xref: /openbmc/qemu/blockdev.c (revision 461a2753)
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_del(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 void drive_del(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 typedef struct {
230     QEMUBH *bh;
231     BlockDriverState *bs;
232 } BDRVPutRefBH;
233 
234 static void bdrv_put_ref_bh(void *opaque)
235 {
236     BDRVPutRefBH *s = opaque;
237 
238     bdrv_unref(s->bs);
239     qemu_bh_delete(s->bh);
240     g_free(s);
241 }
242 
243 /*
244  * Release a BDS reference in a BH
245  *
246  * It is not safe to use bdrv_unref() from a callback function when the callers
247  * still need the BlockDriverState.  In such cases we schedule a BH to release
248  * the reference.
249  */
250 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
251 {
252     BDRVPutRefBH *s;
253 
254     s = g_new(BDRVPutRefBH, 1);
255     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
256     s->bs = bs;
257     qemu_bh_schedule(s->bh);
258 }
259 
260 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
261 {
262     if (!strcmp(buf, "ignore")) {
263         return BLOCKDEV_ON_ERROR_IGNORE;
264     } else if (!is_read && !strcmp(buf, "enospc")) {
265         return BLOCKDEV_ON_ERROR_ENOSPC;
266     } else if (!strcmp(buf, "stop")) {
267         return BLOCKDEV_ON_ERROR_STOP;
268     } else if (!strcmp(buf, "report")) {
269         return BLOCKDEV_ON_ERROR_REPORT;
270     } else {
271         error_setg(errp, "'%s' invalid %s error action",
272                    buf, is_read ? "read" : "write");
273         return -1;
274     }
275 }
276 
277 static inline int parse_enum_option(const char *lookup[], const char *buf,
278                                     int max, int def, Error **errp)
279 {
280     int i;
281 
282     if (!buf) {
283         return def;
284     }
285 
286     for (i = 0; i < max; i++) {
287         if (!strcmp(buf, lookup[i])) {
288             return i;
289         }
290     }
291 
292     error_setg(errp, "invalid parameter value: %s", buf);
293     return def;
294 }
295 
296 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
297 {
298     if (throttle_conflicting(cfg)) {
299         error_setg(errp, "bps/iops/max total values and read/write values"
300                          " cannot be used at the same time");
301         return false;
302     }
303 
304     if (!throttle_is_valid(cfg)) {
305         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
306         return false;
307     }
308 
309     return true;
310 }
311 
312 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
313 
314 /* Takes the ownership of bs_opts */
315 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
316                                 Error **errp)
317 {
318     const char *buf;
319     int ro = 0;
320     int bdrv_flags = 0;
321     int on_read_error, on_write_error;
322     DriveInfo *dinfo;
323     ThrottleConfig cfg;
324     int snapshot = 0;
325     bool copy_on_read;
326     int ret;
327     Error *error = NULL;
328     QemuOpts *opts;
329     const char *id;
330     bool has_driver_specific_opts;
331     BlockdevDetectZeroesOptions detect_zeroes;
332     BlockDriver *drv = NULL;
333 
334     /* Check common options by copying from bs_opts to opts, all other options
335      * stay in bs_opts for processing by bdrv_open(). */
336     id = qdict_get_try_str(bs_opts, "id");
337     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
338     if (error) {
339         error_propagate(errp, error);
340         goto err_no_opts;
341     }
342 
343     qemu_opts_absorb_qdict(opts, bs_opts, &error);
344     if (error) {
345         error_propagate(errp, error);
346         goto early_err;
347     }
348 
349     if (id) {
350         qdict_del(bs_opts, "id");
351     }
352 
353     has_driver_specific_opts = !!qdict_size(bs_opts);
354 
355     /* extract parameters */
356     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
357     ro = qemu_opt_get_bool(opts, "read-only", 0);
358     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
359 
360     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
361         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
362             error_setg(errp, "invalid discard option");
363             goto early_err;
364         }
365     }
366 
367     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
368         bdrv_flags |= BDRV_O_CACHE_WB;
369     }
370     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
371         bdrv_flags |= BDRV_O_NOCACHE;
372     }
373     if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
374         bdrv_flags |= BDRV_O_NO_FLUSH;
375     }
376 
377 #ifdef CONFIG_LINUX_AIO
378     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
379         if (!strcmp(buf, "native")) {
380             bdrv_flags |= BDRV_O_NATIVE_AIO;
381         } else if (!strcmp(buf, "threads")) {
382             /* this is the default */
383         } else {
384            error_setg(errp, "invalid aio option");
385            goto early_err;
386         }
387     }
388 #endif
389 
390     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
391         if (is_help_option(buf)) {
392             error_printf("Supported formats:");
393             bdrv_iterate_format(bdrv_format_print, NULL);
394             error_printf("\n");
395             goto early_err;
396         }
397 
398         drv = bdrv_find_format(buf);
399         if (!drv) {
400             error_setg(errp, "'%s' invalid format", buf);
401             goto early_err;
402         }
403     }
404 
405     /* disk I/O throttling */
406     memset(&cfg, 0, sizeof(cfg));
407     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
408         qemu_opt_get_number(opts, "throttling.bps-total", 0);
409     cfg.buckets[THROTTLE_BPS_READ].avg  =
410         qemu_opt_get_number(opts, "throttling.bps-read", 0);
411     cfg.buckets[THROTTLE_BPS_WRITE].avg =
412         qemu_opt_get_number(opts, "throttling.bps-write", 0);
413     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
414         qemu_opt_get_number(opts, "throttling.iops-total", 0);
415     cfg.buckets[THROTTLE_OPS_READ].avg =
416         qemu_opt_get_number(opts, "throttling.iops-read", 0);
417     cfg.buckets[THROTTLE_OPS_WRITE].avg =
418         qemu_opt_get_number(opts, "throttling.iops-write", 0);
419 
420     cfg.buckets[THROTTLE_BPS_TOTAL].max =
421         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
422     cfg.buckets[THROTTLE_BPS_READ].max  =
423         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
424     cfg.buckets[THROTTLE_BPS_WRITE].max =
425         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
426     cfg.buckets[THROTTLE_OPS_TOTAL].max =
427         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
428     cfg.buckets[THROTTLE_OPS_READ].max =
429         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
430     cfg.buckets[THROTTLE_OPS_WRITE].max =
431         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
432 
433     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
434 
435     if (!check_throttle_config(&cfg, &error)) {
436         error_propagate(errp, error);
437         goto early_err;
438     }
439 
440     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
441     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
442         on_write_error = parse_block_error_action(buf, 0, &error);
443         if (error) {
444             error_propagate(errp, error);
445             goto early_err;
446         }
447     }
448 
449     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
450     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
451         on_read_error = parse_block_error_action(buf, 1, &error);
452         if (error) {
453             error_propagate(errp, error);
