xref: /openbmc/qemu/blockdev.c (revision c13163fba151f0be5176eaf55907bc1dbff3a1d4)
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/blockdev.h"
34 #include "hw/block/block.h"
35 #include "block/blockjob.h"
36 #include "monitor/monitor.h"
37 #include "qapi/qmp/qerror.h"
38 #include "qemu/option.h"
39 #include "qemu/config-file.h"
40 #include "qapi/qmp/types.h"
41 #include "qapi-visit.h"
42 #include "qapi/qmp-output-visitor.h"
43 #include "sysemu/sysemu.h"
44 #include "block/block_int.h"
45 #include "qmp-commands.h"
46 #include "trace.h"
47 #include "sysemu/arch_init.h"
48 
49 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
50 
51 static const char *const if_name[IF_COUNT] = {
52     [IF_NONE] = "none",
53     [IF_IDE] = "ide",
54     [IF_SCSI] = "scsi",
55     [IF_FLOPPY] = "floppy",
56     [IF_PFLASH] = "pflash",
57     [IF_MTD] = "mtd",
58     [IF_SD] = "sd",
59     [IF_VIRTIO] = "virtio",
60     [IF_XEN] = "xen",
61 };
62 
63 static const int if_max_devs[IF_COUNT] = {
64     /*
65      * Do not change these numbers!  They govern how drive option
66      * index maps to unit and bus.  That mapping is ABI.
67      *
68      * All controllers used to imlement if=T drives need to support
69      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
70      * Otherwise, some index values map to "impossible" bus, unit
71      * values.
72      *
73      * For instance, if you change [IF_SCSI] to 255, -drive
74      * if=scsi,index=12 no longer means bus=1,unit=5, but
75      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
76      * the drive can't be set up.  Regression.
77      */
78     [IF_IDE] = 2,
79     [IF_SCSI] = 7,
80 };
81 
82 /*
83  * We automatically delete the drive when a device using it gets
84  * unplugged.  Questionable feature, but we can't just drop it.
85  * Device models call blockdev_mark_auto_del() to schedule the
86  * automatic deletion, and generic qdev code calls blockdev_auto_del()
87  * when deletion is actually safe.
88  */
89 void blockdev_mark_auto_del(BlockDriverState *bs)
90 {
91     DriveInfo *dinfo = drive_get_by_blockdev(bs);
92 
93     if (dinfo && !dinfo->enable_auto_del) {
94         return;
95     }
96 
97     if (bs->job) {
98         block_job_cancel(bs->job);
99     }
100     if (dinfo) {
101         dinfo->auto_del = 1;
102     }
103 }
104 
105 void blockdev_auto_del(BlockDriverState *bs)
106 {
107     DriveInfo *dinfo = drive_get_by_blockdev(bs);
108 
109     if (dinfo && dinfo->auto_del) {
110         drive_put_ref(dinfo);
111     }
112 }
113 
114 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
115 {
116     int max_devs = if_max_devs[type];
117     return max_devs ? index / max_devs : 0;
118 }
119 
120 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
121 {
122     int max_devs = if_max_devs[type];
123     return max_devs ? index % max_devs : index;
124 }
125 
126 QemuOpts *drive_def(const char *optstr)
127 {
128     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
129 }
130 
131 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
132                     const char *optstr)
133 {
134     QemuOpts *opts;
135     char buf[32];
136 
137     opts = drive_def(optstr);
138     if (!opts) {
139         return NULL;
140     }
141     if (type != IF_DEFAULT) {
142         qemu_opt_set(opts, "if", if_name[type]);
143     }
144     if (index >= 0) {
145         snprintf(buf, sizeof(buf), "%d", index);
146         qemu_opt_set(opts, "index", buf);
147     }
148     if (file)
149         qemu_opt_set(opts, "file", file);
150     return opts;
151 }
152 
153 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
154 {
155     DriveInfo *dinfo;
156 
157     /* seek interface, bus and unit */
158 
159     QTAILQ_FOREACH(dinfo, &drives, next) {
160         if (dinfo->type == type &&
161 	    dinfo->bus == bus &&
162 	    dinfo->unit == unit)
163             return dinfo;
164     }
165 
166     return NULL;
167 }
168 
169 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
170 {
171     return drive_get(type,
172                      drive_index_to_bus_id(type, index),
173                      drive_index_to_unit_id(type, index));
174 }
175 
176 int drive_get_max_bus(BlockInterfaceType type)
177 {
178     int max_bus;
179     DriveInfo *dinfo;
180 
181     max_bus = -1;
182     QTAILQ_FOREACH(dinfo, &drives, next) {
183         if(dinfo->type == type &&
184            dinfo->bus > max_bus)
185             max_bus = dinfo->bus;
186     }
187     return max_bus;
188 }
189 
190 /* Get a block device.  This should only be used for single-drive devices
191    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
192    appropriate bus.  */
193 DriveInfo *drive_get_next(BlockInterfaceType type)
194 {
195     static int next_block_unit[IF_COUNT];
196 
197     return drive_get(type, 0, next_block_unit[type]++);
198 }
199 
200 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
201 {
202     DriveInfo *dinfo;
203 
204     QTAILQ_FOREACH(dinfo, &drives, next) {
205         if (dinfo->bdrv == bs) {
206             return dinfo;
207         }
208     }
209     return NULL;
210 }
211 
212 static void bdrv_format_print(void *opaque, const char *name)
213 {
214     error_printf(" %s", name);
215 }
216 
217 static void drive_uninit(DriveInfo *dinfo)
218 {
219     if (dinfo->opts) {
220         qemu_opts_del(dinfo->opts);
221     }
222 
223     bdrv_unref(dinfo->bdrv);
224     g_free(dinfo->id);
225     QTAILQ_REMOVE(&drives, dinfo, next);
226     g_free(dinfo->serial);
227     g_free(dinfo);
228 }
229 
230 void drive_put_ref(DriveInfo *dinfo)
231 {
232     assert(dinfo->refcount);
233     if (--dinfo->refcount == 0) {
234         drive_uninit(dinfo);
235     }
236 }
237 
238 void drive_get_ref(DriveInfo *dinfo)
239 {
240     dinfo->refcount++;
241 }
242 
243 typedef struct {
244     QEMUBH *bh;
245     BlockDriverState *bs;
246 } BDRVPutRefBH;
247 
248 static void bdrv_put_ref_bh(void *opaque)
249 {
250     BDRVPutRefBH *s = opaque;
251 
252     bdrv_unref(s->bs);
253     qemu_bh_delete(s->bh);
254     g_free(s);
255 }
256 
257 /*
258  * Release a BDS reference in a BH
259  *
260  * It is not safe to use bdrv_unref() from a callback function when the callers
261  * still need the BlockDriverState.  In such cases we schedule a BH to release
262  * the reference.
