xref: /qemu/blockdev.c (revision 72ac97cd)
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 "qemu/option.h"
38 #include "qemu/config-file.h"
39 #include "qapi/qmp/types.h"
40 #include "qapi-visit.h"
41 #include "qapi/qmp-output-visitor.h"
42 #include "sysemu/sysemu.h"
43 #include "block/block_int.h"
44 #include "qmp-commands.h"
45 #include "trace.h"
46 #include "sysemu/arch_init.h"
47 
48 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
49 
50 static const char *const if_name[IF_COUNT] = {
51     [IF_NONE] = "none",
52     [IF_IDE] = "ide",
53     [IF_SCSI] = "scsi",
54     [IF_FLOPPY] = "floppy",
55     [IF_PFLASH] = "pflash",
56     [IF_MTD] = "mtd",
57     [IF_SD] = "sd",
58     [IF_VIRTIO] = "virtio",
59     [IF_XEN] = "xen",
60 };
61 
62 static const int if_max_devs[IF_COUNT] = {
63     /*
64      * Do not change these numbers!  They govern how drive option
65      * index maps to unit and bus.  That mapping is ABI.
66      *
67      * All controllers used to imlement if=T drives need to support
68      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
69      * Otherwise, some index values map to "impossible" bus, unit
70      * values.
71      *
72      * For instance, if you change [IF_SCSI] to 255, -drive
73      * if=scsi,index=12 no longer means bus=1,unit=5, but
74      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
75      * the drive can't be set up.  Regression.
76      */
77     [IF_IDE] = 2,
78     [IF_SCSI] = 7,
79 };
80 
81 /*
82  * We automatically delete the drive when a device using it gets
83  * unplugged.  Questionable feature, but we can't just drop it.
84  * Device models call blockdev_mark_auto_del() to schedule the
85  * automatic deletion, and generic qdev code calls blockdev_auto_del()
86  * when deletion is actually safe.
87  */
88 void blockdev_mark_auto_del(BlockDriverState *bs)
89 {
90     DriveInfo *dinfo = drive_get_by_blockdev(bs);
91 
92     if (dinfo && !dinfo->enable_auto_del) {
93         return;
94     }
95 
96     if (bs->job) {
97         block_job_cancel(bs->job);
98     }
99     if (dinfo) {
100         dinfo->auto_del = 1;
101     }
102 }
103 
104 void blockdev_auto_del(BlockDriverState *bs)
105 {
106     DriveInfo *dinfo = drive_get_by_blockdev(bs);
107 
108     if (dinfo && dinfo->auto_del) {
109         drive_put_ref(dinfo);
110     }
111 }
112 
113 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
114 {
115     int max_devs = if_max_devs[type];
116     return max_devs ? index / max_devs : 0;
117 }
118 
119 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
120 {
121     int max_devs = if_max_devs[type];
122     return max_devs ? index % max_devs : index;
123 }
124 
125 QemuOpts *drive_def(const char *optstr)
126 {
127     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
128 }
129 
130 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
131                     const char *optstr)
132 {
133     QemuOpts *opts;
134     char buf[32];
135 
136     opts = drive_def(optstr);
137     if (!opts) {
138         return NULL;
139     }
140     if (type != IF_DEFAULT) {
141         qemu_opt_set(opts, "if", if_name[type]);
142     }
143     if (index >= 0) {
144         snprintf(buf, sizeof(buf), "%d", index);
145         qemu_opt_set(opts, "index", buf);
146     }
147     if (file)
148         qemu_opt_set(opts, "file", file);
149     return opts;
150 }
151 
152 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
153 {
154     DriveInfo *dinfo;
155 
156     /* seek interface, bus and unit */
157 
158     QTAILQ_FOREACH(dinfo, &drives, next) {
159         if (dinfo->type == type &&
160 	    dinfo->bus == bus &&
161 	    dinfo->unit == unit)
162             return dinfo;
163     }
164 
165     return NULL;
166 }
167 
168 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
169 {
170     return drive_get(type,
171                      drive_index_to_bus_id(type, index),
172                      drive_index_to_unit_id(type, index));
173 }
174 
175 int drive_get_max_bus(BlockInterfaceType type)
176 {
177     int max_bus;
178     DriveInfo *dinfo;
179 
180     max_bus = -1;
181     QTAILQ_FOREACH(dinfo, &drives, next) {
182         if(dinfo->type == type &&
183            dinfo->bus > max_bus)
184             max_bus = dinfo->bus;
185     }
186     return max_bus;
187 }
188 
189 /* Get a block device.  This should only be used for single-drive devices
190    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
191    appropriate bus.  */
192 DriveInfo *drive_get_next(BlockInterfaceType type)
193 {
194     static int next_block_unit[IF_COUNT];
195 
196     return drive_get(type, 0, next_block_unit[type]++);
197 }
198 
199 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
200 {
201     DriveInfo *dinfo;
202 
203     QTAILQ_FOREACH(dinfo, &drives, next) {
204         if (dinfo->bdrv == bs) {
205             return dinfo;
206         }
207     }
208     return NULL;
209 }
210 
211 static void bdrv_format_print(void *opaque, const char *name)
212 {
213     error_printf(" %s", name);
214 }
215 
216 static void drive_uninit(DriveInfo *dinfo)
217 {
218     if (dinfo->opts) {
219         qemu_opts_del(dinfo->opts);
220     }
221 
222     bdrv_unref(dinfo->bdrv);
223     g_free(dinfo->id);
224     QTAILQ_REMOVE(&drives, dinfo, next);
225     g_free(dinfo->serial);
226     g_free(dinfo);
227 }
228 
229 void drive_put_ref(DriveInfo *dinfo)
230 {
231     assert(dinfo->refcount);
232     if (--dinfo->refcount == 0) {
233         drive_uninit(dinfo);
234     }
235 }
236 
237 void drive_get_ref(DriveInfo *dinfo)
238 {
239     dinfo->refcount++;
240 }
241 
242 typedef struct {
243     QEMUBH *bh;
244     BlockDriverState *bs;
245 } BDRVPutRefBH;
246 
247 static void bdrv_put_ref_bh(void *opaque)
248 {
249     BDRVPutRefBH *s = opaque;
250 
251     bdrv_unref(s->bs);
252     qemu_bh_delete(s->bh);
253     g_free(s);
254 }
255 
256 /*
257  * Release a BDS reference in a BH
258  *
259  * It is not safe to use bdrv_unref() from a callback function when the callers
260  * still need the BlockDriverState.  In such cases we schedule a BH to release
261  * the reference.