454             goto early_err;
455         }
456     }
457 
458     detect_zeroes =
459         parse_enum_option(BlockdevDetectZeroesOptions_lookup,
460                           qemu_opt_get(opts, "detect-zeroes"),
461                           BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
462                           BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
463                           &error);
464     if (error) {
465         error_propagate(errp, error);
466         goto early_err;
467     }
468 
469     if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
470         !(bdrv_flags & BDRV_O_UNMAP)) {
471         error_setg(errp, "setting detect-zeroes to unmap is not allowed "
472                          "without setting discard operation to unmap");
473         goto early_err;
474     }
475 
476     /* init */
477     dinfo = g_malloc0(sizeof(*dinfo));
478     dinfo->id = g_strdup(qemu_opts_id(opts));
479     dinfo->bdrv = bdrv_new(dinfo->id, &error);
480     if (error) {
481         error_propagate(errp, error);
482         goto bdrv_new_err;
483     }
484     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
485     dinfo->bdrv->read_only = ro;
486     dinfo->bdrv->detect_zeroes = detect_zeroes;
487     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
488 
489     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
490 
491     /* disk I/O throttling */
492     if (throttle_enabled(&cfg)) {
493         bdrv_io_limits_enable(dinfo->bdrv);
494         bdrv_set_io_limits(dinfo->bdrv, &cfg);
495     }
496 
497     if (!file || !*file) {
498         if (has_driver_specific_opts) {
499             file = NULL;
500         } else {
501             QDECREF(bs_opts);
502             qemu_opts_del(opts);
503             return dinfo;
504         }
505     }
506     if (snapshot) {
507         /* always use cache=unsafe with snapshot */
508         bdrv_flags &= ~BDRV_O_CACHE_MASK;
509         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
510     }
511 
512     if (copy_on_read) {
513         bdrv_flags |= BDRV_O_COPY_ON_READ;
514     }
515 
516     if (runstate_check(RUN_STATE_INMIGRATE)) {
517         bdrv_flags |= BDRV_O_INCOMING;
518     }
519 
520     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
521 
522     QINCREF(bs_opts);
523     ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
524 
525     if (ret < 0) {
526         error_setg(errp, "could not open disk image %s: %s",
527                    file ?: dinfo->id, error_get_pretty(error));
528         error_free(error);
529         goto err;
530     }
531 
532     if (bdrv_key_required(dinfo->bdrv))
533         autostart = 0;
534 
535     QDECREF(bs_opts);
536     qemu_opts_del(opts);
537 
538     return dinfo;
539 
540 err:
541     bdrv_unref(dinfo->bdrv);
542     QTAILQ_REMOVE(&drives, dinfo, next);
543 bdrv_new_err:
544     g_free(dinfo->id);
545     g_free(dinfo);
546 early_err:
547     qemu_opts_del(opts);
548 err_no_opts:
549     QDECREF(bs_opts);
550     return NULL;
551 }
552 
553 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
554 {
555     const char *value;
556 
557     value = qemu_opt_get(opts, from);
558     if (value) {
559         qemu_opt_set(opts, to, value);
560         qemu_opt_unset(opts, from);
561     }
562 }
563 
564 QemuOptsList qemu_legacy_drive_opts = {
565     .name = "drive",
566     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
567     .desc = {
568         {
569             .name = "bus",
570             .type = QEMU_OPT_NUMBER,
571             .help = "bus number",
572         },{
573             .name = "unit",
574             .type = QEMU_OPT_NUMBER,
575             .help = "unit number (i.e. lun for scsi)",
576         },{
577             .name = "index",
578             .type = QEMU_OPT_NUMBER,
579             .help = "index number",
580         },{
581             .name = "media",
582             .type = QEMU_OPT_STRING,
583             .help = "media type (disk, cdrom)",
584         },{
585             .name = "if",
586             .type = QEMU_OPT_STRING,
587             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
588         },{
589             .name = "cyls",
590             .type = QEMU_OPT_NUMBER,
591             .help = "number of cylinders (ide disk geometry)",
592         },{
593             .name = "heads",
594             .type = QEMU_OPT_NUMBER,
595             .help = "number of heads (ide disk geometry)",
596         },{
597             .name = "secs",
598             .type = QEMU_OPT_NUMBER,
599             .help = "number of sectors (ide disk geometry)",
600         },{
601             .name = "trans",
602             .type = QEMU_OPT_STRING,
603             .help = "chs translation (auto, lba, none)",
604         },{
605             .name = "boot",
606             .type = QEMU_OPT_BOOL,
607             .help = "(deprecated, ignored)",
608         },{
609             .name = "addr",
610             .type = QEMU_OPT_STRING,
611             .help = "pci address (virtio only)",
612         },{
613             .name = "serial",
614             .type = QEMU_OPT_STRING,
615             .help = "disk serial number",
616         },{
617             .name = "file",
618             .type = QEMU_OPT_STRING,
619             .help = "file name",
620         },
621 
622         /* Options that are passed on, but have special semantics with -drive */
623         {
624             .name = "read-only",
625             .type = QEMU_OPT_BOOL,
626             .help = "open drive file as read-only",
627         },{
628             .name = "rerror",
629             .type = QEMU_OPT_STRING,
630             .help = "read error action",
631         },{
632             .name = "werror",
633             .type = QEMU_OPT_STRING,
634             .help = "write error action",
635         },{
636             .name = "copy-on-read",
637             .type = QEMU_OPT_BOOL,
638             .help = "copy read data from backing file into image file",
639         },
640 
641         { /* end of list */ }
642     },
643 };
644 
645 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
646 {
647     const char *value;
648     DriveInfo *dinfo = NULL;
649     QDict *bs_opts;
650     QemuOpts *legacy_opts;
651     DriveMediaType media = MEDIA_DISK;
652     BlockInterfaceType type;
653     int cyls, heads, secs, translation;
654     int max_devs, bus_id, unit_id, index;
655     const char *devaddr;
656     const char *werror, *rerror;
657     bool read_only = false;
658     bool copy_on_read;
659     const char *serial;
660     const char *filename;
661     Error *local_err = NULL;
662 
663     /* Change legacy command line options into QMP ones */
664     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
665     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
666     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
667 
668     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
669     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
670     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
671 
672     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
673     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
674     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
675 
676     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
677     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
678     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
679 
680     qemu_opt_rename(all_opts,
681                     "iops_size", "throttling.iops-size");
682 
683     qemu_opt_rename(all_opts, "readonly", "read-only");
684 
685     value = qemu_opt_get(all_opts, "cache");
686     if (value) {
687         int flags = 0;
688 
689         if (bdrv_parse_cache_flags(value, &flags) != 0) {
690             error_report("invalid cache option");
