xref: /qemu/blockdev.c (revision 67cc32eb)
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     aio_context = bdrv_get_aio_context(bs);
128     aio_context_acquire(aio_context);
129 
130     if (bs->job) {
131         block_job_cancel(bs->job);
132     }
133 
134     aio_context_release(aio_context);
135 
136     dinfo->auto_del = 1;
137 }
138 
139 void blockdev_auto_del(BlockBackend *blk)
140 {
141     DriveInfo *dinfo = blk_legacy_dinfo(blk);
142 
143     if (dinfo && dinfo->auto_del) {
144         blk_unref(blk);
145     }
146 }
147 
148 /**
149  * Returns the current mapping of how many units per bus
150  * a particular interface can support.
151  *
152  *  A positive integer indicates n units per bus.
153  *  0 implies the mapping has not been established.
154  * -1 indicates an invalid BlockInterfaceType was given.
155  */
156 int drive_get_max_devs(BlockInterfaceType type)
157 {
158     if (type >= IF_IDE && type < IF_COUNT) {
159         return if_max_devs[type];
160     }
161 
162     return -1;
163 }
164 
165 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
166 {
167     int max_devs = if_max_devs[type];
168     return max_devs ? index / max_devs : 0;
169 }
170 
171 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
172 {
173     int max_devs = if_max_devs[type];
174     return max_devs ? index % max_devs : index;
175 }
176 
177 QemuOpts *drive_def(const char *optstr)
178 {
179     return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
180 }
181 
182 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
183                     const char *optstr)
184 {
185     QemuOpts *opts;
186 
187     opts = drive_def(optstr);
188     if (!opts) {
189         return NULL;
190     }
191     if (type != IF_DEFAULT) {
192         qemu_opt_set(opts, "if", if_name[type], &error_abort);
193     }
194     if (index >= 0) {
195         qemu_opt_set_number(opts, "index", index, &error_abort);
196     }
197     if (file)
198         qemu_opt_set(opts, "file", file, &error_abort);
199     return opts;
200 }
201 
202 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
203 {
204     BlockBackend *blk;
205     DriveInfo *dinfo;
206 
207     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
208         dinfo = blk_legacy_dinfo(blk);
209         if (dinfo && dinfo->type == type
210             && dinfo->bus == bus && dinfo->unit == unit) {
211             return dinfo;
212         }
213     }
214 
215     return NULL;
216 }
217 
218 bool drive_check_orphaned(void)
219 {
220     BlockBackend *blk;
221     DriveInfo *dinfo;
222     bool rs = false;
223 
224     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
225         dinfo = blk_legacy_dinfo(blk);
226         /* If dinfo->bdrv->dev is NULL, it has no device attached. */
227         /* Unless this is a default drive, this may be an oversight. */
228         if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
229             dinfo->type != IF_NONE) {
230             fprintf(stderr, "Warning: Orphaned drive without device: "
231                     "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
232                     blk_name(blk), blk_bs(blk)->filename, if_name[dinfo->type],
233                     dinfo->bus, dinfo->unit);
234             rs = true;
235         }
236     }
237 
238     return rs;
239 }
240 
241 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
242 {
243     return drive_get(type,
244                      drive_index_to_bus_id(type, index),
245                      drive_index_to_unit_id(type, index));
246 }
247 
248 int drive_get_max_bus(BlockInterfaceType type)
249 {
250     int max_bus;
251     BlockBackend *blk;
252     DriveInfo *dinfo;
253 
254     max_bus = -1;
255     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
256         dinfo = blk_legacy_dinfo(blk);
257         if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
258             max_bus = dinfo->bus;
259         }
260     }
261     return max_bus;
262 }
263 
264 /* Get a block device.  This should only be used for single-drive devices
265    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
266    appropriate bus.  */
267 DriveInfo *drive_get_next(BlockInterfaceType type)
268 {
269     static int next_block_unit[IF_COUNT];
270 
271     return drive_get(type, 0, next_block_unit[type]++);
272 }
273 
274 static void bdrv_format_print(void *opaque, const char *name)
275 {
276     error_printf(" %s", name);
277 }
278 
279 typedef struct {
280     QEMUBH *bh;
281     BlockDriverState *bs;
282 } BDRVPutRefBH;
283 
284 static void bdrv_put_ref_bh(void *opaque)
285 {
286     BDRVPutRefBH *s = opaque;
287 
288     bdrv_unref(s->bs);
289     qemu_bh_delete(s->bh);
290     g_free(s);
291 }
292 
293 /*
294  * Release a BDS reference in a BH
295  *
296  * It is not safe to use bdrv_unref() from a callback function when the callers
297  * still need the BlockDriverState.  In such cases we schedule a BH to release
298  * the reference.
299  */
300 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
301 {
302     BDRVPutRefBH *s;
303 
304     s = g_new(BDRVPutRefBH, 1);
305     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
306     s->bs = bs;
307     qemu_bh_schedule(s->bh);
308 }
309 
310 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
311 {
312     if (!strcmp(buf, "ignore")) {
313         return BLOCKDEV_ON_ERROR_IGNORE;
314     } else if (!is_read && !strcmp(buf, "enospc")) {
315         return BLOCKDEV_ON_ERROR_ENOSPC;
316     } else if (!strcmp(buf, "stop")) {
317         return BLOCKDEV_ON_ERROR_STOP;
318     } else if (!strcmp(buf, "report")) {
319         return BLOCKDEV_ON_ERROR_REPORT;
320     } else {
321         error_setg(errp, "'%s' invalid %s error action",
322                    buf, is_read ? "read" : "write");
323         return -1;
324     }
325 }
326 
327 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
328 {
329     if (throttle_conflicting(cfg)) {
330         error_setg(errp, "bps/iops/max total values and read/write values"
331                          " cannot be used at the same time");
332         return false;
333     }
334 
335     if (!throttle_is_valid(cfg)) {
336         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
337         return false;
338     }
339 
340     if (throttle_max_is_missing_limit(cfg)) {
341         error_setg(errp, "bps_max/iops_max require corresponding"
342                          " bps/iops values");
343         return false;
344     }
345 
346     return true;
347 }
348 
349 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
350 
351 /* Takes the ownership of bs_opts */
352 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
353                                    Error **errp)
354 {
355     const char *buf;
356     int ro = 0;
357     int bdrv_flags = 0;
358     int on_read_error, on_write_error;
359     BlockBackend *blk;
360     BlockDriverState *bs;
361     ThrottleConfig cfg;
362     int snapshot = 0;
363     bool copy_on_read;
364     Error *error = NULL;
365     QemuOpts *opts;
366     const char *id;
367     bool has_driver_specific_opts;
368     BlockdevDetectZeroesOptions detect_zeroes;
369     const char *throttling_group;
370 
371     /* Check common options by copying from bs_opts to opts, all other options
372      * stay in bs_opts for processing by bdrv_open(). */
373     id = qdict_get_try_str(bs_opts, "id");
374     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
375     if (error) {
376         error_propagate(errp, error);
377         goto err_no_opts;
378     }
379 
380     qemu_opts_absorb_qdict(opts, bs_opts, &error);
381     if (error) {
382         error_propagate(errp, error);
383         goto early_err;
384     }
385 
386     if (id) {
387         qdict_del(bs_opts, "id");
388     }
389 
390     has_driver_specific_opts = !!qdict_size(bs_opts);
391 
392     /* extract parameters */
393     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
394     ro = qemu_opt_get_bool(opts, "read-only", 0);
395     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
396 
397     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
398         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
399             error_setg(errp, "invalid discard option");
400             goto early_err;
401         }
402     }
403 
404     if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true)) {
405         bdrv_flags |= BDRV_O_CACHE_WB;
406     }
407     if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_DIRECT, false)) {
408         bdrv_flags |= BDRV_O_NOCACHE;
409     }
410     if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) {
411         bdrv_flags |= BDRV_O_NO_FLUSH;
412     }
413 
414 #ifdef CONFIG_LINUX_AIO
415     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
416         if (!strcmp(buf, "native")) {
417             bdrv_flags |= BDRV_O_NATIVE_AIO;
418         } else if (!strcmp(buf, "threads")) {
419             /* this is the default */
420         } else {
421            error_setg(errp, "invalid aio option");
422            goto early_err;
423         }
424     }
425 #endif
426 
427     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
428         if (is_help_option(buf)) {
429             error_printf("Supported formats:");
430             bdrv_iterate_format(bdrv_format_print, NULL);
431             error_printf("\n");
432             goto early_err;
433         }
434 
435         if (qdict_haskey(bs_opts, "driver")) {
436             error_setg(errp, "Cannot specify both 'driver' and 'format'");
437             goto early_err;
438         }
439         qdict_put(bs_opts, "driver", qstring_from_str(buf));
440     }
441 
442     /* disk I/O throttling */
443     memset(&cfg, 0, sizeof(cfg));
444     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
445         qemu_opt_get_number(opts, "throttling.bps-total", 0);
446     cfg.buckets[THROTTLE_BPS_READ].avg  =
447         qemu_opt_get_number(opts, "throttling.bps-read", 0);
448     cfg.buckets[THROTTLE_BPS_WRITE].avg =
449         qemu_opt_get_number(opts, "throttling.bps-write", 0);
450     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
451         qemu_opt_get_number(opts, "throttling.iops-total", 0);
452     cfg.buckets[THROTTLE_OPS_READ].avg =
453         qemu_opt_get_number(opts, "throttling.iops-read", 0);
454     cfg.buckets[THROTTLE_OPS_WRITE].avg =
455         qemu_opt_get_number(opts, "throttling.iops-write", 0);
456 
457     cfg.buckets[THROTTLE_BPS_TOTAL].max =
458         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
459     cfg.buckets[THROTTLE_BPS_READ].max  =
460         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
461     cfg.buckets[THROTTLE_BPS_WRITE].max =
462         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
463     cfg.buckets[THROTTLE_OPS_TOTAL].max =
464         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
465     cfg.buckets[THROTTLE_OPS_READ].max =
466         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
467     cfg.buckets[THROTTLE_OPS_WRITE].max =
468         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
469 
470     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
471 
472     throttling_group = qemu_opt_get(opts, "throttling.group");
473 
474     if (!check_throttle_config(&cfg, &error)) {
475         error_propagate(errp, error);
476         goto early_err;
477     }
478 
479     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
480     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
481         on_write_error = parse_block_error_action(buf, 0, &error);
482         if (error) {
483             error_propagate(errp, error);
484             goto early_err;
485         }
486     }
487 
488     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
489     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
490         on_read_error = parse_block_error_action(buf, 1, &error);
491         if (error) {
492             error_propagate(errp, error);
493             goto early_err;
494         }
495     }
496 
497     detect_zeroes =
498         qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
499                         qemu_opt_get(opts, "detect-zeroes"),
500                         BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
501                         BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
502                         &error);
503     if (error) {
504         error_propagate(errp, error);
505         goto early_err;
506     }
507 
508     if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
509         !(bdrv_flags & BDRV_O_UNMAP)) {
