xref: /openbmc/qemu/blockdev.c (revision 91bfcdb0)
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     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
415         if (!strcmp(buf, "native")) {
416             bdrv_flags |= BDRV_O_NATIVE_AIO;
417         } else if (!strcmp(buf, "threads")) {
418             /* this is the default */
419         } else {
420            error_setg(errp, "invalid aio option");
421            goto early_err;
422         }
423     }
424 
425     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
426         if (is_help_option(buf)) {
427             error_printf("Supported formats:");
428             bdrv_iterate_format(bdrv_format_print, NULL);
429             error_printf("\n");
430             goto early_err;
431         }
432 
433         if (qdict_haskey(bs_opts, "driver")) {
434             error_setg(errp, "Cannot specify both 'driver' and 'format'");
435             goto early_err;
436         }
437         qdict_put(bs_opts, "driver", qstring_from_str(buf));
438     }
439 
440     /* disk I/O throttling */
441     memset(&cfg, 0, sizeof(cfg));
442     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
443         qemu_opt_get_number(opts, "throttling.bps-total", 0);
444     cfg.buckets[THROTTLE_BPS_READ].avg  =
445         qemu_opt_get_number(opts, "throttling.bps-read", 0);
446     cfg.buckets[THROTTLE_BPS_WRITE].avg =
447         qemu_opt_get_number(opts, "throttling.bps-write", 0);
448     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
449         qemu_opt_get_number(opts, "throttling.iops-total", 0);
450     cfg.buckets[THROTTLE_OPS_READ].avg =
451         qemu_opt_get_number(opts, "throttling.iops-read", 0);
452     cfg.buckets[THROTTLE_OPS_WRITE].avg =
453         qemu_opt_get_number(opts, "throttling.iops-write", 0);
454 
455     cfg.buckets[THROTTLE_BPS_TOTAL].max =
456         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
457     cfg.buckets[THROTTLE_BPS_READ].max  =
458         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
459     cfg.buckets[THROTTLE_BPS_WRITE].max =
460         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
461     cfg.buckets[THROTTLE_OPS_TOTAL].max =
462         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
463     cfg.buckets[THROTTLE_OPS_READ].max =
464         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
465     cfg.buckets[THROTTLE_OPS_WRITE].max =
466         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
467 
468     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
469 
470     throttling_group = qemu_opt_get(opts, "throttling.group");
471 
472     if (!check_throttle_config(&cfg, &error)) {
473         error_propagate(errp, error);
474         goto early_err;
475     }
476 
477     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
478     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
479         on_write_error = parse_block_error_action(buf, 0, &error);
480         if (error) {
481             error_propagate(errp, error);
482             goto early_err;
483         }
484     }
485 
486     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
487     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
488         on_read_error = parse_block_error_action(buf, 1, &error);
489         if (error) {
490             error_propagate(errp, error);
491             goto early_err;
492         }
493     }
494 
495     detect_zeroes =
496         qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
497                         qemu_opt_get(opts, "detect-zeroes"),
498                         BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
499                         BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
500                         &error);
501     if (error) {
502         error_propagate(errp, error);
503         goto early_err;
504     }
505 
506     if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
507         !(bdrv_flags & BDRV_O_UNMAP)) {
508         error_setg(errp, "setting detect-zeroes to unmap is not allowed "
509                          "without setting discard operation to unmap");
510         goto early_err;
511     }
512 
513     /* init */
514     if ((!file || !*file) && !has_driver_specific_opts) {
515         blk = blk_new_with_bs(qemu_opts_id(opts), errp);
516         if (!blk) {
517             goto early_err;
518         }
519 
520         bs = blk_bs(blk);
521         bs->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
522         bs->read_only = ro;
523 
524         QDECREF(bs_opts);
525     } else {
526         if (file && !*file) {
527             file = NULL;
528         }
529 
530         if (snapshot) {
531             /* always use cache=unsafe with snapshot */
532             bdrv_flags &= ~BDRV_O_CACHE_MASK;
533             bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
534         }
535 
536         if (copy_on_read) {
537             bdrv_flags |= BDRV_O_COPY_ON_READ;
538         }
539 
540         if (runstate_check(RUN_STATE_INMIGRATE)) {
541             bdrv_flags |= BDRV_O_INCOMING;
542         }
543 
544         bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
545 
546         blk = blk_new_open(qemu_opts_id(opts), file, NULL, bs_opts, bdrv_flags,
547                            errp);
548         if (!blk) {
549             goto err_no_bs_opts;
550         }
551         bs = blk_bs(blk);
552     }
553 
554     bs->detect_zeroes = detect_zeroes;
555 
556     bdrv_set_on_error(bs, on_read_error, on_write_error);
557 
558     /* disk I/O throttling */
559     if (throttle_enabled(&cfg)) {
560         if (!throttling_group) {
561             throttling_group = blk_name(blk);
562         }
563         bdrv_io_limits_enable(bs, throttling_group);
564         bdrv_set_io_limits(bs, &cfg);
565     }
566 
567     if (bdrv_key_required(bs)) {
568         autostart = 0;
569     }
570 
571 err_no_bs_opts:
572     qemu_opts_del(opts);
573     return blk;
574 
575 early_err:
576     qemu_opts_del(opts);
577 err_no_opts:
578     QDECREF(bs_opts);
579     return NULL;
580 }
581 
582 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
583                             Error **errp)
584 {
585     const char *value;
586 
587     value = qemu_opt_get(opts, from);
588     if (value) {
589         if (qemu_opt_find(opts, to)) {
590             error_setg(errp, "'%s' and its alias '%s' can't be used at the "
591                        "same time", to, from);
592             return;
593         }
594     }
595 
596     /* rename all items in opts */
597     while ((value = qemu_opt_get(opts, from))) {
598         qemu_opt_set(opts, to, value, &error_abort);
599         qemu_opt_unset(opts, from);
600     }
601 }
602 
603 QemuOptsList qemu_legacy_drive_opts = {
604     .name = "drive",
605     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
606     .desc = {
607         {
608             .name = "bus",
609             .type = QEMU_OPT_NUMBER,
610             .help = "bus number",
611         },{
612             .name = "unit",
613             .type = QEMU_OPT_NUMBER,
614             .help = "unit number (i.e. lun for scsi)",
615         },{
616             .name = "index",
617             .type = QEMU_OPT_NUMBER,
618             .help = "index number",
619         },{
620             .name = "media",
621             .type = QEMU_OPT_STRING,
622             .help = "media type (disk, cdrom)",
623         },{
624             .name = "if",
625             .type = QEMU_OPT_STRING,
626             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
627         },{
628             .name = "cyls",
629             .type = QEMU_OPT_NUMBER,
630             .help = "number of cylinders (ide disk geometry)",
631         },{
632             .name = "heads",
633             .type = QEMU_OPT_NUMBER,
634             .help = "number of heads (ide disk geometry)",
635         },{
636             .name = "secs",
637             .type = QEMU_OPT_NUMBER,
638             .help = "number of sectors (ide disk geometry)",
639         },{
640             .name = "trans",
641             .type = QEMU_OPT_STRING,
642             .help = "chs translation (auto, lba, none)",
643         },{
644             .name = "boot",
645             .type = QEMU_OPT_BOOL,
646             .help = "(deprecated, ignored)",
647         },{
648             .name = "addr",
649             .type = QEMU_OPT_STRING,
650             .help = "pci address (virtio only)",
651         },{
652             .name = "serial",
653             .type = QEMU_OPT_STRING,
654             .help = "disk serial number",
655         },{
656             .name = "file",
657             .type = QEMU_OPT_STRING,
658             .help = "file name",
659         },
660 
661         /* Options that are passed on, but have special semantics with -drive */
662         {
663             .name = "read-only",
664             .type = QEMU_OPT_BOOL,
665             .help = "open drive file as read-only",
666         },{
667             .name = "rerror",
668             .type = QEMU_OPT_STRING,
669             .help = "read error action",
670         },{
671             .name = "werror",
672             .type = QEMU_OPT_STRING,
673             .help = "write error action",
674         },{
675             .name = "copy-on-read",
676             .type = QEMU_OPT_BOOL,
677             .help = "copy read data from backing file into image file",
678         },
679 
680         { /* end of list */ }
681     },
682 };
683 
684 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
685 {
686     const char *value;
687     BlockBackend *blk;
688     DriveInfo *dinfo = NULL;
689     QDict *bs_opts;
690     QemuOpts *legacy_opts;
691     DriveMediaType media = MEDIA_DISK;
692     BlockInterfaceType type;
693     int cyls, heads, secs, translation;
694     int max_devs, bus_id, unit_id, index;
695     const char *devaddr;
696     const char *werror, *rerror;
697     bool read_only = false;
698     bool copy_on_read;
699     const char *serial;
700     const char *filename;
701     Error *local_err = NULL;
702     int i;
703 
704     /* Change legacy command line options into QMP ones */
705     static const struct {
706         const char *from;
707         const char *to;
708     } opt_renames[] = {
709         { "iops",           "throttling.iops-total" },
710         { "iops_rd",        "throttling.iops-read" },
711         { "iops_wr",        "throttling.iops-write" },
712 
713         { "bps",            "throttling.bps-total" },
714         { "bps_rd",         "throttling.bps-read" },
715         { "bps_wr",         "throttling.bps-write" },
716 
717         { "iops_max",       "throttling.iops-total-max" },
718         { "iops_rd_max",    "throttling.iops-read-max" },
719         { "iops_wr_max",    "throttling.iops-write-max" },
720 
721         { "bps_max",        "throttling.bps-total-max" },
722         { "bps_rd_max",     "throttling.bps-read-max" },
723         { "bps_wr_max",     "throttling.bps-write-max" },
724 
725         { "iops_size",      "throttling.iops-size" },
726 
727         { "group",          "throttling.group" },
728 
729         { "readonly",       "read-only" },
730     };
731 
732     for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
733         qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
734                         &local_err);
735         if (local_err) {
736             error_report_err(local_err);
737             return NULL;
738         }
739     }
740 
741     value = qemu_opt_get(all_opts, "cache");
742     if (value) {
743         int flags = 0;
744 
745         if (bdrv_parse_cache_flags(value, &flags) != 0) {
746             error_report("invalid cache option");
747             return NULL;
748         }
749 
750         /* Specific options take precedence */
751         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
752             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
753                               !!(flags & BDRV_O_CACHE_WB), &error_abort);
754         }
755         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
756             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
757                               !!(flags & BDRV_O_NOCACHE), &error_abort);
758         }
759         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
760             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
761                               !!(flags & BDRV_O_NO_FLUSH), &error_abort);
762         }
763         qemu_opt_unset(all_opts, "cache");
764     }
765 
766     /* Get a QDict for processing the options */
767     bs_opts = qdict_new();
768     qemu_opts_to_qdict(all_opts, bs_opts);
769 
770     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
771                                    &error_abort);
772     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
773     if (local_err) {
774         error_report_err(local_err);
775         goto fail;
776     }
777 
778     /* Deprecated option boot=[on|off] */
779     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
780         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
781                 "ignored. Future versions will reject this parameter. Please "
782                 "update your scripts.\n");
783     }
784 
785     /* Media type */
786     value = qemu_opt_get(legacy_opts, "media");
787     if (value) {
788         if (!strcmp(value, "disk")) {
789             media = MEDIA_DISK;
790         } else if (!strcmp(value, "cdrom")) {