262  */
263 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
264 {
265     BDRVPutRefBH *s;
266 
267     s = g_new(BDRVPutRefBH, 1);
268     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
269     s->bs = bs;
270     qemu_bh_schedule(s->bh);
271 }
272 
273 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
274 {
275     if (!strcmp(buf, "ignore")) {
276         return BLOCKDEV_ON_ERROR_IGNORE;
277     } else if (!is_read && !strcmp(buf, "enospc")) {
278         return BLOCKDEV_ON_ERROR_ENOSPC;
279     } else if (!strcmp(buf, "stop")) {
280         return BLOCKDEV_ON_ERROR_STOP;
281     } else if (!strcmp(buf, "report")) {
282         return BLOCKDEV_ON_ERROR_REPORT;
283     } else {
284         error_setg(errp, "'%s' invalid %s error action",
285                    buf, is_read ? "read" : "write");
286         return -1;
287     }
288 }
289 
290 static inline int parse_enum_option(const char *lookup[], const char *buf,
291                                     int max, int def, Error **errp)
292 {
293     int i;
294 
295     if (!buf) {
296         return def;
297     }
298 
299     for (i = 0; i < max; i++) {
300         if (!strcmp(buf, lookup[i])) {
301             return i;
302         }
303     }
304 
305     error_setg(errp, "invalid parameter value: %s", buf);
306     return def;
307 }
308 
309 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
310 {
311     if (throttle_conflicting(cfg)) {
312         error_setg(errp, "bps/iops/max total values and read/write values"
313                          " cannot be used at the same time");
314         return false;
315     }
316 
317     if (!throttle_is_valid(cfg)) {
318         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
319         return false;
320     }
321 
322     return true;
323 }
324 
325 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
326 
327 /* Takes the ownership of bs_opts */
328 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
329                                 Error **errp)
330 {
331     const char *buf;
332     const char *serial;
333     int ro = 0;
334     int bdrv_flags = 0;
335     int on_read_error, on_write_error;
336     DriveInfo *dinfo;
337     ThrottleConfig cfg;
338     int snapshot = 0;
339     bool copy_on_read;
340     int ret;
341     Error *error = NULL;
342     QemuOpts *opts;
343     const char *id;
344     bool has_driver_specific_opts;
345     BlockdevDetectZeroesOptions detect_zeroes;
346     BlockDriver *drv = NULL;
347 
348     /* Check common options by copying from bs_opts to opts, all other options
349      * stay in bs_opts for processing by bdrv_open(). */
350     id = qdict_get_try_str(bs_opts, "id");
351     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
352     if (error) {
353         error_propagate(errp, error);
354         goto err_no_opts;
355     }
356 
357     qemu_opts_absorb_qdict(opts, bs_opts, &error);
358     if (error) {
359         error_propagate(errp, error);
360         goto early_err;
361     }
362 
363     if (id) {
364         qdict_del(bs_opts, "id");
365     }
366 
367     has_driver_specific_opts = !!qdict_size(bs_opts);
368 
369     /* extract parameters */
370     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
371     ro = qemu_opt_get_bool(opts, "read-only", 0);
372     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
373 
374     serial = qemu_opt_get(opts, "serial");
375 
376     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
377         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
378             error_setg(errp, "invalid discard option");
379             goto early_err;
380         }
381     }
382 
383     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
384         bdrv_flags |= BDRV_O_CACHE_WB;
385     }
386     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
387         bdrv_flags |= BDRV_O_NOCACHE;
388     }
389     if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
390         bdrv_flags |= BDRV_O_NO_FLUSH;
391     }
392 
393 #ifdef CONFIG_LINUX_AIO
394     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
395         if (!strcmp(buf, "native")) {
396             bdrv_flags |= BDRV_O_NATIVE_AIO;
397         } else if (!strcmp(buf, "threads")) {
398             /* this is the default */
399         } else {
400            error_setg(errp, "invalid aio option");
401            goto early_err;
402         }
403     }
404 #endif
405 
406     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
407         if (is_help_option(buf)) {
408             error_printf("Supported formats:");
409             bdrv_iterate_format(bdrv_format_print, NULL);
410             error_printf("\n");
411             goto early_err;
412         }
413 
414         drv = bdrv_find_format(buf);
415         if (!drv) {
416             error_setg(errp, "'%s' invalid format", buf);
417             goto early_err;
418         }
419     }
420 
421     /* disk I/O throttling */
422     memset(&cfg, 0, sizeof(cfg));
423     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
424         qemu_opt_get_number(opts, "throttling.bps-total", 0);
425     cfg.buckets[THROTTLE_BPS_READ].avg  =
426         qemu_opt_get_number(opts, "throttling.bps-read", 0);
427     cfg.buckets[THROTTLE_BPS_WRITE].avg =
428         qemu_opt_get_number(opts, "throttling.bps-write", 0);
429     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
430         qemu_opt_get_number(opts, "throttling.iops-total", 0);
431     cfg.buckets[THROTTLE_OPS_READ].avg =
432         qemu_opt_get_number(opts, "throttling.iops-read", 0);
433     cfg.buckets[THROTTLE_OPS_WRITE].avg =
434         qemu_opt_get_number(opts, "throttling.iops-write", 0);
435 
436     cfg.buckets[THROTTLE_BPS_TOTAL].max =
437         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
438     cfg.buckets[THROTTLE_BPS_READ].max  =
439         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
440     cfg.buckets[THROTTLE_BPS_WRITE].max =
441         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
442     cfg.buckets[THROTTLE_OPS_TOTAL].max =
443         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
444     cfg.buckets[THROTTLE_OPS_READ].max =
445         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
446     cfg.buckets[THROTTLE_OPS_WRITE].max =
447         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
448 
449     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
450 
451     if (!check_throttle_config(&cfg, &error)) {
452         error_propagate(errp, error);
453         goto early_err;
454     }
455 
456     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
457     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
458         on_write_error = parse_block_error_action(buf, 0, &error);
459         if (error) {
460             error_propagate(errp, error);
461             goto early_err;
462         }
463     }
464 
465     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
466     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
467         on_read_error = parse_block_error_action(buf, 1, &error);
468         if (error) {
469             error_propagate(errp, error);
470             goto early_err;
471         }
472     }
473 
474     detect_zeroes =
475         parse_enum_option(BlockdevDetectZeroesOptions_lookup,
476                           qemu_opt_get(opts, "detect-zeroes"),
477                           BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
478                           BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
479                           &error);
480     if (error) {
481         error_propagate(errp, error);
482         goto early_err;
483     }
484 
485     if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
486         !(bdrv_flags & BDRV_O_UNMAP)) {
487         error_setg(errp, "setting detect-zeroes to unmap is not allowed "
488                          "without setting discard operation to unmap");
489         goto early_err;
490     }
491 
492     /* init */
493     dinfo = g_malloc0(sizeof(*dinfo));
494     dinfo->id = g_strdup(qemu_opts_id(opts));
495     dinfo->bdrv = bdrv_new(dinfo->id, &error);
496     if (error) {
497         error_propagate(errp, error);
498         goto bdrv_new_err;
499     }
500     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
501     dinfo->bdrv->read_only = ro;
502     dinfo->bdrv->detect_zeroes = detect_zeroes;
503     dinfo->refcount = 1;
504     if (serial != NULL) {
505         dinfo->serial = g_strdup(serial);
506     }
507     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
508 
509     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
510 
511     /* disk I/O throttling */
512     if (throttle_enabled(&cfg)) {
513         bdrv_io_limits_enable(dinfo->bdrv);
514         bdrv_set_io_limits(dinfo->bdrv, &cfg);
515     }
516 
517     if (!file || !*file) {
518         if (has_driver_specific_opts) {
519             file = NULL;
520         } else {
521             QDECREF(bs_opts);
522             qemu_opts_del(opts);
523             return dinfo;
524         }
525     }
526     if (snapshot) {
527         /* always use cache=unsafe with snapshot */
528         bdrv_flags &= ~BDRV_O_CACHE_MASK;
529         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
530     }
531 
532     if (copy_on_read) {
533         bdrv_flags |= BDRV_O_COPY_ON_READ;
534     }
535 
536     if (runstate_check(RUN_STATE_INMIGRATE)) {
537         bdrv_flags |= BDRV_O_INCOMING;
538     }
539 
540     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
541 
542     QINCREF(bs_opts);
543     ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
544 
545     if (ret < 0) {
546         error_setg(errp, "could not open disk image %s: %s",
547                    file ?: dinfo->id, error_get_pretty(error));
548         error_free(error);
549         goto err;
550     }
551 
552     if (bdrv_key_required(dinfo->bdrv))
553         autostart = 0;
554 
555     QDECREF(bs_opts);
556     qemu_opts_del(opts);
557 
558     return dinfo;
559 
560 err:
561     bdrv_unref(dinfo->bdrv);
562     QTAILQ_REMOVE(&drives, dinfo, next);
563 bdrv_new_err:
564     g_free(dinfo->id);
565     g_free(dinfo);
566 early_err:
567     qemu_opts_del(opts);
568 err_no_opts:
569     QDECREF(bs_opts);
570     return NULL;
571 }
572 
573 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
574 {
575     const char *value;
576 
577     value = qemu_opt_get(opts, from);
578     if (value) {
579         qemu_opt_set(opts, to, value);
580         qemu_opt_unset(opts, from);
581     }
582 }
583 
584 QemuOptsList qemu_legacy_drive_opts = {
585     .name = "drive",
586     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
587     .desc = {
588         {
589             .name = "bus",
590             .type = QEMU_OPT_NUMBER,
591             .help = "bus number",
592         },{
593             .name = "unit",
594             .type = QEMU_OPT_NUMBER,
595             .help = "unit number (i.e. lun for scsi)",
596         },{
597             .name = "index",
598             .type = QEMU_OPT_NUMBER,