510         error_setg(errp, "setting detect-zeroes to unmap is not allowed "
511                          "without setting discard operation to unmap");
512         goto early_err;
513     }
514 
515     /* init */
516     if ((!file || !*file) && !has_driver_specific_opts) {
517         blk = blk_new_with_bs(qemu_opts_id(opts), errp);
518         if (!blk) {
519             goto early_err;
520         }
521 
522         bs = blk_bs(blk);
523         bs->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
524         bs->read_only = ro;
525 
526         QDECREF(bs_opts);
527     } else {
528         if (file && !*file) {
529             file = NULL;
530         }
531 
532         if (snapshot) {
533             /* always use cache=unsafe with snapshot */
534             bdrv_flags &= ~BDRV_O_CACHE_MASK;
535             bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
536         }
537 
538         if (copy_on_read) {
539             bdrv_flags |= BDRV_O_COPY_ON_READ;
540         }
541 
542         if (runstate_check(RUN_STATE_INMIGRATE)) {
543             bdrv_flags |= BDRV_O_INCOMING;
544         }
545 
546         bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
547 
548         blk = blk_new_open(qemu_opts_id(opts), file, NULL, bs_opts, bdrv_flags,
549                            errp);
550         if (!blk) {
551             goto err_no_bs_opts;
552         }
553         bs = blk_bs(blk);
554     }
555 
556     bs->detect_zeroes = detect_zeroes;
557 
558     bdrv_set_on_error(bs, on_read_error, on_write_error);
559 
560     /* disk I/O throttling */
561     if (throttle_enabled(&cfg)) {
562         if (!throttling_group) {
563             throttling_group = blk_name(blk);
564         }
565         bdrv_io_limits_enable(bs, throttling_group);
566         bdrv_set_io_limits(bs, &cfg);
567     }
568 
569     if (bdrv_key_required(bs)) {
570         autostart = 0;
571     }
572 
573 err_no_bs_opts:
574     qemu_opts_del(opts);
575     return blk;
576 
577 early_err:
578     qemu_opts_del(opts);
579 err_no_opts:
580     QDECREF(bs_opts);
581     return NULL;
582 }
583 
584 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
585                             Error **errp)
586 {
587     const char *value;
588 
589     value = qemu_opt_get(opts, from);
590     if (value) {
591         if (qemu_opt_find(opts, to)) {
592             error_setg(errp, "'%s' and its alias '%s' can't be used at the "
593                        "same time", to, from);
594             return;
595         }
596     }
597 
598     /* rename all items in opts */
599     while ((value = qemu_opt_get(opts, from))) {
600         qemu_opt_set(opts, to, value, &error_abort);
601         qemu_opt_unset(opts, from);
602     }
603 }
604 
605 QemuOptsList qemu_legacy_drive_opts = {
606     .name = "drive",
607     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
608     .desc = {
609         {
610             .name = "bus",
611             .type = QEMU_OPT_NUMBER,
612             .help = "bus number",
613         },{
614             .name = "unit",
615             .type = QEMU_OPT_NUMBER,
616             .help = "unit number (i.e. lun for scsi)",
617         },{
618             .name = "index",
619             .type = QEMU_OPT_NUMBER,
620             .help = "index number",
621         },{
622             .name = "media",
623             .type = QEMU_OPT_STRING,
624             .help = "media type (disk, cdrom)",
625         },{
626             .name = "if",
627             .type = QEMU_OPT_STRING,
628             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
629         },{
630             .name = "cyls",
631             .type = QEMU_OPT_NUMBER,
632             .help = "number of cylinders (ide disk geometry)",
633         },{
634             .name = "heads",
635             .type = QEMU_OPT_NUMBER,
636             .help = "number of heads (ide disk geometry)",
637         },{
638             .name = "secs",
639             .type = QEMU_OPT_NUMBER,
640             .help = "number of sectors (ide disk geometry)",
641         },{
642             .name = "trans",
643             .type = QEMU_OPT_STRING,
644             .help = "chs translation (auto, lba, none)",
645         },{
646             .name = "boot",
647             .type = QEMU_OPT_BOOL,
648             .help = "(deprecated, ignored)",
649         },{
650             .name = "addr",
651             .type = QEMU_OPT_STRING,
652             .help = "pci address (virtio only)",
653         },{
654             .name = "serial",
655             .type = QEMU_OPT_STRING,
656             .help = "disk serial number",
657         },{
658             .name = "file",
659             .type = QEMU_OPT_STRING,
660             .help = "file name",
661         },
662 
663         /* Options that are passed on, but have special semantics with -drive */
664         {
665             .name = "read-only",
666             .type = QEMU_OPT_BOOL,
667             .help = "open drive file as read-only",
668         },{
669             .name = "rerror",
670             .type = QEMU_OPT_STRING,
671             .help = "read error action",
672         },{
673             .name = "werror",
674             .type = QEMU_OPT_STRING,
675             .help = "write error action",
676         },{
677             .name = "copy-on-read",
678             .type = QEMU_OPT_BOOL,
679             .help = "copy read data from backing file into image file",
680         },
681 
682         { /* end of list */ }
683     },
684 };
685 
686 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
687 {
688     const char *value;
689     BlockBackend *blk;
690     DriveInfo *dinfo = NULL;
691     QDict *bs_opts;
692     QemuOpts *legacy_opts;
693     DriveMediaType media = MEDIA_DISK;
694     BlockInterfaceType type;
695     int cyls, heads, secs, translation;
696     int max_devs, bus_id, unit_id, index;
697     const char *devaddr;
698     const char *werror, *rerror;
699     bool read_only = false;
700     bool copy_on_read;
701     const char *serial;
702     const char *filename;
703     Error *local_err = NULL;
704     int i;
705 
706     /* Change legacy command line options into QMP ones */
707     static const struct {
708         const char *from;
709         const char *to;
710     } opt_renames[] = {
711         { "iops",           "throttling.iops-total" },
712         { "iops_rd",        "throttling.iops-read" },
713         { "iops_wr",        "throttling.iops-write" },
714 
715         { "bps",            "throttling.bps-total" },
716         { "bps_rd",         "throttling.bps-read" },
717         { "bps_wr",         "throttling.bps-write" },
718 
719         { "iops_max",       "throttling.iops-total-max" },
720         { "iops_rd_max",    "throttling.iops-read-max" },
721         { "iops_wr_max",    "throttling.iops-write-max" },
722 
723         { "bps_max",        "throttling.bps-total-max" },
724         { "bps_rd_max",     "throttling.bps-read-max" },
725         { "bps_wr_max",     "throttling.bps-write-max" },
726 
727         { "iops_size",      "throttling.iops-size" },
728 
729         { "group",          "throttling.group" },
730 
731         { "readonly",       "read-only" },
732     };
733 
734     for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
735         qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
736                         &local_err);
737         if (local_err) {
738             error_report_err(local_err);
739             return NULL;
740         }
741     }
742 
743     value = qemu_opt_get(all_opts, "cache");
744     if (value) {
745         int flags = 0;
746 
747         if (bdrv_parse_cache_flags(value, &flags) != 0) {
748             error_report("invalid cache option");
749             return NULL;
750         }
751 
752         /* Specific options take precedence */
753         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
754             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
755                               !!(flags & BDRV_O_CACHE_WB), &error_abort);
756         }
757         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
758             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
759                               !!(flags & BDRV_O_NOCACHE), &error_abort);
760         }
761         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
762             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
763                               !!(flags & BDRV_O_NO_FLUSH), &error_abort);
764         }
765         qemu_opt_unset(all_opts, "cache");
766     }
767 
768     /* Get a QDict for processing the options */
769     bs_opts = qdict_new();
770     qemu_opts_to_qdict(all_opts, bs_opts);
771 
772     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
773                                    &error_abort);
774     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
775     if (local_err) {
776         error_report_err(local_err);
777         goto fail;
778     }
779 
780     /* Deprecated option boot=[on|off] */
781     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
782         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
783                 "ignored. Future versions will reject this parameter. Please "
784                 "update your scripts.\n");
785     }
786 
787     /* Media type */
788     value = qemu_opt_get(legacy_opts, "media");
789     if (value) {
790         if (!strcmp(value, "disk")) {
791             media = MEDIA_DISK;
792         } else if (!strcmp(value, "cdrom")) {
793             media = MEDIA_CDROM;
794             read_only = true;
795         } else {
796             error_report("'%s' invalid media", value);
797             goto fail;
798         }
799     }
800 
801     /* copy-on-read is disabled with a warning for read-only devices */
802     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
803     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
804 
805     if (read_only && copy_on_read) {
806         error_report("warning: disabling copy-on-read on read-only drive");
807         copy_on_read = false;
808     }
809 
810     qdict_put(bs_opts, "read-only",
811               qstring_from_str(read_only ? "on" : "off"));
812     qdict_put(bs_opts, "copy-on-read",
813               qstring_from_str(copy_on_read ? "on" :"off"));
814 
815     /* Controller type */
816     value = qemu_opt_get(legacy_opts, "if");
817     if (value) {
818         for (type = 0;
819              type < IF_COUNT && strcmp(value, if_name[type]);
820              type++) {
821         }
822         if (type == IF_COUNT) {
823             error_report("unsupported bus type '%s'", value);
824             goto fail;
825         }
826     } else {
827         type = block_default_type;
828     }
829 
830     /* Geometry */
831     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
832     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
833     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
834 
835     if (cyls || heads || secs) {
836         if (cyls < 1) {
837             error_report("invalid physical cyls number");
838             goto fail;
839         }
840         if (heads < 1) {
841             error_report("invalid physical heads number");
842             goto fail;
843         }
844         if (secs < 1) {
845             error_report("invalid physical secs number");
846             goto fail;
847         }
848     }
849 
850     translation = BIOS_ATA_TRANSLATION_AUTO;
851     value = qemu_opt_get(legacy_opts, "trans");
852     if (value != NULL) {
853         if (!cyls) {
854             error_report("'%s' trans must be used with cyls, heads and secs",
855                          value);
856             goto fail;
857         }
858         if (!strcmp(value, "none")) {
859             translation = BIOS_ATA_TRANSLATION_NONE;
860         } else if (!strcmp(value, "lba")) {
861             translation = BIOS_ATA_TRANSLATION_LBA;
862         } else if (!strcmp(value, "large")) {
863             translation = BIOS_ATA_TRANSLATION_LARGE;
864         } else if (!strcmp(value, "rechs")) {
865             translation = BIOS_ATA_TRANSLATION_RECHS;
866         } else if (!strcmp(value, "auto")) {
867             translation = BIOS_ATA_TRANSLATION_AUTO;
868         } else {
869             error_report("'%s' invalid translation type", value);
870             goto fail;
871         }
872     }
873 
874     if (media == MEDIA_CDROM) {
875         if (cyls || secs || heads) {
876             error_report("CHS can't be set with media=cdrom");
877             goto fail;
878         }
879     }
880 
881     /* Device address specified by bus/unit or index.