791             media = MEDIA_CDROM;
792             read_only = true;
793         } else {
794             error_report("'%s' invalid media", value);
795             goto fail;
796         }
797     }
798 
799     /* copy-on-read is disabled with a warning for read-only devices */
800     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
801     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
802 
803     if (read_only && copy_on_read) {
804         error_report("warning: disabling copy-on-read on read-only drive");
805         copy_on_read = false;
806     }
807 
808     qdict_put(bs_opts, "read-only",
809               qstring_from_str(read_only ? "on" : "off"));
810     qdict_put(bs_opts, "copy-on-read",
811               qstring_from_str(copy_on_read ? "on" :"off"));
812 
813     /* Controller type */
814     value = qemu_opt_get(legacy_opts, "if");
815     if (value) {
816         for (type = 0;
817              type < IF_COUNT && strcmp(value, if_name[type]);
818              type++) {
819         }
820         if (type == IF_COUNT) {
821             error_report("unsupported bus type '%s'", value);
822             goto fail;
823         }
824     } else {
825         type = block_default_type;
826     }
827 
828     /* Geometry */
829     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
830     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
831     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
832 
833     if (cyls || heads || secs) {
834         if (cyls < 1) {
835             error_report("invalid physical cyls number");
836             goto fail;
837         }
838         if (heads < 1) {
839             error_report("invalid physical heads number");
840             goto fail;
841         }
842         if (secs < 1) {
843             error_report("invalid physical secs number");
844             goto fail;
845         }
846     }
847 
848     translation = BIOS_ATA_TRANSLATION_AUTO;
849     value = qemu_opt_get(legacy_opts, "trans");
850     if (value != NULL) {
851         if (!cyls) {
852             error_report("'%s' trans must be used with cyls, heads and secs",
853                          value);
854             goto fail;
855         }
856         if (!strcmp(value, "none")) {
857             translation = BIOS_ATA_TRANSLATION_NONE;
858         } else if (!strcmp(value, "lba")) {
859             translation = BIOS_ATA_TRANSLATION_LBA;
860         } else if (!strcmp(value, "large")) {
861             translation = BIOS_ATA_TRANSLATION_LARGE;
862         } else if (!strcmp(value, "rechs")) {
863             translation = BIOS_ATA_TRANSLATION_RECHS;
864         } else if (!strcmp(value, "auto")) {
865             translation = BIOS_ATA_TRANSLATION_AUTO;
866         } else {
867             error_report("'%s' invalid translation type", value);
868             goto fail;
869         }
870     }
871 
872     if (media == MEDIA_CDROM) {
873         if (cyls || secs || heads) {
874             error_report("CHS can't be set with media=cdrom");
875             goto fail;
876         }
877     }
878 
879     /* Device address specified by bus/unit or index.
880      * If none was specified, try to find the first free one. */
881     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
882     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
883     index   = qemu_opt_get_number(legacy_opts, "index", -1);
884 
885     max_devs = if_max_devs[type];
886 
887     if (index != -1) {
888         if (bus_id != 0 || unit_id != -1) {
889             error_report("index cannot be used with bus and unit");
890             goto fail;
891         }
892         bus_id = drive_index_to_bus_id(type, index);
893         unit_id = drive_index_to_unit_id(type, index);
894     }
895 
896     if (unit_id == -1) {
897        unit_id = 0;
898        while (drive_get(type, bus_id, unit_id) != NULL) {
899            unit_id++;
900            if (max_devs && unit_id >= max_devs) {
901                unit_id -= max_devs;
902                bus_id++;
903            }
904        }
905     }
906 
907     if (max_devs && unit_id >= max_devs) {
908         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
909         goto fail;
910     }
911 
912     if (drive_get(type, bus_id, unit_id) != NULL) {
913         error_report("drive with bus=%d, unit=%d (index=%d) exists",
914                      bus_id, unit_id, index);
915         goto fail;
916     }
917 
918     /* Serial number */
919     serial = qemu_opt_get(legacy_opts, "serial");
920 
921     /* no id supplied -> create one */
922     if (qemu_opts_id(all_opts) == NULL) {
923         char *new_id;
924         const char *mediastr = "";
925         if (type == IF_IDE || type == IF_SCSI) {
926             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
927         }
928         if (max_devs) {
929             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
930                                      mediastr, unit_id);
931         } else {
932             new_id = g_strdup_printf("%s%s%i", if_name[type],
933                                      mediastr, unit_id);
934         }
935         qdict_put(bs_opts, "id", qstring_from_str(new_id));
936         g_free(new_id);
937     }
938 
939     /* Add virtio block device */
940     devaddr = qemu_opt_get(legacy_opts, "addr");
941     if (devaddr && type != IF_VIRTIO) {
942         error_report("addr is not supported by this bus type");
943         goto fail;
944     }
945 
946     if (type == IF_VIRTIO) {
947         QemuOpts *devopts;
948         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
949                                    &error_abort);
950         if (arch_type == QEMU_ARCH_S390X) {
951             qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
952         } else {
953             qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
954         }
955         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
956                      &error_abort);
957         if (devaddr) {
958             qemu_opt_set(devopts, "addr", devaddr, &error_abort);
959         }
960     }
961 
962     filename = qemu_opt_get(legacy_opts, "file");
963 
964     /* Check werror/rerror compatibility with if=... */
965     werror = qemu_opt_get(legacy_opts, "werror");
966     if (werror != NULL) {
967         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
968             type != IF_NONE) {
969             error_report("werror is not supported by this bus type");
970             goto fail;
971         }
972         qdict_put(bs_opts, "werror", qstring_from_str(werror));
973     }
974 
975     rerror = qemu_opt_get(legacy_opts, "rerror");
976     if (rerror != NULL) {
977         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
978             type != IF_NONE) {
979             error_report("rerror is not supported by this bus type");
980             goto fail;
981         }
982         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
983     }
984 
985     /* Actual block device init: Functionality shared with blockdev-add */
986     blk = blockdev_init(filename, bs_opts, &local_err);
987     bs_opts = NULL;
988     if (!blk) {
989         if (local_err) {
990             error_report_err(local_err);
991         }
992         goto fail;
993     } else {
994         assert(!local_err);
995     }
996 
997     /* Create legacy DriveInfo */
998     dinfo = g_malloc0(sizeof(*dinfo));
999     dinfo->opts = all_opts;
1000 
1001     dinfo->cyls = cyls;
1002     dinfo->heads = heads;
1003     dinfo->secs = secs;
1004     dinfo->trans = translation;
1005 
1006     dinfo->type = type;
1007     dinfo->bus = bus_id;
1008     dinfo->unit = unit_id;
1009     dinfo->devaddr = devaddr;
1010     dinfo->serial = g_strdup(serial);
1011 
1012     blk_set_legacy_dinfo(blk, dinfo);
1013 
1014     switch(type) {
1015     case IF_IDE:
1016     case IF_SCSI:
1017     case IF_XEN:
1018     case IF_NONE:
1019         dinfo->media_cd = media == MEDIA_CDROM;
1020         break;
1021     default:
1022         break;
1023     }
1024 
1025 fail:
1026     qemu_opts_del(legacy_opts);
1027     QDECREF(bs_opts);
1028     return dinfo;
1029 }
1030 
1031 void hmp_commit(Monitor *mon, const QDict *qdict)
1032 {
1033     const char *device = qdict_get_str(qdict, "device");
1034     BlockBackend *blk;
1035     int ret;
1036 
1037     if (!strcmp(device, "all")) {
1038         ret = bdrv_commit_all();
1039     } else {
1040         blk = blk_by_name(device);
1041         if (!blk) {
1042             monitor_printf(mon, "Device '%s' not found\n", device);
1043             return;
1044         }
1045         ret = bdrv_commit(blk_bs(blk));
1046     }
1047     if (ret < 0) {
1048         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1049                        strerror(-ret));
1050     }
1051 }
1052 
1053 static void blockdev_do_action(int kind, void *data, Error **errp)
1054 {
1055     TransactionAction action;
1056     TransactionActionList list;
1057 
1058     action.kind = kind;
1059     action.data = data;
1060     list.value = &action;
1061     list.next = NULL;
1062     qmp_transaction(&list, errp);
1063 }
1064 
1065 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1066                                 bool has_node_name, const char *node_name,
1067                                 const char *snapshot_file,
1068                                 bool has_snapshot_node_name,
1069                                 const char *snapshot_node_name,
1070                                 bool has_format, const char *format,
1071                                 bool has_mode, NewImageMode mode, Error **errp)
1072 {
1073     BlockdevSnapshot snapshot = {
1074         .has_device = has_device,
1075         .device = (char *) device,
1076         .has_node_name = has_node_name,
1077         .node_name = (char *) node_name,
1078         .snapshot_file = (char *) snapshot_file,
1079         .has_snapshot_node_name = has_snapshot_node_name,
1080         .snapshot_node_name = (char *) snapshot_node_name,
1081         .has_format = has_format,
1082         .format = (char *) format,
1083         .has_mode = has_mode,
1084         .mode = mode,
1085     };
1086     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1087                        &snapshot, errp);
1088 }
1089 
1090 void qmp_blockdev_snapshot_internal_sync(const char *device,
1091                                          const char *name,
1092                                          Error **errp)
1093 {
1094     BlockdevSnapshotInternal snapshot = {
1095         .device = (char *) device,
1096         .name = (char *) name
1097     };
1098 
1099     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1100                        &snapshot, errp);
1101 }
1102 
1103 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1104                                                          bool has_id,
1105                                                          const char *id,
1106                                                          bool has_name,
1107                                                          const char *name,
1108                                                          Error **errp)
1109 {
1110     BlockDriverState *bs;
1111     BlockBackend *blk;
1112     AioContext *aio_context;
1113     QEMUSnapshotInfo sn;
1114     Error *local_err = NULL;
1115     SnapshotInfo *info = NULL;
1116     int ret;
1117 
1118     blk = blk_by_name(device);
1119     if (!blk) {
1120         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1121                   "Device '%s' not found", device);
1122         return NULL;
1123     }
1124     bs = blk_bs(blk);
1125 
1126     if (!has_id) {
1127         id = NULL;
1128     }
1129 
1130     if (!has_name) {
1131         name = NULL;
1132     }
1133 
1134     if (!id && !name) {
1135         error_setg(errp, "Name or id must be provided");
1136         return NULL;
1137     }
1138 
1139     aio_context = bdrv_get_aio_context(bs);
1140     aio_context_acquire(aio_context);
1141 
1142     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1143         goto out_aio_context;
1144     }
1145 
1146     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1147     if (local_err) {
1148         error_propagate(errp, local_err);
1149         goto out_aio_context;
1150     }
1151     if (!ret) {
1152         error_setg(errp,
1153                    "Snapshot with id '%s' and name '%s' does not exist on "
1154                    "device '%s'",
1155                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1156         goto out_aio_context;
1157     }
1158 
1159     bdrv_snapshot_delete(bs, id, name, &local_err);
1160     if (local_err) {
1161         error_propagate(errp, local_err);
1162         goto out_aio_context;
1163     }
1164 
1165     aio_context_release(aio_context);
1166 
1167     info = g_new0(SnapshotInfo, 1);
1168     info->id = g_strdup(sn.id_str);
1169     info->name = g_strdup(sn.name);
1170     info->date_nsec = sn.date_nsec;
1171     info->date_sec = sn.date_sec;
1172     info->vm_state_size = sn.vm_state_size;
1173     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1174     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1175 
1176     return info;
1177 
1178 out_aio_context:
1179     aio_context_release(aio_context);
1180     return NULL;
1181 }
1182 
1183 /**
1184  * block_dirty_bitmap_lookup:
1185  * Return a dirty bitmap (if present), after validating
1186  * the node reference and bitmap names.