691             return NULL;
692         }
693 
694         /* Specific options take precedence */
695         if (!qemu_opt_get(all_opts, "cache.writeback")) {
696             qemu_opt_set_bool(all_opts, "cache.writeback",
697                               !!(flags & BDRV_O_CACHE_WB));
698         }
699         if (!qemu_opt_get(all_opts, "cache.direct")) {
700             qemu_opt_set_bool(all_opts, "cache.direct",
701                               !!(flags & BDRV_O_NOCACHE));
702         }
703         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
704             qemu_opt_set_bool(all_opts, "cache.no-flush",
705                               !!(flags & BDRV_O_NO_FLUSH));
706         }
707         qemu_opt_unset(all_opts, "cache");
708     }
709 
710     /* Get a QDict for processing the options */
711     bs_opts = qdict_new();
712     qemu_opts_to_qdict(all_opts, bs_opts);
713 
714     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
715                                    &error_abort);
716     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
717     if (local_err) {
718         error_report("%s", error_get_pretty(local_err));
719         error_free(local_err);
720         goto fail;
721     }
722 
723     /* Deprecated option boot=[on|off] */
724     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
725         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
726                 "ignored. Future versions will reject this parameter. Please "
727                 "update your scripts.\n");
728     }
729 
730     /* Media type */
731     value = qemu_opt_get(legacy_opts, "media");
732     if (value) {
733         if (!strcmp(value, "disk")) {
734             media = MEDIA_DISK;
735         } else if (!strcmp(value, "cdrom")) {
736             media = MEDIA_CDROM;
737             read_only = true;
738         } else {
739             error_report("'%s' invalid media", value);
740             goto fail;
741         }
742     }
743 
744     /* copy-on-read is disabled with a warning for read-only devices */
745     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
746     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
747 
748     if (read_only && copy_on_read) {
749         error_report("warning: disabling copy-on-read on read-only drive");
750         copy_on_read = false;
751     }
752 
753     qdict_put(bs_opts, "read-only",
754               qstring_from_str(read_only ? "on" : "off"));
755     qdict_put(bs_opts, "copy-on-read",
756               qstring_from_str(copy_on_read ? "on" :"off"));
757 
758     /* Controller type */
759     value = qemu_opt_get(legacy_opts, "if");
760     if (value) {
761         for (type = 0;
762              type < IF_COUNT && strcmp(value, if_name[type]);
763              type++) {
764         }
765         if (type == IF_COUNT) {
766             error_report("unsupported bus type '%s'", value);
767             goto fail;
768         }
769     } else {
770         type = block_default_type;
771     }
772 
773     /* Geometry */
774     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
775     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
776     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
777 
778     if (cyls || heads || secs) {
779         if (cyls < 1) {
780             error_report("invalid physical cyls number");
781             goto fail;
782         }
783         if (heads < 1) {
784             error_report("invalid physical heads number");
785             goto fail;
786         }
787         if (secs < 1) {
788             error_report("invalid physical secs number");
789             goto fail;
790         }
791     }
792 
793     translation = BIOS_ATA_TRANSLATION_AUTO;
794     value = qemu_opt_get(legacy_opts, "trans");
795     if (value != NULL) {
796         if (!cyls) {
797             error_report("'%s' trans must be used with cyls, heads and secs",
798                          value);
799             goto fail;
800         }
801         if (!strcmp(value, "none")) {
802             translation = BIOS_ATA_TRANSLATION_NONE;
803         } else if (!strcmp(value, "lba")) {
804             translation = BIOS_ATA_TRANSLATION_LBA;
805         } else if (!strcmp(value, "large")) {
806             translation = BIOS_ATA_TRANSLATION_LARGE;
807         } else if (!strcmp(value, "rechs")) {
808             translation = BIOS_ATA_TRANSLATION_RECHS;
809         } else if (!strcmp(value, "auto")) {
810             translation = BIOS_ATA_TRANSLATION_AUTO;
811         } else {
812             error_report("'%s' invalid translation type", value);
813             goto fail;
814         }
815     }
816 
817     if (media == MEDIA_CDROM) {
818         if (cyls || secs || heads) {
819             error_report("CHS can't be set with media=cdrom");
820             goto fail;
821         }
822     }
823 
824     /* Device address specified by bus/unit or index.
825      * If none was specified, try to find the first free one. */
826     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
827     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
828     index   = qemu_opt_get_number(legacy_opts, "index", -1);
829 
830     max_devs = if_max_devs[type];
831 
832     if (index != -1) {
833         if (bus_id != 0 || unit_id != -1) {
834             error_report("index cannot be used with bus and unit");
835             goto fail;
836         }
837         bus_id = drive_index_to_bus_id(type, index);
838         unit_id = drive_index_to_unit_id(type, index);
839     }
840 
841     if (unit_id == -1) {
842        unit_id = 0;
843        while (drive_get(type, bus_id, unit_id) != NULL) {
844            unit_id++;
845            if (max_devs && unit_id >= max_devs) {
846                unit_id -= max_devs;
847                bus_id++;
848            }
849        }
850     }
851 
852     if (max_devs && unit_id >= max_devs) {
853         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
854         goto fail;
855     }
856 
857     if (drive_get(type, bus_id, unit_id) != NULL) {
858         error_report("drive with bus=%d, unit=%d (index=%d) exists",
859                      bus_id, unit_id, index);
860         goto fail;
861     }
862 
863     /* Serial number */
864     serial = qemu_opt_get(legacy_opts, "serial");
865 
866     /* no id supplied -> create one */
867     if (qemu_opts_id(all_opts) == NULL) {
868         char *new_id;
869         const char *mediastr = "";
870         if (type == IF_IDE || type == IF_SCSI) {
871             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
872         }
873         if (max_devs) {
874             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
875                                      mediastr, unit_id);
876         } else {
877             new_id = g_strdup_printf("%s%s%i", if_name[type],
878                                      mediastr, unit_id);
879         }
880         qdict_put(bs_opts, "id", qstring_from_str(new_id));
881         g_free(new_id);
882     }
883 
884     /* Add virtio block device */
885     devaddr = qemu_opt_get(legacy_opts, "addr");
886     if (devaddr && type != IF_VIRTIO) {
887         error_report("addr is not supported by this bus type");
888         goto fail;
889     }
890 
891     if (type == IF_VIRTIO) {
892         QemuOpts *devopts;
893         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
894                                    &error_abort);
895         if (arch_type == QEMU_ARCH_S390X) {
896             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
897         } else {
898             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
899         }
900         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
901         if (devaddr) {
902             qemu_opt_set(devopts, "addr", devaddr);