599             .help = "index number",
600         },{
601             .name = "media",
602             .type = QEMU_OPT_STRING,
603             .help = "media type (disk, cdrom)",
604         },{
605             .name = "if",
606             .type = QEMU_OPT_STRING,
607             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
608         },{
609             .name = "cyls",
610             .type = QEMU_OPT_NUMBER,
611             .help = "number of cylinders (ide disk geometry)",
612         },{
613             .name = "heads",
614             .type = QEMU_OPT_NUMBER,
615             .help = "number of heads (ide disk geometry)",
616         },{
617             .name = "secs",
618             .type = QEMU_OPT_NUMBER,
619             .help = "number of sectors (ide disk geometry)",
620         },{
621             .name = "trans",
622             .type = QEMU_OPT_STRING,
623             .help = "chs translation (auto, lba, none)",
624         },{
625             .name = "boot",
626             .type = QEMU_OPT_BOOL,
627             .help = "(deprecated, ignored)",
628         },{
629             .name = "addr",
630             .type = QEMU_OPT_STRING,
631             .help = "pci address (virtio only)",
632         },{
633             .name = "file",
634             .type = QEMU_OPT_STRING,
635             .help = "file name",
636         },
637 
638         /* Options that are passed on, but have special semantics with -drive */
639         {
640             .name = "read-only",
641             .type = QEMU_OPT_BOOL,
642             .help = "open drive file as read-only",
643         },{
644             .name = "rerror",
645             .type = QEMU_OPT_STRING,
646             .help = "read error action",
647         },{
648             .name = "werror",
649             .type = QEMU_OPT_STRING,
650             .help = "write error action",
651         },{
652             .name = "copy-on-read",
653             .type = QEMU_OPT_BOOL,
654             .help = "copy read data from backing file into image file",
655         },
656 
657         { /* end of list */ }
658     },
659 };
660 
661 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
662 {
663     const char *value;
664     DriveInfo *dinfo = NULL;
665     QDict *bs_opts;
666     QemuOpts *legacy_opts;
667     DriveMediaType media = MEDIA_DISK;
668     BlockInterfaceType type;
669     int cyls, heads, secs, translation;
670     int max_devs, bus_id, unit_id, index;
671     const char *devaddr;
672     const char *werror, *rerror;
673     bool read_only = false;
674     bool copy_on_read;
675     const char *filename;
676     Error *local_err = NULL;
677 
678     /* Change legacy command line options into QMP ones */
679     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
680     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
681     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
682 
683     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
684     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
685     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
686 
687     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
688     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
689     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
690 
691     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
692     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
693     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
694 
695     qemu_opt_rename(all_opts,
696                     "iops_size", "throttling.iops-size");
697 
698     qemu_opt_rename(all_opts, "readonly", "read-only");
699 
700     value = qemu_opt_get(all_opts, "cache");
701     if (value) {
702         int flags = 0;
703 
704         if (bdrv_parse_cache_flags(value, &flags) != 0) {
705             error_report("invalid cache option");
706             return NULL;
707         }
708 
709         /* Specific options take precedence */
710         if (!qemu_opt_get(all_opts, "cache.writeback")) {
711             qemu_opt_set_bool(all_opts, "cache.writeback",
712                               !!(flags & BDRV_O_CACHE_WB));
713         }
714         if (!qemu_opt_get(all_opts, "cache.direct")) {
715             qemu_opt_set_bool(all_opts, "cache.direct",
716                               !!(flags & BDRV_O_NOCACHE));
717         }
718         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
719             qemu_opt_set_bool(all_opts, "cache.no-flush",
720                               !!(flags & BDRV_O_NO_FLUSH));
721         }
722         qemu_opt_unset(all_opts, "cache");
723     }
724 
725     /* Get a QDict for processing the options */
726     bs_opts = qdict_new();
727     qemu_opts_to_qdict(all_opts, bs_opts);
728 
729     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
730                                    &error_abort);
731     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
732     if (local_err) {
733         error_report("%s", error_get_pretty(local_err));
734         error_free(local_err);
735         goto fail;
736     }
737 
738     /* Deprecated option boot=[on|off] */
739     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
740         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
741                 "ignored. Future versions will reject this parameter. Please "
742                 "update your scripts.\n");
743     }
744 
745     /* Media type */
746     value = qemu_opt_get(legacy_opts, "media");
747     if (value) {
748         if (!strcmp(value, "disk")) {
749             media = MEDIA_DISK;
750         } else if (!strcmp(value, "cdrom")) {
751             media = MEDIA_CDROM;
752             read_only = true;
753         } else {
754             error_report("'%s' invalid media", value);
755             goto fail;
756         }
757     }
758 
759     /* copy-on-read is disabled with a warning for read-only devices */
760     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
761     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
762 
763     if (read_only && copy_on_read) {
764         error_report("warning: disabling copy-on-read on read-only drive");
765         copy_on_read = false;
766     }
767 
768     qdict_put(bs_opts, "read-only",
769               qstring_from_str(read_only ? "on" : "off"));
770     qdict_put(bs_opts, "copy-on-read",
771               qstring_from_str(copy_on_read ? "on" :"off"));
772 
773     /* Controller type */
774     value = qemu_opt_get(legacy_opts, "if");
775     if (value) {
776         for (type = 0;
777              type < IF_COUNT && strcmp(value, if_name[type]);
778              type++) {
779         }
780         if (type == IF_COUNT) {
781             error_report("unsupported bus type '%s'", value);
782             goto fail;
783         }
784     } else {
785         type = block_default_type;
786     }
787 
788     /* Geometry */
789     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
790     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
791     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
792 
793     if (cyls || heads || secs) {
794         if (cyls < 1) {
795             error_report("invalid physical cyls number");
796             goto fail;
797         }
798         if (heads < 1) {
799             error_report("invalid physical heads number");
800             goto fail;
801         }
802         if (secs < 1) {
803             error_report("invalid physical secs number");
804             goto fail;
805         }
806     }
807 
808     translation = BIOS_ATA_TRANSLATION_AUTO;
809     value = qemu_opt_get(legacy_opts, "trans");
810     if (value != NULL) {
811         if (!cyls) {
812             error_report("'%s' trans must be used with cyls, heads and secs",
813                          value);
814             goto fail;
815         }
816         if (!strcmp(value, "none")) {
817             translation = BIOS_ATA_TRANSLATION_NONE;
818         } else if (!strcmp(value, "lba")) {
819             translation = BIOS_ATA_TRANSLATION_LBA;
820         } else if (!strcmp(value, "large")) {
821             translation = BIOS_ATA_TRANSLATION_LARGE;
822         } else if (!strcmp(value, "rechs")) {
823             translation = BIOS_ATA_TRANSLATION_RECHS;
824         } else if (!strcmp(value, "auto")) {
825             translation = BIOS_ATA_TRANSLATION_AUTO;
826         } else {
827             error_report("'%s' invalid translation type", value);
828             goto fail;
829         }
830     }
831 
832     if (media == MEDIA_CDROM) {
833         if (cyls || secs || heads) {
834             error_report("CHS can't be set with media=cdrom");
835             goto fail;
836         }
837     }
838 
839     /* Device address specified by bus/unit or index.
840      * If none was specified, try to find the first free one. */
841     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
842     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
843     index   = qemu_opt_get_number(legacy_opts, "index", -1);
844 
845     max_devs = if_max_devs[type];
846 
847     if (index != -1) {
848         if (bus_id != 0 || unit_id != -1) {
849             error_report("index cannot be used with bus and unit");
850             goto fail;
851         }
852         bus_id = drive_index_to_bus_id(type, index);
853         unit_id = drive_index_to_unit_id(type, index);
854     }
855 
856     if (unit_id == -1) {
857        unit_id = 0;
858        while (drive_get(type, bus_id, unit_id) != NULL) {
859            unit_id++;
860            if (max_devs && unit_id >= max_devs) {
861                unit_id -= max_devs;
862                bus_id++;
863            }
864        }
865     }
866 
867     if (max_devs && unit_id >= max_devs) {
868         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
869         goto fail;
870     }
871 
872     if (drive_get(type, bus_id, unit_id) != NULL) {
873         error_report("drive with bus=%d, unit=%d (index=%d) exists",
874                      bus_id, unit_id, index);
875         goto fail;
876     }
877 
878     /* no id supplied -> create one */
879     if (qemu_opts_id(all_opts) == NULL) {
880         char *new_id;
881         const char *mediastr = "";
882         if (type == IF_IDE || type == IF_SCSI) {
883             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
884         }
885         if (max_devs) {
886             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
887                                      mediastr, unit_id);
888         } else {
889             new_id = g_strdup_printf("%s%s%i", if_name[type],
890                                      mediastr, unit_id);
891         }
892         qdict_put(bs_opts, "id", qstring_from_str(new_id));
893         g_free(new_id);
894     }
895 
896     /* Add virtio block device */
897     devaddr = qemu_opt_get(legacy_opts, "addr");
898     if (devaddr && type != IF_VIRTIO) {
899         error_report("addr is not supported by this bus type");
900         goto fail;
901     }
902 
903     if (type == IF_VIRTIO) {
904         QemuOpts *devopts;
