKind type, void *data,
1151                                Error **errp)
1152 {
1153     TransactionAction action;
1154     TransactionActionList list;
1155 
1156     action.type = type;
1157     action.u.data = data;
1158     list.value = &action;
1159     list.next = NULL;
1160     qmp_transaction(&list, errp);
1161 }
1162 
1163 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1164                                 bool has_node_name, const char *node_name,
1165                                 const char *snapshot_file,
1166                                 bool has_snapshot_node_name,
1167                                 const char *snapshot_node_name,
1168                                 bool has_format, const char *format,
1169                                 bool has_mode, NewImageMode mode, Error **errp)
1170 {
1171     BlockdevSnapshotSync snapshot = {
1172         .has_device = has_device,
1173         .device = (char *) device,
1174         .has_node_name = has_node_name,
1175         .node_name = (char *) node_name,
1176         .snapshot_file = (char *) snapshot_file,
1177         .has_snapshot_node_name = has_snapshot_node_name,
1178         .snapshot_node_name = (char *) snapshot_node_name,
1179         .has_format = has_format,
1180         .format = (char *) format,
1181         .has_mode = has_mode,
1182         .mode = mode,
1183     };
1184     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1185                        &snapshot, errp);
1186 }
1187 
1188 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1189                            Error **errp)
1190 {
1191     BlockdevSnapshot snapshot_data = {
1192         .node = (char *) node,
1193         .overlay = (char *) overlay
1194     };
1195 
1196     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1197                        &snapshot_data, errp);
1198 }
1199 
1200 void qmp_blockdev_snapshot_internal_sync(const char *device,
1201                                          const char *name,
1202                                          Error **errp)
1203 {
1204     BlockdevSnapshotInternal snapshot = {
1205         .device = (char *) device,
1206         .name = (char *) name
1207     };
1208 
1209     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1210                        &snapshot, errp);
1211 }
1212 
1213 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1214                                                          bool has_id,
1215                                                          const char *id,
1216                                                          bool has_name,
1217                                                          const char *name,
1218                                                          Error **errp)
1219 {
1220     BlockDriverState *bs;
1221     BlockBackend *blk;
1222     AioContext *aio_context;
1223     QEMUSnapshotInfo sn;
1224     Error *local_err = NULL;
1225     SnapshotInfo *info = NULL;
1226     int ret;
1227 
1228     blk = blk_by_name(device);
1229     if (!blk) {
1230         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1231                   "Device '%s' not found", device);
1232         return NULL;
1233     }
1234 
1235     aio_context = blk_get_aio_context(blk);
1236     aio_context_acquire(aio_context);
1237 
1238     if (!has_id) {
1239         id = NULL;
1240     }
1241 
1242     if (!has_name) {
1243         name = NULL;
1244     }
1245 
1246     if (!id && !name) {
1247         error_setg(errp, "Name or id must be provided");
1248         goto out_aio_context;
1249     }
1250 
1251     if (!blk_is_available(blk)) {
1252         error_setg(errp, "Device '%s' has no medium", device);
1253         goto out_aio_context;
1254     }
1255     bs = blk_bs(blk);
1256 
1257     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1258         goto out_aio_context;
1259     }
1260 
1261     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1262     if (local_err) {
1263         error_propagate(errp, local_err);
1264         goto out_aio_context;
1265     }
1266     if (!ret) {
1267         error_setg(errp,
1268                    "Snapshot with id '%s' and name '%s' does not exist on "
1269                    "device '%s'",
1270                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1271         goto out_aio_context;
1272     }
1273 
1274     bdrv_snapshot_delete(bs, id, name, &local_err);
1275     if (local_err) {
1276         error_propagate(errp, local_err);
1277         goto out_aio_context;
1278     }
1279 
1280     aio_context_release(aio_context);
1281 
1282     info = g_new0(SnapshotInfo, 1);
1283     info->id = g_strdup(sn.id_str);
1284     info->name = g_strdup(sn.name);
1285     info->date_nsec = sn.date_nsec;
1286     info->date_sec = sn.date_sec;
1287     info->vm_state_size = sn.vm_state_size;
1288     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1289     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1290 
1291     return info;
1292 
1293 out_aio_context:
1294     aio_context_release(aio_context);
1295     return NULL;
1296 }
1297 
1298 /**
1299  * block_dirty_bitmap_lookup:
1300  * Return a dirty bitmap (if present), after validating
1301  * the node reference and bitmap names.
1302  *
1303  * @node: The name of the BDS node to search for bitmaps
1304  * @name: The name of the bitmap to search for
1305  * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1306  * @paio: Output pointer for aio_context acquisition, if desired. Can be NULL.
1307  * @errp: Output pointer for error information. Can be NULL.
1308  *
1309  * @return: A bitmap object on success, or NULL on failure.
1310  */
1311 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1312                                                   const char *name,
1313                                                   BlockDriverState **pbs,
1314                                                   AioContext **paio,
1315                                                   Error **errp)
1316 {
1317     BlockDriverState *bs;
1318     BdrvDirtyBitmap *bitmap;
1319     AioContext *aio_context;
1320 
1321     if (!node) {
1322         error_setg(errp, "Node cannot be NULL");
1323         return NULL;
1324     }
1325     if (!name) {
1326         error_setg(errp, "Bitmap name cannot be NULL");
1327         return NULL;
1328     }
1329     bs = bdrv_lookup_bs(node, node, NULL);
1330     if (!bs) {
1331         error_setg(errp, "Node '%s' not found", node);
1332         return NULL;
1333     }
1334 
1335     aio_context = bdrv_get_aio_context(bs);
1336     aio_context_acquire(aio_context);
1337 
1338     bitmap = bdrv_find_dirty_bitmap(bs, name);
1339     if (!bitmap) {
1340         error_setg(errp, "Dirty bitmap '%s' not found", name);
1341         goto fail;
1342     }
1343 
1344     if (pbs) {
1345         *pbs = bs;
1346     }
1347     if (paio) {
1348         *paio = aio_context;
1349     } else {
1350         aio_context_release(aio_context);
1351     }
1352 
1353     return bitmap;
1354 
1355  fail:
1356     aio_context_release(aio_context);
1357     return NULL;
1358 }
1359 
1360 /* New and old BlockDriverState structs for atomic group operations */
1361 
1362 typedef struct BlkTransactionState BlkTransactionState;
1363 
1364 /* Only prepare() may fail. In a single transaction, only one of commit() or
1365    abort() will be called, clean() will always be called if it present. */
1366 typedef struct BdrvActionOps {
1367     /* Size of state struct, in bytes. */
1368     size_t instance_size;
1369     /* Prepare the work, must NOT be NULL. */
1370     void (*prepare)(BlkTransactionState *common, Error **errp);
1371     /* Commit the changes, can be NULL. */
1372     void (*commit)(BlkTransactionState *common);
1373     /* Abort the changes on fail, can be NULL. */
1374     void (*abort)(BlkTransactionState *common);
1375     /* Clean up resource in the end, can be NULL. */
1376     void (*clean)(BlkTransactionState *common);
1377 } BdrvActionOps;
1378 
1379 /*
1380  * This structure must be arranged as first member in child type, assuming
1381  * that compiler will also arrange it to the same address with parent instance.
1382  * Later it will be used in free().