903         }
904     }
905 
906     filename = qemu_opt_get(legacy_opts, "file");
907 
908     /* Check werror/rerror compatibility with if=... */
909     werror = qemu_opt_get(legacy_opts, "werror");
910     if (werror != NULL) {
911         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
912             type != IF_NONE) {
913             error_report("werror is not supported by this bus type");
914             goto fail;
915         }
916         qdict_put(bs_opts, "werror", qstring_from_str(werror));
917     }
918 
919     rerror = qemu_opt_get(legacy_opts, "rerror");
920     if (rerror != NULL) {
921         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
922             type != IF_NONE) {
923             error_report("rerror is not supported by this bus type");
924             goto fail;
925         }
926         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
927     }
928 
929     /* Actual block device init: Functionality shared with blockdev-add */
930     dinfo = blockdev_init(filename, bs_opts, &local_err);
931     bs_opts = NULL;
932     if (dinfo == NULL) {
933         if (local_err) {
934             error_report("%s", error_get_pretty(local_err));
935             error_free(local_err);
936         }
937         goto fail;
938     } else {
939         assert(!local_err);
940     }
941 
942     /* Set legacy DriveInfo fields */
943     dinfo->enable_auto_del = true;
944     dinfo->opts = all_opts;
945 
946     dinfo->cyls = cyls;
947     dinfo->heads = heads;
948     dinfo->secs = secs;
949     dinfo->trans = translation;
950 
951     dinfo->type = type;
952     dinfo->bus = bus_id;
953     dinfo->unit = unit_id;
954     dinfo->devaddr = devaddr;
955 
956     dinfo->serial = g_strdup(serial);
957 
958     switch(type) {
959     case IF_IDE:
960     case IF_SCSI:
961     case IF_XEN:
962     case IF_NONE:
963         dinfo->media_cd = media == MEDIA_CDROM;
964         break;
965     default:
966         break;
967     }
968 
969 fail:
970     qemu_opts_del(legacy_opts);
971     QDECREF(bs_opts);
972     return dinfo;
973 }
974 
975 void do_commit(Monitor *mon, const QDict *qdict)
976 {
977     const char *device = qdict_get_str(qdict, "device");
978     BlockDriverState *bs;
979     int ret;
980 
981     if (!strcmp(device, "all")) {
982         ret = bdrv_commit_all();
983     } else {
984         bs = bdrv_find(device);
985         if (!bs) {
986             monitor_printf(mon, "Device '%s' not found\n", device);
987             return;
988         }
989         ret = bdrv_commit(bs);
990     }
991     if (ret < 0) {
992         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
993                        strerror(-ret));
994     }
995 }
996 
997 static void blockdev_do_action(int kind, void *data, Error **errp)
998 {
999     TransactionAction action;
1000     TransactionActionList list;
1001 
1002     action.kind = kind;
1003     action.data = data;
1004     list.value = &action;
1005     list.next = NULL;
1006     qmp_transaction(&list, errp);
1007 }
1008 
1009 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1010                                 bool has_node_name, const char *node_name,
1011                                 const char *snapshot_file,
1012                                 bool has_snapshot_node_name,
1013                                 const char *snapshot_node_name,
1014                                 bool has_format, const char *format,
1015                                 bool has_mode, NewImageMode mode, Error **errp)
1016 {
1017     BlockdevSnapshot snapshot = {
1018         .has_device = has_device,
1019         .device = (char *) device,
1020         .has_node_name = has_node_name,
1021         .node_name = (char *) node_name,
1022         .snapshot_file = (char *) snapshot_file,
1023         .has_snapshot_node_name = has_snapshot_node_name,
1024         .snapshot_node_name = (char *) snapshot_node_name,
1025         .has_format = has_format,
1026         .format = (char *) format,
1027         .has_mode = has_mode,
1028         .mode = mode,
1029     };
1030     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1031                        &snapshot, errp);
1032 }
1033 
1034 void qmp_blockdev_snapshot_internal_sync(const char *device,
1035                                          const char *name,
1036                                          Error **errp)
1037 {
1038     BlockdevSnapshotInternal snapshot = {
1039         .device = (char *) device,
1040         .name = (char *) name
1041     };
1042 
1043     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1044                        &snapshot, errp);
1045 }
1046 
1047 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1048                                                          bool has_id,
1049                                                          const char *id,
1050                                                          bool has_name,
1051                                                          const char *name,
1052                                                          Error **errp)
1053 {
1054     BlockDriverState *bs = bdrv_find(device);
1055     QEMUSnapshotInfo sn;
1056     Error *local_err = NULL;
1057     SnapshotInfo *info = NULL;
1058     int ret;
1059 
1060     if (!bs) {
1061         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1062         return NULL;
1063     }
1064 
1065     if (!has_id) {
1066         id = NULL;
1067     }
1068 
1069     if (!has_name) {
1070         name = NULL;
1071     }
1072 
1073     if (!id && !name) {
1074         error_setg(errp, "Name or id must be provided");
1075         return NULL;
1076     }
1077 
1078     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1079     if (local_err) {
1080         error_propagate(errp, local_err);
1081         return NULL;
1082     }
1083     if (!ret) {
1084         error_setg(errp,
1085                    "Snapshot with id '%s' and name '%s' does not exist on "
1086                    "device '%s'",
1087                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1088         return NULL;
1089     }
1090 
1091     bdrv_snapshot_delete(bs, id, name, &local_err);
1092     if (local_err) {
1093         error_propagate(errp, local_err);
1094         return NULL;
1095     }
1096 
1097     info = g_new0(SnapshotInfo, 1);
1098     info->id = g_strdup(sn.id_str);
1099     info->name = g_strdup(sn.name);
1100     info->date_nsec = sn.date_nsec;
1101     info->date_sec = sn.date_sec;
1102     info->vm_state_size = sn.vm_state_size;
1103     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1104     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1105 
1106     return info;
1107 }
1108 
1109 /* New and old BlockDriverState structs for group snapshots */
1110 
1111 typedef struct BlkTransactionState BlkTransactionState;
1112 
1113 /* Only prepare() may fail. In a single transaction, only one of commit() or
1114    abort() will be called, clean() will always be called if it present. */
1115 typedef struct BdrvActionOps {
1116     /* Size of state struct, in bytes. */
1117     size_t instance_size;
1118     /* Prepare the work, must NOT be NULL. */
1119     void (*prepare)(BlkTransactionState *common, Error **errp);
1120     /* Commit the changes, can be NULL. */
1121     void (*commit)(BlkTransactionState *common);
1122     /* Abort the changes on fail, can be NULL. */
1123     void (*abort)(BlkTransactionState *common);
1124     /* Clean up resource in the end, can be NULL. */
1125     void (*clean)(BlkTransactionState *common);
1126 } BdrvActionOps;
1127 
1128 /*
1129  * This structure must be arranged as first member in child type, assuming
1130  * that compiler will also arrange it to the same address with parent instance.
1131  * Later it will be used in free().