263  */
264 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
265 {
266     BDRVPutRefBH *s;
267 
268     s = g_new(BDRVPutRefBH, 1);
269     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
270     s->bs = bs;
271     qemu_bh_schedule(s->bh);
272 }
273 
274 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
275 {
276     if (!strcmp(buf, "ignore")) {
277         return BLOCKDEV_ON_ERROR_IGNORE;
278     } else if (!is_read && !strcmp(buf, "enospc")) {
279         return BLOCKDEV_ON_ERROR_ENOSPC;
280     } else if (!strcmp(buf, "stop")) {
281         return BLOCKDEV_ON_ERROR_STOP;
282     } else if (!strcmp(buf, "report")) {
283         return BLOCKDEV_ON_ERROR_REPORT;
284     } else {
285         error_setg(errp, "'%s' invalid %s error action",
286                    buf, is_read ? "read" : "write");
287         return -1;
288     }
289 }
290 
291 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
292 {
293     if (throttle_conflicting(cfg)) {
294         error_setg(errp, "bps/iops/max total values and read/write values"
295                          " cannot be used at the same time");
296         return false;
297     }
298 
299     if (!throttle_is_valid(cfg)) {
300         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
301         return false;
302     }
303 
304     return true;
305 }
306 
307 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
308 
309 /* Takes the ownership of bs_opts */
310 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
311                                 BlockInterfaceType type,
312                                 Error **errp)
313 {
314     const char *buf;
315     const char *serial;
316     int ro = 0;
317     int bdrv_flags = 0;
318     int on_read_error, on_write_error;
319     DriveInfo *dinfo;
320     ThrottleConfig cfg;
321     int snapshot = 0;
322     bool copy_on_read;
323     int ret;
324     Error *error = NULL;
325     QemuOpts *opts;
326     const char *id;
327     bool has_driver_specific_opts;
328     BlockDriver *drv = NULL;
329 
330     /* Check common options by copying from bs_opts to opts, all other options
331      * stay in bs_opts for processing by bdrv_open(). */
332     id = qdict_get_try_str(bs_opts, "id");
333     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
334     if (error_is_set(&error)) {
335         error_propagate(errp, error);
336         return NULL;
337     }
338 
339     qemu_opts_absorb_qdict(opts, bs_opts, &error);
340     if (error_is_set(&error)) {
341         error_propagate(errp, error);
342         goto early_err;
343     }
344 
345     if (id) {
346         qdict_del(bs_opts, "id");
347     }
348 
349     has_driver_specific_opts = !!qdict_size(bs_opts);
350 
351     /* extract parameters */
352     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
353     ro = qemu_opt_get_bool(opts, "read-only", 0);
354     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
355 
356     serial = qemu_opt_get(opts, "serial");
357 
358     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
359         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
360             error_setg(errp, "invalid discard option");
361             goto early_err;
362         }
363     }
364 
365     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
366         bdrv_flags |= BDRV_O_CACHE_WB;
367     }
368     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
369         bdrv_flags |= BDRV_O_NOCACHE;
370     }
371     if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
372         bdrv_flags |= BDRV_O_NO_FLUSH;
373     }
374 
375 #ifdef CONFIG_LINUX_AIO
376     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
377         if (!strcmp(buf, "native")) {
378             bdrv_flags |= BDRV_O_NATIVE_AIO;
379         } else if (!strcmp(buf, "threads")) {
380             /* this is the default */
381         } else {
382            error_setg(errp, "invalid aio option");
383            goto early_err;
384         }
385     }
386 #endif
387 
388     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
389         if (is_help_option(buf)) {
390             error_printf("Supported formats:");
391             bdrv_iterate_format(bdrv_format_print, NULL);
392             error_printf("\n");
393             goto early_err;
394         }
395 
396         drv = bdrv_find_format(buf);
397         if (!drv) {
398             error_setg(errp, "'%s' invalid format", buf);
399             goto early_err;
400         }
401     }
402 
403     /* disk I/O throttling */
404     memset(&cfg, 0, sizeof(cfg));
405     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
406         qemu_opt_get_number(opts, "throttling.bps-total", 0);
407     cfg.buckets[THROTTLE_BPS_READ].avg  =
408         qemu_opt_get_number(opts, "throttling.bps-read", 0);
409     cfg.buckets[THROTTLE_BPS_WRITE].avg =
410         qemu_opt_get_number(opts, "throttling.bps-write", 0);
411     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
412         qemu_opt_get_number(opts, "throttling.iops-total", 0);
413     cfg.buckets[THROTTLE_OPS_READ].avg =
414         qemu_opt_get_number(opts, "throttling.iops-read", 0);
415     cfg.buckets[THROTTLE_OPS_WRITE].avg =
416         qemu_opt_get_number(opts, "throttling.iops-write", 0);
417 
418     cfg.buckets[THROTTLE_BPS_TOTAL].max =
419         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
420     cfg.buckets[THROTTLE_BPS_READ].max  =
421         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
422     cfg.buckets[THROTTLE_BPS_WRITE].max =
423         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
424     cfg.buckets[THROTTLE_OPS_TOTAL].max =
425         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
426     cfg.buckets[THROTTLE_OPS_READ].max =
427         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
428     cfg.buckets[THROTTLE_OPS_WRITE].max =
429         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
430 
431     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
432 
433     if (!check_throttle_config(&cfg, &error)) {
434         error_propagate(errp, error);
435         goto early_err;
436     }
437 
438     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
439     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
440         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
441             error_setg(errp, "werror is not supported by this bus type");
442             goto early_err;
443         }
444 
445         on_write_error = parse_block_error_action(buf, 0, &error);
446         if (error_is_set(&error)) {
447             error_propagate(errp, error);
448             goto early_err;
449         }
450     }
451 
452     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
453     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
454         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
455             error_report("rerror is not supported by this bus type");
456             goto early_err;
457         }
458 
459         on_read_error = parse_block_error_action(buf, 1, &error);
460         if (error_is_set(&error)) {
461             error_propagate(errp, error);
462             goto early_err;
463         }
464     }
465 
466     /* init */
467     dinfo = g_malloc0(sizeof(*dinfo));
468     dinfo->id = g_strdup(qemu_opts_id(opts));
469     dinfo->bdrv = bdrv_new(dinfo->id);
470     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
471     dinfo->bdrv->read_only = ro;
472     dinfo->type = type;
473     dinfo->refcount = 1;
474     if (serial != NULL) {
475         dinfo->serial = g_strdup(serial);
476     }
477     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
478 
479     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
480 
481     /* disk I/O throttling */
482     if (throttle_enabled(&cfg)) {
483         bdrv_io_limits_enable(dinfo->bdrv);
484         bdrv_set_io_limits(dinfo->bdrv, &cfg);
485     }
486 
487     if (!file || !*file) {
488         if (has_driver_specific_opts) {
489             file = NULL;
490         } else {
491             QDECREF(bs_opts);
492             qemu_opts_del(opts);
493             return dinfo;
494         }
495     }
496     if (snapshot) {
497         /* always use cache=unsafe with snapshot */
498         bdrv_flags &= ~BDRV_O_CACHE_MASK;
499         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
500     }
501 
502     if (copy_on_read) {
503         bdrv_flags |= BDRV_O_COPY_ON_READ;
504     }
505 
506     if (runstate_check(RUN_STATE_INMIGRATE)) {
507         bdrv_flags |= BDRV_O_INCOMING;
508     }
509 
510     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
511 
512     QINCREF(bs_opts);
513     ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv, &error);
514 
515     if (ret < 0) {
516         error_setg(errp, "could not open disk image %s: %s",
517                    file ?: dinfo->id, error_get_pretty(error));
518         error_free(error);
519         goto err;
520     }
521 
522     if (bdrv_key_required(dinfo->bdrv))
523         autostart = 0;
524 
525     QDECREF(bs_opts);
526     qemu_opts_del(opts);
527 
528     return dinfo;
529 
530 err:
531     bdrv_unref(dinfo->bdrv);
532     g_free(dinfo->id);
533     QTAILQ_REMOVE(&drives, dinfo, next);
534     g_free(dinfo);
535 early_err:
536     QDECREF(bs_opts);
537     qemu_opts_del(opts);
538     return NULL;
539 }
540 
541 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
542 {
543     const char *value;
544 
545     value = qemu_opt_get(opts, from);
546     if (value) {
547         qemu_opt_set(opts, to, value);
548         qemu_opt_unset(opts, from);
549     }
550 }
551 
552 QemuOptsList qemu_legacy_drive_opts = {
553     .name = "drive",
554     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
555     .desc = {
556         {
557             .name = "bus",
558             .type = QEMU_OPT_NUMBER,
559             .help = "bus number",
560         },{
561             .name = "unit",
562             .type = QEMU_OPT_NUMBER,
563             .help = "unit number (i.e. lun for scsi)",
564         },{
565             .name = "index",
566             .type = QEMU_OPT_NUMBER,
567             .help = "index number",
568         },{
569             .name = "media",