1383  */
1384 struct BlkTransactionState {
1385     TransactionAction *action;
1386     const BdrvActionOps *ops;
1387     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1388 };
1389 
1390 /* internal snapshot private data */
1391 typedef struct InternalSnapshotState {
1392     BlkTransactionState common;
1393     BlockDriverState *bs;
1394     AioContext *aio_context;
1395     QEMUSnapshotInfo sn;
1396     bool created;
1397 } InternalSnapshotState;
1398 
1399 static void internal_snapshot_prepare(BlkTransactionState *common,
1400                                       Error **errp)
1401 {
1402     Error *local_err = NULL;
1403     const char *device;
1404     const char *name;
1405     BlockBackend *blk;
1406     BlockDriverState *bs;
1407     QEMUSnapshotInfo old_sn, *sn;
1408     bool ret;
1409     qemu_timeval tv;
1410     BlockdevSnapshotInternal *internal;
1411     InternalSnapshotState *state;
1412     int ret1;
1413 
1414     g_assert(common->action->type ==
1415              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1416     internal = common->action->u.blockdev_snapshot_internal_sync;
1417     state = DO_UPCAST(InternalSnapshotState, common, common);
1418 
1419     /* 1. parse input */
1420     device = internal->device;
1421     name = internal->name;
1422 
1423     /* 2. check for validation */
1424     blk = blk_by_name(device);
1425     if (!blk) {
1426         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1427                   "Device '%s' not found", device);
1428         return;
1429     }
1430 
1431     /* AioContext is released in .clean() */
1432     state->aio_context = blk_get_aio_context(blk);
1433     aio_context_acquire(state->aio_context);
1434 
1435     if (!blk_is_available(blk)) {
1436         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1437         return;
1438     }
1439     bs = blk_bs(blk);
1440 
1441     state->bs = bs;
1442     bdrv_drained_begin(bs);
1443 
1444     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1445         return;
1446     }
1447 
1448     if (bdrv_is_read_only(bs)) {
1449         error_setg(errp, "Device '%s' is read only", device);
1450         return;
1451     }
1452 
1453     if (!bdrv_can_snapshot(bs)) {
1454         error_setg(errp, "Block format '%s' used by device '%s' "
1455                    "does not support internal snapshots",
1456                    bs->drv->format_name, device);
1457         return;
1458     }
1459 
1460     if (!strlen(name)) {
1461         error_setg(errp, "Name is empty");
1462         return;
1463     }
1464 
1465     /* check whether a snapshot with name exist */
1466     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1467                                             &local_err);
1468     if (local_err) {
1469         error_propagate(errp, local_err);
1470         return;
1471     } else if (ret) {
1472         error_setg(errp,
1473                    "Snapshot with name '%s' already exists on device '%s'",
1474                    name, device);
1475         return;
1476     }
1477 
1478     /* 3. take the snapshot */
1479     sn = &state->sn;
1480     pstrcpy(sn->name, sizeof(sn->name), name);
1481     qemu_gettimeofday(&tv);
1482     sn->date_sec = tv.tv_sec;
1483     sn->date_nsec = tv.tv_usec * 1000;
1484     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1485 
1486     ret1 = bdrv_snapshot_create(bs, sn);
1487     if (ret1 < 0) {
1488         error_setg_errno(errp, -ret1,
1489                          "Failed to create snapshot '%s' on device '%s'",
1490                          name, device);
1491         return;
1492     }
1493 
1494     /* 4. succeed, mark a snapshot is created */
1495     state->created = true;
1496 }
1497 
1498 static void internal_snapshot_abort(BlkTransactionState *common)
1499 {
1500     InternalSnapshotState *state =
1501                              DO_UPCAST(InternalSnapshotState, common, common);
1502     BlockDriverState *bs = state->bs;
1503     QEMUSnapshotInfo *sn = &state->sn;
1504     Error *local_error = NULL;
1505 
1506     if (!state->created) {
1507         return;
1508     }
1509 
1510     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1511         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1512                      "device '%s' in abort: %s",
1513                      sn->id_str,
1514                      sn->name,
1515                      bdrv_get_device_name(bs),
1516                      error_get_pretty(local_error));
1517         error_free(local_error);
1518     }
1519 }
1520 
1521 static void internal_snapshot_clean(BlkTransactionState *common)
1522 {
1523     InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1524                                              common, common);
1525 
1526     if (state->aio_context) {
1527         if (state->bs) {
1528             bdrv_drained_end(state->bs);
1529         }
1530         aio_context_release(state->aio_context);
1531     }
1532 }
1533 
1534 /* external snapshot private data */
1535 typedef struct ExternalSnapshotState {
1536     BlkTransactionState common;
1537     BlockDriverState *old_bs;
1538     BlockDriverState *new_bs;
1539     AioContext *aio_context;
1540 } ExternalSnapshotState;
1541 
1542 static void external_snapshot_prepare(BlkTransactionState *common,
1543                                       Error **errp)
1544 {
1545     int flags = 0, ret;
1546     QDict *options = NULL;
1547     Error *local_err = NULL;
1548     /* Device and node name of the image to generate the snapshot from */
1549     const char *device;
1550     const char *node_name;
1551     /* Reference to the new image (for 'blockdev-snapshot') */
1552     const char *snapshot_ref;
1553     /* File name of the new image (for 'blockdev-snapshot-sync') */
1554     const char *new_image_file;
1555     ExternalSnapshotState *state =
1556                              DO_UPCAST(ExternalSnapshotState, common, common);
1557     TransactionAction *action = common->action;
1558 
1559     /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1560      * purpose but a different set of parameters */
1561     switch (action->type) {
1562     case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1563         {
1564             BlockdevSnapshot *s = action->u.blockdev_snapshot;
1565             device = s->node;
1566             node_name = s->node;
1567             new_image_file = NULL;
1568             snapshot_ref = s->overlay;
1569         }
1570         break;
1571     case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1572         {
1573             BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync;
1574             device = s->has_device ? s->device : NULL;
1575             node_name = s->has_node_name ? s->node_name : NULL;
1576             new_image_file = s->snapshot_file;
1577             snapshot_ref = NULL;
1578         }
1579         break;
1580     default:
1581         g_assert_not_reached();
1582     }
1583 
1584     /* start processing */
1585     state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1586     if (!state->old_bs) {
1587         return;
1588     }
1589 
1590     /* Acquire AioContext now so any threads operating on old_bs stop */
1591     state->aio_context = bdrv_get_aio_context(state->old_bs);
1592     aio_context_acquire(state->aio_context);
1593     bdrv_drained_begin(state->old_bs);
1594 
1595     if (!bdrv_is_inserted(state->old_bs)) {
1596         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1597         return;
1598     }
1599 
1600     if (bdrv_op_is_blocked(state->old_bs,
1601                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1602         return;
1603     }
1604 
1605     if (!bdrv_is_read_only(state->old_bs)) {
1606         if (bdrv_flush(state->old_bs)) {
1607             error_setg(errp, QERR_IO_ERROR);
1608             return;
1609         }
1610     }
1611 
1612     if (!bdrv_is_first_non_filter(state->old_bs)) {
1613         error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1614         return;
1615     }
1616 
1617     if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1618         BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync;
1619         const char *format = s->has_format ? s->format : "qcow2";
1620         enum NewImageMode mode;
1621         const char *snapshot_node_name =
1622             s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1623 
1624         if (node_name && !snapshot_node_name) {
1625             error_setg(errp, "New snapshot node name missing");
1626             return;
1627         }
1628 
1629         if (snapshot_node_name &&
1630             bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1631             error_setg(errp, "New snapshot node name already in use");
1632             return;
1633         }
1634 
1635         flags = state->old_bs->open_flags;
1636 
1637         /* create new image w/backing file */
1638         mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1639         if (mode != NEW_IMAGE_MODE_EXISTING) {
1640             bdrv_img_create(new_image_file, format,
1641                             state->old_bs->filename,
1642                             state->old_bs->drv->format_name,
1643                             NULL, -1, flags, &local_err, false);
1644             if (local_err) {
1645                 error_propagate(errp, local_err);
1646                 return;
1647             }
1648         }
1649 
1650         options = qdict_new();
1651         if (s->has_snapshot_node_name) {
1652             qdict_put(options, "node-name",
1653                       qstring_from_str(snapshot_node_name));
1654         }
1655         qdict_put(options, "driver", qstring_from_str(format));
1656 
1657         flags |= BDRV_O_NO_BACKING;
1658     }
1659 
1660     assert(state->new_bs == NULL);
1661     ret = bdrv_open(&state->new_bs, new_image_file, snapshot_ref, options,
1662                     flags, errp);
1663     /* We will manually add the backing_hd field to the bs later */
1664     if (ret != 0) {
1665         return;
1666     }
1667 
1668     if (state->new_bs->blk != NULL) {
1669         error_setg(errp, "The snapshot is already in use by %s",
1670                    blk_name(state->new_bs->blk));
1671         return;
1672     }
1673 
1674     if (bdrv_op_is_blocked(state->new_bs, BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
1675                            errp)) {
1676         return;
1677     }
1678 
1679     if (state->new_bs->backing != NULL) {
1680         error_setg(errp, "The snapshot already has a backing image");
1681         return;
1682     }
1683 
1684     if (!state->new_bs->drv->supports_backing) {
1685         error_setg(errp, "The snapshot does not support backing images");
1686     }
1687 }
1688 
1689 static void external_snapshot_commit(BlkTransactionState *common)
1690 {
1691     ExternalSnapshotState *state =
1692                              DO_UPCAST(ExternalSnapshotState, common, common);
1693 
1694     bdrv_set_aio_context(state->new_bs, state->aio_context);
1695 
1696     /* This removes our old bs and adds the new bs */
1697     bdrv_append(state->new_bs, state->old_bs);
1698     /* We don't need (or want) to use the transactional
1699      * bdrv_reopen_multiple() across all the entries at once, because we
1700      * don't want to abort all of them if one of them fails the reopen */
1701     bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR,
1702                 NULL);
1703 }
1704 
1705 static void external_snapshot_abort(BlkTransactionState *common)
1706 {
1707     ExternalSnapshotState *state =
1708                              DO_UPCAST(ExternalSnapshotState, common, common);
1709     if (state->new_bs) {
1710         bdrv_unref(state->new_bs);
1711     }
1712 }
1713 
1714 static void external_snapshot_clean(BlkTransactionState *common)
1715 {
1716     ExternalSnapshotState *state =
1717                              DO_UPCAST(ExternalSnapshotState, common, common);
1718     if (state->aio_context) {