905         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
906                                    &error_abort);
907         if (arch_type == QEMU_ARCH_S390X) {
908             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
909         } else {
910             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
911         }
912         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
913         if (devaddr) {
914             qemu_opt_set(devopts, "addr", devaddr);
915         }
916     }
917 
918     filename = qemu_opt_get(legacy_opts, "file");
919 
920     /* Check werror/rerror compatibility with if=... */
921     werror = qemu_opt_get(legacy_opts, "werror");
922     if (werror != NULL) {
923         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
924             type != IF_NONE) {
925             error_report("werror is not supported by this bus type");
926             goto fail;
927         }
928         qdict_put(bs_opts, "werror", qstring_from_str(werror));
929     }
930 
931     rerror = qemu_opt_get(legacy_opts, "rerror");
932     if (rerror != NULL) {
933         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
934             type != IF_NONE) {
935             error_report("rerror is not supported by this bus type");
936             goto fail;
937         }
938         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
939     }
940 
941     /* Actual block device init: Functionality shared with blockdev-add */
942     dinfo = blockdev_init(filename, bs_opts, &local_err);
943     bs_opts = NULL;
944     if (dinfo == NULL) {
945         if (local_err) {
946             error_report("%s", error_get_pretty(local_err));
947             error_free(local_err);
948         }
949         goto fail;
950     } else {
951         assert(!local_err);
952     }
953 
954     /* Set legacy DriveInfo fields */
955     dinfo->enable_auto_del = true;
956     dinfo->opts = all_opts;
957 
958     dinfo->cyls = cyls;
959     dinfo->heads = heads;
960     dinfo->secs = secs;
961     dinfo->trans = translation;
962 
963     dinfo->type = type;
964     dinfo->bus = bus_id;
965     dinfo->unit = unit_id;
966     dinfo->devaddr = devaddr;
967 
968     switch(type) {
969     case IF_IDE:
970     case IF_SCSI:
971     case IF_XEN:
972     case IF_NONE:
973         dinfo->media_cd = media == MEDIA_CDROM;
974         break;
975     default:
976         break;
977     }
978 
979 fail:
980     qemu_opts_del(legacy_opts);
981     QDECREF(bs_opts);
982     return dinfo;
983 }
984 
985 void do_commit(Monitor *mon, const QDict *qdict)
986 {
987     const char *device = qdict_get_str(qdict, "device");
988     BlockDriverState *bs;
989     int ret;
990 
991     if (!strcmp(device, "all")) {
992         ret = bdrv_commit_all();
993     } else {
994         bs = bdrv_find(device);
995         if (!bs) {
996             monitor_printf(mon, "Device '%s' not found\n", device);
997             return;
998         }
999         ret = bdrv_commit(bs);
1000     }
1001     if (ret < 0) {
1002         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1003                        strerror(-ret));
1004     }
1005 }
1006 
1007 static void blockdev_do_action(int kind, void *data, Error **errp)
1008 {
1009     TransactionAction action;
1010     TransactionActionList list;
1011 
1012     action.kind = kind;
1013     action.data = data;
1014     list.value = &action;
1015     list.next = NULL;
1016     qmp_transaction(&list, errp);
1017 }
1018 
1019 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1020                                 bool has_node_name, const char *node_name,
1021                                 const char *snapshot_file,
1022                                 bool has_snapshot_node_name,
1023                                 const char *snapshot_node_name,
1024                                 bool has_format, const char *format,
1025                                 bool has_mode, NewImageMode mode, Error **errp)
1026 {
1027     BlockdevSnapshot snapshot = {
1028         .has_device = has_device,
1029         .device = (char *) device,
1030         .has_node_name = has_node_name,
1031         .node_name = (char *) node_name,
1032         .snapshot_file = (char *) snapshot_file,
1033         .has_snapshot_node_name = has_snapshot_node_name,
1034         .snapshot_node_name = (char *) snapshot_node_name,
1035         .has_format = has_format,
1036         .format = (char *) format,
1037         .has_mode = has_mode,
1038         .mode = mode,
1039     };
1040     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1041                        &snapshot, errp);
1042 }
1043 
1044 void qmp_blockdev_snapshot_internal_sync(const char *device,
1045                                          const char *name,
1046                                          Error **errp)
1047 {
1048     BlockdevSnapshotInternal snapshot = {
1049         .device = (char *) device,
1050         .name = (char *) name
1051     };
1052 
1053     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1054                        &snapshot, errp);
1055 }
1056 
1057 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1058                                                          bool has_id,
1059                                                          const char *id,
1060                                                          bool has_name,
1061                                                          const char *name,
1062                                                          Error **errp)
1063 {
1064     BlockDriverState *bs = bdrv_find(device);
1065     QEMUSnapshotInfo sn;
1066     Error *local_err = NULL;
1067     SnapshotInfo *info = NULL;
1068     int ret;
1069 
1070     if (!bs) {
1071         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1072         return NULL;
1073     }
1074 
1075     if (!has_id) {
1076         id = NULL;
1077     }
1078 
1079     if (!has_name) {
1080         name = NULL;
1081     }
1082 
1083     if (!id && !name) {
1084         error_setg(errp, "Name or id must be provided");
1085         return NULL;
1086     }
1087 
1088     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1089     if (local_err) {
1090         error_propagate(errp, local_err);
1091         return NULL;
1092     }
1093     if (!ret) {
1094         error_setg(errp,
1095                    "Snapshot with id '%s' and name '%s' does not exist on "
1096                    "device '%s'",
1097                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1098         return NULL;
1099     }
1100 
1101     bdrv_snapshot_delete(bs, id, name, &local_err);
1102     if (local_err) {
1103         error_propagate(errp, local_err);
1104         return NULL;
1105     }
1106 
1107     info = g_malloc0(sizeof(SnapshotInfo));
1108     info->id = g_strdup(sn.id_str);
1109     info->name = g_strdup(sn.name);
1110     info->date_nsec = sn.date_nsec;
1111     info->date_sec = sn.date_sec;
1112     info->vm_state_size = sn.vm_state_size;
1113     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1114     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1115 
1116     return info;
1117 }
1118 
1119 /* New and old BlockDriverState structs for group snapshots */
1120 
1121 typedef struct BlkTransactionState BlkTransactionState;
1122 
1123 /* Only prepare() may fail. In a single transaction, only one of commit() or
1124    abort() will be called, clean() will always be called if it present. */
1125 typedef struct BdrvActionOps {
1126     /* Size of state struct, in bytes. */
1127     size_t instance_size;
1128     /* Prepare the work, must NOT be NULL. */
1129     void (*prepare)(BlkTransactionState *common, Error **errp);
1130     /* Commit the changes, can be NULL. */
1131     void (*commit)(BlkTransactionState *common);
1132     /* Abort the changes on fail, can be NULL. */
1133     void (*abort)(BlkTransactionState *common);
1134     /* Clean up resource in the end, can be NULL. */
1135     void (*clean)(BlkTransactionState *common);
1136 } BdrvActionOps;
1137 
1138 /*
1139  * This structure must be arranged as first member in child type, assuming
1140  * that compiler will also arrange it to the same address with parent instance.
1141  * Later it will be used in free().
1142  */
1143 struct BlkTransactionState {
1144     TransactionAction *action;
1145     const BdrvActionOps *ops;
1146     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1147 };
1148 
1149 /* internal snapshot private data */
1150 typedef struct InternalSnapshotState {
1151     BlkTransactionState common;
1152     BlockDriverState *bs;
1153     QEMUSnapshotInfo sn;
1154 } InternalSnapshotState;
1155 
1156 static void internal_snapshot_prepare(BlkTransactionState *common,
1157                                       Error **errp)
1158 {
1159     Error *local_err = NULL;
1160     const char *device;
1161     const char *name;
1162     BlockDriverState *bs;
1163     QEMUSnapshotInfo old_sn, *sn;
1164     bool ret;
1165     qemu_timeval tv;
1166     BlockdevSnapshotInternal *internal;
1167     InternalSnapshotState *state;
1168     int ret1;
1169 
1170     g_assert(common->action->kind ==
1171              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1172     internal = common->action->blockdev_snapshot_internal_sync;
1173     state = DO_UPCAST(InternalSnapshotState, common, common);
1174 
1175     /* 1. parse input */
1176     device = internal->device;
1177     name = internal->name;
1178 
1179     /* 2. check for validation */
1180     bs = bdrv_find(device);
1181     if (!bs) {
1182         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1183         return;
1184     }
1185 
1186     if (!bdrv_is_inserted(bs)) {
1187         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1188         return;
1189     }
1190 
1191     if (bdrv_is_read_only(bs)) {
1192         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1193         return;
1194     }
1195 
1196     if (!bdrv_can_snapshot(bs)) {
1197         error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1198                   bs->drv->format_name, device, "internal snapshot");
1199         return;
1200     }
1201 
1202     if (!strlen(name)) {
1203         error_setg(errp, "Name is empty");
1204         return;
1205     }
1206 
1207     /* check whether a snapshot with name exist */
1208     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1209                                             &local_err);
1210     if (local_err) {
1211         error_propagate(errp, local_err);
1212         return;
1213     } else if (ret) {
1214         error_setg(errp,
1215                    "Snapshot with name '%s' already exists on device '%s'",
1216                    name, device);
1217         return;
1218     }
1219 
1220     /* 3. take the snapshot */
1221     sn = &state->sn;
1222     pstrcpy(sn->name, sizeof(sn->name), name);
1223     qemu_gettimeofday(&tv);
1224     sn->date_sec = tv.tv_sec;
1225     sn->date_nsec = tv.tv_usec * 1000;
1226     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1227 