570             .type = QEMU_OPT_STRING,
571             .help = "media type (disk, cdrom)",
572         },{
573             .name = "if",
574             .type = QEMU_OPT_STRING,
575             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
576         },{
577             .name = "cyls",
578             .type = QEMU_OPT_NUMBER,
579             .help = "number of cylinders (ide disk geometry)",
580         },{
581             .name = "heads",
582             .type = QEMU_OPT_NUMBER,
583             .help = "number of heads (ide disk geometry)",
584         },{
585             .name = "secs",
586             .type = QEMU_OPT_NUMBER,
587             .help = "number of sectors (ide disk geometry)",
588         },{
589             .name = "trans",
590             .type = QEMU_OPT_STRING,
591             .help = "chs translation (auto, lba, none)",
592         },{
593             .name = "boot",
594             .type = QEMU_OPT_BOOL,
595             .help = "(deprecated, ignored)",
596         },{
597             .name = "addr",
598             .type = QEMU_OPT_STRING,
599             .help = "pci address (virtio only)",
600         },{
601             .name = "file",
602             .type = QEMU_OPT_STRING,
603             .help = "file name",
604         },
605 
606         /* Options that are passed on, but have special semantics with -drive */
607         {
608             .name = "read-only",
609             .type = QEMU_OPT_BOOL,
610             .help = "open drive file as read-only",
611         },{
612             .name = "copy-on-read",
613             .type = QEMU_OPT_BOOL,
614             .help = "copy read data from backing file into image file",
615         },
616 
617         { /* end of list */ }
618     },
619 };
620 
621 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
622 {
623     const char *value;
624     DriveInfo *dinfo = NULL;
625     QDict *bs_opts;
626     QemuOpts *legacy_opts;
627     DriveMediaType media = MEDIA_DISK;
628     BlockInterfaceType type;
629     int cyls, heads, secs, translation;
630     int max_devs, bus_id, unit_id, index;
631     const char *devaddr;
632     bool read_only = false;
633     bool copy_on_read;
634     const char *filename;
635     Error *local_err = NULL;
636 
637     /* Change legacy command line options into QMP ones */
638     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
639     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
640     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
641 
642     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
643     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
644     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
645 
646     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
647     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
648     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
649 
650     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
651     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
652     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
653 
654     qemu_opt_rename(all_opts,
655                     "iops_size", "throttling.iops-size");
656 
657     qemu_opt_rename(all_opts, "readonly", "read-only");
658 
659     value = qemu_opt_get(all_opts, "cache");
660     if (value) {
661         int flags = 0;
662 
663         if (bdrv_parse_cache_flags(value, &flags) != 0) {
664             error_report("invalid cache option");
665             return NULL;
666         }
667 
668         /* Specific options take precedence */
669         if (!qemu_opt_get(all_opts, "cache.writeback")) {
670             qemu_opt_set_bool(all_opts, "cache.writeback",
671                               !!(flags & BDRV_O_CACHE_WB));
672         }
673         if (!qemu_opt_get(all_opts, "cache.direct")) {
674             qemu_opt_set_bool(all_opts, "cache.direct",
675                               !!(flags & BDRV_O_NOCACHE));
676         }
677         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
678             qemu_opt_set_bool(all_opts, "cache.no-flush",
679                               !!(flags & BDRV_O_NO_FLUSH));
680         }
681         qemu_opt_unset(all_opts, "cache");
682     }
683 
684     /* Get a QDict for processing the options */
685     bs_opts = qdict_new();
686     qemu_opts_to_qdict(all_opts, bs_opts);
687 
688     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
689                                    &error_abort);
690     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
691     if (error_is_set(&local_err)) {
692         qerror_report_err(local_err);
693         error_free(local_err);
694         goto fail;
695     }
696 
697     /* Deprecated option boot=[on|off] */
698     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
699         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
700                 "ignored. Future versions will reject this parameter. Please "
701                 "update your scripts.\n");
702     }
703 
704     /* Media type */
705     value = qemu_opt_get(legacy_opts, "media");
706     if (value) {
707         if (!strcmp(value, "disk")) {
708             media = MEDIA_DISK;
709         } else if (!strcmp(value, "cdrom")) {
710             media = MEDIA_CDROM;
711             read_only = true;
712         } else {
713             error_report("'%s' invalid media", value);
714             goto fail;
715         }
716     }
717 
718     /* copy-on-read is disabled with a warning for read-only devices */
719     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
720     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
721 
722     if (read_only && copy_on_read) {
723         error_report("warning: disabling copy-on-read on read-only drive");
724         copy_on_read = false;
725     }
726 
727     qdict_put(bs_opts, "read-only",
728               qstring_from_str(read_only ? "on" : "off"));
729     qdict_put(bs_opts, "copy-on-read",
730               qstring_from_str(copy_on_read ? "on" :"off"));
731 
732     /* Controller type */
733     value = qemu_opt_get(legacy_opts, "if");
734     if (value) {
735         for (type = 0;
736              type < IF_COUNT && strcmp(value, if_name[type]);
737              type++) {
738         }
739         if (type == IF_COUNT) {
740             error_report("unsupported bus type '%s'", value);
741             goto fail;
742         }
743     } else {
744         type = block_default_type;
745     }
746 
747     /* Geometry */
748     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
749     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
750     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
751 
752     if (cyls || heads || secs) {
753         if (cyls < 1) {
754             error_report("invalid physical cyls number");
755             goto fail;
756         }
757         if (heads < 1) {
758             error_report("invalid physical heads number");
759             goto fail;
760         }
761         if (secs < 1) {
762             error_report("invalid physical secs number");
763             goto fail;
764         }
765     }
766 
767     translation = BIOS_ATA_TRANSLATION_AUTO;
768     value = qemu_opt_get(legacy_opts, "trans");
769     if (value != NULL) {
770         if (!cyls) {
771             error_report("'%s' trans must be used with cyls, heads and secs",
772                          value);
773             goto fail;
774         }
775         if (!strcmp(value, "none")) {
776             translation = BIOS_ATA_TRANSLATION_NONE;
777         } else if (!strcmp(value, "lba")) {
778             translation = BIOS_ATA_TRANSLATION_LBA;
779         } else if (!strcmp(value, "auto")) {
780             translation = BIOS_ATA_TRANSLATION_AUTO;
781         } else {
782             error_report("'%s' invalid translation type", value);
783             goto fail;
784         }
785     }
786 
787     if (media == MEDIA_CDROM) {
788         if (cyls || secs || heads) {
789             error_report("CHS can't be set with media=cdrom");
790             goto fail;
791         }
792     }
793 
794     /* Device address specified by bus/unit or index.
795      * If none was specified, try to find the first free one. */
796     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
797     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
798     index   = qemu_opt_get_number(legacy_opts, "index", -1);
799 
800     max_devs = if_max_devs[type];
801 
802     if (index != -1) {
803         if (bus_id != 0 || unit_id != -1) {
804             error_report("index cannot be used with bus and unit");
805             goto fail;
806         }
807         bus_id = drive_index_to_bus_id(type, index);
808         unit_id = drive_index_to_unit_id(type, index);
809     }
810 
811     if (unit_id == -1) {
812        unit_id = 0;
813        while (drive_get(type, bus_id, unit_id) != NULL) {
814            unit_id++;
815            if (max_devs && unit_id >= max_devs) {
816                unit_id -= max_devs;
817                bus_id++;
818            }
819        }
820     }
821 
822     if (max_devs && unit_id >= max_devs) {
823         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
824         goto fail;
825     }
826 
827     if (drive_get(type, bus_id, unit_id) != NULL) {
828         error_report("drive with bus=%d, unit=%d (index=%d) exists",
829                      bus_id, unit_id, index);
830         goto fail;
831     }
832 
833     /* no id supplied -> create one */
834     if (qemu_opts_id(all_opts) == NULL) {
835         char *new_id;
836         const char *mediastr = "";
837         if (type == IF_IDE || type == IF_SCSI) {
838             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
839         }
840         if (max_devs) {
841             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
842                                      mediastr, unit_id);
843         } else {
844             new_id = g_strdup_printf("%s%s%i", if_name[type],
845                                      mediastr, unit_id);
846         }
847         qdict_put(bs_opts, "id", qstring_from_str(new_id));
848         g_free(new_id);
849     }
850 