1132  */
1133 struct BlkTransactionState {
1134     TransactionAction *action;
1135     const BdrvActionOps *ops;
1136     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1137 };
1138 
1139 /* internal snapshot private data */
1140 typedef struct InternalSnapshotState {
1141     BlkTransactionState common;
1142     BlockDriverState *bs;
1143     QEMUSnapshotInfo sn;
1144 } InternalSnapshotState;
1145 
1146 static void internal_snapshot_prepare(BlkTransactionState *common,
1147                                       Error **errp)
1148 {
1149     Error *local_err = NULL;
1150     const char *device;
1151     const char *name;
1152     BlockDriverState *bs;
1153     QEMUSnapshotInfo old_sn, *sn;
1154     bool ret;
1155     qemu_timeval tv;
1156     BlockdevSnapshotInternal *internal;
1157     InternalSnapshotState *state;
1158     int ret1;
1159 
1160     g_assert(common->action->kind ==
1161              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1162     internal = common->action->blockdev_snapshot_internal_sync;
1163     state = DO_UPCAST(InternalSnapshotState, common, common);
1164 
1165     /* 1. parse input */
1166     device = internal->device;
1167     name = internal->name;
1168 
1169     /* 2. check for validation */
1170     bs = bdrv_find(device);
1171     if (!bs) {
1172         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1173         return;
1174     }
1175 
1176     if (!bdrv_is_inserted(bs)) {
1177         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1178         return;
1179     }
1180 
1181     if (bdrv_is_read_only(bs)) {
1182         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1183         return;
1184     }
1185 
1186     if (!bdrv_can_snapshot(bs)) {
1187         error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1188                   bs->drv->format_name, device, "internal snapshot");
1189         return;
1190     }
1191 
1192     if (!strlen(name)) {
1193         error_setg(errp, "Name is empty");
1194         return;
1195     }
1196 
1197     /* check whether a snapshot with name exist */
1198     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1199                                             &local_err);
1200     if (local_err) {
1201         error_propagate(errp, local_err);
1202         return;
1203     } else if (ret) {
1204         error_setg(errp,
1205                    "Snapshot with name '%s' already exists on device '%s'",
1206                    name, device);
1207         return;
1208     }
1209 
1210     /* 3. take the snapshot */
1211     sn = &state->sn;
1212     pstrcpy(sn->name, sizeof(sn->name), name);
1213     qemu_gettimeofday(&tv);
1214     sn->date_sec = tv.tv_sec;
1215     sn->date_nsec = tv.tv_usec * 1000;
1216     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1217 
1218     ret1 = bdrv_snapshot_create(bs, sn);
1219     if (ret1 < 0) {
1220         error_setg_errno(errp, -ret1,
1221                          "Failed to create snapshot '%s' on device '%s'",
1222                          name, device);
1223         return;
1224     }
1225 
1226     /* 4. succeed, mark a snapshot is created */
1227     state->bs = bs;
1228 }
1229 
1230 static void internal_snapshot_abort(BlkTransactionState *common)
1231 {
1232     InternalSnapshotState *state =
1233                              DO_UPCAST(InternalSnapshotState, common, common);
1234     BlockDriverState *bs = state->bs;
1235     QEMUSnapshotInfo *sn = &state->sn;
1236     Error *local_error = NULL;
1237 
1238     if (!bs) {
1239         return;
1240     }
1241 
1242     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1243         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1244                      "device '%s' in abort: %s",
1245                      sn->id_str,
1246                      sn->name,
1247                      bdrv_get_device_name(bs),
1248                      error_get_pretty(local_error));
1249         error_free(local_error);
1250     }
1251 }
1252 
1253 /* external snapshot private data */
1254 typedef struct ExternalSnapshotState {
1255     BlkTransactionState common;
1256     BlockDriverState *old_bs;
1257     BlockDriverState *new_bs;
1258 } ExternalSnapshotState;
1259 
1260 static void external_snapshot_prepare(BlkTransactionState *common,
1261                                       Error **errp)
1262 {
1263     BlockDriver *drv;
1264     int flags, ret;
1265     QDict *options = NULL;
1266     Error *local_err = NULL;
1267     bool has_device = false;
1268     const char *device;
1269     bool has_node_name = false;
1270     const char *node_name;
1271     bool has_snapshot_node_name = false;
1272     const char *snapshot_node_name;
1273     const char *new_image_file;
1274     const char *format = "qcow2";
1275     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1276     ExternalSnapshotState *state =
1277                              DO_UPCAST(ExternalSnapshotState, common, common);
1278     TransactionAction *action = common->action;
1279 
1280     /* get parameters */
1281     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1282 
1283     has_device = action->blockdev_snapshot_sync->has_device;
1284     device = action->blockdev_snapshot_sync->device;
1285     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1286     node_name = action->blockdev_snapshot_sync->node_name;
1287     has_snapshot_node_name =
1288         action->blockdev_snapshot_sync->has_snapshot_node_name;
1289     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1290 
1291     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1292     if (action->blockdev_snapshot_sync->has_format) {
1293         format = action->blockdev_snapshot_sync->format;
1294     }
1295     if (action->blockdev_snapshot_sync->has_mode) {
1296         mode = action->blockdev_snapshot_sync->mode;
1297     }
1298 
1299     /* start processing */
1300     drv = bdrv_find_format(format);
1301     if (!drv) {
1302         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1303         return;
1304     }
1305 
1306     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1307                                    has_node_name ? node_name : NULL,
1308                                    &local_err);
1309     if (local_err) {
1310         error_propagate(errp, local_err);
1311         return;
1312     }
1313 
1314     if (has_node_name && !has_snapshot_node_name) {
1315         error_setg(errp, "New snapshot node name missing");
1316         return;
1317     }
1318 
1319     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1320         error_setg(errp, "New snapshot node name already existing");
1321         return;
1322     }
1323 
1324     if (!bdrv_is_inserted(state->old_bs)) {
1325         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1326         return;
1327     }
1328 
1329     if (bdrv_op_is_blocked(state->old_bs,
1330                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1331         return;
1332     }
1333 
1334     if (!bdrv_is_read_only(state->old_bs)) {
1335         if (bdrv_flush(state->old_bs)) {
1336             error_set(errp, QERR_IO_ERROR);
1337             return;
1338         }
1339     }
1340 
1341     if (!bdrv_is_first_non_filter(state->old_bs)) {
1342         error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1343         return;
1344     }
1345 
1346     flags = state->old_bs->open_flags;
1347 
1348     /* create new image w/backing file */
1349     if (mode != NEW_IMAGE_MODE_EXISTING) {