1228     ret1 = bdrv_snapshot_create(bs, sn);
1229     if (ret1 < 0) {
1230         error_setg_errno(errp, -ret1,
1231                          "Failed to create snapshot '%s' on device '%s'",
1232                          name, device);
1233         return;
1234     }
1235 
1236     /* 4. succeed, mark a snapshot is created */
1237     state->bs = bs;
1238 }
1239 
1240 static void internal_snapshot_abort(BlkTransactionState *common)
1241 {
1242     InternalSnapshotState *state =
1243                              DO_UPCAST(InternalSnapshotState, common, common);
1244     BlockDriverState *bs = state->bs;
1245     QEMUSnapshotInfo *sn = &state->sn;
1246     Error *local_error = NULL;
1247 
1248     if (!bs) {
1249         return;
1250     }
1251 
1252     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1253         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1254                      "device '%s' in abort: %s",
1255                      sn->id_str,
1256                      sn->name,
1257                      bdrv_get_device_name(bs),
1258                      error_get_pretty(local_error));
1259         error_free(local_error);
1260     }
1261 }
1262 
1263 /* external snapshot private data */
1264 typedef struct ExternalSnapshotState {
1265     BlkTransactionState common;
1266     BlockDriverState *old_bs;
1267     BlockDriverState *new_bs;
1268 } ExternalSnapshotState;
1269 
1270 static void external_snapshot_prepare(BlkTransactionState *common,
1271                                       Error **errp)
1272 {
1273     BlockDriver *drv;
1274     int flags, ret;
1275     QDict *options = NULL;
1276     Error *local_err = NULL;
1277     bool has_device = false;
1278     const char *device;
1279     bool has_node_name = false;
1280     const char *node_name;
1281     bool has_snapshot_node_name = false;
1282     const char *snapshot_node_name;
1283     const char *new_image_file;
1284     const char *format = "qcow2";
1285     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1286     ExternalSnapshotState *state =
1287                              DO_UPCAST(ExternalSnapshotState, common, common);
1288     TransactionAction *action = common->action;
1289 
1290     /* get parameters */
1291     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1292 
1293     has_device = action->blockdev_snapshot_sync->has_device;
1294     device = action->blockdev_snapshot_sync->device;
1295     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1296     node_name = action->blockdev_snapshot_sync->node_name;
1297     has_snapshot_node_name =
1298         action->blockdev_snapshot_sync->has_snapshot_node_name;
1299     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1300 
1301     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1302     if (action->blockdev_snapshot_sync->has_format) {
1303         format = action->blockdev_snapshot_sync->format;
1304     }
1305     if (action->blockdev_snapshot_sync->has_mode) {
1306         mode = action->blockdev_snapshot_sync->mode;
1307     }
1308 
1309     /* start processing */
1310     drv = bdrv_find_format(format);
1311     if (!drv) {
1312         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1313         return;
1314     }
1315 
1316     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1317                                    has_node_name ? node_name : NULL,
1318                                    &local_err);
1319     if (local_err) {
1320         error_propagate(errp, local_err);
1321         return;
1322     }
1323 
1324     if (has_node_name && !has_snapshot_node_name) {
1325         error_setg(errp, "New snapshot node name missing");
1326         return;
1327     }
1328 
1329     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1330         error_setg(errp, "New snapshot node name already existing");
1331         return;
1332     }
1333 
1334     if (!bdrv_is_inserted(state->old_bs)) {
1335         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1336         return;
1337     }
1338 
1339     if (bdrv_op_is_blocked(state->old_bs,
1340                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1341         return;
1342     }
1343 
1344     if (!bdrv_is_read_only(state->old_bs)) {
1345         if (bdrv_flush(state->old_bs)) {
1346             error_set(errp, QERR_IO_ERROR);
1347             return;
1348         }
1349     }
1350 
1351     if (!bdrv_is_first_non_filter(state->old_bs)) {
1352         error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1353         return;
1354     }
1355 
1356     flags = state->old_bs->open_flags;
1357 
1358     /* create new image w/backing file */
1359     if (mode != NEW_IMAGE_MODE_EXISTING) {
1360         bdrv_img_create(new_image_file, format,
1361                         state->old_bs->filename,
1362                         state->old_bs->drv->format_name,
1363                         NULL, -1, flags, &local_err, false);
1364         if (local_err) {
1365             error_propagate(errp, local_err);
1366             return;
1367         }
1368     }
1369 
1370     if (has_snapshot_node_name) {
1371         options = qdict_new();
1372         qdict_put(options, "node-name",
1373                   qstring_from_str(snapshot_node_name));
1374     }
1375 
1376     /* TODO Inherit bs->options or only take explicit options with an
1377      * extended QMP command? */
1378     assert(state->new_bs == NULL);
1379     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1380                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1381     /* We will manually add the backing_hd field to the bs later */
1382     if (ret != 0) {
1383         error_propagate(errp, local_err);
1384     }
1385 }
1386 
1387 static void external_snapshot_commit(BlkTransactionState *common)
1388 {
1389     ExternalSnapshotState *state =
1390                              DO_UPCAST(ExternalSnapshotState, common, common);
1391 
1392     /* This removes our old bs and adds the new bs */
1393     bdrv_append(state->new_bs, state->old_bs);
1394     /* We don't need (or want) to use the transactional
1395      * bdrv_reopen_multiple() across all the entries at once, because we
1396      * don't want to abort all of them if one of them fails the reopen */
1397     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1398                 NULL);
1399 }
1400 
1401 static void external_snapshot_abort(BlkTransactionState *common)
1402 {
1403     ExternalSnapshotState *state =
1404                              DO_UPCAST(ExternalSnapshotState, common, common);
1405     if (state->new_bs) {
1406         bdrv_unref(state->new_bs);
1407     }
1408 }
1409 
1410 typedef struct DriveBackupState {
1411     BlkTransactionState common;
1412     BlockDriverState *bs;
1413     BlockJob *job;
1414 } DriveBackupState;
1415 
1416 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1417 {
1418     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1419     DriveBackup *backup;
1420     Error *local_err = NULL;
1421 
1422     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1423     backup = common->action->drive_backup;
1424 
1425     qmp_drive_backup(backup->device, backup->target,
1426                      backup->has_format, backup->format,
1427                      backup->sync,
1428                      backup->has_mode, backup->mode,
1429                      backup->has_speed, backup->speed,
1430                      backup->has_on_source_error, backup->on_source_error,
1431                      backup->has_on_target_error, backup->on_target_error,
1432                      &local_err);
1433     if (local_err) {
1434         error_propagate(errp, local_err);
1435         state->bs = NULL;
1436         state->job = NULL;
1437         return;
1438     }
1439 
1440     state->bs = bdrv_find(backup->device);
1441     state->job = state->bs->job;
1442 }
1443 
1444 static void drive_backup_abort(BlkTransactionState *common)
1445 {
1446     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1447     BlockDriverState *bs = state->bs;
1448 
1449     /* Only cancel if it's the job we started */
1450     if (bs && bs->job && bs->job == state->job) {
1451         block_job_cancel_sync(bs->job);
1452     }
1453 }
1454 
1455 static void abort_prepare(BlkTransactionState *common, Error **errp)
1456 {
1457     error_setg(errp, "Transaction aborted using Abort action");
1458 }
1459 
1460 static void abort_commit(BlkTransactionState *common)
1461 {
1462     g_assert_not_reached(); /* this action never succeeds */
1463 }
1464 
1465 static const BdrvActionOps actions[] = {
1466     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1467         .instance_size = sizeof(ExternalSnapshotState),
1468         .prepare  = external_snapshot_prepare,
1469         .commit   = external_snapshot_commit,
1470         .abort = external_snapshot_abort,
1471     },
1472     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1473         .instance_size = sizeof(DriveBackupState),
1474         .prepare = drive_backup_prepare,
1475         .abort = drive_backup_abort,
1476     },
1477     [TRANSACTION_ACTION_KIND_ABORT] = {
1478         .instance_size = sizeof(BlkTransactionState),
1479         .prepare = abort_prepare,
1480         .commit = abort_commit,
1481     },
1482     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1483         .instance_size = sizeof(InternalSnapshotState),
1484         .prepare  = internal_snapshot_prepare,
1485         .abort = internal_snapshot_abort,
1486     },
1487 };
1488 
1489 /*
1490  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1491  *  then we do not pivot any of the devices in the group, and abandon the
1492  *  snapshots
1493  */
1494 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1495 {
1496     TransactionActionList *dev_entry = dev_list;
1497     BlkTransactionState *state, *next;
1498     Error *local_err = NULL;
1499 
1500     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1501     QSIMPLEQ_INIT(&snap_bdrv_states);
1502 
1503     /* drain all i/o before any snapshots */
1504     bdrv_drain_all();
1505 
1506     /* We don't do anything in this loop that commits us to the snapshot */
1507     while (NULL != dev_entry) {
1508         TransactionAction *dev_info = NULL;
1509         const BdrvActionOps *ops;
1510 
1511         dev_info = dev_entry->value;
1512         dev_entry = dev_entry->next;
1513 
1514         assert(dev_info->kind < ARRAY_SIZE(actions));
1515 
1516         ops = &actions[dev_info->kind];
1517         assert(ops->instance_size > 0);
1518 
1519         state = g_malloc0(ops->instance_size);
1520         state->ops = ops;
1521         state->action = dev_info;
1522         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1523 
1524         state->ops->prepare(state, &local_err);
1525         if (local_err) {
1526             error_propagate(errp, local_err);
1527             goto delete_and_fail;
1528         }
1529     }
1530 
1531     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1532         if (state->ops->commit) {
1533             state->ops->commit(state);
1534         }