851     /* Add virtio block device */
852     devaddr = qemu_opt_get(legacy_opts, "addr");
853     if (devaddr && type != IF_VIRTIO) {
854         error_report("addr is not supported by this bus type");
855         goto fail;
856     }
857 
858     if (type == IF_VIRTIO) {
859         QemuOpts *devopts;
860         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
861                                    &error_abort);
862         if (arch_type == QEMU_ARCH_S390X) {
863             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
864         } else {
865             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
866         }
867         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
868         if (devaddr) {
869             qemu_opt_set(devopts, "addr", devaddr);
870         }
871     }
872 
873     filename = qemu_opt_get(legacy_opts, "file");
874 
875     /* Actual block device init: Functionality shared with blockdev-add */
876     dinfo = blockdev_init(filename, bs_opts, type, &local_err);
877     if (dinfo == NULL) {
878         if (error_is_set(&local_err)) {
879             qerror_report_err(local_err);
880             error_free(local_err);
881         }
882         goto fail;
883     } else {
884         assert(!error_is_set(&local_err));
885     }
886 
887     /* Set legacy DriveInfo fields */
888     dinfo->enable_auto_del = true;
889     dinfo->opts = all_opts;
890 
891     dinfo->cyls = cyls;
892     dinfo->heads = heads;
893     dinfo->secs = secs;
894     dinfo->trans = translation;
895 
896     dinfo->bus = bus_id;
897     dinfo->unit = unit_id;
898     dinfo->devaddr = devaddr;
899 
900     switch(type) {
901     case IF_IDE:
902     case IF_SCSI:
903     case IF_XEN:
904     case IF_NONE:
905         dinfo->media_cd = media == MEDIA_CDROM;
906         break;
907     default:
908         break;
909     }
910 
911 fail:
912     qemu_opts_del(legacy_opts);
913     return dinfo;
914 }
915 
916 void do_commit(Monitor *mon, const QDict *qdict)
917 {
918     const char *device = qdict_get_str(qdict, "device");
919     BlockDriverState *bs;
920     int ret;
921 
922     if (!strcmp(device, "all")) {
923         ret = bdrv_commit_all();
924     } else {
925         bs = bdrv_find(device);
926         if (!bs) {
927             monitor_printf(mon, "Device '%s' not found\n", device);
928             return;
929         }
930         ret = bdrv_commit(bs);
931     }
932     if (ret < 0) {
933         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
934                        strerror(-ret));
935     }
936 }
937 
938 static void blockdev_do_action(int kind, void *data, Error **errp)
939 {
940     TransactionAction action;
941     TransactionActionList list;
942 
943     action.kind = kind;
944     action.data = data;
945     list.value = &action;
946     list.next = NULL;
947     qmp_transaction(&list, errp);
948 }
949 
950 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
951                                 bool has_format, const char *format,
952                                 bool has_mode, enum NewImageMode mode,
953                                 Error **errp)
954 {
955     BlockdevSnapshot snapshot = {
956         .device = (char *) device,
957         .snapshot_file = (char *) snapshot_file,
958         .has_format = has_format,
959         .format = (char *) format,
960         .has_mode = has_mode,
961         .mode = mode,
962     };
963     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
964                        &snapshot, errp);
965 }
966 
967 void qmp_blockdev_snapshot_internal_sync(const char *device,
968                                          const char *name,
969                                          Error **errp)
970 {
971     BlockdevSnapshotInternal snapshot = {
972         .device = (char *) device,
973         .name = (char *) name
974     };
975 
976     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
977                        &snapshot, errp);
978 }
979 
980 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
981                                                          bool has_id,
982                                                          const char *id,
983                                                          bool has_name,
984                                                          const char *name,
985                                                          Error **errp)
986 {
987     BlockDriverState *bs = bdrv_find(device);
988     QEMUSnapshotInfo sn;
989     Error *local_err = NULL;
990     SnapshotInfo *info = NULL;
991     int ret;
992 
993     if (!bs) {
994         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
995         return NULL;
996     }
997 
998     if (!has_id) {
999         id = NULL;
1000     }
1001 
1002     if (!has_name) {
1003         name = NULL;
1004     }
1005 
1006     if (!id && !name) {
1007         error_setg(errp, "Name or id must be provided");
1008         return NULL;
1009     }
1010 
1011     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1012     if (error_is_set(&local_err)) {
1013         error_propagate(errp, local_err);
1014         return NULL;
1015     }
1016     if (!ret) {
1017         error_setg(errp,
1018                    "Snapshot with id '%s' and name '%s' does not exist on "
1019                    "device '%s'",
1020                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1021         return NULL;
1022     }
1023 
1024     bdrv_snapshot_delete(bs, id, name, &local_err);
1025     if (error_is_set(&local_err)) {
1026         error_propagate(errp, local_err);
1027         return NULL;
1028     }
1029 
1030     info = g_malloc0(sizeof(SnapshotInfo));
1031     info->id = g_strdup(sn.id_str);
1032     info->name = g_strdup(sn.name);
1033     info->date_nsec = sn.date_nsec;
1034     info->date_sec = sn.date_sec;
1035     info->vm_state_size = sn.vm_state_size;
1036     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1037     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1038 
1039     return info;
1040 }
1041 
1042 /* New and old BlockDriverState structs for group snapshots */
1043 
1044 typedef struct BlkTransactionState BlkTransactionState;
1045 
1046 /* Only prepare() may fail. In a single transaction, only one of commit() or
1047    abort() will be called, clean() will always be called if it present. */
1048 typedef struct BdrvActionOps {
1049     /* Size of state struct, in bytes. */
1050     size_t instance_size;
1051     /* Prepare the work, must NOT be NULL. */
1052     void (*prepare)(BlkTransactionState *common, Error **errp);
1053     /* Commit the changes, can be NULL. */
1054     void (*commit)(BlkTransactionState *common);
1055     /* Abort the changes on fail, can be NULL. */
1056     void (*abort)(BlkTransactionState *common);
1057     /* Clean up resource in the end, can be NULL. */
1058     void (*clean)(BlkTransactionState *common);
1059 } BdrvActionOps;
1060 
1061 /*
1062  * This structure must be arranged as first member in child type, assuming
1063  * that compiler will also arrange it to the same address with parent instance.
1064  * Later it will be used in free().
1065  */
1066 struct BlkTransactionState {
1067     TransactionAction *action;
1068     const BdrvActionOps *ops;
1069     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1070 };
1071 
1072 /* internal snapshot private data */
1073 typedef struct InternalSnapshotState {
1074     BlkTransactionState common;
1075     BlockDriverState *bs;
1076     QEMUSnapshotInfo sn;
1077 } InternalSnapshotState;
1078 
1079 static void internal_snapshot_prepare(BlkTransactionState *common,
1080                                       Error **errp)
1081 {
1082     const char *device;
1083     const char *name;
1084     BlockDriverState *bs;
1085     QEMUSnapshotInfo old_sn, *sn;
1086     bool ret;
1087     qemu_timeval tv;
1088     BlockdevSnapshotInternal *internal;
1089     InternalSnapshotState *state;
1090     int ret1;
1091 
1092     g_assert(common->action->kind ==
1093              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1094     internal = common->action->blockdev_snapshot_internal_sync;
1095     state = DO_UPCAST(InternalSnapshotState, common, common);
1096 
1097     /* 1. parse input */
1098     device = internal->device;
1099     name = internal->name;
1100 
1101     /* 2. check for validation */
1102     bs = bdrv_find(device);
1103     if (!bs) {
1104         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1105         return;
1106     }
1107 
1108     if (!bdrv_is_inserted(bs)) {
1109         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1110         return;
1111     }
1112 
1113     if (bdrv_is_read_only(bs)) {
1114         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1115         return;
1116     }
1117 
1118     if (!bdrv_can_snapshot(bs)) {
1119         error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1120                   bs->drv->format_name, device, "internal snapshot");
1121         return;
1122     }
1123 
1124     if (!strlen(name)) {
1125         error_setg(errp, "Name is empty");
1126         return;
1127     }
1128 
1129     /* check whether a snapshot with name exist */
1130     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn, errp);
1131     if (error_is_set(errp)) {
1132         return;
1133     } else if (ret) {
1134         error_setg(errp,
1135                    "Snapshot with name '%s' already exists on device '%s'",
1136                    name, device);
1137         return;
1138     }
1139 
1140     /* 3. take the snapshot */
1141     sn = &state->sn;
1142     pstrcpy(sn->name, sizeof(sn->name), name);
1143     qemu_gettimeofday(&tv);
1144     sn->date_sec = tv.tv_sec;
1145     sn->date_nsec = tv.tv_usec * 1000;
1146     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1147 
1148     ret1 = bdrv_snapshot_create(bs, sn);
1149     if (ret1 < 0) {
1150         error_setg_errno(errp, -ret1,
1151                          "Failed to create snapshot '%s' on device '%s'",
1152                          name, device);
1153         return;
1154     }
1155 
1156     /* 4. succeed, mark a snapshot is created */
1157     state->bs = bs;
1158 }