1350         bdrv_img_create(new_image_file, format,
1351                         state->old_bs->filename,
1352                         state->old_bs->drv->format_name,
1353                         NULL, -1, flags, &local_err, false);
1354         if (local_err) {
1355             error_propagate(errp, local_err);
1356             return;
1357         }
1358     }
1359 
1360     if (has_snapshot_node_name) {
1361         options = qdict_new();
1362         qdict_put(options, "node-name",
1363                   qstring_from_str(snapshot_node_name));
1364     }
1365 
1366     /* TODO Inherit bs->options or only take explicit options with an
1367      * extended QMP command? */
1368     assert(state->new_bs == NULL);
1369     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1370                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1371     /* We will manually add the backing_hd field to the bs later */
1372     if (ret != 0) {
1373         error_propagate(errp, local_err);
1374     }
1375 }
1376 
1377 static void external_snapshot_commit(BlkTransactionState *common)
1378 {
1379     ExternalSnapshotState *state =
1380                              DO_UPCAST(ExternalSnapshotState, common, common);
1381 
1382     /* This removes our old bs and adds the new bs */
1383     bdrv_append(state->new_bs, state->old_bs);
1384     /* We don't need (or want) to use the transactional
1385      * bdrv_reopen_multiple() across all the entries at once, because we
1386      * don't want to abort all of them if one of them fails the reopen */
1387     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1388                 NULL);
1389 }
1390 
1391 static void external_snapshot_abort(BlkTransactionState *common)
1392 {
1393     ExternalSnapshotState *state =
1394                              DO_UPCAST(ExternalSnapshotState, common, common);
1395     if (state->new_bs) {
1396         bdrv_unref(state->new_bs);
1397     }
1398 }
1399 
1400 typedef struct DriveBackupState {
1401     BlkTransactionState common;
1402     BlockDriverState *bs;
1403     BlockJob *job;
1404 } DriveBackupState;
1405 
1406 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1407 {
1408     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1409     DriveBackup *backup;
1410     Error *local_err = NULL;
1411 
1412     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1413     backup = common->action->drive_backup;
1414 
1415     qmp_drive_backup(backup->device, backup->target,
1416                      backup->has_format, backup->format,
1417                      backup->sync,
1418                      backup->has_mode, backup->mode,
1419                      backup->has_speed, backup->speed,
1420                      backup->has_on_source_error, backup->on_source_error,
1421                      backup->has_on_target_error, backup->on_target_error,
1422                      &local_err);
1423     if (local_err) {
1424         error_propagate(errp, local_err);
1425         state->bs = NULL;
1426         state->job = NULL;
1427         return;
1428     }
1429 
1430     state->bs = bdrv_find(backup->device);
1431     state->job = state->bs->job;
1432 }
1433 
1434 static void drive_backup_abort(BlkTransactionState *common)
1435 {
1436     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1437     BlockDriverState *bs = state->bs;
1438 
1439     /* Only cancel if it's the job we started */
1440     if (bs && bs->job && bs->job == state->job) {
1441         block_job_cancel_sync(bs->job);
1442     }
1443 }
1444 
1445 static void abort_prepare(BlkTransactionState *common, Error **errp)
1446 {
1447     error_setg(errp, "Transaction aborted using Abort action");
1448 }
1449 
1450 static void abort_commit(BlkTransactionState *common)
1451 {
1452     g_assert_not_reached(); /* this action never succeeds */
1453 }
1454 
1455 static const BdrvActionOps actions[] = {
1456     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1457         .instance_size = sizeof(ExternalSnapshotState),
1458         .prepare  = external_snapshot_prepare,
1459         .commit   = external_snapshot_commit,
1460         .abort = external_snapshot_abort,
1461     },
1462     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1463         .instance_size = sizeof(DriveBackupState),
1464         .prepare = drive_backup_prepare,
1465         .abort = drive_backup_abort,
1466     },
1467     [TRANSACTION_ACTION_KIND_ABORT] = {
1468         .instance_size = sizeof(BlkTransactionState),
1469         .prepare = abort_prepare,
1470         .commit = abort_commit,
1471     },
1472     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1473         .instance_size = sizeof(InternalSnapshotState),
1474         .prepare  = internal_snapshot_prepare,
1475         .abort = internal_snapshot_abort,
1476     },
1477 };
1478 
1479 /*
1480  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1481  *  then we do not pivot any of the devices in the group, and abandon the
1482  *  snapshots
1483  */
1484 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1485 {
1486     TransactionActionList *dev_entry = dev_list;
1487     BlkTransactionState *state, *next;
1488     Error *local_err = NULL;
1489 
1490     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1491     QSIMPLEQ_INIT(&snap_bdrv_states);
1492 
1493     /* drain all i/o before any snapshots */
1494     bdrv_drain_all();
1495 
1496     /* We don't do anything in this loop that commits us to the snapshot */
1497     while (NULL != dev_entry) {
1498         TransactionAction *dev_info = NULL;
1499         const BdrvActionOps *ops;
1500 
1501         dev_info = dev_entry->value;
1502         dev_entry = dev_entry->next;
1503 
1504         assert(dev_info->kind < ARRAY_SIZE(actions));
1505 
1506         ops = &actions[dev_info->kind];
1507         assert(ops->instance_size > 0);
1508 
1509         state = g_malloc0(ops->instance_size);
1510         state->ops = ops;
1511         state->action = dev_info;
1512         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1513 
1514         state->ops->prepare(state, &local_err);
1515         if (local_err) {
1516             error_propagate(errp, local_err);
1517             goto delete_and_fail;
1518         }
1519     }
1520 
1521     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1522         if (state->ops->commit) {
1523             state->ops->commit(state);
1524         }
1525     }
1526 
1527     /* success */
1528     goto exit;
1529 
1530 delete_and_fail:
1531     /*
1532     * failure, and it is all-or-none; abandon each new bs, and keep using
1533     * the original bs for all images
1534     */
1535     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1536         if (state->ops->abort) {
1537             state->ops->abort(state);
1538         }
1539     }
1540 exit:
1541     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1542         if (state->ops->clean) {
1543             state->ops->clean(state);
1544         }
1545         g_free(state);
1546     }
1547 }
1548 
1549 
1550 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1551 {
1552     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1553         return;
1554     }
1555     if (!bdrv_dev_has_removable_media(bs)) {
1556         error_setg(errp, "Device '%s' is not removable",
1557                    bdrv_get_device_name(bs));
1558         return;
1559     }
1560 
1561     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1562         bdrv_dev_eject_request(bs, force);
1563         if (!force) {