1187  *
1188  * @node: The name of the BDS node to search for bitmaps
1189  * @name: The name of the bitmap to search for
1190  * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1191  * @paio: Output pointer for aio_context acquisition, if desired. Can be NULL.
1192  * @errp: Output pointer for error information. Can be NULL.
1193  *
1194  * @return: A bitmap object on success, or NULL on failure.
1195  */
1196 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1197                                                   const char *name,
1198                                                   BlockDriverState **pbs,
1199                                                   AioContext **paio,
1200                                                   Error **errp)
1201 {
1202     BlockDriverState *bs;
1203     BdrvDirtyBitmap *bitmap;
1204     AioContext *aio_context;
1205 
1206     if (!node) {
1207         error_setg(errp, "Node cannot be NULL");
1208         return NULL;
1209     }
1210     if (!name) {
1211         error_setg(errp, "Bitmap name cannot be NULL");
1212         return NULL;
1213     }
1214     bs = bdrv_lookup_bs(node, node, NULL);
1215     if (!bs) {
1216         error_setg(errp, "Node '%s' not found", node);
1217         return NULL;
1218     }
1219 
1220     aio_context = bdrv_get_aio_context(bs);
1221     aio_context_acquire(aio_context);
1222 
1223     bitmap = bdrv_find_dirty_bitmap(bs, name);
1224     if (!bitmap) {
1225         error_setg(errp, "Dirty bitmap '%s' not found", name);
1226         goto fail;
1227     }
1228 
1229     if (pbs) {
1230         *pbs = bs;
1231     }
1232     if (paio) {
1233         *paio = aio_context;
1234     } else {
1235         aio_context_release(aio_context);
1236     }
1237 
1238     return bitmap;
1239 
1240  fail:
1241     aio_context_release(aio_context);
1242     return NULL;
1243 }
1244 
1245 /* New and old BlockDriverState structs for atomic group operations */
1246 
1247 typedef struct BlkTransactionState BlkTransactionState;
1248 
1249 /* Only prepare() may fail. In a single transaction, only one of commit() or
1250    abort() will be called, clean() will always be called if it present. */
1251 typedef struct BdrvActionOps {
1252     /* Size of state struct, in bytes. */
1253     size_t instance_size;
1254     /* Prepare the work, must NOT be NULL. */
1255     void (*prepare)(BlkTransactionState *common, Error **errp);
1256     /* Commit the changes, can be NULL. */
1257     void (*commit)(BlkTransactionState *common);
1258     /* Abort the changes on fail, can be NULL. */
1259     void (*abort)(BlkTransactionState *common);
1260     /* Clean up resource in the end, can be NULL. */
1261     void (*clean)(BlkTransactionState *common);
1262 } BdrvActionOps;
1263 
1264 /*
1265  * This structure must be arranged as first member in child type, assuming
1266  * that compiler will also arrange it to the same address with parent instance.
1267  * Later it will be used in free().
1268  */
1269 struct BlkTransactionState {
1270     TransactionAction *action;
1271     const BdrvActionOps *ops;
1272     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1273 };
1274 
1275 /* internal snapshot private data */
1276 typedef struct InternalSnapshotState {
1277     BlkTransactionState common;
1278     BlockDriverState *bs;
1279     AioContext *aio_context;
1280     QEMUSnapshotInfo sn;
1281 } InternalSnapshotState;
1282 
1283 static void internal_snapshot_prepare(BlkTransactionState *common,
1284                                       Error **errp)
1285 {
1286     Error *local_err = NULL;
1287     const char *device;
1288     const char *name;
1289     BlockBackend *blk;
1290     BlockDriverState *bs;
1291     QEMUSnapshotInfo old_sn, *sn;
1292     bool ret;
1293     qemu_timeval tv;
1294     BlockdevSnapshotInternal *internal;
1295     InternalSnapshotState *state;
1296     int ret1;
1297 
1298     g_assert(common->action->kind ==
1299              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1300     internal = common->action->blockdev_snapshot_internal_sync;
1301     state = DO_UPCAST(InternalSnapshotState, common, common);
1302 
1303     /* 1. parse input */
1304     device = internal->device;
1305     name = internal->name;
1306 
1307     /* 2. check for validation */
1308     blk = blk_by_name(device);
1309     if (!blk) {
1310         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1311                   "Device '%s' not found", device);
1312         return;
1313     }
1314     bs = blk_bs(blk);
1315 
1316     /* AioContext is released in .clean() */
1317     state->aio_context = bdrv_get_aio_context(bs);
1318     aio_context_acquire(state->aio_context);
1319 
1320     if (!bdrv_is_inserted(bs)) {
1321         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1322         return;
1323     }
1324 
1325     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1326         return;
1327     }
1328 
1329     if (bdrv_is_read_only(bs)) {
1330         error_setg(errp, "Device '%s' is read only", device);
1331         return;
1332     }
1333 
1334     if (!bdrv_can_snapshot(bs)) {
1335         error_setg(errp, "Block format '%s' used by device '%s' "
1336                    "does not support internal snapshots",
1337                    bs->drv->format_name, device);
1338         return;
1339     }
1340 
1341     if (!strlen(name)) {
1342         error_setg(errp, "Name is empty");
1343         return;
1344     }
1345 
1346     /* check whether a snapshot with name exist */
1347     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1348                                             &local_err);
1349     if (local_err) {
1350         error_propagate(errp, local_err);
1351         return;
1352     } else if (ret) {
1353         error_setg(errp,
1354                    "Snapshot with name '%s' already exists on device '%s'",
1355                    name, device);
1356         return;
1357     }
1358 
1359     /* 3. take the snapshot */
1360     sn = &state->sn;
1361     pstrcpy(sn->name, sizeof(sn->name), name);
1362     qemu_gettimeofday(&tv);
1363     sn->date_sec = tv.tv_sec;
1364     sn->date_nsec = tv.tv_usec * 1000;
1365     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1366 
1367     ret1 = bdrv_snapshot_create(bs, sn);
1368     if (ret1 < 0) {
1369         error_setg_errno(errp, -ret1,
1370                          "Failed to create snapshot '%s' on device '%s'",
1371                          name, device);
1372         return;
1373     }
1374 
1375     /* 4. succeed, mark a snapshot is created */
1376     state->bs = bs;
1377 }
1378 
1379 static void internal_snapshot_abort(BlkTransactionState *common)
1380 {
1381     InternalSnapshotState *state =
1382                              DO_UPCAST(InternalSnapshotState, common, common);
1383     BlockDriverState *bs = state->bs;
1384     QEMUSnapshotInfo *sn = &state->sn;
1385     Error *local_error = NULL;
1386 
1387     if (!bs) {
1388         return;
1389     }
1390 
1391     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1392         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1393                      "device '%s' in abort: %s",
1394                      sn->id_str,
1395                      sn->name,
1396                      bdrv_get_device_name(bs),
1397                      error_get_pretty(local_error));
1398         error_free(local_error);
1399     }
1400 }
1401 
1402 static void internal_snapshot_clean(BlkTransactionState *common)
1403 {
1404     InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1405                                              common, common);
1406 
1407     if (state->aio_context) {
1408         aio_context_release(state->aio_context);
1409     }
1410 }
1411 
1412 /* external snapshot private data */
1413 typedef struct ExternalSnapshotState {
1414     BlkTransactionState common;
1415     BlockDriverState *old_bs;
1416     BlockDriverState *new_bs;
1417     AioContext *aio_context;
1418 } ExternalSnapshotState;
1419 
1420 static void external_snapshot_prepare(BlkTransactionState *common,
1421                                       Error **errp)
1422 {
1423     int flags, ret;
1424     QDict *options;
1425     Error *local_err = NULL;
1426     bool has_device = false;
1427     const char *device;
1428     bool has_node_name = false;
1429     const char *node_name;
1430     bool has_snapshot_node_name = false;
1431     const char *snapshot_node_name;
1432     const char *new_image_file;
1433     const char *format = "qcow2";
1434     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1435     ExternalSnapshotState *state =
1436                              DO_UPCAST(ExternalSnapshotState, common, common);
1437     TransactionAction *action = common->action;
1438 
1439     /* get parameters */
1440     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1441 
1442     has_device = action->blockdev_snapshot_sync->has_device;
1443     device = action->blockdev_snapshot_sync->device;
1444     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1445     node_name = action->blockdev_snapshot_sync->node_name;
1446     has_snapshot_node_name =
1447         action->blockdev_snapshot_sync->has_snapshot_node_name;
1448     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1449 
1450     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1451     if (action->blockdev_snapshot_sync->has_format) {
1452         format = action->blockdev_snapshot_sync->format;
1453     }
1454     if (action->blockdev_snapshot_sync->has_mode) {
1455         mode = action->blockdev_snapshot_sync->mode;
1456     }
1457 
1458     /* start processing */
1459     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1460                                    has_node_name ? node_name : NULL,
1461                                    &local_err);
1462     if (local_err) {
1463         error_propagate(errp, local_err);
1464         return;
1465     }
1466 
1467     if (has_node_name && !has_snapshot_node_name) {
1468         error_setg(errp, "New snapshot node name missing");
1469         return;
1470     }
1471 