882      * If none was specified, try to find the first free one. */
883     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
884     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
885     index   = qemu_opt_get_number(legacy_opts, "index", -1);
886 
887     max_devs = if_max_devs[type];
888 
889     if (index != -1) {
890         if (bus_id != 0 || unit_id != -1) {
891             error_report("index cannot be used with bus and unit");
892             goto fail;
893         }
894         bus_id = drive_index_to_bus_id(type, index);
895         unit_id = drive_index_to_unit_id(type, index);
896     }
897 
898     if (unit_id == -1) {
899        unit_id = 0;
900        while (drive_get(type, bus_id, unit_id) != NULL) {
901            unit_id++;
902            if (max_devs && unit_id >= max_devs) {
903                unit_id -= max_devs;
904                bus_id++;
905            }
906        }
907     }
908 
909     if (max_devs && unit_id >= max_devs) {
910         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
911         goto fail;
912     }
913 
914     if (drive_get(type, bus_id, unit_id) != NULL) {
915         error_report("drive with bus=%d, unit=%d (index=%d) exists",
916                      bus_id, unit_id, index);
917         goto fail;
918     }
919 
920     /* Serial number */
921     serial = qemu_opt_get(legacy_opts, "serial");
922 
923     /* no id supplied -> create one */
924     if (qemu_opts_id(all_opts) == NULL) {
925         char *new_id;
926         const char *mediastr = "";
927         if (type == IF_IDE || type == IF_SCSI) {
928             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
929         }
930         if (max_devs) {
931             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
932                                      mediastr, unit_id);
933         } else {
934             new_id = g_strdup_printf("%s%s%i", if_name[type],
935                                      mediastr, unit_id);
936         }
937         qdict_put(bs_opts, "id", qstring_from_str(new_id));
938         g_free(new_id);
939     }
940 
941     /* Add virtio block device */
942     devaddr = qemu_opt_get(legacy_opts, "addr");
943     if (devaddr && type != IF_VIRTIO) {
944         error_report("addr is not supported by this bus type");
945         goto fail;
946     }
947 
948     if (type == IF_VIRTIO) {
949         QemuOpts *devopts;
950         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
951                                    &error_abort);
952         if (arch_type == QEMU_ARCH_S390X) {
953             qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
954         } else {
955             qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
956         }
957         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
958                      &error_abort);
959         if (devaddr) {
960             qemu_opt_set(devopts, "addr", devaddr, &error_abort);
961         }
962     }
963 
964     filename = qemu_opt_get(legacy_opts, "file");
965 
966     /* Check werror/rerror compatibility with if=... */
967     werror = qemu_opt_get(legacy_opts, "werror");
968     if (werror != NULL) {
969         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
970             type != IF_NONE) {
971             error_report("werror is not supported by this bus type");
972             goto fail;
973         }
974         qdict_put(bs_opts, "werror", qstring_from_str(werror));
975     }
976 
977     rerror = qemu_opt_get(legacy_opts, "rerror");
978     if (rerror != NULL) {
979         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
980             type != IF_NONE) {
981             error_report("rerror is not supported by this bus type");
982             goto fail;
983         }
984         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
985     }
986 
987     /* Actual block device init: Functionality shared with blockdev-add */
988     blk = blockdev_init(filename, bs_opts, &local_err);
989     bs_opts = NULL;
990     if (!blk) {
991         if (local_err) {
992             error_report_err(local_err);
993         }
994         goto fail;
995     } else {
996         assert(!local_err);
997     }
998 
999     /* Create legacy DriveInfo */
1000     dinfo = g_malloc0(sizeof(*dinfo));
1001     dinfo->opts = all_opts;
1002 
1003     dinfo->cyls = cyls;
1004     dinfo->heads = heads;
1005     dinfo->secs = secs;
1006     dinfo->trans = translation;
1007 
1008     dinfo->type = type;
1009     dinfo->bus = bus_id;
1010     dinfo->unit = unit_id;
1011     dinfo->devaddr = devaddr;
1012     dinfo->serial = g_strdup(serial);
1013 
1014     blk_set_legacy_dinfo(blk, dinfo);
1015 
1016     switch(type) {
1017     case IF_IDE:
1018     case IF_SCSI:
1019     case IF_XEN:
1020     case IF_NONE:
1021         dinfo->media_cd = media == MEDIA_CDROM;
1022         break;
1023     default:
1024         break;
1025     }
1026 
1027 fail:
1028     qemu_opts_del(legacy_opts);
1029     QDECREF(bs_opts);
1030     return dinfo;
1031 }
1032 
1033 void hmp_commit(Monitor *mon, const QDict *qdict)
1034 {
1035     const char *device = qdict_get_str(qdict, "device");
1036     BlockBackend *blk;
1037     int ret;
1038 
1039     if (!strcmp(device, "all")) {
1040         ret = bdrv_commit_all();
1041     } else {
1042         blk = blk_by_name(device);
1043         if (!blk) {
1044             monitor_printf(mon, "Device '%s' not found\n", device);
1045             return;
1046         }
1047         ret = bdrv_commit(blk_bs(blk));
1048     }
1049     if (ret < 0) {
1050         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1051                        strerror(-ret));
1052     }
1053 }
1054 
1055 static void blockdev_do_action(int kind, void *data, Error **errp)
1056 {
1057     TransactionAction action;
1058     TransactionActionList list;
1059 
1060     action.kind = kind;
1061     action.data = data;
1062     list.value = &action;
1063     list.next = NULL;
1064     qmp_transaction(&list, errp);
1065 }
1066 
1067 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1068                                 bool has_node_name, const char *node_name,
1069                                 const char *snapshot_file,
1070                                 bool has_snapshot_node_name,
1071                                 const char *snapshot_node_name,
1072                                 bool has_format, const char *format,
1073                                 bool has_mode, NewImageMode mode, Error **errp)
1074 {
1075     BlockdevSnapshot snapshot = {
1076         .has_device = has_device,
1077         .device = (char *) device,
1078         .has_node_name = has_node_name,
1079         .node_name = (char *) node_name,
1080         .snapshot_file = (char *) snapshot_file,
1081         .has_snapshot_node_name = has_snapshot_node_name,
1082         .snapshot_node_name = (char *) snapshot_node_name,
1083         .has_format = has_format,
1084         .format = (char *) format,
1085         .has_mode = has_mode,
1086         .mode = mode,
1087     };
1088     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1089                        &snapshot, errp);
1090 }
1091 
1092 void qmp_blockdev_snapshot_internal_sync(const char *device,
1093                                          const char *name,
1094                                          Error **errp)
1095 {
1096     BlockdevSnapshotInternal snapshot = {
1097         .device = (char *) device,
1098         .name = (char *) name
1099     };
1100 
1101     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1102                        &snapshot, errp);
1103 }
1104 
1105 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1106                                                          bool has_id,
1107                                                          const char *id,
1108                                                          bool has_name,
1109                                                          const char *name,
1110                                                          Error **errp)
1111 {
1112     BlockDriverState *bs;
1113     BlockBackend *blk;
1114     AioContext *aio_context;
1115     QEMUSnapshotInfo sn;
1116     Error *local_err = NULL;
1117     SnapshotInfo *info = NULL;
1118     int ret;
1119 
1120     blk = blk_by_name(device);
1121     if (!blk) {
1122         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1123                   "Device '%s' not found", device);
1124         return NULL;
1125     }
1126     bs = blk_bs(blk);
1127 
1128     if (!has_id) {
1129         id = NULL;
1130     }
1131 
1132     if (!has_name) {
1133         name = NULL;
1134     }
1135 
1136     if (!id && !name) {
1137         error_setg(errp, "Name or id must be provided");
1138         return NULL;
1139     }
1140 
1141     aio_context = bdrv_get_aio_context(bs);
1142     aio_context_acquire(aio_context);
1143 
1144     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1145         goto out_aio_context;
1146     }
1147 
1148     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1149     if (local_err) {
1150         error_propagate(errp, local_err);
1151         goto out_aio_context;
1152     }
1153     if (!ret) {
1154         error_setg(errp,
1155                    "Snapshot with id '%s' and name '%s' does not exist on "
1156                    "device '%s'",
1157                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1158         goto out_aio_context;
1159     }
1160 
1161     bdrv_snapshot_delete(bs, id, name, &local_err);
1162     if (local_err) {
1163         error_propagate(errp, local_err);
1164         goto out_aio_context;
1165     }
1166 
1167     aio_context_release(aio_context);
1168 
1169     info = g_new0(SnapshotInfo, 1);
1170     info->id = g_strdup(sn.id_str);
1171     info->name = g_strdup(sn.name);
1172     info->date_nsec = sn.date_nsec;
1173     info->date_sec = sn.date_sec;
1174     info->vm_state_size = sn.vm_state_size;
1175     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1176     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1177 
1178     return info;
1179 
1180 out_aio_context:
1181     aio_context_release(aio_context);
1182     return NULL;
1183 }
1184 
1185 /**
1186  * block_dirty_bitmap_lookup:
1187  * Return a dirty bitmap (if present), after validating
1188  * the node reference and bitmap names.
1189  *
1190  * @node: The name of the BDS node to search for bitmaps
1191  * @name: The name of the bitmap to search for
1192  * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1193  * @paio: Output pointer for aio_context acquisition, if desired. Can be NULL.
1194  * @errp: Output pointer for error information. Can be NULL.
1195  *
1196  * @return: A bitmap object on success, or NULL on failure.
1197  */
1198 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1199                                                   const char *name,
1200                                                   BlockDriverState **pbs,
1201                                                   AioContext **paio,
1202                                                   Error **errp)
1203 {
1204     BlockDriverState *bs;
1205     BdrvDirtyBitmap *bitmap;
1206     AioContext *aio_context;
1207 
1208     if (!node) {
1209         error_setg(errp, "Node cannot be NULL");
1210         return NULL;
1211     }
1212     if (!name) {
1213         error_setg(errp, "Bitmap name cannot be NULL");
1214         return NULL;
1215     }
1216     bs = bdrv_lookup_bs(node, node, NULL);
1217     if (!bs) {
1218         error_setg(errp, "Node '%s' not found", node);
1219         return NULL;
1220     }
1221 
1222     aio_context = bdrv_get_aio_context(bs);
1223     aio_context_acquire(aio_context);
1224 
1225     bitmap = bdrv_find_dirty_bitmap(bs, name);
1226     if (!bitmap) {
1227         error_setg(errp, "Dirty bitmap '%s' not found", name);
1228         goto fail;
1229     }
1230 
1231     if (pbs) {
1232         *pbs = bs;
1233     }
1234     if (paio) {
1235         *paio = aio_context;
1236     } else {
1237         aio_context_release(aio_context);
1238     }
1239 
1240     return bitmap;
1241 
1242  fail:
1243     aio_context_release(aio_context);
1244     return NULL;
1245 }
1246 
1247 /* New and old BlockDriverState structs for atomic group operations */
1248 
1249 typedef struct BlkTransactionState BlkTransactionState;
1250 
1251 /* Only prepare() may fail. In a single transaction, only one of commit() or
1252    abort() will be called, clean() will always be called if it present. */
1253 typedef struct BdrvActionOps {
1254     /* Size of state struct, in bytes. */
1255     size_t instance_size;
1256     /* Prepare the work, must NOT be NULL. */
1257     void (*prepare)(BlkTransactionState *common, Error **errp);
1258     /* Commit the changes, can be NULL. */
1259     void (*commit)(BlkTransactionState *common);
1260     /* Abort the changes on fail, can be NULL. */
1261     void (*abort)(BlkTransactionState *common);
1262     /* Clean up resource in the end, can be NULL. */
1263     void (*clean)(BlkTransactionState *common);
1264 } BdrvActionOps;
1265 
1266 /*
1267  * This structure must be arranged as first member in child type, assuming
1268  * that compiler will also arrange it to the same address with parent instance.
1269  * Later it will be used in free().