1564             error_setg(errp, "Device '%s' is locked",
1565                        bdrv_get_device_name(bs));
1566             return;
1567         }
1568     }
1569 
1570     bdrv_close(bs);
1571 }
1572 
1573 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1574 {
1575     BlockDriverState *bs;
1576 
1577     bs = bdrv_find(device);
1578     if (!bs) {
1579         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1580         return;
1581     }
1582 
1583     eject_device(bs, force, errp);
1584 }
1585 
1586 void qmp_block_passwd(bool has_device, const char *device,
1587                       bool has_node_name, const char *node_name,
1588                       const char *password, Error **errp)
1589 {
1590     Error *local_err = NULL;
1591     BlockDriverState *bs;
1592     int err;
1593 
1594     bs = bdrv_lookup_bs(has_device ? device : NULL,
1595                         has_node_name ? node_name : NULL,
1596                         &local_err);
1597     if (local_err) {
1598         error_propagate(errp, local_err);
1599         return;
1600     }
1601 
1602     err = bdrv_set_key(bs, password);
1603     if (err == -EINVAL) {
1604         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1605         return;
1606     } else if (err < 0) {
1607         error_set(errp, QERR_INVALID_PASSWORD);
1608         return;
1609     }
1610 }
1611 
1612 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1613                                     int bdrv_flags, BlockDriver *drv,
1614                                     const char *password, Error **errp)
1615 {
1616     Error *local_err = NULL;
1617     int ret;
1618 
1619     ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1620     if (ret < 0) {
1621         error_propagate(errp, local_err);
1622         return;
1623     }
1624 
1625     if (bdrv_key_required(bs)) {
1626         if (password) {
1627             if (bdrv_set_key(bs, password) < 0) {
1628                 error_set(errp, QERR_INVALID_PASSWORD);
1629             }
1630         } else {
1631             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1632                       bdrv_get_encrypted_filename(bs));
1633         }
1634     } else if (password) {
1635         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1636     }
1637 }
1638 
1639 void qmp_change_blockdev(const char *device, const char *filename,
1640                          const char *format, Error **errp)
1641 {
1642     BlockDriverState *bs;
1643     BlockDriver *drv = NULL;
1644     int bdrv_flags;
1645     Error *err = NULL;
1646 
1647     bs = bdrv_find(device);
1648     if (!bs) {
1649         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1650         return;
1651     }
1652 
1653     if (format) {
1654         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1655         if (!drv) {
1656             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1657             return;
1658         }
1659     }
1660 
1661     eject_device(bs, 0, &err);
1662     if (err) {
1663         error_propagate(errp, err);
1664         return;
1665     }
1666 
1667     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1668     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1669 
1670     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1671 }
1672 
1673 /* throttling disk I/O limits */
1674 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1675                                int64_t bps_wr,
1676                                int64_t iops,
1677                                int64_t iops_rd,
1678                                int64_t iops_wr,
1679                                bool has_bps_max,
1680                                int64_t bps_max,
1681                                bool has_bps_rd_max,
1682                                int64_t bps_rd_max,
1683                                bool has_bps_wr_max,
1684                                int64_t bps_wr_max,
1685                                bool has_iops_max,
1686                                int64_t iops_max,
1687                                bool has_iops_rd_max,
1688                                int64_t iops_rd_max,
1689                                bool has_iops_wr_max,
1690                                int64_t iops_wr_max,
1691                                bool has_iops_size,
1692                                int64_t iops_size, Error **errp)
1693 {
1694     ThrottleConfig cfg;
1695     BlockDriverState *bs;
1696     AioContext *aio_context;
1697 
1698     bs = bdrv_find(device);
1699     if (!bs) {
1700         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1701         return;
1702     }
1703 
1704     memset(&cfg, 0, sizeof(cfg));
1705     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1706     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1707     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1708 
1709     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1710     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1711     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1712 
1713     if (has_bps_max) {
1714         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1715     }
1716     if (has_bps_rd_max) {
1717         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1718     }
1719     if (has_bps_wr_max) {
1720         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1721     }
1722     if (has_iops_max) {
1723         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1724     }
1725     if (has_iops_rd_max) {
1726         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1727     }
1728     if (has_iops_wr_max) {
1729         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1730     }
1731 
1732     if (has_iops_size) {
1733         cfg.op_size = iops_size;
1734     }
1735 
1736     if (!check_throttle_config(&cfg, errp)) {
1737         return;
1738     }
1739 
1740     aio_context = bdrv_get_aio_context(bs);
1741     aio_context_acquire(aio_context);
1742 
1743     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1744         bdrv_io_limits_enable(bs);
1745     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1746         bdrv_io_limits_disable(bs);
1747     }
1748 
1749     if (bs->io_limits_enabled) {
1750         bdrv_set_io_limits(bs, &cfg);
1751     }
1752 
1753     aio_context_release(aio_context);
1754 }
1755 
1756 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1757 {
1758     const char *id = qdict_get_str(qdict, "id");
1759     BlockDriverState *bs;
1760     AioContext *aio_context;
1761     Error *local_err = NULL;
1762 
1763     bs = bdrv_find(id);
1764     if (!bs) {
1765         error_report("Device '%s' not found", id);
1766         return -1;
1767     }
1768 
1769     aio_context = bdrv_get_aio_context(bs);
1770     aio_context_acquire(aio_context);
1771 
1772     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1773         error_report("%s", error_get_pretty(local_err));
1774         error_free(local_err);
1775         aio_context_release(aio_context);
1776         return -1;
1777     }
1778 
1779     /* quiesce block driver; prevent further io */
1780     bdrv_drain_all();
1781     bdrv_flush(bs);
1782     bdrv_close(bs);
1783 
1784     /* if we have a device attached to this BlockDriverState
1785      * then we need to make the drive anonymous until the device
1786      * can be removed.  If this is a drive with no device backing
1787      * then we can just get rid of the block driver state right here.