1472     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1473         error_setg(errp, "New snapshot node name already existing");
1474         return;
1475     }
1476 
1477     /* Acquire AioContext now so any threads operating on old_bs stop */
1478     state->aio_context = bdrv_get_aio_context(state->old_bs);
1479     aio_context_acquire(state->aio_context);
1480 
1481     if (!bdrv_is_inserted(state->old_bs)) {
1482         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1483         return;
1484     }
1485 
1486     if (bdrv_op_is_blocked(state->old_bs,
1487                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1488         return;
1489     }
1490 
1491     if (!bdrv_is_read_only(state->old_bs)) {
1492         if (bdrv_flush(state->old_bs)) {
1493             error_setg(errp, QERR_IO_ERROR);
1494             return;
1495         }
1496     }
1497 
1498     if (!bdrv_is_first_non_filter(state->old_bs)) {
1499         error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1500         return;
1501     }
1502 
1503     flags = state->old_bs->open_flags;
1504 
1505     /* create new image w/backing file */
1506     if (mode != NEW_IMAGE_MODE_EXISTING) {
1507         bdrv_img_create(new_image_file, format,
1508                         state->old_bs->filename,
1509                         state->old_bs->drv->format_name,
1510                         NULL, -1, flags, &local_err, false);
1511         if (local_err) {
1512             error_propagate(errp, local_err);
1513             return;
1514         }
1515     }
1516 
1517     options = qdict_new();
1518     if (has_snapshot_node_name) {
1519         qdict_put(options, "node-name",
1520                   qstring_from_str(snapshot_node_name));
1521     }
1522     qdict_put(options, "driver", qstring_from_str(format));
1523 
1524     /* TODO Inherit bs->options or only take explicit options with an
1525      * extended QMP command? */
1526     assert(state->new_bs == NULL);
1527     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1528                     flags | BDRV_O_NO_BACKING, &local_err);
1529     /* We will manually add the backing_hd field to the bs later */
1530     if (ret != 0) {
1531         error_propagate(errp, local_err);
1532     }
1533 }
1534 
1535 static void external_snapshot_commit(BlkTransactionState *common)
1536 {
1537     ExternalSnapshotState *state =
1538                              DO_UPCAST(ExternalSnapshotState, common, common);
1539 
1540     bdrv_set_aio_context(state->new_bs, state->aio_context);
1541 
1542     /* This removes our old bs and adds the new bs */
1543     bdrv_append(state->new_bs, state->old_bs);
1544     /* We don't need (or want) to use the transactional
1545      * bdrv_reopen_multiple() across all the entries at once, because we
1546      * don't want to abort all of them if one of them fails the reopen */
1547     bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR,
1548                 NULL);
1549 
1550     aio_context_release(state->aio_context);
1551 }
1552 
1553 static void external_snapshot_abort(BlkTransactionState *common)
1554 {
1555     ExternalSnapshotState *state =
1556                              DO_UPCAST(ExternalSnapshotState, common, common);
1557     if (state->new_bs) {
1558         bdrv_unref(state->new_bs);
1559     }
1560     if (state->aio_context) {
1561         aio_context_release(state->aio_context);
1562     }
1563 }
1564 
1565 typedef struct DriveBackupState {
1566     BlkTransactionState common;
1567     BlockDriverState *bs;
1568     AioContext *aio_context;
1569     BlockJob *job;
1570 } DriveBackupState;
1571 
1572 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1573 {
1574     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1575     BlockDriverState *bs;
1576     BlockBackend *blk;
1577     DriveBackup *backup;
1578     Error *local_err = NULL;
1579 
1580     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1581     backup = common->action->drive_backup;
1582 
1583     blk = blk_by_name(backup->device);
1584     if (!blk) {
1585         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1586                   "Device '%s' not found", backup->device);
1587         return;
1588     }
1589     bs = blk_bs(blk);
1590 
1591     /* AioContext is released in .clean() */
1592     state->aio_context = bdrv_get_aio_context(bs);
1593     aio_context_acquire(state->aio_context);
1594 
1595     qmp_drive_backup(backup->device, backup->target,
1596                      backup->has_format, backup->format,
1597                      backup->sync,
1598                      backup->has_mode, backup->mode,
1599                      backup->has_speed, backup->speed,
1600                      backup->has_bitmap, backup->bitmap,
1601                      backup->has_on_source_error, backup->on_source_error,
1602                      backup->has_on_target_error, backup->on_target_error,
1603                      &local_err);
1604     if (local_err) {
1605         error_propagate(errp, local_err);
1606         return;
1607     }
1608 
1609     state->bs = bs;
1610     state->job = state->bs->job;
1611 }
1612 
1613 static void drive_backup_abort(BlkTransactionState *common)
1614 {
1615     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1616     BlockDriverState *bs = state->bs;
1617 
1618     /* Only cancel if it's the job we started */
1619     if (bs && bs->job && bs->job == state->job) {
1620         block_job_cancel_sync(bs->job);
1621     }
1622 }
1623 
1624 static void drive_backup_clean(BlkTransactionState *common)
1625 {
1626     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1627 
1628     if (state->aio_context) {
1629         aio_context_release(state->aio_context);
1630     }
1631 }
1632 
1633 typedef struct BlockdevBackupState {
1634     BlkTransactionState common;
1635     BlockDriverState *bs;
1636     BlockJob *job;
1637     AioContext *aio_context;
1638 } BlockdevBackupState;
1639 
1640 static void blockdev_backup_prepare(BlkTransactionState *common, Error **errp)
1641 {
1642     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1643     BlockdevBackup *backup;
1644     BlockDriverState *bs, *target;
1645     BlockBackend *blk;
1646     Error *local_err = NULL;
1647 
1648     assert(common->action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1649     backup = common->action->blockdev_backup;
1650 
1651     blk = blk_by_name(backup->device);
1652     if (!blk) {
1653         error_setg(errp, "Device '%s' not found", backup->device);
1654         return;
1655     }
1656     bs = blk_bs(blk);
1657 
1658     blk = blk_by_name(backup->target);
1659     if (!blk) {
1660         error_setg(errp, "Device '%s' not found", backup->target);
1661         return;
1662     }
1663     target = blk_bs(blk);
1664 
1665     /* AioContext is released in .clean() */
1666     state->aio_context = bdrv_get_aio_context(bs);
1667     if (state->aio_context != bdrv_get_aio_context(target)) {
1668         state->aio_context = NULL;
1669         error_setg(errp, "Backup between two IO threads is not implemented");
1670         return;
1671     }
1672     aio_context_acquire(state->aio_context);
1673 
1674     qmp_blockdev_backup(backup->device, backup->target,
1675                         backup->sync,
1676                         backup->has_speed, backup->speed,
1677                         backup->has_on_source_error, backup->on_source_error,
1678                         backup->has_on_target_error, backup->on_target_error,
1679                         &local_err);
1680     if (local_err) {
1681         error_propagate(errp, local_err);
1682         return;
1683     }
1684 
1685     state->bs = bs;
1686     state->job = state->bs->job;
1687 }
1688 
1689 static void blockdev_backup_abort(BlkTransactionState *common)
1690 {
1691     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1692     BlockDriverState *bs = state->bs;
1693 
1694     /* Only cancel if it's the job we started */
1695     if (bs && bs->job && bs->job == state->job) {
1696         block_job_cancel_sync(bs->job);
1697     }
1698 }
1699 
1700 static void blockdev_backup_clean(BlkTransactionState *common)
1701 {
1702     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1703 
1704     if (state->aio_context) {
1705         aio_context_release(state->aio_context);
1706     }
1707 }
1708 
1709 static void abort_prepare(BlkTransactionState *common, Error **errp)
1710 {
1711     error_setg(errp, "Transaction aborted using Abort action");
1712 }
1713 
1714 static void abort_commit(BlkTransactionState *common)
1715 {
1716     g_assert_not_reached(); /* this action never succeeds */
1717 }
1718 
1719 static const BdrvActionOps actions[] = {
1720     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1721         .instance_size = sizeof(ExternalSnapshotState),
1722         .prepare  = external_snapshot_prepare,
1723         .commit   = external_snapshot_commit,
1724         .abort = external_snapshot_abort,
1725     },
1726     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1727         .instance_size = sizeof(DriveBackupState),
1728         .prepare = drive_backup_prepare,
1729         .abort = drive_backup_abort,
1730         .clean = drive_backup_clean,
1731     },
1732     [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
1733         .instance_size = sizeof(BlockdevBackupState),
1734         .prepare = blockdev_backup_prepare,
1735         .abort = blockdev_backup_abort,
1736         .clean = blockdev_backup_clean,
1737     },
1738     [TRANSACTION_ACTION_KIND_ABORT] = {
1739         .instance_size = sizeof(BlkTransactionState),
1740         .prepare = abort_prepare,
1741         .commit = abort_commit,
1742     },
1743     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1744         .instance_size = sizeof(InternalSnapshotState),
1745         .prepare  = internal_snapshot_prepare,
1746         .abort = internal_snapshot_abort,
1747         .clean = internal_snapshot_clean,
1748     },
1749 };
1750 
1751 /*
1752  * 'Atomic' group operations.  The operations are performed as a set, and if
1753  * any fail then we roll back all operations in the group.