1270  */
1271 struct BlkTransactionState {
1272     TransactionAction *action;
1273     const BdrvActionOps *ops;
1274     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1275 };
1276 
1277 /* internal snapshot private data */
1278 typedef struct InternalSnapshotState {
1279     BlkTransactionState common;
1280     BlockDriverState *bs;
1281     AioContext *aio_context;
1282     QEMUSnapshotInfo sn;
1283 } InternalSnapshotState;
1284 
1285 static void internal_snapshot_prepare(BlkTransactionState *common,
1286                                       Error **errp)
1287 {
1288     Error *local_err = NULL;
1289     const char *device;
1290     const char *name;
1291     BlockBackend *blk;
1292     BlockDriverState *bs;
1293     QEMUSnapshotInfo old_sn, *sn;
1294     bool ret;
1295     qemu_timeval tv;
1296     BlockdevSnapshotInternal *internal;
1297     InternalSnapshotState *state;
1298     int ret1;
1299 
1300     g_assert(common->action->kind ==
1301              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1302     internal = common->action->blockdev_snapshot_internal_sync;
1303     state = DO_UPCAST(InternalSnapshotState, common, common);
1304 
1305     /* 1. parse input */
1306     device = internal->device;
1307     name = internal->name;
1308 
1309     /* 2. check for validation */
1310     blk = blk_by_name(device);
1311     if (!blk) {
1312         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1313                   "Device '%s' not found", device);
1314         return;
1315     }
1316     bs = blk_bs(blk);
1317 
1318     /* AioContext is released in .clean() */
1319     state->aio_context = bdrv_get_aio_context(bs);
1320     aio_context_acquire(state->aio_context);
1321 
1322     if (!bdrv_is_inserted(bs)) {
1323         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1324         return;
1325     }
1326 
1327     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1328         return;
1329     }
1330 
1331     if (bdrv_is_read_only(bs)) {
1332         error_setg(errp, "Device '%s' is read only", device);
1333         return;
1334     }
1335 
1336     if (!bdrv_can_snapshot(bs)) {
1337         error_setg(errp, "Block format '%s' used by device '%s' "
1338                    "does not support internal snapshots",
1339                    bs->drv->format_name, device);
1340         return;
1341     }
1342 
1343     if (!strlen(name)) {
1344         error_setg(errp, "Name is empty");
1345         return;
1346     }
1347 
1348     /* check whether a snapshot with name exist */
1349     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1350                                             &local_err);
1351     if (local_err) {
1352         error_propagate(errp, local_err);
1353         return;
1354     } else if (ret) {
1355         error_setg(errp,
1356                    "Snapshot with name '%s' already exists on device '%s'",
1357                    name, device);
1358         return;
1359     }
1360 
1361     /* 3. take the snapshot */
1362     sn = &state->sn;
1363     pstrcpy(sn->name, sizeof(sn->name), name);
1364     qemu_gettimeofday(&tv);
1365     sn->date_sec = tv.tv_sec;
1366     sn->date_nsec = tv.tv_usec * 1000;
1367     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1368 
1369     ret1 = bdrv_snapshot_create(bs, sn);
1370     if (ret1 < 0) {
1371         error_setg_errno(errp, -ret1,
1372                          "Failed to create snapshot '%s' on device '%s'",
1373                          name, device);
1374         return;
1375     }
1376 
1377     /* 4. succeed, mark a snapshot is created */
1378     state->bs = bs;
1379 }
1380 
1381 static void internal_snapshot_abort(BlkTransactionState *common)
1382 {
1383     InternalSnapshotState *state =
1384                              DO_UPCAST(InternalSnapshotState, common, common);
1385     BlockDriverState *bs = state->bs;
1386     QEMUSnapshotInfo *sn = &state->sn;
1387     Error *local_error = NULL;
1388 
1389     if (!bs) {
1390         return;
1391     }
1392 
1393     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1394         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1395                      "device '%s' in abort: %s",
1396                      sn->id_str,
1397                      sn->name,
1398                      bdrv_get_device_name(bs),
1399                      error_get_pretty(local_error));
1400         error_free(local_error);
1401     }
1402 }
1403 
1404 static void internal_snapshot_clean(BlkTransactionState *common)
1405 {
1406     InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1407                                              common, common);
1408 
1409     if (state->aio_context) {
1410         aio_context_release(state->aio_context);
1411     }
1412 }
1413 
1414 /* external snapshot private data */
1415 typedef struct ExternalSnapshotState {
1416     BlkTransactionState common;
1417     BlockDriverState *old_bs;
1418     BlockDriverState *new_bs;
1419     AioContext *aio_context;
1420 } ExternalSnapshotState;
1421 
1422 static void external_snapshot_prepare(BlkTransactionState *common,
1423                                       Error **errp)
1424 {
1425     BlockDriver *drv;
1426     int flags, ret;
1427     QDict *options = NULL;
1428     Error *local_err = NULL;
1429     bool has_device = false;
1430     const char *device;
1431     bool has_node_name = false;
1432     const char *node_name;
1433     bool has_snapshot_node_name = false;
1434     const char *snapshot_node_name;
1435     const char *new_image_file;
1436     const char *format = "qcow2";
1437     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1438     ExternalSnapshotState *state =
1439                              DO_UPCAST(ExternalSnapshotState, common, common);
1440     TransactionAction *action = common->action;
1441 
1442     /* get parameters */
1443     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1444 
1445     has_device = action->blockdev_snapshot_sync->has_device;
1446     device = action->blockdev_snapshot_sync->device;
1447     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1448     node_name = action->blockdev_snapshot_sync->node_name;
1449     has_snapshot_node_name =
1450         action->blockdev_snapshot_sync->has_snapshot_node_name;
1451     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1452 
1453     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1454     if (action->blockdev_snapshot_sync->has_format) {
1455         format = action->blockdev_snapshot_sync->format;
1456     }
1457     if (action->blockdev_snapshot_sync->has_mode) {
1458         mode = action->blockdev_snapshot_sync->mode;
1459     }
1460 
1461     /* start processing */
1462     drv = bdrv_find_format(format);
1463     if (!drv) {
1464         error_setg(errp, QERR_INVALID_BLOCK_FORMAT, format);
1465         return;
1466     }
1467 
1468     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1469                                    has_node_name ? node_name : NULL,
1470                                    &local_err);
1471     if (local_err) {
1472         error_propagate(errp, local_err);
1473         return;
1474     }
1475 
1476     if (has_node_name && !has_snapshot_node_name) {
1477         error_setg(errp, "New snapshot node name missing");
1478         return;
1479     }
1480 
1481     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1482         error_setg(errp, "New snapshot node name already existing");
1483         return;
1484     }
1485 
1486     /* Acquire AioContext now so any threads operating on old_bs stop */
1487     state->aio_context = bdrv_get_aio_context(state->old_bs);
1488     aio_context_acquire(state->aio_context);
1489 
1490     if (!bdrv_is_inserted(state->old_bs)) {
1491         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1492         return;
1493     }
1494 
1495     if (bdrv_op_is_blocked(state->old_bs,
1496                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1497         return;
1498     }
1499 
1500     if (!bdrv_is_read_only(state->old_bs)) {
1501         if (bdrv_flush(state->old_bs)) {
1502             error_setg(errp, QERR_IO_ERROR);
1503             return;
1504         }
1505     }
1506 
1507     if (!bdrv_is_first_non_filter(state->old_bs)) {
1508         error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1509         return;
1510     }
1511 
1512     flags = state->old_bs->open_flags;
1513 
1514     /* create new image w/backing file */
1515     if (mode != NEW_IMAGE_MODE_EXISTING) {
1516         bdrv_img_create(new_image_file, format,
1517                         state->old_bs->filename,
1518                         state->old_bs->drv->format_name,
1519                         NULL, -1, flags, &local_err, false);
1520         if (local_err) {
1521             error_propagate(errp, local_err);
1522             return;
1523         }
1524     }
1525 
1526     if (has_snapshot_node_name) {
1527         options = qdict_new();
1528         qdict_put(options, "node-name",
1529                   qstring_from_str(snapshot_node_name));
1530     }
1531 
1532     /* TODO Inherit bs->options or only take explicit options with an
1533      * extended QMP command? */
1534     assert(state->new_bs == NULL);
1535     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1536                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1537     /* We will manually add the backing_hd field to the bs later */
1538     if (ret != 0) {
1539         error_propagate(errp, local_err);
1540     }
1541 }
1542 
1543 static void external_snapshot_commit(BlkTransactionState *common)
1544 {
1545     ExternalSnapshotState *state =
1546                              DO_UPCAST(ExternalSnapshotState, common, common);
1547 
1548     bdrv_set_aio_context(state->new_bs, state->aio_context);
1549 
1550     /* This removes our old bs and adds the new bs */
1551     bdrv_append(state->new_bs, state->old_bs);
1552     /* We don't need (or want) to use the transactional
1553      * bdrv_reopen_multiple() across all the entries at once, because we
1554      * don't want to abort all of them if one of them fails the reopen */
1555     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1556                 NULL);
1557 
1558     aio_context_release(state->aio_context);
1559 }
1560 
1561 static void external_snapshot_abort(BlkTransactionState *common)
1562 {
1563     ExternalSnapshotState *state =
1564                              DO_UPCAST(ExternalSnapshotState, common, common);
1565     if (state->new_bs) {
1566         bdrv_unref(state->new_bs);
1567     }
1568     if (state->aio_context) {
1569         aio_context_release(state->aio_context);
1570     }
1571 }
1572 
1573 typedef struct DriveBackupState {
1574     BlkTransactionState common;
1575     BlockDriverState *bs;
1576     AioContext *aio_context;
1577     BlockJob *job;
1578 } DriveBackupState;
1579 
1580 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1581 {
1582     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1583     BlockDriverState *bs;
1584     BlockBackend *blk;
1585     DriveBackup *backup;
1586     Error *local_err = NULL;
1587 
1588     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1589     backup = common->action->drive_backup;
1590 
1591     blk = blk_by_name(backup->device);
1592     if (!blk) {
1593         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1594                   "Device '%s' not found", backup->device);
1595         return;
1596     }
1597     bs = blk_bs(blk);
1598 
1599     /* AioContext is released in .clean() */
1600     state->aio_context = bdrv_get_aio_context(bs);
1601     aio_context_acquire(state->aio_context);
1602 
1603     qmp_drive_backup(backup->device, backup->target,
1604                      backup->has_format, backup->format,
1605                      backup->sync,
1606                      backup->has_mode, backup->mode,
1607                      backup->has_speed, backup->speed,
1608                      backup->has_bitmap, backup->bitmap,
1609                      backup->has_on_source_error, backup->on_source_error,
1610                      backup->has_on_target_error, backup->on_target_error,
1611                      &local_err);
1612     if (local_err) {
1613         error_propagate(errp, local_err);
1614         return;
1615     }
1616 
1617     state->bs = bs;
1618     state->job = state->bs->job;
1619 }
1620 
1621 static void drive_backup_abort(BlkTransactionState *common)
1622 {
1623     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1624     BlockDriverState *bs = state->bs;
1625 
1626     /* Only cancel if it's the job we started */
1627     if (bs && bs->job && bs->job == state->job) {
1628         block_job_cancel_sync(bs->job);
1629     }
1630 }
1631 
1632 static void drive_backup_clean(BlkTransactionState *common)
1633 {
1634     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1635 
1636     if (state->aio_context) {
1637         aio_context_release(state->aio_context);
1638     }
1639 }
1640 
1641 typedef struct BlockdevBackupState {
1642     BlkTransactionState common;
1643     BlockDriverState *bs;
1644     BlockJob *job;
1645     AioContext *aio_context;
1646 } BlockdevBackupState;
1647 
1648 static void blockdev_backup_prepare(BlkTransactionState *common, Error **errp)
1649 {
1650     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1651     BlockdevBackup *backup;
1652     BlockDriverState *bs, *target;
1653     BlockBackend *blk;
1654     Error *local_err = NULL;
1655 
1656     assert(common->action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1657     backup = common->action->blockdev_backup;
1658 
1659     blk = blk_by_name(backup->device);
1660     if (!blk) {
1661         error_setg(errp, "Device '%s' not found", backup->device);
1662         return;
1663     }
1664     bs = blk_bs(blk);
1665 
1666     blk = blk_by_name(backup->target);
1667     if (!blk) {
1668         error_setg(errp, "Device '%s' not found", backup->target);
1669         return;
1670     }
1671     target = blk_bs(blk);
1672 
1673     /* AioContext is released in .clean() */
1674     state->aio_context = bdrv_get_aio_context(bs);
1675     if (state->aio_context != bdrv_get_aio_context(target)) {
1676         state->aio_context = NULL;
1677         error_setg(errp, "Backup between two IO threads is not implemented");
1678         return;
1679     }
1680     aio_context_acquire(state->aio_context);
1681 
1682     qmp_blockdev_backup(backup->device, backup->target,
1683                         backup->sync,
1684                         backup->has_speed, backup->speed,
1685                         backup->has_on_source_error, backup->on_source_error,
1686                         backup->has_on_target_error, backup->on_target_error,
1687                         &local_err);
1688     if (local_err) {
1689         error_propagate(errp, local_err);
1690         return;
1691     }
1692 
1693     state->bs = bs;
1694     state->job = state->bs->job;
1695 }
1696 
1697 static void blockdev_backup_abort(BlkTransactionState *common)
1698 {
1699     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1700     BlockDriverState *bs = state->bs;
1701 
1702     /* Only cancel if it's the job we started */
1703     if (bs && bs->job && bs->job == state->job) {
1704         block_job_cancel_sync(bs->job);
1705     }
1706 }
1707 
1708 static void blockdev_backup_clean(BlkTransactionState *common)
1709 {
1710     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1711 
1712     if (state->aio_context) {
1713         aio_context_release(state->aio_context);
1714     }
1715 }
1716 
1717 static void abort_prepare(BlkTransactionState *common, Error **errp)
1718 {
1719     error_setg(errp, "Transaction aborted using Abort action");
1720 }
1721 
1722 static void abort_commit(BlkTransactionState *common)
1723 {
1724     g_assert_not_reached(); /* this action never succeeds */
1725 }
1726 
1727 static const BdrvActionOps actions[] = {
1728     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1729         .instance_size = sizeof(ExternalSnapshotState),
1730         .prepare  = external_snapshot_prepare,
1731         .commit   = external_snapshot_commit,
1732         .abort = external_snapshot_abort,
1733     },
1734     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1735         .instance_size = sizeof(DriveBackupState),
1736         .prepare = drive_backup_prepare,
1737         .abort = drive_backup_abort,
1738         .clean = drive_backup_clean,
1739     },
1740     [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
1741         .instance_size = sizeof(BlockdevBackupState),
1742         .prepare = blockdev_backup_prepare,
1743         .abort = blockdev_backup_abort,
1744         .clean = blockdev_backup_clean,
1745     },
1746     [TRANSACTION_ACTION_KIND_ABORT] = {
1747         .instance_size = sizeof(BlkTransactionState),
1748         .prepare = abort_prepare,
1749         .commit = abort_commit,
1750     },
1751     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1752         .instance_size = sizeof(InternalSnapshotState),
1753         .prepare  = internal_snapshot_prepare,
1754         .abort = internal_snapshot_abort,
1755         .clean = internal_snapshot_clean,
1756     },
1757 };
1758 
1759 /*
1760  * 'Atomic' group operations.  The operations are performed as a set, and if
1761  * any fail then we roll back all operations in the group.