1535     }
1536 
1537     /* success */
1538     goto exit;
1539 
1540 delete_and_fail:
1541     /*
1542     * failure, and it is all-or-none; abandon each new bs, and keep using
1543     * the original bs for all images
1544     */
1545     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1546         if (state->ops->abort) {
1547             state->ops->abort(state);
1548         }
1549     }
1550 exit:
1551     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1552         if (state->ops->clean) {
1553             state->ops->clean(state);
1554         }
1555         g_free(state);
1556     }
1557 }
1558 
1559 
1560 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1561 {
1562     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1563         return;
1564     }
1565     if (!bdrv_dev_has_removable_media(bs)) {
1566         error_setg(errp, "Device '%s' is not removable",
1567                    bdrv_get_device_name(bs));
1568         return;
1569     }
1570 
1571     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1572         bdrv_dev_eject_request(bs, force);
1573         if (!force) {
1574             error_setg(errp, "Device '%s' is locked",
1575                        bdrv_get_device_name(bs));
1576             return;
1577         }
1578     }
1579 
1580     bdrv_close(bs);
1581 }
1582 
1583 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1584 {
1585     BlockDriverState *bs;
1586 
1587     bs = bdrv_find(device);
1588     if (!bs) {
1589         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1590         return;
1591     }
1592 
1593     eject_device(bs, force, errp);
1594 }
1595 
1596 void qmp_block_passwd(bool has_device, const char *device,
1597                       bool has_node_name, const char *node_name,
1598                       const char *password, Error **errp)
1599 {
1600     Error *local_err = NULL;
1601     BlockDriverState *bs;
1602     int err;
1603 
1604     bs = bdrv_lookup_bs(has_device ? device : NULL,
1605                         has_node_name ? node_name : NULL,
1606                         &local_err);
1607     if (local_err) {
1608         error_propagate(errp, local_err);
1609         return;
1610     }
1611 
1612     err = bdrv_set_key(bs, password);
1613     if (err == -EINVAL) {
1614         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1615         return;
1616     } else if (err < 0) {
1617         error_set(errp, QERR_INVALID_PASSWORD);
1618         return;
1619     }
1620 }
1621 
1622 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1623                                     int bdrv_flags, BlockDriver *drv,
1624                                     const char *password, Error **errp)
1625 {
1626     Error *local_err = NULL;
1627     int ret;
1628 
1629     ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1630     if (ret < 0) {
1631         error_propagate(errp, local_err);
1632         return;
1633     }
1634 
1635     if (bdrv_key_required(bs)) {
1636         if (password) {
1637             if (bdrv_set_key(bs, password) < 0) {
1638                 error_set(errp, QERR_INVALID_PASSWORD);
1639             }
1640         } else {
1641             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1642                       bdrv_get_encrypted_filename(bs));
1643         }
1644     } else if (password) {
1645         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1646     }
1647 }
1648 
1649 void qmp_change_blockdev(const char *device, const char *filename,
1650                          const char *format, Error **errp)
1651 {
1652     BlockDriverState *bs;
1653     BlockDriver *drv = NULL;
1654     int bdrv_flags;
1655     Error *err = NULL;
1656 
1657     bs = bdrv_find(device);
1658     if (!bs) {
1659         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1660         return;
1661     }
1662 
1663     if (format) {
1664         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1665         if (!drv) {
1666             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1667             return;
1668         }
1669     }
1670 
1671     eject_device(bs, 0, &err);
1672     if (err) {
1673         error_propagate(errp, err);
1674         return;
1675     }
1676 
1677     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1678     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1679 
1680     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1681 }
1682 
1683 /* throttling disk I/O limits */
1684 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1685                                int64_t bps_wr,
1686                                int64_t iops,
1687                                int64_t iops_rd,
1688                                int64_t iops_wr,
1689                                bool has_bps_max,
1690                                int64_t bps_max,
1691                                bool has_bps_rd_max,
1692                                int64_t bps_rd_max,
1693                                bool has_bps_wr_max,
1694                                int64_t bps_wr_max,
1695                                bool has_iops_max,
1696                                int64_t iops_max,
1697                                bool has_iops_rd_max,
1698                                int64_t iops_rd_max,
1699                                bool has_iops_wr_max,
1700                                int64_t iops_wr_max,
1701                                bool has_iops_size,
1702                                int64_t iops_size, Error **errp)
1703 {
1704     ThrottleConfig cfg;
1705     BlockDriverState *bs;
1706     AioContext *aio_context;
1707 
1708     bs = bdrv_find(device);
1709     if (!bs) {
1710         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1711         return;
1712     }
1713 
1714     memset(&cfg, 0, sizeof(cfg));
1715     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1716     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1717     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1718 
1719     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1720     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1721     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1722 
1723     if (has_bps_max) {
1724         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1725     }
1726     if (has_bps_rd_max) {
1727         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1728     }
1729     if (has_bps_wr_max) {
1730         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1731     }
1732     if (has_iops_max) {
1733         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1734     }
1735     if (has_iops_rd_max) {
1736         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1737     }
1738     if (has_iops_wr_max) {
1739         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1740     }
1741 
1742     if (has_iops_size) {
1743         cfg.op_size = iops_size;
1744     }
1745 
1746     if (!check_throttle_config(&cfg, errp)) {
1747         return;
1748     }
1749 
1750     aio_context = bdrv_get_aio_context(bs);
1751     aio_context_acquire(aio_context);
1752 
1753     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1754         bdrv_io_limits_enable(bs);
1755     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1756         bdrv_io_limits_disable(bs);
1757     }
1758 
1759     if (bs->io_limits_enabled) {
1760         bdrv_set_io_limits(bs, &cfg);
1761     }
1762 
1763     aio_context_release(aio_context);
1764 }
1765 
1766 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1767 {
1768     const char *id = qdict_get_str(qdict, "id");
1769     BlockDriverState *bs;
1770     Error *local_err = NULL;
1771 
1772     bs = bdrv_find(id);
1773     if (!bs) {
1774         error_report("Device '%s' not found", id);
1775         return -1;
1776     }
1777     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1778         error_report("%s", error_get_pretty(local_err));
1779         error_free(local_err);
1780         return -1;
1781     }
1782 
1783     /* quiesce block driver; prevent further io */
1784     bdrv_drain_all();
1785     bdrv_flush(bs);
1786     bdrv_close(bs);
1787 
1788     /* if we have a device attached to this BlockDriverState
1789      * then we need to make the drive anonymous until the device
1790      * can be removed.  If this is a drive with no device backing
1791      * then we can just get rid of the block driver state right here.
1792      */
1793     if (bdrv_get_attached_dev(bs)) {
1794         bdrv_make_anon(bs);
1795 
1796         /* Further I/O must not pause the guest */
1797         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1798                           BLOCKDEV_ON_ERROR_REPORT);
1799     } else {
1800         drive_uninit(drive_get_by_blockdev(bs));
1801     }
1802 
1803     return 0;
1804 }
1805 
1806 void qmp_block_resize(bool has_device, const char *device,
1807                       bool has_node_name, const char *node_name,
1808                       int64_t size, Error **errp)
1809 {
1810     Error *local_err = NULL;
1811     BlockDriverState *bs;
1812     int ret;
1813 
1814     bs = bdrv_lookup_bs(has_device ? device : NULL,
1815                         has_node_name ? node_name : NULL,
1816                         &local_err);
1817     if (local_err) {
1818         error_propagate(errp, local_err);
1819         return;
1820     }
1821 
1822     if (!bdrv_is_first_non_filter(bs)) {
1823         error_set(errp, QERR_FEATURE_DISABLED, "resize");
1824         return;
1825     }
1826 
1827     if (size < 0) {
1828         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1829         return;
1830     }
1831 
1832     /* complete all in-flight operations before resizing the device */
1833     bdrv_drain_all();
1834 
1835     ret = bdrv_truncate(bs, size);
1836     switch (ret) {
1837     case 0:
1838         break;
1839     case -ENOMEDIUM:
1840         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1841         break;
1842     case -ENOTSUP:
1843         error_set(errp, QERR_UNSUPPORTED);
1844         break;
1845     case -EACCES:
1846         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1847         break;
1848     case -EBUSY:
1849         error_set(errp, QERR_DEVICE_IN_USE, device);
1850         break;
1851     default:
1852         error_setg_errno(errp, -ret, "Could not resize");
1853         break;
1854     }
1855 }
1856 
1857 static void block_job_cb(void *opaque, int ret)
1858 {
1859     BlockDriverState *bs = opaque;
1860     QObject *obj;
1861 
1862     trace_block_job_cb(bs, bs->job, ret);
1863 
1864     assert(bs->job);
1865     obj = qobject_from_block_job(bs->job);
1866     if (ret < 0) {
1867         QDict *dict = qobject_to_qdict(obj);
1868         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1869     }
1870 
1871     if (block_job_is_cancelled(bs->job)) {
1872         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1873     } else {
1874         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1875     }
1876     qobject_decref(obj);
1877 