1754  */
1755 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1756 {
1757     TransactionActionList *dev_entry = dev_list;
1758     BlkTransactionState *state, *next;
1759     Error *local_err = NULL;
1760 
1761     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1762     QSIMPLEQ_INIT(&snap_bdrv_states);
1763 
1764     /* drain all i/o before any operations */
1765     bdrv_drain_all();
1766 
1767     /* We don't do anything in this loop that commits us to the operations */
1768     while (NULL != dev_entry) {
1769         TransactionAction *dev_info = NULL;
1770         const BdrvActionOps *ops;
1771 
1772         dev_info = dev_entry->value;
1773         dev_entry = dev_entry->next;
1774 
1775         assert(dev_info->kind < ARRAY_SIZE(actions));
1776 
1777         ops = &actions[dev_info->kind];
1778         assert(ops->instance_size > 0);
1779 
1780         state = g_malloc0(ops->instance_size);
1781         state->ops = ops;
1782         state->action = dev_info;
1783         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1784 
1785         state->ops->prepare(state, &local_err);
1786         if (local_err) {
1787             error_propagate(errp, local_err);
1788             goto delete_and_fail;
1789         }
1790     }
1791 
1792     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1793         if (state->ops->commit) {
1794             state->ops->commit(state);
1795         }
1796     }
1797 
1798     /* success */
1799     goto exit;
1800 
1801 delete_and_fail:
1802     /* failure, and it is all-or-none; roll back all operations */
1803     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1804         if (state->ops->abort) {
1805             state->ops->abort(state);
1806         }
1807     }
1808 exit:
1809     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1810         if (state->ops->clean) {
1811             state->ops->clean(state);
1812         }
1813         g_free(state);
1814     }
1815 }
1816 
1817 
1818 static void eject_device(BlockBackend *blk, int force, Error **errp)
1819 {
1820     BlockDriverState *bs = blk_bs(blk);
1821     AioContext *aio_context;
1822 
1823     aio_context = bdrv_get_aio_context(bs);
1824     aio_context_acquire(aio_context);
1825 
1826     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1827         goto out;
1828     }
1829     if (!blk_dev_has_removable_media(blk)) {
1830         error_setg(errp, "Device '%s' is not removable",
1831                    bdrv_get_device_name(bs));
1832         goto out;
1833     }
1834 
1835     if (blk_dev_is_medium_locked(blk) && !blk_dev_is_tray_open(blk)) {
1836         blk_dev_eject_request(blk, force);
1837         if (!force) {
1838             error_setg(errp, "Device '%s' is locked",
1839                        bdrv_get_device_name(bs));
1840             goto out;
1841         }
1842     }
1843 
1844     bdrv_close(bs);
1845 
1846 out:
1847     aio_context_release(aio_context);
1848 }
1849 
1850 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1851 {
1852     BlockBackend *blk;
1853 
1854     blk = blk_by_name(device);
1855     if (!blk) {
1856         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1857                   "Device '%s' not found", device);
1858         return;
1859     }
1860 
1861     eject_device(blk, force, errp);
1862 }
1863 
1864 void qmp_block_passwd(bool has_device, const char *device,
1865                       bool has_node_name, const char *node_name,
1866                       const char *password, Error **errp)
1867 {
1868     Error *local_err = NULL;
1869     BlockDriverState *bs;
1870     AioContext *aio_context;
1871 
1872     bs = bdrv_lookup_bs(has_device ? device : NULL,
1873                         has_node_name ? node_name : NULL,
1874                         &local_err);
1875     if (local_err) {
1876         error_propagate(errp, local_err);
1877         return;
1878     }
1879 
1880     aio_context = bdrv_get_aio_context(bs);
1881     aio_context_acquire(aio_context);
1882 
1883     bdrv_add_key(bs, password, errp);
1884 
1885     aio_context_release(aio_context);
1886 }
1887 
1888 /* Assumes AioContext is held */
1889 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1890                                     int bdrv_flags, const char *format,
1891                                     const char *password, Error **errp)
1892 {
1893     Error *local_err = NULL;
1894     QDict *options = NULL;
1895     int ret;
1896 
1897     if (format) {
1898         options = qdict_new();
1899         qdict_put(options, "driver", qstring_from_str(format));
1900     }
1901 
1902     ret = bdrv_open(&bs, filename, NULL, options, bdrv_flags, &local_err);
1903     if (ret < 0) {
1904         error_propagate(errp, local_err);
1905         return;
1906     }
1907 
1908     bdrv_add_key(bs, password, errp);
1909 }
1910 
1911 void qmp_change_blockdev(const char *device, const char *filename,
1912                          const char *format, Error **errp)
1913 {
1914     BlockBackend *blk;
1915     BlockDriverState *bs;
1916     AioContext *aio_context;
1917     int bdrv_flags;
1918     Error *err = NULL;
1919 
1920     blk = blk_by_name(device);
1921     if (!blk) {
1922         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1923                   "Device '%s' not found", device);
1924         return;
1925     }
1926     bs = blk_bs(blk);
1927 
1928     aio_context = bdrv_get_aio_context(bs);
1929     aio_context_acquire(aio_context);
1930 
1931     eject_device(blk, 0, &err);
1932     if (err) {
1933         error_propagate(errp, err);
1934         goto out;
1935     }
1936 
1937     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1938     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1939 
1940     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, format, NULL, errp);
1941 
1942 out:
1943     aio_context_release(aio_context);
1944 }
1945 
1946 /* throttling disk I/O limits */
1947 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1948                                int64_t bps_wr,
1949                                int64_t iops,
1950                                int64_t iops_rd,
1951                                int64_t iops_wr,
1952                                bool has_bps_max,
1953                                int64_t bps_max,
1954                                bool has_bps_rd_max,
1955                                int64_t bps_rd_max,
1956                                bool has_bps_wr_max,
1957                                int64_t bps_wr_max,
1958                                bool has_iops_max,
1959                                int64_t iops_max,
1960                                bool has_iops_rd_max,
1961                                int64_t iops_rd_max,
1962                                bool has_iops_wr_max,
1963                                int64_t iops_wr_max,
1964                                bool has_iops_size,
1965                                int64_t iops_size,
1966                                bool has_group,
1967                                const char *group, Error **errp)
1968 {
1969     ThrottleConfig cfg;
1970     BlockDriverState *bs;
1971     BlockBackend *blk;
1972     AioContext *aio_context;
1973 
1974     blk = blk_by_name(device);
1975     if (!blk) {
1976         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1977                   "Device '%s' not found", device);
1978         return;
1979     }
1980     bs = blk_bs(blk);
1981 
1982     memset(&cfg, 0, sizeof(cfg));
1983     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1984     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1985     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1986 
1987     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1988     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1989     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1990 
1991     if (has_bps_max) {
1992         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1993     }
1994     if (has_bps_rd_max) {
1995         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1996     }
1997     if (has_bps_wr_max) {
1998         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1999     }
2000     if (has_iops_max) {
2001         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
2002     }
2003     if (has_iops_rd_max) {
2004         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
2005     }
2006     if (has_iops_wr_max) {
2007         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
2008     }
2009 
2010     if (has_iops_size) {
2011         cfg.op_size = iops_size;
2012     }
2013 
2014     if (!check_throttle_config(&cfg, errp)) {
2015         return;
2016     }
2017 
2018     aio_context = bdrv_get_aio_context(bs);
2019     aio_context_acquire(aio_context);
2020 
2021     if (throttle_enabled(&cfg)) {
2022         /* Enable I/O limits if they're not enabled yet, otherwise
2023          * just update the throttling group. */
2024         if (!bs->io_limits_enabled) {
2025             bdrv_io_limits_enable(bs, has_group ? group : device);
2026         } else if (has_group) {
2027             bdrv_io_limits_update_group(bs, group);
2028         }
2029         /* Set the new throttling configuration */
2030         bdrv_set_io_limits(bs, &cfg);
2031     } else if (bs->io_limits_enabled) {
2032         /* If all throttling settings are set to 0, disable I/O limits */
2033         bdrv_io_limits_disable(bs);
2034     }
2035 
2036     aio_context_release(aio_context);
2037 }
2038 
2039 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2040                                 bool has_granularity, uint32_t granularity,
2041                                 Error **errp)
2042 {
2043     AioContext *aio_context;
2044     BlockDriverState *bs;
2045 
2046     if (!name || name[0] == '\0') {
2047         error_setg(errp, "Bitmap name cannot be empty");
2048         return;
2049     }
2050 
2051     bs = bdrv_lookup_bs(node, node, errp);
2052     if (!bs) {
2053         return;
2054     }
2055 
2056     aio_context = bdrv_get_aio_context(bs);
2057     aio_context_acquire(aio_context);
2058 
2059     if (has_granularity) {
2060         if (granularity < 512 || !is_power_of_2(granularity)) {
2061             error_setg(errp, "Granularity must be power of 2 "
2062                              "and at least 512");
2063             goto out;
2064         }
2065     } else {
2066         /* Default to cluster size, if available: */
2067         granularity = bdrv_get_default_bitmap_granularity(bs);
2068     }
2069 
2070     bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2071 
2072  out:
2073     aio_context_release(aio_context);
2074 }
2075 
2076 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2077                                    Error **errp)
2078 {
2079     AioContext *aio_context;
2080     BlockDriverState *bs;
2081     BdrvDirtyBitmap *bitmap;
2082 
2083     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2084     if (!bitmap || !bs) {
2085         return;
2086     }
2087 
2088     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2089         error_setg(errp,
2090                    "Bitmap '%s' is currently frozen and cannot be removed",
2091                    name);
2092         goto out;
2093     }
2094     bdrv_dirty_bitmap_make_anon(bitmap);
2095     bdrv_release_dirty_bitmap(bs, bitmap);
2096 
2097  out:
2098     aio_context_release(aio_context);
2099 }
2100 
2101 /**
2102  * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2103  * immediately after a full backup operation.
2104  */
2105 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2106                                   Error **errp)
2107 {
2108     AioContext *aio_context;
2109     BdrvDirtyBitmap *bitmap;
2110     BlockDriverState *bs;
2111 
2112     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2113     if (!bitmap || !bs) {
2114         return;
2115     }
2116 
2117     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2118         error_setg(errp,
2119                    "Bitmap '%s' is currently frozen and cannot be modified",
2120                    name);
2121         goto out;
2122     } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
2123         error_setg(errp,
2124                    "Bitmap '%s' is currently disabled and cannot be cleared",
2125                    name);
2126         goto out;
2127     }
2128 
2129     bdrv_clear_dirty_bitmap(bitmap);
2130 
2131  out:
2132     aio_context_release(aio_context);
2133 }
2134 
2135 void hmp_drive_del(Monitor *mon, const QDict *qdict)
2136 {
2137     const char *id = qdict_get_str(qdict, "id");
2138     BlockBackend *blk;
2139     BlockDriverState *bs;
2140     AioContext *aio_context;
2141     Error *local_err = NULL;
2142 
2143     blk = blk_by_name(id);
2144     if (!blk) {
2145         error_report("Device '%s' not found", id);
2146         return;
2147     }
2148     bs = blk_bs(blk);
2149 
2150     if (!blk_legacy_dinfo(blk)) {
2151         error_report("Deleting device added with blockdev-add"
2152                      " is not supported");
2153         return;
2154     }
2155 
2156     aio_context = bdrv_get_aio_context(bs);
2157     aio_context_acquire(aio_context);
2158 
2159     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2160         error_report_err(local_err);
2161         aio_context_release(aio_context);
2162         return;
2163     }
2164 
2165     bdrv_close(bs);
2166 
2167     /* if we have a device attached to this BlockDriverState
2168      * then we need to make the drive anonymous until the device
2169      * can be removed.  If this is a drive with no device backing
2170      * then we can just get rid of the block driver state right here.