1719         bdrv_drained_end(state->old_bs);
1720         aio_context_release(state->aio_context);
1721     }
1722 }
1723 
1724 typedef struct DriveBackupState {
1725     BlkTransactionState common;
1726     BlockDriverState *bs;
1727     AioContext *aio_context;
1728     BlockJob *job;
1729 } DriveBackupState;
1730 
1731 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1732 {
1733     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1734     BlockBackend *blk;
1735     DriveBackup *backup;
1736     Error *local_err = NULL;
1737 
1738     assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1739     backup = common->action->u.drive_backup;
1740 
1741     blk = blk_by_name(backup->device);
1742     if (!blk) {
1743         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1744                   "Device '%s' not found", backup->device);
1745         return;
1746     }
1747 
1748     if (!blk_is_available(blk)) {
1749         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1750         return;
1751     }
1752 
1753     /* AioContext is released in .clean() */
1754     state->aio_context = blk_get_aio_context(blk);
1755     aio_context_acquire(state->aio_context);
1756     bdrv_drained_begin(blk_bs(blk));
1757     state->bs = blk_bs(blk);
1758 
1759     qmp_drive_backup(backup->device, backup->target,
1760                      backup->has_format, backup->format,
1761                      backup->sync,
1762                      backup->has_mode, backup->mode,
1763                      backup->has_speed, backup->speed,
1764                      backup->has_bitmap, backup->bitmap,
1765                      backup->has_on_source_error, backup->on_source_error,
1766                      backup->has_on_target_error, backup->on_target_error,
1767                      &local_err);
1768     if (local_err) {
1769         error_propagate(errp, local_err);
1770         return;
1771     }
1772 
1773     state->job = state->bs->job;
1774 }
1775 
1776 static void drive_backup_abort(BlkTransactionState *common)
1777 {
1778     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1779     BlockDriverState *bs = state->bs;
1780 
1781     /* Only cancel if it's the job we started */
1782     if (bs && bs->job && bs->job == state->job) {
1783         block_job_cancel_sync(bs->job);
1784     }
1785 }
1786 
1787 static void drive_backup_clean(BlkTransactionState *common)
1788 {
1789     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1790 
1791     if (state->aio_context) {
1792         bdrv_drained_end(state->bs);
1793         aio_context_release(state->aio_context);
1794     }
1795 }
1796 
1797 typedef struct BlockdevBackupState {
1798     BlkTransactionState common;
1799     BlockDriverState *bs;
1800     BlockJob *job;
1801     AioContext *aio_context;
1802 } BlockdevBackupState;
1803 
1804 static void blockdev_backup_prepare(BlkTransactionState *common, Error **errp)
1805 {
1806     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1807     BlockdevBackup *backup;
1808     BlockBackend *blk, *target;
1809     Error *local_err = NULL;
1810 
1811     assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1812     backup = common->action->u.blockdev_backup;
1813 
1814     blk = blk_by_name(backup->device);
1815     if (!blk) {
1816         error_setg(errp, "Device '%s' not found", backup->device);
1817         return;
1818     }
1819 
1820     if (!blk_is_available(blk)) {
1821         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1822         return;
1823     }
1824 
1825     target = blk_by_name(backup->target);
1826     if (!target) {
1827         error_setg(errp, "Device '%s' not found", backup->target);
1828         return;
1829     }
1830 
1831     /* AioContext is released in .clean() */
1832     state->aio_context = blk_get_aio_context(blk);
1833     if (state->aio_context != blk_get_aio_context(target)) {
1834         state->aio_context = NULL;
1835         error_setg(errp, "Backup between two IO threads is not implemented");
1836         return;
1837     }
1838     aio_context_acquire(state->aio_context);
1839     state->bs = blk_bs(blk);
1840     bdrv_drained_begin(state->bs);
1841 
1842     qmp_blockdev_backup(backup->device, backup->target,
1843                         backup->sync,
1844                         backup->has_speed, backup->speed,
1845                         backup->has_on_source_error, backup->on_source_error,
1846                         backup->has_on_target_error, backup->on_target_error,
1847                         &local_err);
1848     if (local_err) {
1849         error_propagate(errp, local_err);
1850         return;
1851     }
1852 
1853     state->job = state->bs->job;
1854 }
1855 
1856 static void blockdev_backup_abort(BlkTransactionState *common)
1857 {
1858     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1859     BlockDriverState *bs = state->bs;
1860 
1861     /* Only cancel if it's the job we started */
1862     if (bs && bs->job && bs->job == state->job) {
1863         block_job_cancel_sync(bs->job);
1864     }
1865 }
1866 
1867 static void blockdev_backup_clean(BlkTransactionState *common)
1868 {
1869     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1870 
1871     if (state->aio_context) {
1872         bdrv_drained_end(state->bs);
1873         aio_context_release(state->aio_context);
1874     }
1875 }
1876 
1877 static void abort_prepare(BlkTransactionState *common, Error **errp)
1878 {
1879     error_setg(errp, "Transaction aborted using Abort action");
1880 }
1881 
1882 static void abort_commit(BlkTransactionState *common)
1883 {
1884     g_assert_not_reached(); /* this action never succeeds */
1885 }
1886 
1887 static const BdrvActionOps actions[] = {
1888     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
1889         .instance_size = sizeof(ExternalSnapshotState),
1890         .prepare  = external_snapshot_prepare,
1891         .commit   = external_snapshot_commit,
1892         .abort = external_snapshot_abort,
1893     },
1894     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1895         .instance_size = sizeof(ExternalSnapshotState),
1896         .prepare  = external_snapshot_prepare,
1897         .commit   = external_snapshot_commit,
1898         .abort = external_snapshot_abort,
1899         .clean = external_snapshot_clean,
1900     },
1901     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1902         .instance_size = sizeof(DriveBackupState),
1903         .prepare = drive_backup_prepare,
1904         .abort = drive_backup_abort,
1905         .clean = drive_backup_clean,
1906     },
1907     [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
1908         .instance_size = sizeof(BlockdevBackupState),
1909         .prepare = blockdev_backup_prepare,
1910         .abort = blockdev_backup_abort,
1911         .clean = blockdev_backup_clean,
1912     },
1913     [TRANSACTION_ACTION_KIND_ABORT] = {
1914         .instance_size = sizeof(BlkTransactionState),
1915         .prepare = abort_prepare,
1916         .commit = abort_commit,
1917     },
1918     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1919         .instance_size = sizeof(InternalSnapshotState),
1920         .prepare  = internal_snapshot_prepare,
1921         .abort = internal_snapshot_abort,
1922         .clean = internal_snapshot_clean,
1923     },
1924 };
1925 
1926 /*
1927  * 'Atomic' group operations.  The operations are performed as a set, and if
1928  * any fail then we roll back all operations in the group.
1929  */
1930 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1931 {
1932     TransactionActionList *dev_entry = dev_list;
1933     BlkTransactionState *state, *next;
1934     Error *local_err = NULL;
1935 
1936     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1937     QSIMPLEQ_INIT(&snap_bdrv_states);
1938 
1939     /* drain all i/o before any operations */
1940     bdrv_drain_all();
1941 
1942     /* We don't do anything in this loop that commits us to the operations */
1943     while (NULL != dev_entry) {
1944         TransactionAction *dev_info = NULL;
1945         const BdrvActionOps *ops;
1946 
1947         dev_info = dev_entry->value;
1948         dev_entry = dev_entry->next;
1949 
1950         assert(dev_info->type < ARRAY_SIZE(actions));
1951 
1952         ops = &actions[dev_info->type];
1953         assert(ops->instance_size > 0);
1954 
1955         state = g_malloc0(ops->instance_size);
1956         state->ops = ops;
1957         state->action = dev_info;
1958         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1959 
1960         state->ops->prepare(state, &local_err);
1961         if (local_err) {
1962             error_propagate(errp, local_err);
1963             goto delete_and_fail;
1964         }
1965     }
1966 
1967     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1968         if (state->ops->commit) {
1969             state->ops->commit(state);
1970         }
1971     }
1972 
1973     /* success */
1974     goto exit;
1975 
1976 delete_and_fail:
1977     /* failure, and it is all-or-none; roll back all operations */
1978     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1979         if (state->ops->abort) {
1980             state->ops->abort(state);
1981         }
1982     }
1983 exit:
1984     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1985         if (state->ops->clean) {
1986             state->ops->clean(state);
1987         }
1988         g_free(state);
1989     }
1990 }
1991 
1992 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1993 {
1994     Error *local_err = NULL;
1995 
1996     qmp_blockdev_open_tray(device, has_force, force, &local_err);
1997     if (local_err) {
1998         error_propagate(errp, local_err);
1999         return;
2000     }
2001 
2002     qmp_blockdev_remove_medium(device, errp);
2003 }
2004 
2005 void qmp_block_passwd(bool has_device, const char *device,
2006                       bool has_node_name, const char *node_name,
2007                       const char *password, Error **errp)
2008 {
2009     Error *local_err = NULL;
2010     BlockDriverState *bs;
2011     AioContext *aio_context;
2012 
2013     bs = bdrv_lookup_bs(has_device ? device : NULL,
2014                         has_node_name ? node_name : NULL,
2015                         &local_err);
2016     if (local_err) {
2017         error_propagate(errp, local_err);
2018         return;
2019     }
2020 
2021     aio_context = bdrv_get_aio_context(bs);
2022     aio_context_acquire(aio_context);
2023 
2024     bdrv_add_key(bs, password, errp);
2025 
2026     aio_context_release(aio_context);
2027 }
2028 
2029 void qmp_blockdev_open_tray(const char *device, bool has_force, bool force,
2030                             Error **errp)
2031 {
2032     BlockBackend *blk;
2033     bool locked;
2034 
2035     if (!has_force) {
2036         force = false;
2037     }
2038 
2039     blk = blk_by_name(device);
2040     if (!blk) {
2041         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2042                   "Device '%s' not found", device);
2043         return;
2044     }
2045 
2046     if (!blk_dev_has_removable_media(blk)) {
2047         error_setg(errp, "Device '%s' is not removable", device);
2048         return;
2049     }
2050 
2051     if (blk_dev_is_tray_open(blk)) {
2052         return;
2053     }
2054 
2055     locked = blk_dev_is_medium_locked(blk);
2056     if (locked) {
2057         blk_dev_eject_request(blk, force);
2058     }
2059 
2060     if (!locked || force) {
2061         blk_dev_change_media_cb(blk, false);
2062     }
2063 }
2064 
2065 void qmp_blockdev_close_tray(const char *device, Error **errp)
2066 {
2067     BlockBackend *blk;
2068 
2069     blk = blk_by_name(device);
2070     if (!blk) {
2071         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2072                   "Device '%s' not found", device);
2073         return;
2074     }
2075 
2076     if (!blk_dev_has_removable_media(blk)) {
2077         error_setg(errp, "Device '%s' is not removable", device);
2078         return;
2079     }
2080 
2081     if (!blk_dev_is_tray_open(blk)) {