1159 
1160 static void internal_snapshot_abort(BlkTransactionState *common)
1161 {
1162     InternalSnapshotState *state =
1163                              DO_UPCAST(InternalSnapshotState, common, common);
1164     BlockDriverState *bs = state->bs;
1165     QEMUSnapshotInfo *sn = &state->sn;
1166     Error *local_error = NULL;
1167 
1168     if (!bs) {
1169         return;
1170     }
1171 
1172     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1173         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1174                      "device '%s' in abort: %s",
1175                      sn->id_str,
1176                      sn->name,
1177                      bdrv_get_device_name(bs),
1178                      error_get_pretty(local_error));
1179         error_free(local_error);
1180     }
1181 }
1182 
1183 /* external snapshot private data */
1184 typedef struct ExternalSnapshotState {
1185     BlkTransactionState common;
1186     BlockDriverState *old_bs;
1187     BlockDriverState *new_bs;
1188 } ExternalSnapshotState;
1189 
1190 static void external_snapshot_prepare(BlkTransactionState *common,
1191                                       Error **errp)
1192 {
1193     BlockDriver *drv;
1194     int flags, ret;
1195     Error *local_err = NULL;
1196     const char *device;
1197     const char *new_image_file;
1198     const char *format = "qcow2";
1199     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1200     ExternalSnapshotState *state =
1201                              DO_UPCAST(ExternalSnapshotState, common, common);
1202     TransactionAction *action = common->action;
1203 
1204     /* get parameters */
1205     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1206 
1207     device = action->blockdev_snapshot_sync->device;
1208     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1209     if (action->blockdev_snapshot_sync->has_format) {
1210         format = action->blockdev_snapshot_sync->format;
1211     }
1212     if (action->blockdev_snapshot_sync->has_mode) {
1213         mode = action->blockdev_snapshot_sync->mode;
1214     }
1215 
1216     /* start processing */
1217     drv = bdrv_find_format(format);
1218     if (!drv) {
1219         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1220         return;
1221     }
1222 
1223     state->old_bs = bdrv_find(device);
1224     if (!state->old_bs) {
1225         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1226         return;
1227     }
1228 
1229     if (!bdrv_is_inserted(state->old_bs)) {
1230         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1231         return;
1232     }
1233 
1234     if (bdrv_in_use(state->old_bs)) {
1235         error_set(errp, QERR_DEVICE_IN_USE, device);
1236         return;
1237     }
1238 
1239     if (!bdrv_is_read_only(state->old_bs)) {
1240         if (bdrv_flush(state->old_bs)) {
1241             error_set(errp, QERR_IO_ERROR);
1242             return;
1243         }
1244     }
1245 
1246     if (bdrv_check_ext_snapshot(state->old_bs) != EXT_SNAPSHOT_ALLOWED) {
1247         error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1248         return;
1249     }
1250 
1251     flags = state->old_bs->open_flags;
1252 
1253     /* create new image w/backing file */
1254     if (mode != NEW_IMAGE_MODE_EXISTING) {
1255         bdrv_img_create(new_image_file, format,
1256                         state->old_bs->filename,
1257                         state->old_bs->drv->format_name,
1258                         NULL, -1, flags, &local_err, false);
1259         if (error_is_set(&local_err)) {
1260             error_propagate(errp, local_err);
1261             return;
1262         }
1263     }
1264 
1265     /* We will manually add the backing_hd field to the bs later */
1266     state->new_bs = bdrv_new("");
1267     /* TODO Inherit bs->options or only take explicit options with an
1268      * extended QMP command? */
1269     ret = bdrv_open(state->new_bs, new_image_file, NULL,
1270                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1271     if (ret != 0) {
1272         error_propagate(errp, local_err);
1273     }
1274 }
1275 
1276 static void external_snapshot_commit(BlkTransactionState *common)
1277 {
1278     ExternalSnapshotState *state =
1279                              DO_UPCAST(ExternalSnapshotState, common, common);
1280 
1281     /* This removes our old bs and adds the new bs */
1282     bdrv_append(state->new_bs, state->old_bs);
1283     /* We don't need (or want) to use the transactional
1284      * bdrv_reopen_multiple() across all the entries at once, because we
1285      * don't want to abort all of them if one of them fails the reopen */
1286     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1287                 NULL);
1288 }
1289 
1290 static void external_snapshot_abort(BlkTransactionState *common)
1291 {
1292     ExternalSnapshotState *state =
1293                              DO_UPCAST(ExternalSnapshotState, common, common);
1294     if (state->new_bs) {
1295         bdrv_unref(state->new_bs);
1296     }
1297 }
1298 
1299 typedef struct DriveBackupState {
1300     BlkTransactionState common;
1301     BlockDriverState *bs;
1302     BlockJob *job;
1303 } DriveBackupState;
1304 
1305 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1306 {
1307     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1308     DriveBackup *backup;
1309     Error *local_err = NULL;
1310 
1311     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1312     backup = common->action->drive_backup;
1313 
1314     qmp_drive_backup(backup->device, backup->target,
1315                      backup->has_format, backup->format,
1316                      backup->sync,
1317                      backup->has_mode, backup->mode,
1318                      backup->has_speed, backup->speed,
1319                      backup->has_on_source_error, backup->on_source_error,
1320                      backup->has_on_target_error, backup->on_target_error,
1321                      &local_err);
1322     if (error_is_set(&local_err)) {
1323         error_propagate(errp, local_err);
1324         state->bs = NULL;
1325         state->job = NULL;
1326         return;
1327     }
1328 
1329     state->bs = bdrv_find(backup->device);
1330     state->job = state->bs->job;
1331 }
1332 
1333 static void drive_backup_abort(BlkTransactionState *common)
1334 {
1335     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1336     BlockDriverState *bs = state->bs;
1337 
1338     /* Only cancel if it's the job we started */
1339     if (bs && bs->job && bs->job == state->job) {
1340         block_job_cancel_sync(bs->job);
1341     }
1342 }
1343 
1344 static void abort_prepare(BlkTransactionState *common, Error **errp)
1345 {
1346     error_setg(errp, "Transaction aborted using Abort action");
1347 }
1348 
1349 static void abort_commit(BlkTransactionState *common)
1350 {
1351     g_assert_not_reached(); /* this action never succeeds */
1352 }
1353 
1354 static const BdrvActionOps actions[] = {
1355     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1356         .instance_size = sizeof(ExternalSnapshotState),
1357         .prepare  = external_snapshot_prepare,
1358         .commit   = external_snapshot_commit,
1359         .abort = external_snapshot_abort,
1360     },
1361     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1362         .instance_size = sizeof(DriveBackupState),
1363         .prepare = drive_backup_prepare,
1364         .abort = drive_backup_abort,
1365     },
1366     [TRANSACTION_ACTION_KIND_ABORT] = {
1367         .instance_size = sizeof(BlkTransactionState),
1368         .prepare = abort_prepare,
1369         .commit = abort_commit,
1370     },
1371     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1372         .instance_size = sizeof(InternalSnapshotState),
1373         .prepare  = internal_snapshot_prepare,
1374         .abort = internal_snapshot_abort,
1375     },
1376 };
1377 
1378 /*
1379  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1380  *  then we do not pivot any of the devices in the group, and abandon the
1381  *  snapshots
1382  */
1383 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1384 {
1385     TransactionActionList *dev_entry = dev_list;
1386     BlkTransactionState *state, *next;
1387     Error *local_err = NULL;
1388 
1389     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1390     QSIMPLEQ_INIT(&snap_bdrv_states);
1391 
1392     /* drain all i/o before any snapshots */
1393     bdrv_drain_all();
1394 
1395     /* We don't do anything in this loop that commits us to the snapshot */
1396     while (NULL != dev_entry) {
1397         TransactionAction *dev_info = NULL;
1398         const BdrvActionOps *ops;
1399 
1400         dev_info = dev_entry->value;
1401         dev_entry = dev_entry->next;
1402 
1403         assert(dev_info->kind < ARRAY_SIZE(actions));
1404 
1405         ops = &actions[dev_info->kind];
1406         assert(ops->instance_size > 0);
1407 
1408         state = g_malloc0(ops->instance_size);
1409         state->ops = ops;
1410         state->action = dev_info;
1411         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1412 
1413         state->ops->prepare(state, &local_err);
1414         if (error_is_set(&local_err)) {
1415             error_propagate(errp, local_err);
1416             goto delete_and_fail;
1417         }
1418     }
1419 
1420     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1421         if (state->ops->commit) {
1422             state->ops->commit(state);
1423         }
1424     }
1425 
1426     /* success */
1427     goto exit;
1428 
1429 delete_and_fail:
1430     /*
1431     * failure, and it is all-or-none; abandon each new bs, and keep using
1432     * the original bs for all images
1433     */
1434     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1435         if (state->ops->abort) {
1436             state->ops->abort(state);
1437         }
1438     }
1439 exit:
1440     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1441         if (state->ops->clean) {
1442             state->ops->clean(state);
1443         }
1444         g_free(state);
1445     }
1446 }
1447 
1448 