2171      */
2172     if (blk_get_attached_dev(blk)) {
2173         blk_hide_on_behalf_of_hmp_drive_del(blk);
2174         /* Further I/O must not pause the guest */
2175         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
2176                           BLOCKDEV_ON_ERROR_REPORT);
2177     } else {
2178         blk_unref(blk);
2179     }
2180 
2181     aio_context_release(aio_context);
2182 }
2183 
2184 void qmp_block_resize(bool has_device, const char *device,
2185                       bool has_node_name, const char *node_name,
2186                       int64_t size, Error **errp)
2187 {
2188     Error *local_err = NULL;
2189     BlockDriverState *bs;
2190     AioContext *aio_context;
2191     int ret;
2192 
2193     bs = bdrv_lookup_bs(has_device ? device : NULL,
2194                         has_node_name ? node_name : NULL,
2195                         &local_err);
2196     if (local_err) {
2197         error_propagate(errp, local_err);
2198         return;
2199     }
2200 
2201     aio_context = bdrv_get_aio_context(bs);
2202     aio_context_acquire(aio_context);
2203 
2204     if (!bdrv_is_first_non_filter(bs)) {
2205         error_setg(errp, QERR_FEATURE_DISABLED, "resize");
2206         goto out;
2207     }
2208 
2209     if (size < 0) {
2210         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2211         goto out;
2212     }
2213 
2214     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2215         error_setg(errp, QERR_DEVICE_IN_USE, device);
2216         goto out;
2217     }
2218 
2219     /* complete all in-flight operations before resizing the device */
2220     bdrv_drain_all();
2221 
2222     ret = bdrv_truncate(bs, size);
2223     switch (ret) {
2224     case 0:
2225         break;
2226     case -ENOMEDIUM:
2227         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2228         break;
2229     case -ENOTSUP:
2230         error_setg(errp, QERR_UNSUPPORTED);
2231         break;
2232     case -EACCES:
2233         error_setg(errp, "Device '%s' is read only", device);
2234         break;
2235     case -EBUSY:
2236         error_setg(errp, QERR_DEVICE_IN_USE, device);
2237         break;
2238     default:
2239         error_setg_errno(errp, -ret, "Could not resize");
2240         break;
2241     }
2242 
2243 out:
2244     aio_context_release(aio_context);
2245 }
2246 
2247 static void block_job_cb(void *opaque, int ret)
2248 {
2249     /* Note that this function may be executed from another AioContext besides
2250      * the QEMU main loop.  If you need to access anything that assumes the
2251      * QEMU global mutex, use a BH or introduce a mutex.
2252      */
2253 
2254     BlockDriverState *bs = opaque;
2255     const char *msg = NULL;
2256 
2257     trace_block_job_cb(bs, bs->job, ret);
2258 
2259     assert(bs->job);
2260 
2261     if (ret < 0) {
2262         msg = strerror(-ret);
2263     }
2264 
2265     if (block_job_is_cancelled(bs->job)) {
2266         block_job_event_cancelled(bs->job);
2267     } else {
2268         block_job_event_completed(bs->job, msg);
2269     }
2270 
2271     bdrv_put_ref_bh_schedule(bs);
2272 }
2273 
2274 void qmp_block_stream(const char *device,
2275                       bool has_base, const char *base,
2276                       bool has_backing_file, const char *backing_file,
2277                       bool has_speed, int64_t speed,
2278                       bool has_on_error, BlockdevOnError on_error,
2279                       Error **errp)
2280 {
2281     BlockBackend *blk;
2282     BlockDriverState *bs;
2283     BlockDriverState *base_bs = NULL;
2284     AioContext *aio_context;
2285     Error *local_err = NULL;
2286     const char *base_name = NULL;
2287 
2288     if (!has_on_error) {
2289         on_error = BLOCKDEV_ON_ERROR_REPORT;
2290     }
2291 
2292     blk = blk_by_name(device);
2293     if (!blk) {
2294         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2295                   "Device '%s' not found", device);
2296         return;
2297     }
2298     bs = blk_bs(blk);
2299 
2300     aio_context = bdrv_get_aio_context(bs);
2301     aio_context_acquire(aio_context);
2302 
2303     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
2304         goto out;
2305     }
2306 
2307     if (has_base) {
2308         base_bs = bdrv_find_backing_image(bs, base);
2309         if (base_bs == NULL) {
2310             error_setg(errp, QERR_BASE_NOT_FOUND, base);
2311             goto out;
2312         }
2313         assert(bdrv_get_aio_context(base_bs) == aio_context);
2314         base_name = base;
2315     }
2316 
2317     /* if we are streaming the entire chain, the result will have no backing
2318      * file, and specifying one is therefore an error */
2319     if (base_bs == NULL && has_backing_file) {
2320         error_setg(errp, "backing file specified, but streaming the "
2321                          "entire chain");
2322         goto out;
2323     }
2324 
2325     /* backing_file string overrides base bs filename */
2326     base_name = has_backing_file ? backing_file : base_name;
2327 
2328     stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2329                  on_error, block_job_cb, bs, &local_err);
2330     if (local_err) {
2331         error_propagate(errp, local_err);
2332         goto out;
2333     }
2334 
2335     trace_qmp_block_stream(bs, bs->job);
2336 
2337 out:
2338     aio_context_release(aio_context);
2339 }
2340 
2341 void qmp_block_commit(const char *device,
2342                       bool has_base, const char *base,
2343                       bool has_top, const char *top,
2344                       bool has_backing_file, const char *backing_file,
2345                       bool has_speed, int64_t speed,
2346                       Error **errp)
2347 {
2348     BlockBackend *blk;
2349     BlockDriverState *bs;
2350     BlockDriverState *base_bs, *top_bs;
2351     AioContext *aio_context;
2352     Error *local_err = NULL;
2353     /* This will be part of the QMP command, if/when the
2354      * BlockdevOnError change for blkmirror makes it in
2355      */
2356     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2357 
2358     if (!has_speed) {
2359         speed = 0;
2360     }
2361 
2362     /* Important Note:
2363      *  libvirt relies on the DeviceNotFound error class in order to probe for
2364      *  live commit feature versions; for this to work, we must make sure to
2365      *  perform the device lookup before any generic errors that may occur in a
2366      *  scenario in which all optional arguments are omitted. */
2367     blk = blk_by_name(device);
2368     if (!blk) {
2369         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2370                   "Device '%s' not found", device);
2371         return;
2372     }
2373     bs = blk_bs(blk);
2374 
2375     aio_context = bdrv_get_aio_context(bs);
2376     aio_context_acquire(aio_context);
2377 
2378     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2379         goto out;
2380     }
2381 
2382     /* default top_bs is the active layer */
2383     top_bs = bs;
2384 
2385     if (has_top && top) {
2386         if (strcmp(bs->filename, top) != 0) {
2387             top_bs = bdrv_find_backing_image(bs, top);
2388         }
2389     }
2390 
2391     if (top_bs == NULL) {
2392         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2393         goto out;
2394     }
2395 
2396     assert(bdrv_get_aio_context(top_bs) == aio_context);
2397 
2398     if (has_base && base) {
2399         base_bs = bdrv_find_backing_image(top_bs, base);
2400     } else {
2401         base_bs = bdrv_find_base(top_bs);
2402     }
2403 
2404     if (base_bs == NULL) {
2405         error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2406         goto out;
2407     }
2408 
2409     assert(bdrv_get_aio_context(base_bs) == aio_context);
2410 
2411     if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2412         goto out;
2413     }
2414 
2415     /* Do not allow attempts to commit an image into itself */
2416     if (top_bs == base_bs) {
2417         error_setg(errp, "cannot commit an image into itself");
2418         goto out;
2419     }
2420 
2421     if (top_bs == bs) {
2422         if (has_backing_file) {
2423             error_setg(errp, "'backing-file' specified,"
2424                              " but 'top' is the active layer");
2425             goto out;
2426         }
2427         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2428                             bs, &local_err);
2429     } else {
2430         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2431                      has_backing_file ? backing_file : NULL, &local_err);
2432     }
2433     if (local_err != NULL) {
2434         error_propagate(errp, local_err);
2435         goto out;
2436     }
2437 
2438 out:
2439     aio_context_release(aio_context);
2440 }
2441 
2442 void qmp_drive_backup(const char *device, const char *target,
2443                       bool has_format, const char *format,
2444                       enum MirrorSyncMode sync,
2445                       bool has_mode, enum NewImageMode mode,
2446                       bool has_speed, int64_t speed,
2447                       bool has_bitmap, const char *bitmap,
2448                       bool has_on_source_error, BlockdevOnError on_source_error,
2449                       bool has_on_target_error, BlockdevOnError on_target_error,
2450                       Error **errp)
2451 {
2452     BlockBackend *blk;
2453     BlockDriverState *bs;
2454     BlockDriverState *target_bs;
2455     BlockDriverState *source = NULL;
2456     BdrvDirtyBitmap *bmap = NULL;
2457     AioContext *aio_context;
2458     QDict *options = NULL;
2459     Error *local_err = NULL;
2460     int flags;
2461     int64_t size;
2462     int ret;
2463 
2464     if (!has_speed) {
2465         speed = 0;
2466     }
2467     if (!has_on_source_error) {
2468         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2469     }
2470     if (!has_on_target_error) {
2471         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2472     }
2473     if (!has_mode) {
2474         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2475     }
2476 
2477     blk = blk_by_name(device);
2478     if (!blk) {
2479         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2480                   "Device '%s' not found", device);
2481         return;
2482     }
2483     bs = blk_bs(blk);
2484 
2485     aio_context = bdrv_get_aio_context(bs);