1762  */
1763 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1764 {
1765     TransactionActionList *dev_entry = dev_list;
1766     BlkTransactionState *state, *next;
1767     Error *local_err = NULL;
1768 
1769     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1770     QSIMPLEQ_INIT(&snap_bdrv_states);
1771 
1772     /* drain all i/o before any operations */
1773     bdrv_drain_all();
1774 
1775     /* We don't do anything in this loop that commits us to the operations */
1776     while (NULL != dev_entry) {
1777         TransactionAction *dev_info = NULL;
1778         const BdrvActionOps *ops;
1779 
1780         dev_info = dev_entry->value;
1781         dev_entry = dev_entry->next;
1782 
1783         assert(dev_info->kind < ARRAY_SIZE(actions));
1784 
1785         ops = &actions[dev_info->kind];
1786         assert(ops->instance_size > 0);
1787 
1788         state = g_malloc0(ops->instance_size);
1789         state->ops = ops;
1790         state->action = dev_info;
1791         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1792 
1793         state->ops->prepare(state, &local_err);
1794         if (local_err) {
1795             error_propagate(errp, local_err);
1796             goto delete_and_fail;
1797         }
1798     }
1799 
1800     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1801         if (state->ops->commit) {
1802             state->ops->commit(state);
1803         }
1804     }
1805 
1806     /* success */
1807     goto exit;
1808 
1809 delete_and_fail:
1810     /* failure, and it is all-or-none; roll back all operations */
1811     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1812         if (state->ops->abort) {
1813             state->ops->abort(state);
1814         }
1815     }
1816 exit:
1817     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1818         if (state->ops->clean) {
1819             state->ops->clean(state);
1820         }
1821         g_free(state);
1822     }
1823 }
1824 
1825 
1826 static void eject_device(BlockBackend *blk, int force, Error **errp)
1827 {
1828     BlockDriverState *bs = blk_bs(blk);
1829     AioContext *aio_context;
1830 
1831     aio_context = bdrv_get_aio_context(bs);
1832     aio_context_acquire(aio_context);
1833 
1834     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1835         goto out;
1836     }
1837     if (!blk_dev_has_removable_media(blk)) {
1838         error_setg(errp, "Device '%s' is not removable",
1839                    bdrv_get_device_name(bs));
1840         goto out;
1841     }
1842 
1843     if (blk_dev_is_medium_locked(blk) && !blk_dev_is_tray_open(blk)) {
1844         blk_dev_eject_request(blk, force);
1845         if (!force) {
1846             error_setg(errp, "Device '%s' is locked",
1847                        bdrv_get_device_name(bs));
1848             goto out;
1849         }
1850     }
1851 
1852     bdrv_close(bs);
1853 
1854 out:
1855     aio_context_release(aio_context);
1856 }
1857 
1858 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1859 {
1860     BlockBackend *blk;
1861 
1862     blk = blk_by_name(device);
1863     if (!blk) {
1864         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1865                   "Device '%s' not found", device);
1866         return;
1867     }
1868 
1869     eject_device(blk, force, errp);
1870 }
1871 
1872 void qmp_block_passwd(bool has_device, const char *device,
1873                       bool has_node_name, const char *node_name,
1874                       const char *password, Error **errp)
1875 {
1876     Error *local_err = NULL;
1877     BlockDriverState *bs;
1878     AioContext *aio_context;
1879 
1880     bs = bdrv_lookup_bs(has_device ? device : NULL,
1881                         has_node_name ? node_name : NULL,
1882                         &local_err);
1883     if (local_err) {
1884         error_propagate(errp, local_err);
1885         return;
1886     }
1887 
1888     aio_context = bdrv_get_aio_context(bs);
1889     aio_context_acquire(aio_context);
1890 
1891     bdrv_add_key(bs, password, errp);
1892 
1893     aio_context_release(aio_context);
1894 }
1895 
1896 /* Assumes AioContext is held */
1897 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1898                                     int bdrv_flags, BlockDriver *drv,
1899                                     const char *password, Error **errp)
1900 {
1901     Error *local_err = NULL;
1902     int ret;
1903 
1904     ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1905     if (ret < 0) {
1906         error_propagate(errp, local_err);
1907         return;
1908     }
1909 
1910     bdrv_add_key(bs, password, errp);
1911 }
1912 
1913 void qmp_change_blockdev(const char *device, const char *filename,
1914                          const char *format, Error **errp)
1915 {
1916     BlockBackend *blk;
1917     BlockDriverState *bs;
1918     AioContext *aio_context;
1919     BlockDriver *drv = NULL;
1920     int bdrv_flags;
1921     Error *err = NULL;
1922 
1923     blk = blk_by_name(device);
1924     if (!blk) {
1925         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1926                   "Device '%s' not found", device);
1927         return;
1928     }
1929     bs = blk_bs(blk);
1930 
1931     aio_context = bdrv_get_aio_context(bs);
1932     aio_context_acquire(aio_context);
1933 
1934     if (format) {
1935         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1936         if (!drv) {
1937             error_setg(errp, QERR_INVALID_BLOCK_FORMAT, format);
1938             goto out;
1939         }
1940     }
1941 
1942     eject_device(blk, 0, &err);
1943     if (err) {
1944         error_propagate(errp, err);
1945         goto out;
1946     }
1947 
1948     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1949     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1950 
1951     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1952 
1953 out:
1954     aio_context_release(aio_context);
1955 }
1956 
1957 /* throttling disk I/O limits */
1958 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1959                                int64_t bps_wr,
1960                                int64_t iops,
1961                                int64_t iops_rd,
1962                                int64_t iops_wr,
1963                                bool has_bps_max,
1964                                int64_t bps_max,
1965                                bool has_bps_rd_max,
1966                                int64_t bps_rd_max,
1967                                bool has_bps_wr_max,
1968                                int64_t bps_wr_max,
1969                                bool has_iops_max,
1970                                int64_t iops_max,
1971                                bool has_iops_rd_max,
1972                                int64_t iops_rd_max,
1973                                bool has_iops_wr_max,
1974                                int64_t iops_wr_max,
1975                                bool has_iops_size,
1976                                int64_t iops_size,
1977                                bool has_group,
1978                                const char *group, Error **errp)
1979 {
1980     ThrottleConfig cfg;
1981     BlockDriverState *bs;
1982     BlockBackend *blk;
1983     AioContext *aio_context;
1984 
1985     blk = blk_by_name(device);
1986     if (!blk) {
1987         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1988                   "Device '%s' not found", device);
1989         return;
1990     }
1991     bs = blk_bs(blk);
1992 
1993     memset(&cfg, 0, sizeof(cfg));
1994     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1995     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1996     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1997 
1998     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1999     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
2000     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
2001 
2002     if (has_bps_max) {
2003         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
2004     }
2005     if (has_bps_rd_max) {
2006         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
2007     }
2008     if (has_bps_wr_max) {
2009         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
2010     }
2011     if (has_iops_max) {
2012         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
2013     }
2014     if (has_iops_rd_max) {
2015         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
2016     }
2017     if (has_iops_wr_max) {
2018         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
2019     }
2020 
2021     if (has_iops_size) {
2022         cfg.op_size = iops_size;
2023     }
2024 
2025     if (!check_throttle_config(&cfg, errp)) {
2026         return;
2027     }
2028 
2029     aio_context = bdrv_get_aio_context(bs);
2030     aio_context_acquire(aio_context);
2031 
2032     if (throttle_enabled(&cfg)) {
2033         /* Enable I/O limits if they're not enabled yet, otherwise
2034          * just update the throttling group. */
2035         if (!bs->io_limits_enabled) {
2036             bdrv_io_limits_enable(bs, has_group ? group : device);
2037         } else if (has_group) {
2038             bdrv_io_limits_update_group(bs, group);
2039         }
2040         /* Set the new throttling configuration */
2041         bdrv_set_io_limits(bs, &cfg);
2042     } else if (bs->io_limits_enabled) {
2043         /* If all throttling settings are set to 0, disable I/O limits */
2044         bdrv_io_limits_disable(bs);
2045     }
2046 
2047     aio_context_release(aio_context);
2048 }
2049 
2050 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2051                                 bool has_granularity, uint32_t granularity,
2052                                 Error **errp)
2053 {
2054     AioContext *aio_context;
2055     BlockDriverState *bs;
2056 
2057     if (!name || name[0] == '\0') {
2058         error_setg(errp, "Bitmap name cannot be empty");
2059         return;
2060     }
2061 
2062     bs = bdrv_lookup_bs(node, node, errp);
2063     if (!bs) {
2064         return;
2065     }
2066 
2067     aio_context = bdrv_get_aio_context(bs);
2068     aio_context_acquire(aio_context);
2069 
2070     if (has_granularity) {
2071         if (granularity < 512 || !is_power_of_2(granularity)) {
2072             error_setg(errp, "Granularity must be power of 2 "
2073                              "and at least 512");
2074             goto out;
2075         }
2076     } else {
2077         /* Default to cluster size, if available: */
2078         granularity = bdrv_get_default_bitmap_granularity(bs);
2079     }
2080 
2081     bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2082 
2083  out:
2084     aio_context_release(aio_context);
2085 }
2086 
2087 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2088                                    Error **errp)
2089 {
2090     AioContext *aio_context;
2091     BlockDriverState *bs;
2092     BdrvDirtyBitmap *bitmap;
2093 
2094     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2095     if (!bitmap || !bs) {
2096         return;
2097     }
2098 
2099     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2100         error_setg(errp,
2101                    "Bitmap '%s' is currently frozen and cannot be removed",
2102                    name);
2103         goto out;
2104     }
2105     bdrv_dirty_bitmap_make_anon(bitmap);
2106     bdrv_release_dirty_bitmap(bs, bitmap);
2107 
2108  out:
2109     aio_context_release(aio_context);
2110 }
2111 
2112 /**
2113  * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2114  * immediately after a full backup operation.