1449 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1450 {
1451     if (bdrv_in_use(bs)) {
1452         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1453         return;
1454     }
1455     if (!bdrv_dev_has_removable_media(bs)) {
1456         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1457         return;
1458     }
1459 
1460     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1461         bdrv_dev_eject_request(bs, force);
1462         if (!force) {
1463             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1464             return;
1465         }
1466     }
1467 
1468     bdrv_close(bs);
1469 }
1470 
1471 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1472 {
1473     BlockDriverState *bs;
1474 
1475     bs = bdrv_find(device);
1476     if (!bs) {
1477         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1478         return;
1479     }
1480 
1481     eject_device(bs, force, errp);
1482 }
1483 
1484 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1485 {
1486     BlockDriverState *bs;
1487     int err;
1488 
1489     bs = bdrv_find(device);
1490     if (!bs) {
1491         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1492         return;
1493     }
1494 
1495     err = bdrv_set_key(bs, password);
1496     if (err == -EINVAL) {
1497         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1498         return;
1499     } else if (err < 0) {
1500         error_set(errp, QERR_INVALID_PASSWORD);
1501         return;
1502     }
1503 }
1504 
1505 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1506                                     int bdrv_flags, BlockDriver *drv,
1507                                     const char *password, Error **errp)
1508 {
1509     Error *local_err = NULL;
1510     int ret;
1511 
1512     ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv, &local_err);
1513     if (ret < 0) {
1514         error_propagate(errp, local_err);
1515         return;
1516     }
1517 
1518     if (bdrv_key_required(bs)) {
1519         if (password) {
1520             if (bdrv_set_key(bs, password) < 0) {
1521                 error_set(errp, QERR_INVALID_PASSWORD);
1522             }
1523         } else {
1524             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1525                       bdrv_get_encrypted_filename(bs));
1526         }
1527     } else if (password) {
1528         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1529     }
1530 }
1531 
1532 void qmp_change_blockdev(const char *device, const char *filename,
1533                          const char *format, Error **errp)
1534 {
1535     BlockDriverState *bs;
1536     BlockDriver *drv = NULL;
1537     int bdrv_flags;
1538     Error *err = NULL;
1539 
1540     bs = bdrv_find(device);
1541     if (!bs) {
1542         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1543         return;
1544     }
1545 
1546     if (format) {
1547         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1548         if (!drv) {
1549             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1550             return;
1551         }
1552     }
1553 
1554     eject_device(bs, 0, &err);
1555     if (error_is_set(&err)) {
1556         error_propagate(errp, err);
1557         return;
1558     }
1559 
1560     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1561     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1562 
1563     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1564 }
1565 
1566 /* throttling disk I/O limits */
1567 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1568                                int64_t bps_wr,
1569                                int64_t iops,
1570                                int64_t iops_rd,
1571                                int64_t iops_wr,
1572                                bool has_bps_max,
1573                                int64_t bps_max,
1574                                bool has_bps_rd_max,
1575                                int64_t bps_rd_max,
1576                                bool has_bps_wr_max,
1577                                int64_t bps_wr_max,
1578                                bool has_iops_max,
1579                                int64_t iops_max,
1580                                bool has_iops_rd_max,
1581                                int64_t iops_rd_max,
1582                                bool has_iops_wr_max,
1583                                int64_t iops_wr_max,
1584                                bool has_iops_size,
1585                                int64_t iops_size, Error **errp)
1586 {
1587     ThrottleConfig cfg;
1588     BlockDriverState *bs;
1589 
1590     bs = bdrv_find(device);
1591     if (!bs) {
1592         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1593         return;
1594     }
1595 
1596     memset(&cfg, 0, sizeof(cfg));
1597     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1598     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1599     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1600 
1601     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1602     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1603     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1604 
1605     if (has_bps_max) {
1606         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1607     }
1608     if (has_bps_rd_max) {
1609         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1610     }
1611     if (has_bps_wr_max) {
1612         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1613     }
1614     if (has_iops_max) {
1615         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1616     }
1617     if (has_iops_rd_max) {
1618         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1619     }
1620     if (has_iops_wr_max) {
1621         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1622     }
1623 
1624     if (has_iops_size) {
1625         cfg.op_size = iops_size;
1626     }
1627 
1628     if (!check_throttle_config(&cfg, errp)) {
1629         return;
1630     }
1631 
1632     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1633         bdrv_io_limits_enable(bs);
1634     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1635         bdrv_io_limits_disable(bs);
1636     }
1637 
1638     if (bs->io_limits_enabled) {
1639         bdrv_set_io_limits(bs, &cfg);
1640     }
1641 }
1642 
1643 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1644 {
1645     const char *id = qdict_get_str(qdict, "id");
1646     BlockDriverState *bs;
1647 
1648     bs = bdrv_find(id);
1649     if (!bs) {
1650         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1651         return -1;
1652     }
1653     if (bdrv_in_use(bs)) {
1654         qerror_report(QERR_DEVICE_IN_USE, id);
1655         return -1;
1656     }
1657 
1658     /* quiesce block driver; prevent further io */
1659     bdrv_drain_all();
1660     bdrv_flush(bs);
1661     bdrv_close(bs);
1662 
1663     /* if we have a device attached to this BlockDriverState
1664      * then we need to make the drive anonymous until the device
1665      * can be removed.  If this is a drive with no device backing
1666      * then we can just get rid of the block driver state right here.
1667      */
1668     if (bdrv_get_attached_dev(bs)) {
1669         bdrv_make_anon(bs);
1670 
1671         /* Further I/O must not pause the guest */
1672         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1673                           BLOCKDEV_ON_ERROR_REPORT);
1674     } else {
1675         drive_uninit(drive_get_by_blockdev(bs));
1676     }
1677 
1678     return 0;
1679 }
1680 
1681 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1682 {
1683     BlockDriverState *bs;
1684     int ret;
1685 
1686     bs = bdrv_find(device);
1687     if (!bs) {
1688         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1689         return;
1690     }
1691 
1692     if (size < 0) {
1693         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1694         return;
1695     }
1696 
1697     /* complete all in-flight operations before resizing the device */
1698     bdrv_drain_all();
1699 
1700     ret = bdrv_truncate(bs, size);
1701     switch (ret) {
1702     case 0:
1703         break;
1704     case -ENOMEDIUM:
1705         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1706         break;
1707     case -ENOTSUP:
1708         error_set(errp, QERR_UNSUPPORTED);
1709         break;
1710     case -EACCES:
1711         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1712         break;
1713     case -EBUSY:
1714         error_set(errp, QERR_DEVICE_IN_USE, device);
1715         break;
1716     default:
1717         error_setg_errno(errp, -ret, "Could not resize");
1718         break;
1719     }
1720 }
1721 
1722 static void block_job_cb(void *opaque, int ret)
1723 {
1724     BlockDriverState *bs = opaque;
1725     QObject *obj;
1726 
1727     trace_block_job_cb(bs, bs->job, ret);
1728 
1729     assert(bs->job);
1730     obj = qobject_from_block_job(bs->job);
1731     if (ret < 0) {
1732         QDict *dict = qobject_to_qdict(obj);
1733         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1734     }
1735 
1736     if (block_job_is_cancelled(bs->job)) {
1737         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1738     } else {
1739         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1740     }
1741     qobject_decref(obj);
1742 
1743     bdrv_put_ref_bh_schedule(bs);
1744 }
1745 
1746 void qmp_block_stream(const char *device, bool has_base,
1747                       const char *base, bool has_speed, int64_t speed,
1748                       bool has_on_error, BlockdevOnError on_error,
1749                       Error **errp)
1750 {
1751     BlockDriverState *bs;
1752     BlockDriverState *base_bs = NULL;
1753     Error *local_err = NULL;
1754 
1755     if (!has_on_error) {
1756         on_error = BLOCKDEV_ON_ERROR_REPORT;
1757     }
1758 
1759     bs = bdrv_find(device);
1760     if (!bs) {
1761         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1762         return;
1763     }
1764 
1765     if (base) {
1766         base_bs = bdrv_find_backing_image(bs, base);
1767         if (base_bs == NULL) {