2082         return;
2083     }
2084 
2085     blk_dev_change_media_cb(blk, true);
2086 }
2087 
2088 void qmp_blockdev_remove_medium(const char *device, Error **errp)
2089 {
2090     BlockBackend *blk;
2091     BlockDriverState *bs;
2092     AioContext *aio_context;
2093     bool has_device;
2094 
2095     blk = blk_by_name(device);
2096     if (!blk) {
2097         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2098                   "Device '%s' not found", device);
2099         return;
2100     }
2101 
2102     /* For BBs without a device, we can exchange the BDS tree at will */
2103     has_device = blk_get_attached_dev(blk);
2104 
2105     if (has_device && !blk_dev_has_removable_media(blk)) {
2106         error_setg(errp, "Device '%s' is not removable", device);
2107         return;
2108     }
2109 
2110     if (has_device && !blk_dev_is_tray_open(blk)) {
2111         error_setg(errp, "Tray of device '%s' is not open", device);
2112         return;
2113     }
2114 
2115     bs = blk_bs(blk);
2116     if (!bs) {
2117         return;
2118     }
2119 
2120     aio_context = bdrv_get_aio_context(bs);
2121     aio_context_acquire(aio_context);
2122 
2123     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
2124         goto out;
2125     }
2126 
2127     /* This follows the convention established by bdrv_make_anon() */
2128     if (bs->device_list.tqe_prev) {
2129         QTAILQ_REMOVE(&bdrv_states, bs, device_list);
2130         bs->device_list.tqe_prev = NULL;
2131     }
2132 
2133     blk_remove_bs(blk);
2134 
2135 out:
2136     aio_context_release(aio_context);
2137 }
2138 
2139 static void qmp_blockdev_insert_anon_medium(const char *device,
2140                                             BlockDriverState *bs, Error **errp)
2141 {
2142     BlockBackend *blk;
2143     bool has_device;
2144 
2145     blk = blk_by_name(device);
2146     if (!blk) {
2147         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2148                   "Device '%s' not found", device);
2149         return;
2150     }
2151 
2152     /* For BBs without a device, we can exchange the BDS tree at will */
2153     has_device = blk_get_attached_dev(blk);
2154 
2155     if (has_device && !blk_dev_has_removable_media(blk)) {
2156         error_setg(errp, "Device '%s' is not removable", device);
2157         return;
2158     }
2159 
2160     if (has_device && !blk_dev_is_tray_open(blk)) {
2161         error_setg(errp, "Tray of device '%s' is not open", device);
2162         return;
2163     }
2164 
2165     if (blk_bs(blk)) {
2166         error_setg(errp, "There already is a medium in device '%s'", device);
2167         return;
2168     }
2169 
2170     blk_insert_bs(blk, bs);
2171 
2172     QTAILQ_INSERT_TAIL(&bdrv_states, bs, device_list);
2173 }
2174 
2175 void qmp_blockdev_insert_medium(const char *device, const char *node_name,
2176                                 Error **errp)
2177 {
2178     BlockDriverState *bs;
2179 
2180     bs = bdrv_find_node(node_name);
2181     if (!bs) {
2182         error_setg(errp, "Node '%s' not found", node_name);
2183         return;
2184     }
2185 
2186     if (bs->blk) {
2187         error_setg(errp, "Node '%s' is already in use by '%s'", node_name,
2188                    blk_name(bs->blk));
2189         return;
2190     }
2191 
2192     qmp_blockdev_insert_anon_medium(device, bs, errp);
2193 }
2194 
2195 void qmp_blockdev_change_medium(const char *device, const char *filename,
2196                                 bool has_format, const char *format,
2197                                 bool has_read_only,
2198                                 BlockdevChangeReadOnlyMode read_only,
2199                                 Error **errp)
2200 {
2201     BlockBackend *blk;
2202     BlockDriverState *medium_bs = NULL;
2203     int bdrv_flags, ret;
2204     QDict *options = NULL;
2205     Error *err = NULL;
2206 
2207     blk = blk_by_name(device);
2208     if (!blk) {
2209         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2210                   "Device '%s' not found", device);
2211         goto fail;
2212     }
2213 
2214     if (blk_bs(blk)) {
2215         blk_update_root_state(blk);
2216     }
2217 
2218     bdrv_flags = blk_get_open_flags_from_root_state(blk);
2219 
2220     if (!has_read_only) {
2221         read_only = BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN;
2222     }
2223 
2224     switch (read_only) {
2225     case BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN:
2226         break;
2227 
2228     case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_ONLY:
2229         bdrv_flags &= ~BDRV_O_RDWR;
2230         break;
2231 
2232     case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_WRITE:
2233         bdrv_flags |= BDRV_O_RDWR;
2234         break;
2235 
2236     default:
2237         abort();
2238     }
2239 
2240     if (has_format) {
2241         options = qdict_new();
2242         qdict_put(options, "driver", qstring_from_str(format));
2243     }
2244 
2245     assert(!medium_bs);
2246     ret = bdrv_open(&medium_bs, filename, NULL, options, bdrv_flags, errp);
2247     if (ret < 0) {
2248         goto fail;
2249     }
2250 
2251     blk_apply_root_state(blk, medium_bs);
2252 
2253     bdrv_add_key(medium_bs, NULL, &err);
2254     if (err) {
2255         error_propagate(errp, err);
2256         goto fail;
2257     }
2258 
2259     qmp_blockdev_open_tray(device, false, false, &err);
2260     if (err) {
2261         error_propagate(errp, err);
2262         goto fail;
2263     }
2264 
2265     qmp_blockdev_remove_medium(device, &err);
2266     if (err) {
2267         error_propagate(errp, err);
2268         goto fail;
2269     }
2270 
2271     qmp_blockdev_insert_anon_medium(device, medium_bs, &err);
2272     if (err) {
2273         error_propagate(errp, err);
2274         goto fail;
2275     }
2276 
2277     qmp_blockdev_close_tray(device, errp);
2278 
2279 fail:
2280     /* If the medium has been inserted, the device has its own reference, so
2281      * ours must be relinquished; and if it has not been inserted successfully,
2282      * the reference must be relinquished anyway */
2283     bdrv_unref(medium_bs);
2284 }
2285 
2286 /* throttling disk I/O limits */
2287 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
2288                                int64_t bps_wr,
2289                                int64_t iops,
2290                                int64_t iops_rd,
2291                                int64_t iops_wr,
2292                                bool has_bps_max,
2293                                int64_t bps_max,
2294                                bool has_bps_rd_max,
2295                                int64_t bps_rd_max,
2296                                bool has_bps_wr_max,
2297                                int64_t bps_wr_max,
2298                                bool has_iops_max,
2299                                int64_t iops_max,
2300                                bool has_iops_rd_max,
2301                                int64_t iops_rd_max,
2302                                bool has_iops_wr_max,
2303                                int64_t iops_wr_max,
2304                                bool has_iops_size,
2305                                int64_t iops_size,
2306                                bool has_group,
2307                                const char *group, Error **errp)
2308 {
2309     ThrottleConfig cfg;
2310     BlockDriverState *bs;
2311     BlockBackend *blk;
2312     AioContext *aio_context;
2313 
2314     blk = blk_by_name(device);
2315     if (!blk) {
2316         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2317                   "Device '%s' not found", device);
2318         return;
2319     }
2320 
2321     aio_context = blk_get_aio_context(blk);
2322     aio_context_acquire(aio_context);
2323 
2324     bs = blk_bs(blk);
2325     if (!bs) {
2326         error_setg(errp, "Device '%s' has no medium", device);
2327         goto out;
2328     }
2329 
2330     memset(&cfg, 0, sizeof(cfg));
2331     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
2332     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
2333     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
2334 
2335     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
2336     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
2337     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
2338 
2339     if (has_bps_max) {
2340         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
2341     }
2342     if (has_bps_rd_max) {
2343         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
2344     }
2345     if (has_bps_wr_max) {
2346         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
2347     }
2348     if (has_iops_max) {
2349         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
2350     }
2351     if (has_iops_rd_max) {
2352         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
2353     }
2354     if (has_iops_wr_max) {
2355         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
2356     }
2357 
2358     if (has_iops_size) {
2359         cfg.op_size = iops_size;
2360     }
2361 
2362     if (!check_throttle_config(&cfg, errp)) {
2363         goto out;
2364     }
2365 
2366     if (throttle_enabled(&cfg)) {
2367         /* Enable I/O limits if they're not enabled yet, otherwise
2368          * just update the throttling group. */
2369         if (!bs->throttle_state) {
2370             bdrv_io_limits_enable(bs, has_group ? group : device);
2371         } else if (has_group) {
2372             bdrv_io_limits_update_group(bs, group);
2373         }
2374         /* Set the new throttling configuration */
2375         bdrv_set_io_limits(bs, &cfg);
2376     } else if (bs->throttle_state) {
2377         /* If all throttling settings are set to 0, disable I/O limits */
2378         bdrv_io_limits_disable(bs);
2379     }
2380 
2381 out:
2382     aio_context_release(aio_context);
2383 }
2384 
2385 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2386                                 bool has_granularity, uint32_t granularity,
2387                                 Error **errp)
2388 {
2389     AioContext *aio_context;
2390     BlockDriverState *bs;
2391 
2392     if (!name || name[0] == '\0') {
2393         error_setg(errp, "Bitmap name cannot be empty");
2394         return;
2395     }
2396 
2397     bs = bdrv_lookup_bs(node, node, errp);
2398     if (!bs) {
2399         return;
2400     }
2401 
2402     aio_context = bdrv_get_aio_context(bs);
2403     aio_context_acquire(aio_context);
2404 
2405     if (has_granularity) {
2406         if (granularity < 512 || !is_power_of_2(granularity)) {
2407             error_setg(errp, "Granularity must be power of 2 "
2408                              "and at least 512");
2409             goto out;
2410         }
2411     } else {
2412         /* Default to cluster size, if available: */
2413         granularity = bdrv_get_default_bitmap_granularity(bs);
2414     }
2415 
2416     bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2417 
2418  out:
2419     aio_context_release(aio_context);
2420 }
2421 
2422 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2423                                    Error **errp)
2424 {
2425     AioContext *aio_context;
2426     BlockDriverState *bs;
2427     BdrvDirtyBitmap *bitmap;
2428 
2429     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2430     if (!bitmap || !bs) {
2431         return;
2432     }
2433 
2434     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2435         error_setg(errp,
2436                    "Bitmap '%s' is currently frozen and cannot be removed",
2437                    name);
2438         goto out;
2439     }
2440     bdrv_dirty_bitmap_make_anon(bitmap);
2441     bdrv_release_dirty_bitmap(bs, bitmap);
2442 
2443  out:
2444     aio_context_release(aio_context);
2445 }
2446 
2447 /**
2448  * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2449  * immediately after a full backup operation.