2115  */
2116 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2117                                   Error **errp)
2118 {
2119     AioContext *aio_context;
2120     BdrvDirtyBitmap *bitmap;
2121     BlockDriverState *bs;
2122 
2123     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2124     if (!bitmap || !bs) {
2125         return;
2126     }
2127 
2128     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2129         error_setg(errp,
2130                    "Bitmap '%s' is currently frozen and cannot be modified",
2131                    name);
2132         goto out;
2133     } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
2134         error_setg(errp,
2135                    "Bitmap '%s' is currently disabled and cannot be cleared",
2136                    name);
2137         goto out;
2138     }
2139 
2140     bdrv_clear_dirty_bitmap(bitmap);
2141 
2142  out:
2143     aio_context_release(aio_context);
2144 }
2145 
2146 void hmp_drive_del(Monitor *mon, const QDict *qdict)
2147 {
2148     const char *id = qdict_get_str(qdict, "id");
2149     BlockBackend *blk;
2150     BlockDriverState *bs;
2151     AioContext *aio_context;
2152     Error *local_err = NULL;
2153 
2154     blk = blk_by_name(id);
2155     if (!blk) {
2156         error_report("Device '%s' not found", id);
2157         return;
2158     }
2159     bs = blk_bs(blk);
2160 
2161     if (!blk_legacy_dinfo(blk)) {
2162         error_report("Deleting device added with blockdev-add"
2163                      " is not supported");
2164         return;
2165     }
2166 
2167     aio_context = bdrv_get_aio_context(bs);
2168     aio_context_acquire(aio_context);
2169 
2170     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2171         error_report_err(local_err);
2172         aio_context_release(aio_context);
2173         return;
2174     }
2175 
2176     bdrv_close(bs);
2177 
2178     /* if we have a device attached to this BlockDriverState
2179      * then we need to make the drive anonymous until the device
2180      * can be removed.  If this is a drive with no device backing
2181      * then we can just get rid of the block driver state right here.
2182      */
2183     if (blk_get_attached_dev(blk)) {
2184         blk_hide_on_behalf_of_hmp_drive_del(blk);
2185         /* Further I/O must not pause the guest */
2186         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
2187                           BLOCKDEV_ON_ERROR_REPORT);
2188     } else {
2189         blk_unref(blk);
2190     }
2191 
2192     aio_context_release(aio_context);
2193 }
2194 
2195 void qmp_block_resize(bool has_device, const char *device,
2196                       bool has_node_name, const char *node_name,
2197                       int64_t size, Error **errp)
2198 {
2199     Error *local_err = NULL;
2200     BlockDriverState *bs;
2201     AioContext *aio_context;
2202     int ret;
2203 
2204     bs = bdrv_lookup_bs(has_device ? device : NULL,
2205                         has_node_name ? node_name : NULL,
2206                         &local_err);
2207     if (local_err) {
2208         error_propagate(errp, local_err);
2209         return;
2210     }
2211 
2212     aio_context = bdrv_get_aio_context(bs);
2213     aio_context_acquire(aio_context);
2214 
2215     if (!bdrv_is_first_non_filter(bs)) {
2216         error_setg(errp, QERR_FEATURE_DISABLED, "resize");
2217         goto out;
2218     }
2219 
2220     if (size < 0) {
2221         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2222         goto out;
2223     }
2224 
2225     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2226         error_setg(errp, QERR_DEVICE_IN_USE, device);
2227         goto out;
2228     }
2229 
2230     /* complete all in-flight operations before resizing the device */
2231     bdrv_drain_all();
2232 
2233     ret = bdrv_truncate(bs, size);
2234     switch (ret) {
2235     case 0:
2236         break;
2237     case -ENOMEDIUM:
2238         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2239         break;
2240     case -ENOTSUP:
2241         error_setg(errp, QERR_UNSUPPORTED);
2242         break;
2243     case -EACCES:
2244         error_setg(errp, "Device '%s' is read only", device);
2245         break;
2246     case -EBUSY:
2247         error_setg(errp, QERR_DEVICE_IN_USE, device);
2248         break;
2249     default:
2250         error_setg_errno(errp, -ret, "Could not resize");
2251         break;
2252     }
2253 
2254 out:
2255     aio_context_release(aio_context);
2256 }
2257 
2258 static void block_job_cb(void *opaque, int ret)
2259 {
2260     /* Note that this function may be executed from another AioContext besides
2261      * the QEMU main loop.  If you need to access anything that assumes the
2262      * QEMU global mutex, use a BH or introduce a mutex.
2263      */
2264 
2265     BlockDriverState *bs = opaque;
2266     const char *msg = NULL;
2267 
2268     trace_block_job_cb(bs, bs->job, ret);
2269 
2270     assert(bs->job);
2271 
2272     if (ret < 0) {
2273         msg = strerror(-ret);
2274     }
2275 
2276     if (block_job_is_cancelled(bs->job)) {
2277         block_job_event_cancelled(bs->job);
2278     } else {
2279         block_job_event_completed(bs->job, msg);
2280     }
2281 
2282     bdrv_put_ref_bh_schedule(bs);
2283 }
2284 
2285 void qmp_block_stream(const char *device,
2286                       bool has_base, const char *base,
2287                       bool has_backing_file, const char *backing_file,
2288                       bool has_speed, int64_t speed,
2289                       bool has_on_error, BlockdevOnError on_error,
2290                       Error **errp)
2291 {
2292     BlockBackend *blk;
2293     BlockDriverState *bs;
2294     BlockDriverState *base_bs = NULL;
2295     AioContext *aio_context;
2296     Error *local_err = NULL;
2297     const char *base_name = NULL;
2298 
2299     if (!has_on_error) {
2300         on_error = BLOCKDEV_ON_ERROR_REPORT;
2301     }
2302 
2303     blk = blk_by_name(device);
2304     if (!blk) {
2305         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2306                   "Device '%s' not found", device);
2307         return;
2308     }
2309     bs = blk_bs(blk);
2310 
2311     aio_context = bdrv_get_aio_context(bs);
2312     aio_context_acquire(aio_context);
2313 
2314     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
2315         goto out;
2316     }
2317 
2318     if (has_base) {
2319         base_bs = bdrv_find_backing_image(bs, base);
2320         if (base_bs == NULL) {
2321             error_setg(errp, QERR_BASE_NOT_FOUND, base);
2322             goto out;
2323         }
2324         assert(bdrv_get_aio_context(base_bs) == aio_context);
2325         base_name = base;
2326     }
2327 
2328     /* if we are streaming the entire chain, the result will have no backing
2329      * file, and specifying one is therefore an error */
2330     if (base_bs == NULL && has_backing_file) {
2331         error_setg(errp, "backing file specified, but streaming the "
2332                          "entire chain");
2333         goto out;
2334     }
2335 
2336     /* backing_file string overrides base bs filename */
2337     base_name = has_backing_file ? backing_file : base_name;
2338 
2339     stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2340                  on_error, block_job_cb, bs, &local_err);
2341     if (local_err) {
2342         error_propagate(errp, local_err);
2343         goto out;
2344     }
2345 
2346     trace_qmp_block_stream(bs, bs->job);
2347 
2348 out:
2349     aio_context_release(aio_context);
2350 }
2351 
2352 void qmp_block_commit(const char *device,
2353                       bool has_base, const char *base,
2354                       bool has_top, const char *top,
2355                       bool has_backing_file, const char *backing_file,
2356                       bool has_speed, int64_t speed,
2357                       Error **errp)
2358 {
2359     BlockBackend *blk;
2360     BlockDriverState *bs;
2361     BlockDriverState *base_bs, *top_bs;
2362     AioContext *aio_context;
2363     Error *local_err = NULL;
2364     /* This will be part of the QMP command, if/when the
2365      * BlockdevOnError change for blkmirror makes it in
2366      */
2367     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2368 
2369     if (!has_speed) {
2370         speed = 0;
2371     }
2372 
2373     /* Important Note:
2374      *  libvirt relies on the DeviceNotFound error class in order to probe for
2375      *  live commit feature versions; for this to work, we must make sure to
2376      *  perform the device lookup before any generic errors that may occur in a
2377      *  scenario in which all optional arguments are omitted. */
2378     blk = blk_by_name(device);
2379     if (!blk) {
2380         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2381                   "Device '%s' not found", device);
2382         return;
2383     }
2384     bs = blk_bs(blk);
2385 
2386     aio_context = bdrv_get_aio_context(bs);
2387     aio_context_acquire(aio_context);
2388 
2389     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2390         goto out;
2391     }
2392 
2393     /* default top_bs is the active layer */
2394     top_bs = bs;
2395 
2396     if (has_top && top) {
2397         if (strcmp(bs->filename, top) != 0) {
2398             top_bs = bdrv_find_backing_image(bs, top);
2399         }
2400     }
2401 
2402     if (top_bs == NULL) {
2403         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2404         goto out;
2405     }
2406 
2407     assert(bdrv_get_aio_context(top_bs) == aio_context);
2408 
2409     if (has_base && base) {
2410         base_bs = bdrv_find_backing_image(top_bs, base);
2411     } else {
2412         base_bs = bdrv_find_base(top_bs);
2413     }
2414 
2415     if (base_bs == NULL) {
2416         error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2417         goto out;
2418     }
2419 
2420     assert(bdrv_get_aio_context(base_bs) == aio_context);
2421 
2422     if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2423         goto out;
2424     }
2425 
2426     /* Do not allow attempts to commit an image into itself */
2427     if (top_bs == base_bs) {
2428         error_setg(errp, "cannot commit an image into itself");
2429         goto out;
2430     }
2431 
2432     if (top_bs == bs) {
2433         if (has_backing_file) {
2434             error_setg(errp, "'backing-file' specified,"
2435                              " but 'top' is the active layer");
2436             goto out;
2437         }
2438         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2439                             bs, &local_err);
2440     } else {
2441         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2442                      has_backing_file ? backing_file : NULL, &local_err);
2443     }
2444     if (local_err != NULL) {
2445         error_propagate(errp, local_err);
2446         goto out;
2447     }
2448 
2449 out:
2450     aio_context_release(aio_context);
2451 }
2452 
2453 void qmp_drive_backup(const char *device, const char *target,
2454                       bool has_format, const char *format,
2455                       enum MirrorSyncMode sync,
2456                       bool has_mode, enum NewImageMode mode,
2457                       bool has_speed, int64_t speed,
2458                       bool has_bitmap, const char *bitmap,
2459                       bool has_on_source_error, BlockdevOnError on_source_error,
2460                       bool has_on_target_error, BlockdevOnError on_target_error,
2461                       Error **errp)
2462 {
2463     BlockBackend *blk;
2464     BlockDriverState *bs;