2486     aio_context_acquire(aio_context);
2487 
2488     /* Although backup_run has this check too, we need to use bs->drv below, so
2489      * do an early check redundantly. */
2490     if (!bdrv_is_inserted(bs)) {
2491         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2492         goto out;
2493     }
2494 
2495     if (!has_format) {
2496         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2497     }
2498 
2499     /* Early check to avoid creating target */
2500     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2501         goto out;
2502     }
2503 
2504     flags = bs->open_flags | BDRV_O_RDWR;
2505 
2506     /* See if we have a backing HD we can use to create our new image
2507      * on top of. */
2508     if (sync == MIRROR_SYNC_MODE_TOP) {
2509         source = backing_bs(bs);
2510         if (!source) {
2511             sync = MIRROR_SYNC_MODE_FULL;
2512         }
2513     }
2514     if (sync == MIRROR_SYNC_MODE_NONE) {
2515         source = bs;
2516     }
2517 
2518     size = bdrv_getlength(bs);
2519     if (size < 0) {
2520         error_setg_errno(errp, -size, "bdrv_getlength failed");
2521         goto out;
2522     }
2523 
2524     if (mode != NEW_IMAGE_MODE_EXISTING) {
2525         assert(format);
2526         if (source) {
2527             bdrv_img_create(target, format, source->filename,
2528                             source->drv->format_name, NULL,
2529                             size, flags, &local_err, false);
2530         } else {
2531             bdrv_img_create(target, format, NULL, NULL, NULL,
2532                             size, flags, &local_err, false);
2533         }
2534     }
2535 
2536     if (local_err) {
2537         error_propagate(errp, local_err);
2538         goto out;
2539     }
2540 
2541     if (format) {
2542         options = qdict_new();
2543         qdict_put(options, "driver", qstring_from_str(format));
2544     }
2545 
2546     target_bs = NULL;
2547     ret = bdrv_open(&target_bs, target, NULL, options, flags, &local_err);
2548     if (ret < 0) {
2549         error_propagate(errp, local_err);
2550         goto out;
2551     }
2552 
2553     bdrv_set_aio_context(target_bs, aio_context);
2554 
2555     if (has_bitmap) {
2556         bmap = bdrv_find_dirty_bitmap(bs, bitmap);
2557         if (!bmap) {
2558             error_setg(errp, "Bitmap '%s' could not be found", bitmap);
2559             goto out;
2560         }
2561     }
2562 
2563     backup_start(bs, target_bs, speed, sync, bmap,
2564                  on_source_error, on_target_error,
2565                  block_job_cb, bs, &local_err);
2566     if (local_err != NULL) {
2567         bdrv_unref(target_bs);
2568         error_propagate(errp, local_err);
2569         goto out;
2570     }
2571 
2572 out:
2573     aio_context_release(aio_context);
2574 }
2575 
2576 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2577 {
2578     return bdrv_named_nodes_list(errp);
2579 }
2580 
2581 void qmp_blockdev_backup(const char *device, const char *target,
2582                          enum MirrorSyncMode sync,
2583                          bool has_speed, int64_t speed,
2584                          bool has_on_source_error,
2585                          BlockdevOnError on_source_error,
2586                          bool has_on_target_error,
2587                          BlockdevOnError on_target_error,
2588                          Error **errp)
2589 {
2590     BlockBackend *blk;
2591     BlockDriverState *bs;
2592     BlockDriverState *target_bs;
2593     Error *local_err = NULL;
2594     AioContext *aio_context;
2595 
2596     if (!has_speed) {
2597         speed = 0;
2598     }
2599     if (!has_on_source_error) {
2600         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2601     }
2602     if (!has_on_target_error) {
2603         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2604     }
2605 
2606     blk = blk_by_name(device);
2607     if (!blk) {
2608         error_setg(errp, "Device '%s' not found", device);
2609         return;
2610     }
2611     bs = blk_bs(blk);
2612 
2613     aio_context = bdrv_get_aio_context(bs);
2614     aio_context_acquire(aio_context);
2615 
2616     blk = blk_by_name(target);
2617     if (!blk) {
2618         error_setg(errp, "Device '%s' not found", target);
2619         goto out;
2620     }
2621     target_bs = blk_bs(blk);
2622 
2623     bdrv_ref(target_bs);
2624     bdrv_set_aio_context(target_bs, aio_context);
2625     backup_start(bs, target_bs, speed, sync, NULL, on_source_error,
2626                  on_target_error, block_job_cb, bs, &local_err);
2627     if (local_err != NULL) {
2628         bdrv_unref(target_bs);
2629         error_propagate(errp, local_err);
2630     }
2631 out:
2632     aio_context_release(aio_context);
2633 }
2634 
2635 void qmp_drive_mirror(const char *device, const char *target,
2636                       bool has_format, const char *format,
2637                       bool has_node_name, const char *node_name,
2638                       bool has_replaces, const char *replaces,
2639                       enum MirrorSyncMode sync,
2640                       bool has_mode, enum NewImageMode mode,
2641                       bool has_speed, int64_t speed,
2642                       bool has_granularity, uint32_t granularity,
2643                       bool has_buf_size, int64_t buf_size,
2644                       bool has_on_source_error, BlockdevOnError on_source_error,
2645                       bool has_on_target_error, BlockdevOnError on_target_error,
2646                       bool has_unmap, bool unmap,
2647                       Error **errp)
2648 {
2649     BlockBackend *blk;
2650     BlockDriverState *bs;
2651     BlockDriverState *source, *target_bs;
2652     AioContext *aio_context;
2653     Error *local_err = NULL;
2654     QDict *options;
2655     int flags;
2656     int64_t size;
2657     int ret;
2658 
2659     if (!has_speed) {
2660         speed = 0;
2661     }
2662     if (!has_on_source_error) {
2663         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2664     }
2665     if (!has_on_target_error) {
2666         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2667     }
2668     if (!has_mode) {
2669         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2670     }
2671     if (!has_granularity) {
2672         granularity = 0;
2673     }
2674     if (!has_buf_size) {
2675         buf_size = 0;
2676     }
2677     if (!has_unmap) {
2678         unmap = true;
2679     }
2680 
2681     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2682         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2683                    "a value in range [512B, 64MB]");
2684         return;
2685     }
2686     if (granularity & (granularity - 1)) {
2687         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2688                    "power of 2");
2689         return;
2690     }
2691 
2692     blk = blk_by_name(device);
2693     if (!blk) {
2694         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2695                   "Device '%s' not found", device);
2696         return;
2697     }
2698     bs = blk_bs(blk);
2699 
2700     aio_context = bdrv_get_aio_context(bs);
2701     aio_context_acquire(aio_context);
2702 
2703     if (!bdrv_is_inserted(bs)) {
2704         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2705         goto out;
2706     }
2707 
2708     if (!has_format) {
2709         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2710     }
2711 
2712     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2713         goto out;
2714     }
2715 
2716     flags = bs->open_flags | BDRV_O_RDWR;
2717     source = backing_bs(bs);
2718     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2719         sync = MIRROR_SYNC_MODE_FULL;
2720     }
2721     if (sync == MIRROR_SYNC_MODE_NONE) {
2722         source = bs;
2723     }
2724 
2725     size = bdrv_getlength(bs);
2726     if (size < 0) {
2727         error_setg_errno(errp, -size, "bdrv_getlength failed");
2728         goto out;
2729     }
2730 
2731     if (has_replaces) {
2732         BlockDriverState *to_replace_bs;
2733         AioContext *replace_aio_context;
2734         int64_t replace_size;
2735 
2736         if (!has_node_name) {
2737             error_setg(errp, "a node-name must be provided when replacing a"
2738                              " named node of the graph");
2739             goto out;
2740         }
2741 
2742         to_replace_bs = check_to_replace_node(bs, replaces, &local_err);
2743 
2744         if (!to_replace_bs) {
2745             error_propagate(errp, local_err);
2746             goto out;
2747         }
2748 
2749         replace_aio_context = bdrv_get_aio_context(to_replace_bs);
2750         aio_context_acquire(replace_aio_context);
2751         replace_size = bdrv_getlength(to_replace_bs);
2752         aio_context_release(replace_aio_context);
2753 
2754         if (size != replace_size) {
2755             error_setg(errp, "cannot replace image with a mirror image of "
2756                              "different size");
2757             goto out;
2758         }
2759     }
2760 
2761     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2762         && mode != NEW_IMAGE_MODE_EXISTING)
2763     {
2764         /* create new image w/o backing file */
2765         assert(format);
2766         bdrv_img_create(target, format,
2767                         NULL, NULL, NULL, size, flags, &local_err, false);
2768     } else {
2769         switch (mode) {
2770         case NEW_IMAGE_MODE_EXISTING:
2771             break;
2772         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2773             /* create new image with backing file */
2774             bdrv_img_create(target, format,
2775                             source->filename,
2776                             source->drv->format_name,
2777                             NULL, size, flags, &local_err, false);
2778             break;
2779         default:
2780             abort();
2781         }
2782     }
2783 
2784     if (local_err) {
2785         error_propagate(errp, local_err);
2786         goto out;
2787     }
2788 
2789     options = qdict_new();
2790     if (has_node_name) {
2791         qdict_put(options, "node-name", qstring_from_str(node_name));
2792     }
2793     if (format) {
2794         qdict_put(options, "driver", qstring_from_str(format));
2795     }
2796 
2797     /* Mirroring takes care of copy-on-write using the source's backing
2798      * file.