2450  */
2451 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2452                                   Error **errp)
2453 {
2454     AioContext *aio_context;
2455     BdrvDirtyBitmap *bitmap;
2456     BlockDriverState *bs;
2457 
2458     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2459     if (!bitmap || !bs) {
2460         return;
2461     }
2462 
2463     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2464         error_setg(errp,
2465                    "Bitmap '%s' is currently frozen and cannot be modified",
2466                    name);
2467         goto out;
2468     } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
2469         error_setg(errp,
2470                    "Bitmap '%s' is currently disabled and cannot be cleared",
2471                    name);
2472         goto out;
2473     }
2474 
2475     bdrv_clear_dirty_bitmap(bitmap);
2476 
2477  out:
2478     aio_context_release(aio_context);
2479 }
2480 
2481 void hmp_drive_del(Monitor *mon, const QDict *qdict)
2482 {
2483     const char *id = qdict_get_str(qdict, "id");
2484     BlockBackend *blk;
2485     BlockDriverState *bs;
2486     AioContext *aio_context;
2487     Error *local_err = NULL;
2488 
2489     blk = blk_by_name(id);
2490     if (!blk) {
2491         error_report("Device '%s' not found", id);
2492         return;
2493     }
2494 
2495     if (!blk_legacy_dinfo(blk)) {
2496         error_report("Deleting device added with blockdev-add"
2497                      " is not supported");
2498         return;
2499     }
2500 
2501     aio_context = blk_get_aio_context(blk);
2502     aio_context_acquire(aio_context);
2503 
2504     bs = blk_bs(blk);
2505     if (bs) {
2506         if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2507             error_report_err(local_err);
2508             aio_context_release(aio_context);
2509             return;
2510         }
2511 
2512         bdrv_close(bs);
2513     }
2514 
2515     /* if we have a device attached to this BlockDriverState
2516      * then we need to make the drive anonymous until the device
2517      * can be removed.  If this is a drive with no device backing
2518      * then we can just get rid of the block driver state right here.
2519      */
2520     if (blk_get_attached_dev(blk)) {
2521         blk_hide_on_behalf_of_hmp_drive_del(blk);
2522         /* Further I/O must not pause the guest */
2523         blk_set_on_error(blk, BLOCKDEV_ON_ERROR_REPORT,
2524                          BLOCKDEV_ON_ERROR_REPORT);
2525     } else {
2526         blk_unref(blk);
2527     }
2528 
2529     aio_context_release(aio_context);
2530 }
2531 
2532 void qmp_block_resize(bool has_device, const char *device,
2533                       bool has_node_name, const char *node_name,
2534                       int64_t size, Error **errp)
2535 {
2536     Error *local_err = NULL;
2537     BlockDriverState *bs;
2538     AioContext *aio_context;
2539     int ret;
2540 
2541     bs = bdrv_lookup_bs(has_device ? device : NULL,
2542                         has_node_name ? node_name : NULL,
2543                         &local_err);
2544     if (local_err) {
2545         error_propagate(errp, local_err);
2546         return;
2547     }
2548 
2549     aio_context = bdrv_get_aio_context(bs);
2550     aio_context_acquire(aio_context);
2551 
2552     if (!bdrv_is_first_non_filter(bs)) {
2553         error_setg(errp, QERR_FEATURE_DISABLED, "resize");
2554         goto out;
2555     }
2556 
2557     if (size < 0) {
2558         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2559         goto out;
2560     }
2561 
2562     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2563         error_setg(errp, QERR_DEVICE_IN_USE, device);
2564         goto out;
2565     }
2566 
2567     /* complete all in-flight operations before resizing the device */
2568     bdrv_drain_all();
2569 
2570     ret = bdrv_truncate(bs, size);
2571     switch (ret) {
2572     case 0:
2573         break;
2574     case -ENOMEDIUM:
2575         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2576         break;
2577     case -ENOTSUP:
2578         error_setg(errp, QERR_UNSUPPORTED);
2579         break;
2580     case -EACCES:
2581         error_setg(errp, "Device '%s' is read only", device);
2582         break;
2583     case -EBUSY:
2584         error_setg(errp, QERR_DEVICE_IN_USE, device);
2585         break;
2586     default:
2587         error_setg_errno(errp, -ret, "Could not resize");
2588         break;
2589     }
2590 
2591 out:
2592     aio_context_release(aio_context);
2593 }
2594 
2595 static void block_job_cb(void *opaque, int ret)
2596 {
2597     /* Note that this function may be executed from another AioContext besides
2598      * the QEMU main loop.  If you need to access anything that assumes the
2599      * QEMU global mutex, use a BH or introduce a mutex.
2600      */
2601 
2602     BlockDriverState *bs = opaque;
2603     const char *msg = NULL;
2604 
2605     trace_block_job_cb(bs, bs->job, ret);
2606 
2607     assert(bs->job);
2608 
2609     if (ret < 0) {
2610         msg = strerror(-ret);
2611     }
2612 
2613     if (block_job_is_cancelled(bs->job)) {
2614         block_job_event_cancelled(bs->job);
2615     } else {
2616         block_job_event_completed(bs->job, msg);
2617     }
2618 
2619     bdrv_put_ref_bh_schedule(bs);
2620 }
2621 
2622 void qmp_block_stream(const char *device,
2623                       bool has_base, const char *base,
2624                       bool has_backing_file, const char *backing_file,
2625                       bool has_speed, int64_t speed,
2626                       bool has_on_error, BlockdevOnError on_error,
2627                       Error **errp)
2628 {
2629     BlockBackend *blk;
2630     BlockDriverState *bs;
2631     BlockDriverState *base_bs = NULL;
2632     AioContext *aio_context;
2633     Error *local_err = NULL;
2634     const char *base_name = NULL;
2635 
2636     if (!has_on_error) {
2637         on_error = BLOCKDEV_ON_ERROR_REPORT;
2638     }
2639 
2640     blk = blk_by_name(device);
2641     if (!blk) {
2642         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2643                   "Device '%s' not found", device);
2644         return;
2645     }
2646 
2647     aio_context = blk_get_aio_context(blk);
2648     aio_context_acquire(aio_context);
2649 
2650     if (!blk_is_available(blk)) {
2651         error_setg(errp, "Device '%s' has no medium", device);
2652         goto out;
2653     }
2654     bs = blk_bs(blk);
2655 
2656     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
2657         goto out;
2658     }
2659 
2660     if (has_base) {
2661         base_bs = bdrv_find_backing_image(bs, base);
2662         if (base_bs == NULL) {
2663             error_setg(errp, QERR_BASE_NOT_FOUND, base);
2664             goto out;
2665         }
2666         assert(bdrv_get_aio_context(base_bs) == aio_context);
2667         base_name = base;
2668     }
2669 
2670     /* if we are streaming the entire chain, the result will have no backing
2671      * file, and specifying one is therefore an error */
2672     if (base_bs == NULL && has_backing_file) {
2673         error_setg(errp, "backing file specified, but streaming the "
2674                          "entire chain");
2675         goto out;
2676     }
2677 
2678     /* backing_file string overrides base bs filename */
2679     base_name = has_backing_file ? backing_file : base_name;
2680 
2681     stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2682                  on_error, block_job_cb, bs, &local_err);
2683     if (local_err) {
2684         error_propagate(errp, local_err);
2685         goto out;
2686     }
2687 
2688     trace_qmp_block_stream(bs, bs->job);
2689 
2690 out:
2691     aio_context_release(aio_context);
2692 }
2693 
2694 void qmp_block_commit(const char *device,
2695                       bool has_base, const char *base,
2696                       bool has_top, const char *top,
2697                       bool has_backing_file, const char *backing_file,
2698                       bool has_speed, int64_t speed,
2699                       Error **errp)
2700 {
2701     BlockBackend *blk;
2702     BlockDriverState *bs;
2703     BlockDriverState *base_bs, *top_bs;
2704     AioContext *aio_context;
2705     Error *local_err = NULL;
2706     /* This will be part of the QMP command, if/when the
2707      * BlockdevOnError change for blkmirror makes it in
2708      */
2709     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2710 
2711     if (!has_speed) {
2712         speed = 0;
2713     }
2714 
2715     /* Important Note:
2716      *  libvirt relies on the DeviceNotFound error class in order to probe for
2717      *  live commit feature versions; for this to work, we must make sure to
2718      *  perform the device lookup before any generic errors that may occur in a
2719      *  scenario in which all optional arguments are omitted. */
2720     blk = blk_by_name(device);
2721     if (!blk) {
2722         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2723                   "Device '%s' not found", device);
2724         return;
2725     }
2726 
2727     aio_context = blk_get_aio_context(blk);
2728     aio_context_acquire(aio_context);
2729 
2730     if (!blk_is_available(blk)) {
2731         error_setg(errp, "Device '%s' has no medium", device);
2732         goto out;
2733     }
2734     bs = blk_bs(blk);
2735 
2736     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2737         goto out;
2738     }
2739 
2740     /* default top_bs is the active layer */
2741     top_bs = bs;
2742 
2743     if (has_top && top) {
2744         if (strcmp(bs->filename, top) != 0) {
2745             top_bs = bdrv_find_backing_image(bs, top);
2746         }
2747     }
2748 
2749     if (top_bs == NULL) {
2750         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2751         goto out;
2752     }
2753 
2754     assert(bdrv_get_aio_context(top_bs) == aio_context);
2755 
2756     if (has_base && base) {
2757         base_bs = bdrv_find_backing_image(top_bs, base);
2758     } else {
2759         base_bs = bdrv_find_base(top_bs);
2760     }
2761 
2762     if (base_bs == NULL) {
2763         error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2764         goto out;
2765     }
2766 
2767     assert(bdrv_get_aio_context(base_bs) == aio_context);
2768 
2769     if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2770         goto out;
2771     }
2772 
2773     /* Do not allow attempts to commit an image into itself */
2774     if (top_bs == base_bs) {
2775         error_setg(errp, "cannot commit an image into itself");
2776         goto out;
2777     }
2778 
2779     if (top_bs == bs) {
2780         if (has_backing_file) {
2781             error_setg(errp, "'backing-file' specified,"
2782                              " but 'top' is the active layer");
2783             goto out;
2784         }
2785         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2786                             bs, &local_err);
2787     } else {
2788         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2789                      has_backing_file ? backing_file : NULL, &local_err);
2790     }
2791     if (local_err != NULL) {
2792         error_propagate(errp, local_err);
2793         goto out;
2794     }
2795 
2796 out:
2797     aio_context_release(aio_context);
2798 }
2799 
2800 void qmp_drive_backup(const char *device, const char *target,
2801                       bool has_format, const char *format,
2802                       enum MirrorSyncMode sync,
2803                       bool has_mode, enum NewImageMode mode,
2804                       bool has_speed, int64_t speed,
2805                       bool has_bitmap, const char *bitmap,
2806                       bool has_on_source_error, BlockdevOnError on_source_error,
2807                       bool has_on_target_error, BlockdevOnError on_target_error,