1768             error_set(errp, QERR_BASE_NOT_FOUND, base);
1769             return;
1770         }
1771     }
1772 
1773     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1774                  on_error, block_job_cb, bs, &local_err);
1775     if (error_is_set(&local_err)) {
1776         error_propagate(errp, local_err);
1777         return;
1778     }
1779 
1780     trace_qmp_block_stream(bs, bs->job);
1781 }
1782 
1783 void qmp_block_commit(const char *device,
1784                       bool has_base, const char *base, const char *top,
1785                       bool has_speed, int64_t speed,
1786                       Error **errp)
1787 {
1788     BlockDriverState *bs;
1789     BlockDriverState *base_bs, *top_bs;
1790     Error *local_err = NULL;
1791     /* This will be part of the QMP command, if/when the
1792      * BlockdevOnError change for blkmirror makes it in
1793      */
1794     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1795 
1796     /* drain all i/o before commits */
1797     bdrv_drain_all();
1798 
1799     bs = bdrv_find(device);
1800     if (!bs) {
1801         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1802         return;
1803     }
1804 
1805     /* default top_bs is the active layer */
1806     top_bs = bs;
1807 
1808     if (top) {
1809         if (strcmp(bs->filename, top) != 0) {
1810             top_bs = bdrv_find_backing_image(bs, top);
1811         }
1812     }
1813 
1814     if (top_bs == NULL) {
1815         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1816         return;
1817     }
1818 
1819     if (has_base && base) {
1820         base_bs = bdrv_find_backing_image(top_bs, base);
1821     } else {
1822         base_bs = bdrv_find_base(top_bs);
1823     }
1824 
1825     if (base_bs == NULL) {
1826         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1827         return;
1828     }
1829 
1830     if (top_bs == bs) {
1831         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1832                             bs, &local_err);
1833     } else {
1834         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1835                     &local_err);
1836     }
1837     if (local_err != NULL) {
1838         error_propagate(errp, local_err);
1839         return;
1840     }
1841 }
1842 
1843 void qmp_drive_backup(const char *device, const char *target,
1844                       bool has_format, const char *format,
1845                       enum MirrorSyncMode sync,
1846                       bool has_mode, enum NewImageMode mode,
1847                       bool has_speed, int64_t speed,
1848                       bool has_on_source_error, BlockdevOnError on_source_error,
1849                       bool has_on_target_error, BlockdevOnError on_target_error,
1850                       Error **errp)
1851 {
1852     BlockDriverState *bs;
1853     BlockDriverState *target_bs;
1854     BlockDriverState *source = NULL;
1855     BlockDriver *drv = NULL;
1856     Error *local_err = NULL;
1857     int flags;
1858     int64_t size;
1859     int ret;
1860 
1861     if (!has_speed) {
1862         speed = 0;
1863     }
1864     if (!has_on_source_error) {
1865         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1866     }
1867     if (!has_on_target_error) {
1868         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1869     }
1870     if (!has_mode) {
1871         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1872     }
1873 
1874     bs = bdrv_find(device);
1875     if (!bs) {
1876         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1877         return;
1878     }
1879 
1880     if (!bdrv_is_inserted(bs)) {
1881         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1882         return;
1883     }
1884 
1885     if (!has_format) {
1886         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1887     }
1888     if (format) {
1889         drv = bdrv_find_format(format);
1890         if (!drv) {
1891             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1892             return;
1893         }
1894     }
1895 
1896     if (bdrv_in_use(bs)) {
1897         error_set(errp, QERR_DEVICE_IN_USE, device);
1898         return;
1899     }
1900 
1901     flags = bs->open_flags | BDRV_O_RDWR;
1902 
1903     /* See if we have a backing HD we can use to create our new image
1904      * on top of. */
1905     if (sync == MIRROR_SYNC_MODE_TOP) {
1906         source = bs->backing_hd;
1907         if (!source) {
1908             sync = MIRROR_SYNC_MODE_FULL;
1909         }
1910     }
1911     if (sync == MIRROR_SYNC_MODE_NONE) {
1912         source = bs;
1913     }
1914 
1915     size = bdrv_getlength(bs);
1916     if (size < 0) {
1917         error_setg_errno(errp, -size, "bdrv_getlength failed");
1918         return;
1919     }
1920 
1921     if (mode != NEW_IMAGE_MODE_EXISTING) {
1922         assert(format && drv);
1923         if (source) {
1924             bdrv_img_create(target, format, source->filename,
1925                             source->drv->format_name, NULL,
1926                             size, flags, &local_err, false);
1927         } else {
1928             bdrv_img_create(target, format, NULL, NULL, NULL,
1929                             size, flags, &local_err, false);
1930         }
1931     }
1932 
1933     if (error_is_set(&local_err)) {
1934         error_propagate(errp, local_err);
1935         return;
1936     }
1937 
1938     target_bs = bdrv_new("");
1939     ret = bdrv_open(target_bs, target, NULL, flags, drv, &local_err);
1940     if (ret < 0) {
1941         bdrv_unref(target_bs);
1942         error_propagate(errp, local_err);
1943         return;
1944     }
1945 
1946     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1947                  block_job_cb, bs, &local_err);
1948     if (local_err != NULL) {
1949         bdrv_unref(target_bs);
1950         error_propagate(errp, local_err);
1951         return;
1952     }
1953 }
1954 
1955 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
1956 {
1957     return bdrv_named_nodes_list();
1958 }
1959 
1960 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1961 
1962 void qmp_drive_mirror(const char *device, const char *target,
1963                       bool has_format, const char *format,
1964                       enum MirrorSyncMode sync,
1965                       bool has_mode, enum NewImageMode mode,
1966                       bool has_speed, int64_t speed,
1967                       bool has_granularity, uint32_t granularity,
1968                       bool has_buf_size, int64_t buf_size,
1969                       bool has_on_source_error, BlockdevOnError on_source_error,
1970                       bool has_on_target_error, BlockdevOnError on_target_error,
1971                       Error **errp)
1972 {
1973     BlockDriverState *bs;
1974     BlockDriverState *source, *target_bs;
1975     BlockDriver *drv = NULL;
1976     Error *local_err = NULL;
1977     int flags;
1978     int64_t size;
1979     int ret;
1980 
1981     if (!has_speed) {
1982         speed = 0;
1983     }
1984     if (!has_on_source_error) {
1985         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1986     }
1987     if (!has_on_target_error) {
1988         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1989     }
1990     if (!has_mode) {
1991         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1992     }
1993     if (!has_granularity) {
1994         granularity = 0;
1995     }
1996     if (!has_buf_size) {
1997         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1998     }
1999 
2000     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2001         error_set(errp, QERR_INVALID_PARAMETER, device);
2002         return;
2003     }
2004     if (granularity & (granularity - 1)) {
2005         error_set(errp, QERR_INVALID_PARAMETER, device);
2006         return;
2007     }
2008 
2009     bs = bdrv_find(device);
2010     if (!bs) {
2011         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2012         return;
2013     }
2014 
2015     if (!bdrv_is_inserted(bs)) {
2016         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2017         return;
2018     }
2019 
2020     if (!has_format) {
2021         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2022     }
2023     if (format) {
2024         drv = bdrv_find_format(format);
2025         if (!drv) {
2026             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2027             return;
2028         }
2029     }
2030 
2031     if (bdrv_in_use(bs)) {
2032         error_set(errp, QERR_DEVICE_IN_USE, device);
2033         return;
2034     }
2035 
2036     flags = bs->open_flags | BDRV_O_RDWR;
2037     source = bs->backing_hd;
2038     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2039         sync = MIRROR_SYNC_MODE_FULL;
2040     }
2041     if (sync == MIRROR_SYNC_MODE_NONE) {
2042         source = bs;
2043     }
2044 
2045     size = bdrv_getlength(bs);
2046     if (size < 0) {
2047         error_setg_errno(errp, -size, "bdrv_getlength failed");
2048         return;
2049     }
2050 
2051     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2052         && mode != NEW_IMAGE_MODE_EXISTING)
2053     {
2054         /* create new image w/o backing file */
2055         assert(format && drv);
2056         bdrv_img_create(target, format,
2057                         NULL, NULL, NULL, size, flags, &local_err, false);
2058     } else {
2059         switch (mode) {
2060         case NEW_IMAGE_MODE_EXISTING:
2061             break;
2062         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2063             /* create new image with backing file */
2064             bdrv_img_create(target, format,
2065                             source->filename,
2066                             source->drv->format_name,
2067                             NULL, size, flags, &local_err, false);
2068             break;
2069         default:
2070             abort();
2071         }
2072     }
2073 
2074     if (error_is_set(&local_err)) {
2075         error_propagate(errp, local_err);
2076         return;
2077     }
2078 
2079     /* Mirroring takes care of copy-on-write using the source's backing
2080      * file.