1878     bdrv_put_ref_bh_schedule(bs);
1879 }
1880 
1881 void qmp_block_stream(const char *device, bool has_base,
1882                       const char *base, bool has_speed, int64_t speed,
1883                       bool has_on_error, BlockdevOnError on_error,
1884                       Error **errp)
1885 {
1886     BlockDriverState *bs;
1887     BlockDriverState *base_bs = NULL;
1888     Error *local_err = NULL;
1889 
1890     if (!has_on_error) {
1891         on_error = BLOCKDEV_ON_ERROR_REPORT;
1892     }
1893 
1894     bs = bdrv_find(device);
1895     if (!bs) {
1896         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1897         return;
1898     }
1899 
1900     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
1901         return;
1902     }
1903 
1904     if (base) {
1905         base_bs = bdrv_find_backing_image(bs, base);
1906         if (base_bs == NULL) {
1907             error_set(errp, QERR_BASE_NOT_FOUND, base);
1908             return;
1909         }
1910     }
1911 
1912     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1913                  on_error, block_job_cb, bs, &local_err);
1914     if (local_err) {
1915         error_propagate(errp, local_err);
1916         return;
1917     }
1918 
1919     trace_qmp_block_stream(bs, bs->job);
1920 }
1921 
1922 void qmp_block_commit(const char *device,
1923                       bool has_base, const char *base, const char *top,
1924                       bool has_speed, int64_t speed,
1925                       Error **errp)
1926 {
1927     BlockDriverState *bs;
1928     BlockDriverState *base_bs, *top_bs;
1929     Error *local_err = NULL;
1930     /* This will be part of the QMP command, if/when the
1931      * BlockdevOnError change for blkmirror makes it in
1932      */
1933     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1934 
1935     if (!has_speed) {
1936         speed = 0;
1937     }
1938 
1939     /* drain all i/o before commits */
1940     bdrv_drain_all();
1941 
1942     bs = bdrv_find(device);
1943     if (!bs) {
1944         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1945         return;
1946     }
1947 
1948     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, errp)) {
1949         return;
1950     }
1951 
1952     /* default top_bs is the active layer */
1953     top_bs = bs;
1954 
1955     if (top) {
1956         if (strcmp(bs->filename, top) != 0) {
1957             top_bs = bdrv_find_backing_image(bs, top);
1958         }
1959     }
1960 
1961     if (top_bs == NULL) {
1962         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1963         return;
1964     }
1965 
1966     if (has_base && base) {
1967         base_bs = bdrv_find_backing_image(top_bs, base);
1968     } else {
1969         base_bs = bdrv_find_base(top_bs);
1970     }
1971 
1972     if (base_bs == NULL) {
1973         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1974         return;
1975     }
1976 
1977     if (top_bs == bs) {
1978         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1979                             bs, &local_err);
1980     } else {
1981         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1982                     &local_err);
1983     }
1984     if (local_err != NULL) {
1985         error_propagate(errp, local_err);
1986         return;
1987     }
1988 }
1989 
1990 void qmp_drive_backup(const char *device, const char *target,
1991                       bool has_format, const char *format,
1992                       enum MirrorSyncMode sync,
1993                       bool has_mode, enum NewImageMode mode,
1994                       bool has_speed, int64_t speed,
1995                       bool has_on_source_error, BlockdevOnError on_source_error,
1996                       bool has_on_target_error, BlockdevOnError on_target_error,
1997                       Error **errp)
1998 {
1999     BlockDriverState *bs;
2000     BlockDriverState *target_bs;
2001     BlockDriverState *source = NULL;
2002     BlockDriver *drv = NULL;
2003     Error *local_err = NULL;
2004     int flags;
2005     int64_t size;
2006     int ret;
2007 
2008     if (!has_speed) {
2009         speed = 0;
2010     }
2011     if (!has_on_source_error) {
2012         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2013     }
2014     if (!has_on_target_error) {
2015         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2016     }
2017     if (!has_mode) {
2018         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2019     }
2020 
2021     bs = bdrv_find(device);
2022     if (!bs) {
2023         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2024         return;
2025     }
2026 
2027     if (!bdrv_is_inserted(bs)) {
2028         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2029         return;
2030     }
2031 
2032     if (!has_format) {
2033         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2034     }
2035     if (format) {
2036         drv = bdrv_find_format(format);
2037         if (!drv) {
2038             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2039             return;
2040         }
2041     }
2042 
2043     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2044         return;
2045     }
2046 
2047     flags = bs->open_flags | BDRV_O_RDWR;
2048 
2049     /* See if we have a backing HD we can use to create our new image
2050      * on top of. */
2051     if (sync == MIRROR_SYNC_MODE_TOP) {
2052         source = bs->backing_hd;
2053         if (!source) {
2054             sync = MIRROR_SYNC_MODE_FULL;
2055         }
2056     }
2057     if (sync == MIRROR_SYNC_MODE_NONE) {
2058         source = bs;
2059     }
2060 
2061     size = bdrv_getlength(bs);
2062     if (size < 0) {
2063         error_setg_errno(errp, -size, "bdrv_getlength failed");
2064         return;
2065     }
2066 
2067     if (mode != NEW_IMAGE_MODE_EXISTING) {
2068         assert(format && drv);
2069         if (source) {
2070             bdrv_img_create(target, format, source->filename,
2071                             source->drv->format_name, NULL,
2072                             size, flags, &local_err, false);
2073         } else {
2074             bdrv_img_create(target, format, NULL, NULL, NULL,
2075                             size, flags, &local_err, false);
2076         }
2077     }
2078 
2079     if (local_err) {
2080         error_propagate(errp, local_err);
2081         return;
2082     }
2083 
2084     target_bs = NULL;
2085     ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2086     if (ret < 0) {
2087         error_propagate(errp, local_err);
2088         return;
2089     }
2090 
2091     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2092                  block_job_cb, bs, &local_err);
2093     if (local_err != NULL) {
2094         bdrv_unref(target_bs);
2095         error_propagate(errp, local_err);
2096         return;
2097     }
2098 }
2099 
2100 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2101 {
2102     return bdrv_named_nodes_list();
2103 }
2104 
2105 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
2106 
2107 void qmp_drive_mirror(const char *device, const char *target,
2108                       bool has_format, const char *format,
2109                       enum MirrorSyncMode sync,
2110                       bool has_mode, enum NewImageMode mode,
2111                       bool has_speed, int64_t speed,
2112                       bool has_granularity, uint32_t granularity,
2113                       bool has_buf_size, int64_t buf_size,
2114                       bool has_on_source_error, BlockdevOnError on_source_error,
2115                       bool has_on_target_error, BlockdevOnError on_target_error,
2116                       Error **errp)
2117 {
2118     BlockDriverState *bs;
2119     BlockDriverState *source, *target_bs;
2120     BlockDriver *drv = NULL;
2121     Error *local_err = NULL;
2122     int flags;
2123     int64_t size;
2124     int ret;
2125 
2126     if (!has_speed) {
2127         speed = 0;
2128     }
2129     if (!has_on_source_error) {
2130         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2131     }
2132     if (!has_on_target_error) {
2133         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2134     }
2135     if (!has_mode) {
2136         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2137     }
2138     if (!has_granularity) {
2139         granularity = 0;
2140     }
2141     if (!has_buf_size) {
2142         buf_size = DEFAULT_MIRROR_BUF_SIZE;
2143     }
2144 
2145     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2146         error_set(errp, QERR_INVALID_PARAMETER, device);
2147         return;
2148     }
2149     if (granularity & (granularity - 1)) {
2150         error_set(errp, QERR_INVALID_PARAMETER, device);
2151         return;
2152     }
2153 
2154     bs = bdrv_find(device);
2155     if (!bs) {
2156         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2157         return;
2158     }
2159 
2160     if (!bdrv_is_inserted(bs)) {
2161         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2162         return;
2163     }
2164 
2165     if (!has_format) {
2166         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2167     }
2168     if (format) {
2169         drv = bdrv_find_format(format);
2170         if (!drv) {
2171             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2172             return;
2173         }
2174     }
2175 
2176     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2177         return;
2178     }
2179 
2180     flags = bs->open_flags | BDRV_O_RDWR;
2181     source = bs->backing_hd;
2182     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2183         sync = MIRROR_SYNC_MODE_FULL;
2184     }
2185     if (sync == MIRROR_SYNC_MODE_NONE) {
2186         source = bs;
2187     }
2188 
2189     size = bdrv_getlength(bs);
2190     if (size < 0) {
2191         error_setg_errno(errp, -size, "bdrv_getlength failed");
2192         return;
2193     }
2194 
2195     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2196         && mode != NEW_IMAGE_MODE_EXISTING)
2197     {
2198         /* create new image w/o backing file */
2199         assert(format && drv);
2200         bdrv_img_create(target, format,
2201                         NULL, NULL, NULL, size, flags, &local_err, false);
2202     } else {
2203         switch (mode) {
2204         case NEW_IMAGE_MODE_EXISTING:
2205             break;
2206         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2207             /* create new image with backing file */
2208             bdrv_img_create(target, format,
2209                             source->filename,
2210                             source->drv->format_name,
2211                             NULL, size, flags, &local_err, false);
2212             break;
2213         default:
2214             abort();
2215         }
2216     }
2217 
2218     if (local_err) {
2219         error_propagate(errp, local_err);
2220         return;
2221     }
2222 
2223     /* Mirroring takes care of copy-on-write using the source's backing
2224      * file.