2465     BlockDriverState *target_bs;
2466     BlockDriverState *source = NULL;
2467     BdrvDirtyBitmap *bmap = NULL;
2468     AioContext *aio_context;
2469     BlockDriver *drv = NULL;
2470     Error *local_err = NULL;
2471     int flags;
2472     int64_t size;
2473     int ret;
2474 
2475     if (!has_speed) {
2476         speed = 0;
2477     }
2478     if (!has_on_source_error) {
2479         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2480     }
2481     if (!has_on_target_error) {
2482         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2483     }
2484     if (!has_mode) {
2485         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2486     }
2487 
2488     blk = blk_by_name(device);
2489     if (!blk) {
2490         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2491                   "Device '%s' not found", device);
2492         return;
2493     }
2494     bs = blk_bs(blk);
2495 
2496     aio_context = bdrv_get_aio_context(bs);
2497     aio_context_acquire(aio_context);
2498 
2499     /* Although backup_run has this check too, we need to use bs->drv below, so
2500      * do an early check redundantly. */
2501     if (!bdrv_is_inserted(bs)) {
2502         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2503         goto out;
2504     }
2505 
2506     if (!has_format) {
2507         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2508     }
2509     if (format) {
2510         drv = bdrv_find_format(format);
2511         if (!drv) {
2512             error_setg(errp, QERR_INVALID_BLOCK_FORMAT, format);
2513             goto out;
2514         }
2515     }
2516 
2517     /* Early check to avoid creating target */
2518     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2519         goto out;
2520     }
2521 
2522     flags = bs->open_flags | BDRV_O_RDWR;
2523 
2524     /* See if we have a backing HD we can use to create our new image
2525      * on top of. */
2526     if (sync == MIRROR_SYNC_MODE_TOP) {
2527         source = bs->backing_hd;
2528         if (!source) {
2529             sync = MIRROR_SYNC_MODE_FULL;
2530         }
2531     }
2532     if (sync == MIRROR_SYNC_MODE_NONE) {
2533         source = bs;
2534     }
2535 
2536     size = bdrv_getlength(bs);
2537     if (size < 0) {
2538         error_setg_errno(errp, -size, "bdrv_getlength failed");
2539         goto out;
2540     }
2541 
2542     if (mode != NEW_IMAGE_MODE_EXISTING) {
2543         assert(format && drv);
2544         if (source) {
2545             bdrv_img_create(target, format, source->filename,
2546                             source->drv->format_name, NULL,
2547                             size, flags, &local_err, false);
2548         } else {
2549             bdrv_img_create(target, format, NULL, NULL, NULL,
2550                             size, flags, &local_err, false);
2551         }
2552     }
2553 
2554     if (local_err) {
2555         error_propagate(errp, local_err);
2556         goto out;
2557     }
2558 
2559     target_bs = NULL;
2560     ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2561     if (ret < 0) {
2562         error_propagate(errp, local_err);
2563         goto out;
2564     }
2565 
2566     bdrv_set_aio_context(target_bs, aio_context);
2567 
2568     if (has_bitmap) {
2569         bmap = bdrv_find_dirty_bitmap(bs, bitmap);
2570         if (!bmap) {
2571             error_setg(errp, "Bitmap '%s' could not be found", bitmap);
2572             goto out;
2573         }
2574     }
2575 
2576     backup_start(bs, target_bs, speed, sync, bmap,
2577                  on_source_error, on_target_error,
2578                  block_job_cb, bs, &local_err);
2579     if (local_err != NULL) {
2580         bdrv_unref(target_bs);
2581         error_propagate(errp, local_err);
2582         goto out;
2583     }
2584 
2585 out:
2586     aio_context_release(aio_context);
2587 }
2588 
2589 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2590 {
2591     return bdrv_named_nodes_list(errp);
2592 }
2593 
2594 void qmp_blockdev_backup(const char *device, const char *target,
2595                          enum MirrorSyncMode sync,
2596                          bool has_speed, int64_t speed,
2597                          bool has_on_source_error,
2598                          BlockdevOnError on_source_error,
2599                          bool has_on_target_error,
2600                          BlockdevOnError on_target_error,
2601                          Error **errp)
2602 {
2603     BlockBackend *blk;
2604     BlockDriverState *bs;
2605     BlockDriverState *target_bs;
2606     Error *local_err = NULL;
2607     AioContext *aio_context;
2608 
2609     if (!has_speed) {
2610         speed = 0;
2611     }
2612     if (!has_on_source_error) {
2613         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2614     }
2615     if (!has_on_target_error) {
2616         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2617     }
2618 
2619     blk = blk_by_name(device);
2620     if (!blk) {
2621         error_setg(errp, "Device '%s' not found", device);
2622         return;
2623     }
2624     bs = blk_bs(blk);
2625 
2626     aio_context = bdrv_get_aio_context(bs);
2627     aio_context_acquire(aio_context);
2628 
2629     blk = blk_by_name(target);
2630     if (!blk) {
2631         error_setg(errp, "Device '%s' not found", target);
2632         goto out;
2633     }
2634     target_bs = blk_bs(blk);
2635 
2636     bdrv_ref(target_bs);
2637     bdrv_set_aio_context(target_bs, aio_context);
2638     backup_start(bs, target_bs, speed, sync, NULL, on_source_error,
2639                  on_target_error, block_job_cb, bs, &local_err);
2640     if (local_err != NULL) {
2641         bdrv_unref(target_bs);
2642         error_propagate(errp, local_err);
2643     }
2644 out:
2645     aio_context_release(aio_context);
2646 }
2647 
2648 void qmp_drive_mirror(const char *device, const char *target,
2649                       bool has_format, const char *format,
2650                       bool has_node_name, const char *node_name,
2651                       bool has_replaces, const char *replaces,
2652                       enum MirrorSyncMode sync,
2653                       bool has_mode, enum NewImageMode mode,
2654                       bool has_speed, int64_t speed,
2655                       bool has_granularity, uint32_t granularity,
2656                       bool has_buf_size, int64_t buf_size,
2657                       bool has_on_source_error, BlockdevOnError on_source_error,
2658                       bool has_on_target_error, BlockdevOnError on_target_error,
2659                       bool has_unmap, bool unmap,
2660                       Error **errp)
2661 {
2662     BlockBackend *blk;
2663     BlockDriverState *bs;
2664     BlockDriverState *source, *target_bs;
2665     AioContext *aio_context;
2666     BlockDriver *drv = NULL;
2667     Error *local_err = NULL;
2668     QDict *options = NULL;
2669     int flags;
2670     int64_t size;
2671     int ret;
2672 
2673     if (!has_speed) {
2674         speed = 0;
2675     }
2676     if (!has_on_source_error) {
2677         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2678     }
2679     if (!has_on_target_error) {
2680         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2681     }
2682     if (!has_mode) {
2683         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2684     }
2685     if (!has_granularity) {
2686         granularity = 0;
2687     }
2688     if (!has_buf_size) {
2689         buf_size = 0;
2690     }
2691     if (!has_unmap) {
2692         unmap = true;
2693     }
2694 
2695     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2696         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2697                    "a value in range [512B, 64MB]");
2698         return;
2699     }
2700     if (granularity & (granularity - 1)) {
2701         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2702                    "power of 2");
2703         return;
2704     }
2705 
2706     blk = blk_by_name(device);
2707     if (!blk) {
2708         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2709                   "Device '%s' not found", device);
2710         return;
2711     }
2712     bs = blk_bs(blk);
2713 
2714     aio_context = bdrv_get_aio_context(bs);
2715     aio_context_acquire(aio_context);
2716 
2717     if (!bdrv_is_inserted(bs)) {
2718         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2719         goto out;
2720     }
2721 
2722     if (!has_format) {
2723         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2724     }
2725     if (format) {
2726         drv = bdrv_find_format(format);
2727         if (!drv) {
2728             error_setg(errp, QERR_INVALID_BLOCK_FORMAT, format);
2729             goto out;
2730         }
2731     }
2732 
2733     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2734         goto out;
2735     }
2736 
2737     flags = bs->open_flags | BDRV_O_RDWR;
2738     source = bs->backing_hd;
2739     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2740         sync = MIRROR_SYNC_MODE_FULL;
2741     }
2742     if (sync == MIRROR_SYNC_MODE_NONE) {
2743         source = bs;
2744     }
2745 
2746     size = bdrv_getlength(bs);
2747     if (size < 0) {
2748         error_setg_errno(errp, -size, "bdrv_getlength failed");
2749         goto out;
2750     }
2751 
2752     if (has_replaces) {
2753         BlockDriverState *to_replace_bs;
2754         AioContext *replace_aio_context;
2755         int64_t replace_size;
2756 
2757         if (!has_node_name) {
2758             error_setg(errp, "a node-name must be provided when replacing a"
2759                              " named node of the graph");
2760             goto out;
2761         }
2762 
2763         to_replace_bs = check_to_replace_node(bs, replaces, &local_err);
2764 
2765         if (!to_replace_bs) {
2766             error_propagate(errp, local_err);
2767             goto out;
2768         }
2769 
2770         replace_aio_context = bdrv_get_aio_context(to_replace_bs);
2771         aio_context_acquire(replace_aio_context);
2772         replace_size = bdrv_getlength(to_replace_bs);
2773         aio_context_release(replace_aio_context);
2774 
2775         if (size != replace_size) {
2776             error_setg(errp, "cannot replace image with a mirror image of "
2777                              "different size");
2778             goto out;
2779         }
2780     }
2781 
2782     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2783         && mode != NEW_IMAGE_MODE_EXISTING)
2784     {
2785         /* create new image w/o backing file */
2786         assert(format && drv);
2787         bdrv_img_create(target, format,
2788                         NULL, NULL, NULL, size, flags, &local_err, false);
2789     } else {
2790         switch (mode) {
2791         case NEW_IMAGE_MODE_EXISTING:
2792             break;
2793         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2794             /* create new image with backing file */
2795             bdrv_img_create(target, format,
2796                             source->filename,
2797                             source->drv->format_name,
2798                             NULL, size, flags, &local_err, false);
2799             break;
2800         default:
2801             abort();
2802         }
2803     }
2804 
2805     if (local_err) {
2806         error_propagate(errp, local_err);
2807         goto out;
2808     }
2809 
2810     if (has_node_name) {
2811         options = qdict_new();
2812         qdict_put(options, "node-name", qstring_from_str(node_name));
2813     }
2814 
2815     /* Mirroring takes care of copy-on-write using the source's backing
2816      * file.