2808                       Error **errp)
2809 {
2810     BlockBackend *blk;
2811     BlockDriverState *bs;
2812     BlockDriverState *target_bs;
2813     BlockDriverState *source = NULL;
2814     BdrvDirtyBitmap *bmap = NULL;
2815     AioContext *aio_context;
2816     QDict *options = NULL;
2817     Error *local_err = NULL;
2818     int flags;
2819     int64_t size;
2820     int ret;
2821 
2822     if (!has_speed) {
2823         speed = 0;
2824     }
2825     if (!has_on_source_error) {
2826         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2827     }
2828     if (!has_on_target_error) {
2829         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2830     }
2831     if (!has_mode) {
2832         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2833     }
2834 
2835     blk = blk_by_name(device);
2836     if (!blk) {
2837         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2838                   "Device '%s' not found", device);
2839         return;
2840     }
2841 
2842     aio_context = blk_get_aio_context(blk);
2843     aio_context_acquire(aio_context);
2844 
2845     /* Although backup_run has this check too, we need to use bs->drv below, so
2846      * do an early check redundantly. */
2847     if (!blk_is_available(blk)) {
2848         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2849         goto out;
2850     }
2851     bs = blk_bs(blk);
2852 
2853     if (!has_format) {
2854         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2855     }
2856 
2857     /* Early check to avoid creating target */
2858     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2859         goto out;
2860     }
2861 
2862     flags = bs->open_flags | BDRV_O_RDWR;
2863 
2864     /* See if we have a backing HD we can use to create our new image
2865      * on top of. */
2866     if (sync == MIRROR_SYNC_MODE_TOP) {
2867         source = backing_bs(bs);
2868         if (!source) {
2869             sync = MIRROR_SYNC_MODE_FULL;
2870         }
2871     }
2872     if (sync == MIRROR_SYNC_MODE_NONE) {
2873         source = bs;
2874     }
2875 
2876     size = bdrv_getlength(bs);
2877     if (size < 0) {
2878         error_setg_errno(errp, -size, "bdrv_getlength failed");
2879         goto out;
2880     }
2881 
2882     if (mode != NEW_IMAGE_MODE_EXISTING) {
2883         assert(format);
2884         if (source) {
2885             bdrv_img_create(target, format, source->filename,
2886                             source->drv->format_name, NULL,
2887                             size, flags, &local_err, false);
2888         } else {
2889             bdrv_img_create(target, format, NULL, NULL, NULL,
2890                             size, flags, &local_err, false);
2891         }
2892     }
2893 
2894     if (local_err) {
2895         error_propagate(errp, local_err);
2896         goto out;
2897     }
2898 
2899     if (format) {
2900         options = qdict_new();
2901         qdict_put(options, "driver", qstring_from_str(format));
2902     }
2903 
2904     target_bs = NULL;
2905     ret = bdrv_open(&target_bs, target, NULL, options, flags, &local_err);
2906     if (ret < 0) {
2907         error_propagate(errp, local_err);
2908         goto out;
2909     }
2910 
2911     bdrv_set_aio_context(target_bs, aio_context);
2912 
2913     if (has_bitmap) {
2914         bmap = bdrv_find_dirty_bitmap(bs, bitmap);
2915         if (!bmap) {
2916             error_setg(errp, "Bitmap '%s' could not be found", bitmap);
2917             goto out;
2918         }
2919     }
2920 
2921     backup_start(bs, target_bs, speed, sync, bmap,
2922                  on_source_error, on_target_error,
2923                  block_job_cb, bs, &local_err);
2924     if (local_err != NULL) {
2925         bdrv_unref(target_bs);
2926         error_propagate(errp, local_err);
2927         goto out;
2928     }
2929 
2930 out:
2931     aio_context_release(aio_context);
2932 }
2933 
2934 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2935 {
2936     return bdrv_named_nodes_list(errp);
2937 }
2938 
2939 void qmp_blockdev_backup(const char *device, const char *target,
2940                          enum MirrorSyncMode sync,
2941                          bool has_speed, int64_t speed,
2942                          bool has_on_source_error,
2943                          BlockdevOnError on_source_error,
2944                          bool has_on_target_error,
2945                          BlockdevOnError on_target_error,
2946                          Error **errp)
2947 {
2948     BlockBackend *blk, *target_blk;
2949     BlockDriverState *bs;
2950     BlockDriverState *target_bs;
2951     Error *local_err = NULL;
2952     AioContext *aio_context;
2953 
2954     if (!has_speed) {
2955         speed = 0;
2956     }
2957     if (!has_on_source_error) {
2958         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2959     }
2960     if (!has_on_target_error) {
2961         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2962     }
2963 
2964     blk = blk_by_name(device);
2965     if (!blk) {
2966         error_setg(errp, "Device '%s' not found", device);
2967         return;
2968     }
2969 
2970     aio_context = blk_get_aio_context(blk);
2971     aio_context_acquire(aio_context);
2972 
2973     if (!blk_is_available(blk)) {
2974         error_setg(errp, "Device '%s' has no medium", device);
2975         goto out;
2976     }
2977     bs = blk_bs(blk);
2978 
2979     target_blk = blk_by_name(target);
2980     if (!target_blk) {
2981         error_setg(errp, "Device '%s' not found", target);
2982         goto out;
2983     }
2984 
2985     if (!blk_is_available(target_blk)) {
2986         error_setg(errp, "Device '%s' has no medium", target);
2987         goto out;
2988     }
2989     target_bs = blk_bs(target_blk);
2990 
2991     bdrv_ref(target_bs);
2992     bdrv_set_aio_context(target_bs, aio_context);
2993     backup_start(bs, target_bs, speed, sync, NULL, on_source_error,
2994                  on_target_error, block_job_cb, bs, &local_err);
2995     if (local_err != NULL) {
2996         bdrv_unref(target_bs);
2997         error_propagate(errp, local_err);
2998     }
2999 out:
3000     aio_context_release(aio_context);
3001 }
3002 
3003 void qmp_drive_mirror(const char *device, const char *target,
3004                       bool has_format, const char *format,
3005                       bool has_node_name, const char *node_name,
3006                       bool has_replaces, const char *replaces,
3007                       enum MirrorSyncMode sync,
3008                       bool has_mode, enum NewImageMode mode,
3009                       bool has_speed, int64_t speed,
3010                       bool has_granularity, uint32_t granularity,
3011                       bool has_buf_size, int64_t buf_size,
3012                       bool has_on_source_error, BlockdevOnError on_source_error,
3013                       bool has_on_target_error, BlockdevOnError on_target_error,
3014                       bool has_unmap, bool unmap,
3015                       Error **errp)
3016 {
3017     BlockBackend *blk;
3018     BlockDriverState *bs;
3019     BlockDriverState *source, *target_bs;
3020     AioContext *aio_context;
3021     Error *local_err = NULL;
3022     QDict *options;
3023     int flags;
3024     int64_t size;
3025     int ret;
3026 
3027     if (!has_speed) {
3028         speed = 0;
3029     }
3030     if (!has_on_source_error) {
3031         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3032     }
3033     if (!has_on_target_error) {
3034         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3035     }
3036     if (!has_mode) {
3037         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3038     }
3039     if (!has_granularity) {
3040         granularity = 0;
3041     }
3042     if (!has_buf_size) {
3043         buf_size = 0;
3044     }
3045     if (!has_unmap) {
3046         unmap = true;
3047     }
3048 
3049     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3050         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3051                    "a value in range [512B, 64MB]");
3052         return;
3053     }
3054     if (granularity & (granularity - 1)) {
3055         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3056                    "power of 2");
3057         return;
3058     }
3059 
3060     blk = blk_by_name(device);
3061     if (!blk) {
3062         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3063                   "Device '%s' not found", device);
3064         return;
3065     }
3066 
3067     aio_context = blk_get_aio_context(blk);
3068     aio_context_acquire(aio_context);
3069 
3070     if (!blk_is_available(blk)) {
3071         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
3072         goto out;
3073     }
3074     bs = blk_bs(blk);
3075 
3076     if (!has_format) {
3077         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
3078     }
3079 
3080     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
3081         goto out;
3082     }
3083 
3084     flags = bs->open_flags | BDRV_O_RDWR;
3085     source = backing_bs(bs);
3086     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
3087         sync = MIRROR_SYNC_MODE_FULL;
3088     }
3089     if (sync == MIRROR_SYNC_MODE_NONE) {
3090         source = bs;
3091     }
3092 
3093     size = bdrv_getlength(bs);
3094     if (size < 0) {
3095         error_setg_errno(errp, -size, "bdrv_getlength failed");
3096         goto out;
3097     }
3098 
3099     if (has_replaces) {
3100         BlockDriverState *to_replace_bs;
3101         AioContext *replace_aio_context;
3102         int64_t replace_size;
3103 
3104         if (!has_node_name) {
3105             error_setg(errp, "a node-name must be provided when replacing a"
3106                              " named node of the graph");
3107             goto out;
3108         }
3109 
3110         to_replace_bs = check_to_replace_node(bs, replaces, &local_err);
3111 
3112         if (!to_replace_bs) {
3113             error_propagate(errp, local_err);
3114             goto out;
3115         }
3116 
3117         replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3118         aio_context_acquire(replace_aio_context);
3119         replace_size = bdrv_getlength(to_replace_bs);
3120         aio_context_release(replace_aio_context);
3121 
3122         if (size != replace_size) {
3123             error_setg(errp, "cannot replace image with a mirror image of "
3124                              "different size");
3125             goto out;
3126         }
3127     }
3128 
3129     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
3130         && mode != NEW_IMAGE_MODE_EXISTING)
3131     {
3132         /* create new image w/o backing file */
3133         assert(format);
3134         bdrv_img_create(target, format,
3135                         NULL, NULL, NULL, size, flags, &local_err, false);
3136     } else {
3137         switch (mode) {
3138         case NEW_IMAGE_MODE_EXISTING:
3139             break;
3140         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3141             /* create new image with backing file */
3142             bdrv_img_create(target, format,
3143                             source->filename,
3144                             source->drv->format_name,
3145                             NULL, size, flags, &local_err, false);
3146             break;
3147         default:
3148             abort();
3149         }
3150     }
3151 
3152     if (local_err) {
3153         error_propagate(errp, local_err);
3154         goto out;
3155     }
3156 
3157     options = qdict_new();
3158     if (has_node_name) {
3159         qdict_put(options, "node-name", qstring_from_str(node_name));
3160     }
3161     if (format) {
3162         qdict_put(options, "driver", qstring_from_str(format));
3163     }
3164 
3165     /* Mirroring takes care of copy-on-write using the source's backing
3166      * file.