2225      */
2226     target_bs = NULL;
2227     ret = bdrv_open(&target_bs, target, NULL, NULL, flags | BDRV_O_NO_BACKING,
2228                     drv, &local_err);
2229     if (ret < 0) {
2230         error_propagate(errp, local_err);
2231         return;
2232     }
2233 
2234     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2235                  on_source_error, on_target_error,
2236                  block_job_cb, bs, &local_err);
2237     if (local_err != NULL) {
2238         bdrv_unref(target_bs);
2239         error_propagate(errp, local_err);
2240         return;
2241     }
2242 }
2243 
2244 static BlockJob *find_block_job(const char *device)
2245 {
2246     BlockDriverState *bs;
2247 
2248     bs = bdrv_find(device);
2249     if (!bs || !bs->job) {
2250         return NULL;
2251     }
2252     return bs->job;
2253 }
2254 
2255 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2256 {
2257     BlockJob *job = find_block_job(device);
2258 
2259     if (!job) {
2260         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2261         return;
2262     }
2263 
2264     block_job_set_speed(job, speed, errp);
2265 }
2266 
2267 void qmp_block_job_cancel(const char *device,
2268                           bool has_force, bool force, Error **errp)
2269 {
2270     BlockJob *job = find_block_job(device);
2271 
2272     if (!has_force) {
2273         force = false;
2274     }
2275 
2276     if (!job) {
2277         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2278         return;
2279     }
2280     if (job->paused && !force) {
2281         error_setg(errp, "The block job for device '%s' is currently paused",
2282                    device);
2283         return;
2284     }
2285 
2286     trace_qmp_block_job_cancel(job);
2287     block_job_cancel(job);
2288 }
2289 
2290 void qmp_block_job_pause(const char *device, Error **errp)
2291 {
2292     BlockJob *job = find_block_job(device);
2293 
2294     if (!job) {
2295         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2296         return;
2297     }
2298 
2299     trace_qmp_block_job_pause(job);
2300     block_job_pause(job);
2301 }
2302 
2303 void qmp_block_job_resume(const char *device, Error **errp)
2304 {
2305     BlockJob *job = find_block_job(device);
2306 
2307     if (!job) {
2308         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2309         return;
2310     }
2311 
2312     trace_qmp_block_job_resume(job);
2313     block_job_resume(job);
2314 }
2315 
2316 void qmp_block_job_complete(const char *device, Error **errp)
2317 {
2318     BlockJob *job = find_block_job(device);
2319 
2320     if (!job) {
2321         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2322         return;
2323     }
2324 
2325     trace_qmp_block_job_complete(job);
2326     block_job_complete(job, errp);
2327 }
2328 
2329 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2330 {
2331     QmpOutputVisitor *ov = qmp_output_visitor_new();
2332     DriveInfo *dinfo;
2333     QObject *obj;
2334     QDict *qdict;
2335     Error *local_err = NULL;
2336 
2337     /* Require an ID in the top level */
2338     if (!options->has_id) {
2339         error_setg(errp, "Block device needs an ID");
2340         goto fail;
2341     }
2342 
2343     /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2344      * cache.direct=false instead of silently switching to aio=threads, except
2345      * if called from drive_init.
2346      *
2347      * For now, simply forbidding the combination for all drivers will do. */
2348     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2349         bool direct = options->has_cache &&
2350                       options->cache->has_direct &&
2351                       options->cache->direct;
2352         if (!direct) {
2353             error_setg(errp, "aio=native requires cache.direct=true");
2354             goto fail;
2355         }
2356     }
2357 
2358     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2359                                &options, NULL, &local_err);
2360     if (local_err) {
2361         error_propagate(errp, local_err);
2362         goto fail;
2363     }
2364 
2365     obj = qmp_output_get_qobject(ov);
2366     qdict = qobject_to_qdict(obj);
2367 
2368     qdict_flatten(qdict);
2369 
2370     dinfo = blockdev_init(NULL, qdict, &local_err);
2371     if (local_err) {
2372         error_propagate(errp, local_err);
2373         goto fail;
2374     }
2375 
2376     if (bdrv_key_required(dinfo->bdrv)) {
2377         drive_uninit(dinfo);
2378         error_setg(errp, "blockdev-add doesn't support encrypted devices");
2379         goto fail;
2380     }
2381 
2382 fail:
2383     qmp_output_visitor_cleanup(ov);
2384 }
2385 
2386 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2387 {
2388     BlockJobInfoList **prev = opaque;
2389     BlockJob *job = bs->job;
2390 
2391     if (job) {
2392         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2393         elem->value = block_job_query(bs->job);
2394         (*prev)->next = elem;
2395         *prev = elem;
2396     }
2397 }
2398 
2399 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2400 {
2401     /* Dummy is a fake list element for holding the head pointer */
2402     BlockJobInfoList dummy = {};
2403     BlockJobInfoList *prev = &dummy;
2404     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2405     return dummy.next;
2406 }
2407 
2408 QemuOptsList qemu_common_drive_opts = {
2409     .name = "drive",
2410     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2411     .desc = {
2412         {
2413             .name = "snapshot",
2414             .type = QEMU_OPT_BOOL,
2415             .help = "enable/disable snapshot mode",
2416         },{
2417             .name = "discard",
2418             .type = QEMU_OPT_STRING,
2419             .help = "discard operation (ignore/off, unmap/on)",
2420         },{
2421             .name = "cache.writeback",
2422             .type = QEMU_OPT_BOOL,
2423             .help = "enables writeback mode for any caches",
2424         },{
2425             .name = "cache.direct",
2426             .type = QEMU_OPT_BOOL,
2427             .help = "enables use of O_DIRECT (bypass the host page cache)",
2428         },{
2429             .name = "cache.no-flush",
2430             .type = QEMU_OPT_BOOL,
2431             .help = "ignore any flush requests for the device",
2432         },{
2433             .name = "aio",
2434             .type = QEMU_OPT_STRING,
2435             .help = "host AIO implementation (threads, native)",
2436         },{
2437             .name = "format",
2438             .type = QEMU_OPT_STRING,
2439             .help = "disk format (raw, qcow2, ...)",
2440         },{
2441             .name = "serial",
2442             .type = QEMU_OPT_STRING,
2443             .help = "disk serial number",
2444         },{
2445             .name = "rerror",
2446             .type = QEMU_OPT_STRING,
2447             .help = "read error action",
2448         },{
2449             .name = "werror",
2450             .type = QEMU_OPT_STRING,
2451             .help = "write error action",
2452         },{
2453             .name = "read-only",
2454             .type = QEMU_OPT_BOOL,
2455             .help = "open drive file as read-only",
2456         },{
2457             .name = "throttling.iops-total",
2458             .type = QEMU_OPT_NUMBER,
2459             .help = "limit total I/O operations per second",
2460         },{
2461             .name = "throttling.iops-read",
2462             .type = QEMU_OPT_NUMBER,
2463             .help = "limit read operations per second",
2464         },{
2465             .name = "throttling.iops-write",
2466             .type = QEMU_OPT_NUMBER,
2467             .help = "limit write operations per second",
2468         },{
2469             .name = "throttling.bps-total",
2470             .type = QEMU_OPT_NUMBER,
2471             .help = "limit total bytes per second",
2472         },{
2473             .name = "throttling.bps-read",
2474             .type = QEMU_OPT_NUMBER,
2475             .help = "limit read bytes per second",
2476         },{
2477             .name = "throttling.bps-write",
2478             .type = QEMU_OPT_NUMBER,
2479             .help = "limit write bytes per second",
2480         },{
2481             .name = "throttling.iops-total-max",
2482             .type = QEMU_OPT_NUMBER,
2483             .help = "I/O operations burst",
2484         },{
2485             .name = "throttling.iops-read-max",
2486             .type = QEMU_OPT_NUMBER,
2487             .help = "I/O operations read burst",
2488         },{
2489             .name = "throttling.iops-write-max",
2490             .type = QEMU_OPT_NUMBER,
2491             .help = "I/O operations write burst",
2492         },{
2493             .name = "throttling.bps-total-max",
2494             .type = QEMU_OPT_NUMBER,
2495             .help = "total bytes burst",
2496         },{
2497             .name = "throttling.bps-read-max",
2498             .type = QEMU_OPT_NUMBER,
2499             .help = "total bytes read burst",
2500         },{
2501             .name = "throttling.bps-write-max",
2502             .type = QEMU_OPT_NUMBER,
2503             .help = "total bytes write burst",
2504         },{
2505             .name = "throttling.iops-size",
2506             .type = QEMU_OPT_NUMBER,
2507             .help = "when limiting by iops max size of an I/O in bytes",
2508         },{
2509             .name = "copy-on-read",
2510             .type = QEMU_OPT_BOOL,
2511             .help = "copy read data from backing file into image file",
2512         },{
2513             .name = "detect-zeroes",
2514             .type = QEMU_OPT_STRING,
2515             .help = "try to optimize zero writes (off, on, unmap)",
2516         },
2517         { /* end of list */ }
2518     },
2519 };
2520 
2521 QemuOptsList qemu_drive_opts = {
2522     .name = "drive",
2523     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2524     .desc = {
2525         /*
2526          * no elements => accept any params
2527          * validation will happen later
2528          */
2529         { /* end of list */ }
2530     },
2531 };
2532