2799      */
2800     target_bs = NULL;
2801     ret = bdrv_open(&target_bs, target, NULL, options,
2802                     flags | BDRV_O_NO_BACKING, &local_err);
2803     if (ret < 0) {
2804         error_propagate(errp, local_err);
2805         goto out;
2806     }
2807 
2808     bdrv_set_aio_context(target_bs, aio_context);
2809 
2810     /* pass the node name to replace to mirror start since it's loose coupling
2811      * and will allow to check whether the node still exist at mirror completion
2812      */
2813     mirror_start(bs, target_bs,
2814                  has_replaces ? replaces : NULL,
2815                  speed, granularity, buf_size, sync,
2816                  on_source_error, on_target_error,
2817                  unmap,
2818                  block_job_cb, bs, &local_err);
2819     if (local_err != NULL) {
2820         bdrv_unref(target_bs);
2821         error_propagate(errp, local_err);
2822         goto out;
2823     }
2824 
2825 out:
2826     aio_context_release(aio_context);
2827 }
2828 
2829 /* Get the block job for a given device name and acquire its AioContext */
2830 static BlockJob *find_block_job(const char *device, AioContext **aio_context,
2831                                 Error **errp)
2832 {
2833     BlockBackend *blk;
2834     BlockDriverState *bs;
2835 
2836     blk = blk_by_name(device);
2837     if (!blk) {
2838         goto notfound;
2839     }
2840     bs = blk_bs(blk);
2841 
2842     *aio_context = bdrv_get_aio_context(bs);
2843     aio_context_acquire(*aio_context);
2844 
2845     if (!bs->job) {
2846         aio_context_release(*aio_context);
2847         goto notfound;
2848     }
2849 
2850     return bs->job;
2851 
2852 notfound:
2853     error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
2854               "No active block job on device '%s'", device);
2855     *aio_context = NULL;
2856     return NULL;
2857 }
2858 
2859 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2860 {
2861     AioContext *aio_context;
2862     BlockJob *job = find_block_job(device, &aio_context, errp);
2863 
2864     if (!job) {
2865         return;
2866     }
2867 
2868     block_job_set_speed(job, speed, errp);
2869     aio_context_release(aio_context);
2870 }
2871 
2872 void qmp_block_job_cancel(const char *device,
2873                           bool has_force, bool force, Error **errp)
2874 {
2875     AioContext *aio_context;
2876     BlockJob *job = find_block_job(device, &aio_context, errp);
2877 
2878     if (!job) {
2879         return;
2880     }
2881 
2882     if (!has_force) {
2883         force = false;
2884     }
2885 
2886     if (job->user_paused && !force) {
2887         error_setg(errp, "The block job for device '%s' is currently paused",
2888                    device);
2889         goto out;
2890     }
2891 
2892     trace_qmp_block_job_cancel(job);
2893     block_job_cancel(job);
2894 out:
2895     aio_context_release(aio_context);
2896 }
2897 
2898 void qmp_block_job_pause(const char *device, Error **errp)
2899 {
2900     AioContext *aio_context;
2901     BlockJob *job = find_block_job(device, &aio_context, errp);
2902 
2903     if (!job || job->user_paused) {
2904         return;
2905     }
2906 
2907     job->user_paused = true;
2908     trace_qmp_block_job_pause(job);
2909     block_job_pause(job);
2910     aio_context_release(aio_context);
2911 }
2912 
2913 void qmp_block_job_resume(const char *device, Error **errp)
2914 {
2915     AioContext *aio_context;
2916     BlockJob *job = find_block_job(device, &aio_context, errp);
2917 
2918     if (!job || !job->user_paused) {
2919         return;
2920     }
2921 
2922     job->user_paused = false;
2923     trace_qmp_block_job_resume(job);
2924     block_job_resume(job);
2925     aio_context_release(aio_context);
2926 }
2927 
2928 void qmp_block_job_complete(const char *device, Error **errp)
2929 {
2930     AioContext *aio_context;
2931     BlockJob *job = find_block_job(device, &aio_context, errp);
2932 
2933     if (!job) {
2934         return;
2935     }
2936 
2937     trace_qmp_block_job_complete(job);
2938     block_job_complete(job, errp);
2939     aio_context_release(aio_context);
2940 }
2941 
2942 void qmp_change_backing_file(const char *device,
2943                              const char *image_node_name,
2944                              const char *backing_file,
2945                              Error **errp)
2946 {
2947     BlockBackend *blk;
2948     BlockDriverState *bs = NULL;
2949     AioContext *aio_context;
2950     BlockDriverState *image_bs = NULL;
2951     Error *local_err = NULL;
2952     bool ro;
2953     int open_flags;
2954     int ret;
2955 
2956     blk = blk_by_name(device);
2957     if (!blk) {
2958         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2959                   "Device '%s' not found", device);
2960         return;
2961     }
2962     bs = blk_bs(blk);
2963 
2964     aio_context = bdrv_get_aio_context(bs);
2965     aio_context_acquire(aio_context);
2966 
2967     image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
2968     if (local_err) {
2969         error_propagate(errp, local_err);
2970         goto out;
2971     }
2972 
2973     if (!image_bs) {
2974         error_setg(errp, "image file not found");
2975         goto out;
2976     }
2977 
2978     if (bdrv_find_base(image_bs) == image_bs) {
2979         error_setg(errp, "not allowing backing file change on an image "
2980                          "without a backing file");
2981         goto out;
2982     }
2983 
2984     /* even though we are not necessarily operating on bs, we need it to
2985      * determine if block ops are currently prohibited on the chain */
2986     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
2987         goto out;
2988     }
2989 
2990     /* final sanity check */
2991     if (!bdrv_chain_contains(bs, image_bs)) {
2992         error_setg(errp, "'%s' and image file are not in the same chain",
2993                    device);
2994         goto out;
2995     }
2996 
2997     /* if not r/w, reopen to make r/w */
2998     open_flags = image_bs->open_flags;
2999     ro = bdrv_is_read_only(image_bs);
3000 
3001     if (ro) {
3002         bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
3003         if (local_err) {
3004             error_propagate(errp, local_err);
3005             goto out;
3006         }
3007     }
3008 
3009     ret = bdrv_change_backing_file(image_bs, backing_file,
3010                                image_bs->drv ? image_bs->drv->format_name : "");
3011 
3012     if (ret < 0) {
3013         error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3014                          backing_file);
3015         /* don't exit here, so we can try to restore open flags if
3016          * appropriate */
3017     }
3018 
3019     if (ro) {
3020         bdrv_reopen(image_bs, open_flags, &local_err);
3021         if (local_err) {
3022             error_propagate(errp, local_err); /* will preserve prior errp */
3023         }
3024     }
3025 
3026 out:
3027     aio_context_release(aio_context);
3028 }
3029 
3030 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3031 {
3032     QmpOutputVisitor *ov = qmp_output_visitor_new();
3033     BlockBackend *blk;
3034     QObject *obj;
3035     QDict *qdict;
3036     Error *local_err = NULL;
3037 
3038     /* Require an ID in the top level */
3039     if (!options->has_id) {
3040         error_setg(errp, "Block device needs an ID");
3041         goto fail;
3042     }
3043 
3044     /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
3045      * cache.direct=false instead of silently switching to aio=threads, except
3046      * when called from drive_new().
3047      *
3048      * For now, simply forbidding the combination for all drivers will do. */
3049     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
3050         bool direct = options->has_cache &&
3051                       options->cache->has_direct &&
3052                       options->cache->direct;
3053         if (!direct) {
3054             error_setg(errp, "aio=native requires cache.direct=true");
3055             goto fail;
3056         }
3057     }
3058 
3059     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
3060                                &options, NULL, &local_err);
3061     if (local_err) {
3062         error_propagate(errp, local_err);
3063         goto fail;
3064     }
3065 
3066     obj = qmp_output_get_qobject(ov);
3067     qdict = qobject_to_qdict(obj);
3068 
3069     qdict_flatten(qdict);
3070 
3071     blk = blockdev_init(NULL, qdict, &local_err);
3072     if (local_err) {
3073         error_propagate(errp, local_err);
3074         goto fail;
3075     }
3076 
3077     if (bdrv_key_required(blk_bs(blk))) {
3078         blk_unref(blk);
3079         error_setg(errp, "blockdev-add doesn't support encrypted devices");
3080         goto fail;
3081     }
3082 
3083 fail:
3084     qmp_output_visitor_cleanup(ov);
3085 }
3086 
3087 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3088 {
3089     BlockJobInfoList *head = NULL, **p_next = &head;
3090     BlockDriverState *bs;
3091 
3092     for (bs = bdrv_next(NULL); bs; bs = bdrv_next(bs)) {
3093         AioContext *aio_context = bdrv_get_aio_context(bs);
3094 
3095         aio_context_acquire(aio_context);
3096 
3097         if (bs->job) {
3098             BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
3099             elem->value = block_job_query(bs->job);
3100             *p_next = elem;
3101             p_next = &elem->next;
3102         }
3103 
3104         aio_context_release(aio_context);
3105     }
3106 
3107     return head;
3108 }
3109 
3110 QemuOptsList qemu_common_drive_opts = {
3111     .name = "drive",
3112     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3113     .desc = {
3114         {
3115             .name = "snapshot",
3116             .type = QEMU_OPT_BOOL,
3117             .help = "enable/disable snapshot mode",
3118         },{
3119             .name = "discard",
3120             .type = QEMU_OPT_STRING,
3121             .help = "discard operation (ignore/off, unmap/on)",
3122         },{
3123             .name = BDRV_OPT_CACHE_WB,
3124             .type = QEMU_OPT_BOOL,
3125             .help = "enables writeback mode for any caches",
3126         },{
3127             .name = BDRV_OPT_CACHE_DIRECT,
3128             .type = QEMU_OPT_BOOL,
3129             .help = "enables use of O_DIRECT (bypass the host page cache)",
3130         },{
3131             .name = BDRV_OPT_CACHE_NO_FLUSH,
3132             .type = QEMU_OPT_BOOL,
3133             .help = "ignore any flush requests for the device",
3134         },{
3135             .name = "aio",
3136             .type = QEMU_OPT_STRING,
3137             .help = "host AIO implementation (threads, native)",
3138         },{
3139             .name = "format",
3140             .type = QEMU_OPT_STRING,
3141             .help = "disk format (raw, qcow2, ...)",
3142         },{
3143             .name = "rerror",
3144             .type = QEMU_OPT_STRING,
3145             .help = "read error action",
3146         },{
3147             .name = "werror",
3148             .type = QEMU_OPT_STRING,
3149             .help = "write error action",
3150         },{
3151             .name = "read-only",
3152             .type = QEMU_OPT_BOOL,
3153             .help = "open drive file as read-only",
3154         },{
3155             .name = "throttling.iops-total",
3156             .type = QEMU_OPT_NUMBER,
3157             .help = "limit total I/O operations per second",
3158         },{
3159             .name = "throttling.iops-read",
3160             .type = QEMU_OPT_NUMBER,
3161             .help = "limit read operations per second",
3162         },{
3163             .name = "throttling.iops-write",
3164             .type = QEMU_OPT_NUMBER,
3165             .help = "limit write operations per second",
3166         },{
3167             .name = "throttling.bps-total",
3168             .type = QEMU_OPT_NUMBER,
3169             .help = "limit total bytes per second",
3170         },{
3171             .name = "throttling.bps-read",
3172             .type = QEMU_OPT_NUMBER,
3173             .help = "limit read bytes per second",
3174         },{
3175             .name = "throttling.bps-write",
3176             .type = QEMU_OPT_NUMBER,
3177             .help = "limit write bytes per second",
3178         },{
3179             .name = "throttling.iops-total-max",
3180             .type = QEMU_OPT_NUMBER,
3181             .help = "I/O operations burst",
3182         },{
3183             .name = "throttling.iops-read-max",
3184             .type = QEMU_OPT_NUMBER,
3185             .help = "I/O operations read burst",
3186         },{
3187             .name = "throttling.iops-write-max",
3188             .type = QEMU_OPT_NUMBER,
3189             .help = "I/O operations write burst",
3190         },{
3191             .name = "throttling.bps-total-max",
3192             .type = QEMU_OPT_NUMBER,
3193             .help = "total bytes burst",
3194         },{
3195             .name = "throttling.bps-read-max",
3196             .type = QEMU_OPT_NUMBER,
3197             .help = "total bytes read burst",
3198         },{
3199             .name = "throttling.bps-write-max",
3200             .type = QEMU_OPT_NUMBER,
3201             .help = "total bytes write burst",
3202         },{
3203             .name = "throttling.iops-size",
3204             .type = QEMU_OPT_NUMBER,
3205             .help = "when limiting by iops max size of an I/O in bytes",
3206         },{
3207             .name = "throttling.group",
3208             .type = QEMU_OPT_STRING,
3209             .help = "name of the block throttling group",
3210         },{
3211             .name = "copy-on-read",
3212             .type = QEMU_OPT_BOOL,
3213             .help = "copy read data from backing file into image file",
3214         },{
3215             .name = "detect-zeroes",
3216             .type = QEMU_OPT_STRING,
3217             .help = "try to optimize zero writes (off, on, unmap)",
3218         },
3219         { /* end of list */ }
3220     },
3221 };
3222 
3223 QemuOptsList qemu_drive_opts = {
3224     .name = "drive",
3225     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3226     .desc = {
3227         /*
3228          * no elements => accept any params
3229          * validation will happen later
3230          */
3231         { /* end of list */ }
3232     },
3233 };
3234