3167      */
3168     target_bs = NULL;
3169     ret = bdrv_open(&target_bs, target, NULL, options,
3170                     flags | BDRV_O_NO_BACKING, &local_err);
3171     if (ret < 0) {
3172         error_propagate(errp, local_err);
3173         goto out;
3174     }
3175 
3176     bdrv_set_aio_context(target_bs, aio_context);
3177 
3178     /* pass the node name to replace to mirror start since it's loose coupling
3179      * and will allow to check whether the node still exist at mirror completion
3180      */
3181     mirror_start(bs, target_bs,
3182                  has_replaces ? replaces : NULL,
3183                  speed, granularity, buf_size, sync,
3184                  on_source_error, on_target_error,
3185                  unmap,
3186                  block_job_cb, bs, &local_err);
3187     if (local_err != NULL) {
3188         bdrv_unref(target_bs);
3189         error_propagate(errp, local_err);
3190         goto out;
3191     }
3192 
3193 out:
3194     aio_context_release(aio_context);
3195 }
3196 
3197 /* Get the block job for a given device name and acquire its AioContext */
3198 static BlockJob *find_block_job(const char *device, AioContext **aio_context,
3199                                 Error **errp)
3200 {
3201     BlockBackend *blk;
3202     BlockDriverState *bs;
3203 
3204     *aio_context = NULL;
3205 
3206     blk = blk_by_name(device);
3207     if (!blk) {
3208         goto notfound;
3209     }
3210 
3211     *aio_context = blk_get_aio_context(blk);
3212     aio_context_acquire(*aio_context);
3213 
3214     if (!blk_is_available(blk)) {
3215         goto notfound;
3216     }
3217     bs = blk_bs(blk);
3218 
3219     if (!bs->job) {
3220         goto notfound;
3221     }
3222 
3223     return bs->job;
3224 
3225 notfound:
3226     error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3227               "No active block job on device '%s'", device);
3228     if (*aio_context) {
3229         aio_context_release(*aio_context);
3230         *aio_context = NULL;
3231     }
3232     return NULL;
3233 }
3234 
3235 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3236 {
3237     AioContext *aio_context;
3238     BlockJob *job = find_block_job(device, &aio_context, errp);
3239 
3240     if (!job) {
3241         return;
3242     }
3243 
3244     block_job_set_speed(job, speed, errp);
3245     aio_context_release(aio_context);
3246 }
3247 
3248 void qmp_block_job_cancel(const char *device,
3249                           bool has_force, bool force, Error **errp)
3250 {
3251     AioContext *aio_context;
3252     BlockJob *job = find_block_job(device, &aio_context, errp);
3253 
3254     if (!job) {
3255         return;
3256     }
3257 
3258     if (!has_force) {
3259         force = false;
3260     }
3261 
3262     if (job->user_paused && !force) {
3263         error_setg(errp, "The block job for device '%s' is currently paused",
3264                    device);
3265         goto out;
3266     }
3267 
3268     trace_qmp_block_job_cancel(job);
3269     block_job_cancel(job);
3270 out:
3271     aio_context_release(aio_context);
3272 }
3273 
3274 void qmp_block_job_pause(const char *device, Error **errp)
3275 {
3276     AioContext *aio_context;
3277     BlockJob *job = find_block_job(device, &aio_context, errp);
3278 
3279     if (!job || job->user_paused) {
3280         return;
3281     }
3282 
3283     job->user_paused = true;
3284     trace_qmp_block_job_pause(job);
3285     block_job_pause(job);
3286     aio_context_release(aio_context);
3287 }
3288 
3289 void qmp_block_job_resume(const char *device, Error **errp)
3290 {
3291     AioContext *aio_context;
3292     BlockJob *job = find_block_job(device, &aio_context, errp);
3293 
3294     if (!job || !job->user_paused) {
3295         return;
3296     }
3297 
3298     job->user_paused = false;
3299     trace_qmp_block_job_resume(job);
3300     block_job_resume(job);
3301     aio_context_release(aio_context);
3302 }
3303 
3304 void qmp_block_job_complete(const char *device, Error **errp)
3305 {
3306     AioContext *aio_context;
3307     BlockJob *job = find_block_job(device, &aio_context, errp);
3308 
3309     if (!job) {
3310         return;
3311     }
3312 
3313     trace_qmp_block_job_complete(job);
3314     block_job_complete(job, errp);
3315     aio_context_release(aio_context);
3316 }
3317 
3318 void qmp_change_backing_file(const char *device,
3319                              const char *image_node_name,
3320                              const char *backing_file,
3321                              Error **errp)
3322 {
3323     BlockBackend *blk;
3324     BlockDriverState *bs = NULL;
3325     AioContext *aio_context;
3326     BlockDriverState *image_bs = NULL;
3327     Error *local_err = NULL;
3328     bool ro;
3329     int open_flags;
3330     int ret;
3331 
3332     blk = blk_by_name(device);
3333     if (!blk) {
3334         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3335                   "Device '%s' not found", device);
3336         return;
3337     }
3338 
3339     aio_context = blk_get_aio_context(blk);
3340     aio_context_acquire(aio_context);
3341 
3342     if (!blk_is_available(blk)) {
3343         error_setg(errp, "Device '%s' has no medium", device);
3344         goto out;
3345     }
3346     bs = blk_bs(blk);
3347 
3348     image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3349     if (local_err) {
3350         error_propagate(errp, local_err);
3351         goto out;
3352     }
3353 
3354     if (!image_bs) {
3355         error_setg(errp, "image file not found");
3356         goto out;
3357     }
3358 
3359     if (bdrv_find_base(image_bs) == image_bs) {
3360         error_setg(errp, "not allowing backing file change on an image "
3361                          "without a backing file");
3362         goto out;
3363     }
3364 
3365     /* even though we are not necessarily operating on bs, we need it to
3366      * determine if block ops are currently prohibited on the chain */
3367     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3368         goto out;
3369     }
3370 
3371     /* final sanity check */
3372     if (!bdrv_chain_contains(bs, image_bs)) {
3373         error_setg(errp, "'%s' and image file are not in the same chain",
3374                    device);
3375         goto out;
3376     }
3377 
3378     /* if not r/w, reopen to make r/w */
3379     open_flags = image_bs->open_flags;
3380     ro = bdrv_is_read_only(image_bs);
3381 
3382     if (ro) {
3383         bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
3384         if (local_err) {
3385             error_propagate(errp, local_err);
3386             goto out;
3387         }
3388     }
3389 
3390     ret = bdrv_change_backing_file(image_bs, backing_file,
3391                                image_bs->drv ? image_bs->drv->format_name : "");
3392 
3393     if (ret < 0) {
3394         error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3395                          backing_file);
3396         /* don't exit here, so we can try to restore open flags if
3397          * appropriate */
3398     }
3399 
3400     if (ro) {
3401         bdrv_reopen(image_bs, open_flags, &local_err);
3402         if (local_err) {
3403             error_propagate(errp, local_err); /* will preserve prior errp */
3404         }
3405     }
3406 
3407 out:
3408     aio_context_release(aio_context);
3409 }
3410 
3411 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3412 {
3413     QmpOutputVisitor *ov = qmp_output_visitor_new();
3414     BlockDriverState *bs;
3415     BlockBackend *blk = NULL;
3416     QObject *obj;
3417     QDict *qdict;
3418     Error *local_err = NULL;
3419 
3420     /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
3421      * cache.direct=false instead of silently switching to aio=threads, except
3422      * when called from drive_new().