2817      */
2818     target_bs = NULL;
2819     ret = bdrv_open(&target_bs, target, NULL, options,
2820                     flags | BDRV_O_NO_BACKING, drv, &local_err);
2821     if (ret < 0) {
2822         error_propagate(errp, local_err);
2823         goto out;
2824     }
2825 
2826     bdrv_set_aio_context(target_bs, aio_context);
2827 
2828     /* pass the node name to replace to mirror start since it's loose coupling
2829      * and will allow to check whether the node still exist at mirror completion
2830      */
2831     mirror_start(bs, target_bs,
2832                  has_replaces ? replaces : NULL,
2833                  speed, granularity, buf_size, sync,
2834                  on_source_error, on_target_error,
2835                  unmap,
2836                  block_job_cb, bs, &local_err);
2837     if (local_err != NULL) {
2838         bdrv_unref(target_bs);
2839         error_propagate(errp, local_err);
2840         goto out;
2841     }
2842 
2843 out:
2844     aio_context_release(aio_context);
2845 }
2846 
2847 /* Get the block job for a given device name and acquire its AioContext */
2848 static BlockJob *find_block_job(const char *device, AioContext **aio_context,
2849                                 Error **errp)
2850 {
2851     BlockBackend *blk;
2852     BlockDriverState *bs;
2853 
2854     blk = blk_by_name(device);
2855     if (!blk) {
2856         goto notfound;
2857     }
2858     bs = blk_bs(blk);
2859 
2860     *aio_context = bdrv_get_aio_context(bs);
2861     aio_context_acquire(*aio_context);
2862 
2863     if (!bs->job) {
2864         aio_context_release(*aio_context);
2865         goto notfound;
2866     }
2867 
2868     return bs->job;
2869 
2870 notfound:
2871     error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
2872               "No active block job on device '%s'", device);
2873     *aio_context = NULL;
2874     return NULL;
2875 }
2876 
2877 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2878 {
2879     AioContext *aio_context;
2880     BlockJob *job = find_block_job(device, &aio_context, errp);
2881 
2882     if (!job) {
2883         return;
2884     }
2885 
2886     block_job_set_speed(job, speed, errp);
2887     aio_context_release(aio_context);
2888 }
2889 
2890 void qmp_block_job_cancel(const char *device,
2891                           bool has_force, bool force, Error **errp)
2892 {
2893     AioContext *aio_context;
2894     BlockJob *job = find_block_job(device, &aio_context, errp);
2895 
2896     if (!job) {
2897         return;
2898     }
2899 
2900     if (!has_force) {
2901         force = false;
2902     }
2903 
2904     if (job->user_paused && !force) {
2905         error_setg(errp, "The block job for device '%s' is currently paused",
2906                    device);
2907         goto out;
2908     }
2909 
2910     trace_qmp_block_job_cancel(job);
2911     block_job_cancel(job);
2912 out:
2913     aio_context_release(aio_context);
2914 }
2915 
2916 void qmp_block_job_pause(const char *device, Error **errp)
2917 {
2918     AioContext *aio_context;
2919     BlockJob *job = find_block_job(device, &aio_context, errp);
2920 
2921     if (!job || job->user_paused) {
2922         return;
2923     }
2924 
2925     job->user_paused = true;
2926     trace_qmp_block_job_pause(job);
2927     block_job_pause(job);
2928     aio_context_release(aio_context);
2929 }
2930 
2931 void qmp_block_job_resume(const char *device, Error **errp)
2932 {
2933     AioContext *aio_context;
2934     BlockJob *job = find_block_job(device, &aio_context, errp);
2935 
2936     if (!job || !job->user_paused) {
2937         return;
2938     }
2939 
2940     job->user_paused = false;
2941     trace_qmp_block_job_resume(job);
2942     block_job_resume(job);
2943     aio_context_release(aio_context);
2944 }
2945 
2946 void qmp_block_job_complete(const char *device, Error **errp)
2947 {
2948     AioContext *aio_context;
2949     BlockJob *job = find_block_job(device, &aio_context, errp);
2950 
2951     if (!job) {
2952         return;
2953     }
2954 
2955     trace_qmp_block_job_complete(job);
2956     block_job_complete(job, errp);
2957     aio_context_release(aio_context);
2958 }
2959 
2960 void qmp_change_backing_file(const char *device,
2961                              const char *image_node_name,
2962                              const char *backing_file,
2963                              Error **errp)
2964 {
2965     BlockBackend *blk;
2966     BlockDriverState *bs = NULL;
2967     AioContext *aio_context;
2968     BlockDriverState *image_bs = NULL;
2969     Error *local_err = NULL;
2970     bool ro;
2971     int open_flags;
2972     int ret;
2973 
2974     blk = blk_by_name(device);
2975     if (!blk) {
2976         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2977                   "Device '%s' not found", device);
2978         return;
2979     }
2980     bs = blk_bs(blk);
2981 
2982     aio_context = bdrv_get_aio_context(bs);
2983     aio_context_acquire(aio_context);
2984 
2985     image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
2986     if (local_err) {
2987         error_propagate(errp, local_err);
2988         goto out;
2989     }
2990 
2991     if (!image_bs) {
2992         error_setg(errp, "image file not found");
2993         goto out;
2994     }
2995 
2996     if (bdrv_find_base(image_bs) == image_bs) {
2997         error_setg(errp, "not allowing backing file change on an image "
2998                          "without a backing file");
2999         goto out;
3000     }
3001 
3002     /* even though we are not necessarily operating on bs, we need it to
3003      * determine if block ops are currently prohibited on the chain */
3004     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3005         goto out;
3006     }
3007 
3008     /* final sanity check */
3009     if (!bdrv_chain_contains(bs, image_bs)) {
3010         error_setg(errp, "'%s' and image file are not in the same chain",
3011                    device);
3012         goto out;
3013     }
3014 
3015     /* if not r/w, reopen to make r/w */
3016     open_flags = image_bs->open_flags;
3017     ro = bdrv_is_read_only(image_bs);
3018 
3019     if (ro) {
3020         bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
3021         if (local_err) {
3022             error_propagate(errp, local_err);
3023             goto out;
3024         }
3025     }
3026 
3027     ret = bdrv_change_backing_file(image_bs, backing_file,
3028                                image_bs->drv ? image_bs->drv->format_name : "");
3029 
3030     if (ret < 0) {
3031         error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3032                          backing_file);
3033         /* don't exit here, so we can try to restore open flags if
3034          * appropriate */
3035     }
3036 
3037     if (ro) {
3038         bdrv_reopen(image_bs, open_flags, &local_err);
3039         if (local_err) {
3040             error_propagate(errp, local_err); /* will preserve prior errp */
3041         }
3042     }
3043 
3044 out:
3045     aio_context_release(aio_context);
3046 }
3047 
3048 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3049 {
3050     QmpOutputVisitor *ov = qmp_output_visitor_new();
3051     BlockBackend *blk;
3052     QObject *obj;
3053     QDict *qdict;
3054     Error *local_err = NULL;
3055 
3056     /* Require an ID in the top level */
3057     if (!options->has_id) {
3058         error_setg(errp, "Block device needs an ID");
3059         goto fail;
3060     }
3061 
3062     /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
3063      * cache.direct=false instead of silently switching to aio=threads, except
3064      * when called from drive_new().
3065      *
3066      * For now, simply forbidding the combination for all drivers will do. */
3067     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
3068         bool direct = options->has_cache &&
3069                       options->cache->has_direct &&
3070                       options->cache->direct;
3071         if (!direct) {
3072             error_setg(errp, "aio=native requires cache.direct=true");
3073             goto fail;
3074         }
3075     }
3076 
3077     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
3078                                &options, NULL, &local_err);
3079     if (local_err) {
3080         error_propagate(errp, local_err);
3081         goto fail;
3082     }
3083 
3084     obj = qmp_output_get_qobject(ov);
3085     qdict = qobject_to_qdict(obj);
3086 
3087     qdict_flatten(qdict);
3088 
3089     blk = blockdev_init(NULL, qdict, &local_err);
3090     if (local_err) {
3091         error_propagate(errp, local_err);
3092         goto fail;
3093     }
3094 
3095     if (bdrv_key_required(blk_bs(blk))) {
3096         blk_unref(blk);
3097         error_setg(errp, "blockdev-add doesn't support encrypted devices");
3098         goto fail;
3099     }
3100 
3101 fail:
3102     qmp_output_visitor_cleanup(ov);
3103 }
3104 
3105 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3106 {
3107     BlockJobInfoList *head = NULL, **p_next = &head;
3108     BlockDriverState *bs;
3109 
3110     for (bs = bdrv_next(NULL); bs; bs = bdrv_next(bs)) {
3111         AioContext *aio_context = bdrv_get_aio_context(bs);
3112 
3113         aio_context_acquire(aio_context);
3114 
3115         if (bs->job) {
3116             BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
3117             elem->value = block_job_query(bs->job);
3118             *p_next = elem;
3119             p_next = &elem->next;
3120         }
3121 
3122         aio_context_release(aio_context);
3123     }
3124 
3125     return head;
3126 }
3127 
3128 QemuOptsList qemu_common_drive_opts = {
3129     .name = "drive",
3130     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3131     .desc = {
3132         {
3133             .name = "snapshot",
3134             .type = QEMU_OPT_BOOL,
3135             .help = "enable/disable snapshot mode",
3136         },{
3137             .name = "discard",
3138             .type = QEMU_OPT_STRING,
3139             .help = "discard operation (ignore/off, unmap/on)",
3140         },{
3141             .name = BDRV_OPT_CACHE_WB,
3142             .type = QEMU_OPT_BOOL,
3143             .help = "enables writeback mode for any caches",
3144         },{
3145             .name = BDRV_OPT_CACHE_DIRECT,
3146             .type = QEMU_OPT_BOOL,
3147             .help = "enables use of O_DIRECT (bypass the host page cache)",
3148         },{
3149             .name = BDRV_OPT_CACHE_NO_FLUSH,
3150             .type = QEMU_OPT_BOOL,
3151             .help = "ignore any flush requests for the device",
3152         },{
3153             .name = "aio",
3154             .type = QEMU_OPT_STRING,
3155             .help = "host AIO implementation (threads, native)",
3156         },{
3157             .name = "format",
3158             .type = QEMU_OPT_STRING,
3159             .help = "disk format (raw, qcow2, ...)",
3160         },{
3161             .name = "rerror",
3162             .type = QEMU_OPT_STRING,
3163             .help = "read error action",
3164         },{
3165             .name = "werror",
3166             .type = QEMU_OPT_STRING,
3167             .help = "write error action",
3168         },{
3169             .name = "read-only",
3170             .type = QEMU_OPT_BOOL,
3171             .help = "open drive file as read-only",
3172         },{
3173             .name = "throttling.iops-total",
3174             .type = QEMU_OPT_NUMBER,
3175             .help = "limit total I/O operations per second",
3176         },{
3177             .name = "throttling.iops-read",
3178             .type = QEMU_OPT_NUMBER,
3179             .help = "limit read operations per second",
3180         },{
3181             .name = "throttling.iops-write",
3182             .type = QEMU_OPT_NUMBER,
3183             .help = "limit write operations per second",
3184         },{
3185             .name = "throttling.bps-total",
3186             .type = QEMU_OPT_NUMBER,
3187             .help = "limit total bytes per second",
3188         },{
3189             .name = "throttling.bps-read",
3190             .type = QEMU_OPT_NUMBER,
3191             .help = "limit read bytes per second",
3192         },{
3193             .name = "throttling.bps-write",
3194             .type = QEMU_OPT_NUMBER,
3195             .help = "limit write bytes per second",
3196         },{
3197             .name = "throttling.iops-total-max",
3198             .type = QEMU_OPT_NUMBER,
3199             .help = "I/O operations burst",
3200         },{
3201             .name = "throttling.iops-read-max",
3202             .type = QEMU_OPT_NUMBER,
3203             .help = "I/O operations read burst",
3204         },{
3205             .name = "throttling.iops-write-max",
3206             .type = QEMU_OPT_NUMBER,
3207             .help = "I/O operations write burst",
3208         },{
3209             .name = "throttling.bps-total-max",
3210             .type = QEMU_OPT_NUMBER,
3211             .help = "total bytes burst",
3212         },{
3213             .name = "throttling.bps-read-max",
3214             .type = QEMU_OPT_NUMBER,
3215             .help = "total bytes read burst",
3216         },{
3217             .name = "throttling.bps-write-max",
3218             .type = QEMU_OPT_NUMBER,
3219             .help = "total bytes write burst",
3220         },{
3221             .name = "throttling.iops-size",
3222             .type = QEMU_OPT_NUMBER,
3223             .help = "when limiting by iops max size of an I/O in bytes",
3224         },{
3225             .name = "throttling.group",
3226             .type = QEMU_OPT_STRING,
3227             .help = "name of the block throttling group",
3228         },{
3229             .name = "copy-on-read",
3230             .type = QEMU_OPT_BOOL,
3231             .help = "copy read data from backing file into image file",
3232         },{
3233             .name = "detect-zeroes",
3234             .type = QEMU_OPT_STRING,
3235             .help = "try to optimize zero writes (off, on, unmap)",
3236         },
3237         { /* end of list */ }
3238     },
3239 };
3240 
3241 QemuOptsList qemu_drive_opts = {
3242     .name = "drive",
3243     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3244     .desc = {
3245         /*
3246          * no elements => accept any params
3247          * validation will happen later
3248          */
3249         { /* end of list */ }
3250     },
3251 };
3252