2081      */
2082     target_bs = bdrv_new("");
2083     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv,
2084                     &local_err);
2085     if (ret < 0) {
2086         bdrv_unref(target_bs);
2087         error_propagate(errp, local_err);
2088         return;
2089     }
2090 
2091     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2092                  on_source_error, on_target_error,
2093                  block_job_cb, bs, &local_err);
2094     if (local_err != NULL) {
2095         bdrv_unref(target_bs);
2096         error_propagate(errp, local_err);
2097         return;
2098     }
2099 }
2100 
2101 static BlockJob *find_block_job(const char *device)
2102 {
2103     BlockDriverState *bs;
2104 
2105     bs = bdrv_find(device);
2106     if (!bs || !bs->job) {
2107         return NULL;
2108     }
2109     return bs->job;
2110 }
2111 
2112 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2113 {
2114     BlockJob *job = find_block_job(device);
2115 
2116     if (!job) {
2117         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2118         return;
2119     }
2120 
2121     block_job_set_speed(job, speed, errp);
2122 }
2123 
2124 void qmp_block_job_cancel(const char *device,
2125                           bool has_force, bool force, Error **errp)
2126 {
2127     BlockJob *job = find_block_job(device);
2128 
2129     if (!has_force) {
2130         force = false;
2131     }
2132 
2133     if (!job) {
2134         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2135         return;
2136     }
2137     if (job->paused && !force) {
2138         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
2139         return;
2140     }
2141 
2142     trace_qmp_block_job_cancel(job);
2143     block_job_cancel(job);
2144 }
2145 
2146 void qmp_block_job_pause(const char *device, Error **errp)
2147 {
2148     BlockJob *job = find_block_job(device);
2149 
2150     if (!job) {
2151         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2152         return;
2153     }
2154 
2155     trace_qmp_block_job_pause(job);
2156     block_job_pause(job);
2157 }
2158 
2159 void qmp_block_job_resume(const char *device, Error **errp)
2160 {
2161     BlockJob *job = find_block_job(device);
2162 
2163     if (!job) {
2164         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2165         return;
2166     }
2167 
2168     trace_qmp_block_job_resume(job);
2169     block_job_resume(job);
2170 }
2171 
2172 void qmp_block_job_complete(const char *device, Error **errp)
2173 {
2174     BlockJob *job = find_block_job(device);
2175 
2176     if (!job) {
2177         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2178         return;
2179     }
2180 
2181     trace_qmp_block_job_complete(job);
2182     block_job_complete(job, errp);
2183 }
2184 
2185 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2186 {
2187     QmpOutputVisitor *ov = qmp_output_visitor_new();
2188     QObject *obj;
2189     QDict *qdict;
2190     Error *local_err = NULL;
2191 
2192     /* Require an ID in the top level */
2193     if (!options->has_id) {
2194         error_setg(errp, "Block device needs an ID");
2195         goto fail;
2196     }
2197 
2198     /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2199      * cache.direct=false instead of silently switching to aio=threads, except
2200      * if called from drive_init.
2201      *
2202      * For now, simply forbidding the combination for all drivers will do. */
2203     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2204         bool direct = options->cache->has_direct && options->cache->direct;
2205         if (!options->has_cache && !direct) {
2206             error_setg(errp, "aio=native requires cache.direct=true");
2207             goto fail;
2208         }
2209     }
2210 
2211     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2212                                &options, NULL, &local_err);
2213     if (error_is_set(&local_err)) {
2214         error_propagate(errp, local_err);
2215         goto fail;
2216     }
2217 
2218     obj = qmp_output_get_qobject(ov);
2219     qdict = qobject_to_qdict(obj);
2220 
2221     qdict_flatten(qdict);
2222 
2223     blockdev_init(NULL, qdict, IF_NONE, &local_err);
2224     if (error_is_set(&local_err)) {
2225         error_propagate(errp, local_err);
2226         goto fail;
2227     }
2228 
2229 fail:
2230     qmp_output_visitor_cleanup(ov);
2231 }
2232 
2233 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2234 {
2235     BlockJobInfoList **prev = opaque;
2236     BlockJob *job = bs->job;
2237 
2238     if (job) {
2239         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2240         elem->value = block_job_query(bs->job);
2241         (*prev)->next = elem;
2242         *prev = elem;
2243     }
2244 }
2245 
2246 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2247 {
2248     /* Dummy is a fake list element for holding the head pointer */
2249     BlockJobInfoList dummy = {};
2250     BlockJobInfoList *prev = &dummy;
2251     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2252     return dummy.next;
2253 }
2254 
2255 QemuOptsList qemu_common_drive_opts = {
2256     .name = "drive",
2257     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2258     .desc = {
2259         {
2260             .name = "snapshot",
2261             .type = QEMU_OPT_BOOL,
2262             .help = "enable/disable snapshot mode",
2263         },{
2264             .name = "discard",
2265             .type = QEMU_OPT_STRING,
2266             .help = "discard operation (ignore/off, unmap/on)",
2267         },{
2268             .name = "cache.writeback",
2269             .type = QEMU_OPT_BOOL,
2270             .help = "enables writeback mode for any caches",
2271         },{
2272             .name = "cache.direct",
2273             .type = QEMU_OPT_BOOL,
2274             .help = "enables use of O_DIRECT (bypass the host page cache)",
2275         },{
2276             .name = "cache.no-flush",
2277             .type = QEMU_OPT_BOOL,
2278             .help = "ignore any flush requests for the device",
2279         },{
2280             .name = "aio",
2281             .type = QEMU_OPT_STRING,
2282             .help = "host AIO implementation (threads, native)",
2283         },{
2284             .name = "format",
2285             .type = QEMU_OPT_STRING,
2286             .help = "disk format (raw, qcow2, ...)",
2287         },{
2288             .name = "serial",
2289             .type = QEMU_OPT_STRING,
2290             .help = "disk serial number",
2291         },{
2292             .name = "rerror",
2293             .type = QEMU_OPT_STRING,
2294             .help = "read error action",
2295         },{
2296             .name = "werror",
2297             .type = QEMU_OPT_STRING,
2298             .help = "write error action",
2299         },{
2300             .name = "read-only",
2301             .type = QEMU_OPT_BOOL,
2302             .help = "open drive file as read-only",
2303         },{
2304             .name = "throttling.iops-total",
2305             .type = QEMU_OPT_NUMBER,
2306             .help = "limit total I/O operations per second",
2307         },{
2308             .name = "throttling.iops-read",
2309             .type = QEMU_OPT_NUMBER,
2310             .help = "limit read operations per second",
2311         },{
2312             .name = "throttling.iops-write",
2313             .type = QEMU_OPT_NUMBER,
2314             .help = "limit write operations per second",
2315         },{
2316             .name = "throttling.bps-total",
2317             .type = QEMU_OPT_NUMBER,
2318             .help = "limit total bytes per second",
2319         },{
2320             .name = "throttling.bps-read",
2321             .type = QEMU_OPT_NUMBER,
2322             .help = "limit read bytes per second",
2323         },{
2324             .name = "throttling.bps-write",
2325             .type = QEMU_OPT_NUMBER,
2326             .help = "limit write bytes per second",
2327         },{
2328             .name = "throttling.iops-total-max",
2329             .type = QEMU_OPT_NUMBER,
2330             .help = "I/O operations burst",
2331         },{
2332             .name = "throttling.iops-read-max",
2333             .type = QEMU_OPT_NUMBER,
2334             .help = "I/O operations read burst",
2335         },{
2336             .name = "throttling.iops-write-max",
2337             .type = QEMU_OPT_NUMBER,
2338             .help = "I/O operations write burst",
2339         },{
2340             .name = "throttling.bps-total-max",
2341             .type = QEMU_OPT_NUMBER,
2342             .help = "total bytes burst",
2343         },{
2344             .name = "throttling.bps-read-max",
2345             .type = QEMU_OPT_NUMBER,
2346             .help = "total bytes read burst",
2347         },{
2348             .name = "throttling.bps-write-max",
2349             .type = QEMU_OPT_NUMBER,
2350             .help = "total bytes write burst",
2351         },{
2352             .name = "throttling.iops-size",
2353             .type = QEMU_OPT_NUMBER,
2354             .help = "when limiting by iops max size of an I/O in bytes",
2355         },{
2356             .name = "copy-on-read",
2357             .type = QEMU_OPT_BOOL,
2358             .help = "copy read data from backing file into image file",
2359         },
2360         { /* end of list */ }
2361     },
2362 };
2363 
2364 QemuOptsList qemu_drive_opts = {
2365     .name = "drive",
2366     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2367     .desc = {
2368         /*
2369          * no elements => accept any params
2370          * validation will happen later
2371          */
2372         { /* end of list */ }
2373     },
2374 };
2375