3423      *
3424      * For now, simply forbidding the combination for all drivers will do. */
3425     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
3426         bool direct = options->has_cache &&
3427                       options->cache->has_direct &&
3428                       options->cache->direct;
3429         if (!direct) {
3430             error_setg(errp, "aio=native requires cache.direct=true");
3431             goto fail;
3432         }
3433     }
3434 
3435     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
3436                                &options, NULL, &local_err);
3437     if (local_err) {
3438         error_propagate(errp, local_err);
3439         goto fail;
3440     }
3441 
3442     obj = qmp_output_get_qobject(ov);
3443     qdict = qobject_to_qdict(obj);
3444 
3445     qdict_flatten(qdict);
3446 
3447     if (options->has_id) {
3448         blk = blockdev_init(NULL, qdict, &local_err);
3449         if (local_err) {
3450             error_propagate(errp, local_err);
3451             goto fail;
3452         }
3453 
3454         bs = blk_bs(blk);
3455     } else {
3456         if (!qdict_get_try_str(qdict, "node-name")) {
3457             error_setg(errp, "'id' and/or 'node-name' need to be specified for "
3458                        "the root node");
3459             goto fail;
3460         }
3461 
3462         bs = bds_tree_init(qdict, errp);
3463         if (!bs) {
3464             goto fail;
3465         }
3466     }
3467 
3468     if (bs && bdrv_key_required(bs)) {
3469         if (blk) {
3470             blk_unref(blk);
3471         } else {
3472             bdrv_unref(bs);
3473         }
3474         error_setg(errp, "blockdev-add doesn't support encrypted devices");
3475         goto fail;
3476     }
3477 
3478 fail:
3479     qmp_output_visitor_cleanup(ov);
3480 }
3481 
3482 void qmp_x_blockdev_del(bool has_id, const char *id,
3483                         bool has_node_name, const char *node_name, Error **errp)
3484 {
3485     AioContext *aio_context;
3486     BlockBackend *blk;
3487     BlockDriverState *bs;
3488 
3489     if (has_id && has_node_name) {
3490         error_setg(errp, "Only one of id and node-name must be specified");
3491         return;
3492     } else if (!has_id && !has_node_name) {
3493         error_setg(errp, "No block device specified");
3494         return;
3495     }
3496 
3497     if (has_id) {
3498         blk = blk_by_name(id);
3499         if (!blk) {
3500             error_setg(errp, "Cannot find block backend %s", id);
3501             return;
3502         }
3503         if (blk_get_refcnt(blk) > 1) {
3504             error_setg(errp, "Block backend %s is in use", id);
3505             return;
3506         }
3507         bs = blk_bs(blk);
3508         aio_context = blk_get_aio_context(blk);
3509     } else {
3510         bs = bdrv_find_node(node_name);
3511         if (!bs) {
3512             error_setg(errp, "Cannot find node %s", node_name);
3513             return;
3514         }
3515         blk = bs->blk;
3516         if (blk) {
3517             error_setg(errp, "Node %s is in use by %s",
3518                        node_name, blk_name(blk));
3519             return;
3520         }
3521         aio_context = bdrv_get_aio_context(bs);
3522     }
3523 
3524     aio_context_acquire(aio_context);
3525 
3526     if (bs) {
3527         if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3528             goto out;
3529         }
3530 
3531         if (bs->refcnt > 1 || !QLIST_EMPTY(&bs->parents)) {
3532             error_setg(errp, "Block device %s is in use",
3533                        bdrv_get_device_or_node_name(bs));
3534             goto out;
3535         }
3536     }
3537 
3538     if (blk) {
3539         blk_unref(blk);
3540     } else {
3541         bdrv_unref(bs);
3542     }
3543 
3544 out:
3545     aio_context_release(aio_context);
3546 }
3547 
3548 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3549 {
3550     BlockJobInfoList *head = NULL, **p_next = &head;
3551     BlockDriverState *bs;
3552 
3553     for (bs = bdrv_next(NULL); bs; bs = bdrv_next(bs)) {
3554         AioContext *aio_context = bdrv_get_aio_context(bs);
3555 
3556         aio_context_acquire(aio_context);
3557 
3558         if (bs->job) {
3559             BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
3560             elem->value = block_job_query(bs->job);
3561             *p_next = elem;
3562             p_next = &elem->next;
3563         }
3564 
3565         aio_context_release(aio_context);
3566     }
3567 
3568     return head;
3569 }
3570 
3571 QemuOptsList qemu_common_drive_opts = {
3572     .name = "drive",
3573     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3574     .desc = {
3575         {
3576             .name = "snapshot",
3577             .type = QEMU_OPT_BOOL,
3578             .help = "enable/disable snapshot mode",
3579         },{
3580             .name = "discard",
3581             .type = QEMU_OPT_STRING,
3582             .help = "discard operation (ignore/off, unmap/on)",
3583         },{
3584             .name = BDRV_OPT_CACHE_WB,
3585             .type = QEMU_OPT_BOOL,
3586             .help = "enables writeback mode for any caches",
3587         },{
3588             .name = BDRV_OPT_CACHE_DIRECT,
3589             .type = QEMU_OPT_BOOL,
3590             .help = "enables use of O_DIRECT (bypass the host page cache)",
3591         },{
3592             .name = BDRV_OPT_CACHE_NO_FLUSH,
3593             .type = QEMU_OPT_BOOL,
3594             .help = "ignore any flush requests for the device",
3595         },{
3596             .name = "aio",
3597             .type = QEMU_OPT_STRING,
3598             .help = "host AIO implementation (threads, native)",
3599         },{
3600             .name = "format",
3601             .type = QEMU_OPT_STRING,
3602             .help = "disk format (raw, qcow2, ...)",
3603         },{
3604             .name = "rerror",
3605             .type = QEMU_OPT_STRING,
3606             .help = "read error action",
3607         },{
3608             .name = "werror",
3609             .type = QEMU_OPT_STRING,
3610             .help = "write error action",
3611         },{
3612             .name = "read-only",
3613             .type = QEMU_OPT_BOOL,
3614             .help = "open drive file as read-only",
3615         },{
3616             .name = "throttling.iops-total",
3617             .type = QEMU_OPT_NUMBER,
3618             .help = "limit total I/O operations per second",
3619         },{
3620             .name = "throttling.iops-read",
3621             .type = QEMU_OPT_NUMBER,
3622             .help = "limit read operations per second",
3623         },{
3624             .name = "throttling.iops-write",
3625             .type = QEMU_OPT_NUMBER,
3626             .help = "limit write operations per second",
3627         },{
3628             .name = "throttling.bps-total",
3629             .type = QEMU_OPT_NUMBER,
3630             .help = "limit total bytes per second",
3631         },{
3632             .name = "throttling.bps-read",
3633             .type = QEMU_OPT_NUMBER,
3634             .help = "limit read bytes per second",
3635         },{
3636             .name = "throttling.bps-write",
3637             .type = QEMU_OPT_NUMBER,
3638             .help = "limit write bytes per second",
3639         },{
3640             .name = "throttling.iops-total-max",
3641             .type = QEMU_OPT_NUMBER,
3642             .help = "I/O operations burst",
3643         },{
3644             .name = "throttling.iops-read-max",
3645             .type = QEMU_OPT_NUMBER,
3646             .help = "I/O operations read burst",
3647         },{
3648             .name = "throttling.iops-write-max",
3649             .type = QEMU_OPT_NUMBER,
3650             .help = "I/O operations write burst",
3651         },{
3652             .name = "throttling.bps-total-max",
3653             .type = QEMU_OPT_NUMBER,
3654             .help = "total bytes burst",
3655         },{
3656             .name = "throttling.bps-read-max",
3657             .type = QEMU_OPT_NUMBER,
3658             .help = "total bytes read burst",
3659         },{
3660             .name = "throttling.bps-write-max",
3661             .type = QEMU_OPT_NUMBER,
3662             .help = "total bytes write burst",
3663         },{
3664             .name = "throttling.iops-size",
3665             .type = QEMU_OPT_NUMBER,
3666             .help = "when limiting by iops max size of an I/O in bytes",
3667         },{
3668             .name = "throttling.group",
3669             .type = QEMU_OPT_STRING,
3670             .help = "name of the block throttling group",
3671         },{
3672             .name = "copy-on-read",
3673             .type = QEMU_OPT_BOOL,
3674             .help = "copy read data from backing file into image file",
3675         },{
3676             .name = "detect-zeroes",
3677             .type = QEMU_OPT_STRING,
3678             .help = "try to optimize zero writes (off, on, unmap)",
3679         },
3680         { /* end of list */ }
3681     },
3682 };
3683 
3684 static QemuOptsList qemu_root_bds_opts = {
3685     .name = "root-bds",
3686     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3687     .desc = {
3688         {
3689             .name = "discard",
3690             .type = QEMU_OPT_STRING,
3691             .help = "discard operation (ignore/off, unmap/on)",
3692         },{
3693             .name = "cache.writeback",
3694             .type = QEMU_OPT_BOOL,
3695             .help = "enables writeback mode for any caches",
3696         },{
3697             .name = "cache.direct",
3698             .type = QEMU_OPT_BOOL,
3699             .help = "enables use of O_DIRECT (bypass the host page cache)",
3700         },{
3701             .name = "cache.no-flush",
3702             .type = QEMU_OPT_BOOL,
3703             .help = "ignore any flush requests for the device",
3704         },{
3705             .name = "aio",
3706             .type = QEMU_OPT_STRING,
3707             .help = "host AIO implementation (threads, native)",
3708         },{
3709             .name = "read-only",
3710             .type = QEMU_OPT_BOOL,
3711             .help = "open drive file as read-only",
3712         },{
3713             .name = "copy-on-read",
3714             .type = QEMU_OPT_BOOL,
3715             .help = "copy read data from backing file into image file",
3716         },{
3717             .name = "detect-zeroes",
3718             .type = QEMU_OPT_STRING,
3719             .help = "try to optimize zero writes (off, on, unmap)",
3720         },
3721         { /* end of list */ }
3722     },
3723 };
3724 
3725 QemuOptsList qemu_drive_opts = {
3726     .name = "drive",
3727     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3728     .desc = {
3729         /*
3730          * no elements => accept any params
3731          * validation will happen later
3732          */
3733         { /* end of list */ }
3734     },
3735 };
3736