1788      */
1789     if (bdrv_get_attached_dev(bs)) {
1790         bdrv_make_anon(bs);
1791 
1792         /* Further I/O must not pause the guest */
1793         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1794                           BLOCKDEV_ON_ERROR_REPORT);
1795     } else {
1796         drive_del(drive_get_by_blockdev(bs));
1797     }
1798 
1799     aio_context_release(aio_context);
1800     return 0;
1801 }
1802 
1803 void qmp_block_resize(bool has_device, const char *device,
1804                       bool has_node_name, const char *node_name,
1805                       int64_t size, Error **errp)
1806 {
1807     Error *local_err = NULL;
1808     BlockDriverState *bs;
1809     AioContext *aio_context;
1810     int ret;
1811 
1812     bs = bdrv_lookup_bs(has_device ? device : NULL,
1813                         has_node_name ? node_name : NULL,
1814                         &local_err);
1815     if (local_err) {
1816         error_propagate(errp, local_err);
1817         return;
1818     }
1819 
1820     aio_context = bdrv_get_aio_context(bs);
1821     aio_context_acquire(aio_context);
1822 
1823     if (!bdrv_is_first_non_filter(bs)) {
1824         error_set(errp, QERR_FEATURE_DISABLED, "resize");
1825         goto out;
1826     }
1827 
1828     if (size < 0) {
1829         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1830         goto out;
1831     }
1832 
1833     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
1834         error_set(errp, QERR_DEVICE_IN_USE, device);
1835         goto out;
1836     }
1837 
1838     /* complete all in-flight operations before resizing the device */
1839     bdrv_drain_all();
1840 
1841     ret = bdrv_truncate(bs, size);
1842     switch (ret) {
1843     case 0:
1844         break;
1845     case -ENOMEDIUM:
1846         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1847         break;
1848     case -ENOTSUP:
1849         error_set(errp, QERR_UNSUPPORTED);
1850         break;
1851     case -EACCES:
1852         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1853         break;
1854     case -EBUSY:
1855         error_set(errp, QERR_DEVICE_IN_USE, device);
1856         break;
1857     default:
1858         error_setg_errno(errp, -ret, "Could not resize");
1859         break;
1860     }
1861 
1862 out:
1863     aio_context_release(aio_context);
1864 }
1865 
1866 static void block_job_cb(void *opaque, int ret)
1867 {
1868     BlockDriverState *bs = opaque;
1869     const char *msg = NULL;
1870 
1871     trace_block_job_cb(bs, bs->job, ret);
1872 
1873     assert(bs->job);
1874 
1875     if (ret < 0) {
1876         msg = strerror(-ret);
1877     }
1878 
1879     if (block_job_is_cancelled(bs->job)) {
1880         block_job_event_cancelled(bs->job);
1881     } else {
1882         block_job_event_completed(bs->job, msg);
1883     }
1884 
1885     bdrv_put_ref_bh_schedule(bs);
1886 }
1887 
1888 void qmp_block_stream(const char *device,
1889                       bool has_base, const char *base,
1890                       bool has_backing_file, const char *backing_file,
1891                       bool has_speed, int64_t speed,
1892                       bool has_on_error, BlockdevOnError on_error,
1893                       Error **errp)
1894 {
1895     BlockDriverState *bs;
1896     BlockDriverState *base_bs = NULL;
1897     Error *local_err = NULL;
1898     const char *base_name = NULL;
1899 
1900     if (!has_on_error) {
1901         on_error = BLOCKDEV_ON_ERROR_REPORT;
1902     }
1903 
1904     bs = bdrv_find(device);
1905     if (!bs) {
1906         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1907         return;
1908     }
1909 
1910     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
1911         return;
1912     }
1913 
1914     if (has_base) {
1915         base_bs = bdrv_find_backing_image(bs, base);
1916         if (base_bs == NULL) {
1917             error_set(errp, QERR_BASE_NOT_FOUND, base);
1918             return;
1919         }
1920         base_name = base;
1921     }
1922 
1923     /* if we are streaming the entire chain, the result will have no backing
1924      * file, and specifying one is therefore an error */
1925     if (base_bs == NULL && has_backing_file) {
1926         error_setg(errp, "backing file specified, but streaming the "
1927                          "entire chain");
1928         return;
1929     }
1930 
1931     /* backing_file string overrides base bs filename */
1932     base_name = has_backing_file ? backing_file : base_name;
1933 
1934     stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
1935                  on_error, block_job_cb, bs, &local_err);
1936     if (local_err) {
1937         error_propagate(errp, local_err);
1938         return;
1939     }
1940 
1941     trace_qmp_block_stream(bs, bs->job);
1942 }
1943 
1944 void qmp_block_commit(const char *device,
1945                       bool has_base, const char *base,
1946                       bool has_top, const char *top,
1947                       bool has_backing_file, const char *backing_file,
1948                       bool has_speed, int64_t speed,
1949                       Error **errp)
1950 {
1951     BlockDriverState *bs;
1952     BlockDriverState *base_bs, *top_bs;
1953     Error *local_err = NULL;
1954     /* This will be part of the QMP command, if/when the
1955      * BlockdevOnError change for blkmirror makes it in
1956      */
1957     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1958 
1959     if (!has_speed) {
1960         speed = 0;
1961     }
1962 
1963     /* drain all i/o before commits */
1964     bdrv_drain_all();
1965 
1966     /* Important Note:
1967      *  libvirt relies on the DeviceNotFound error class in order to probe for
1968      *  live commit feature versions; for this to work, we must make sure to
1969      *  perform the device lookup before any generic errors that may occur in a
1970      *  scenario in which all optional arguments are omitted. */
1971     bs = bdrv_find(device);
1972     if (!bs) {
1973         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1974         return;
1975     }
1976 
1977     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, errp)) {
1978         return;
1979     }
1980 
1981     /* default top_bs is the active layer */
1982     top_bs = bs;
1983 
1984     if (has_top && top) {
1985         if (strcmp(bs->filename, top) != 0) {
1986             top_bs = bdrv_find_backing_image(bs, top);
1987         }
1988     }
1989 
1990     if (top_bs == NULL) {
1991         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1992         return;
1993     }
1994 
1995     if (has_base && base) {
1996         base_bs = bdrv_find_backing_image(top_bs, base);
1997     } else {
1998         base_bs = bdrv_find_base(top_bs);
1999     }
2000 
2001     if (base_bs == NULL) {
2002         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2003         return;
2004     }
2005 
2006     /* Do not allow attempts to commit an image into itself */
2007     if (top_bs == base_bs) {
2008         error_setg(errp, "cannot commit an image into itself");
2009         return;
2010     }
2011 
2012     if (top_bs == bs) {
2013         if (has_backing_file) {
2014             error_setg(errp, "'backing-file' specified,"
2015                              " but 'top' is the active layer");
2016             return;
2017         }
2018         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2019                             bs, &local_err);
2020     } else {
2021         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2022                      has_backing_file ? backing_file : NULL, &local_err);
2023     }
2024     if (local_err != NULL) {
2025         error_propagate(errp, local_err);
2026         return;
2027     }
2028 }
2029 
2030 void qmp_drive_backup(const char *device, const char *target,
2031                       bool has_format, const char *format,
2032                       enum MirrorSyncMode sync,
2033                       bool has_mode, enum NewImageMode mode,
2034                       bool has_speed, int64_t speed,
2035                       bool has_on_source_error, BlockdevOnError on_source_error,
2036                       bool has_on_target_error, BlockdevOnError on_target_error,
2037                       Error **errp)
2038 {
2039     BlockDriverState *bs;
2040     BlockDriverState *target_bs;
2041     BlockDriverState *source = NULL;
2042     BlockDriver *drv = NULL;
2043     Error *local_err = NULL;
2044     int flags;
2045     int64_t size;
2046     int ret;
2047 
2048     if (!has_speed) {
2049         speed = 0;
2050     }
2051     if (!has_on_source_error) {
2052         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2053     }
2054     if (!has_on_target_error) {
2055         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2056     }
2057     if (!has_mode) {
2058         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2059     }
2060 
2061     bs = bdrv_find(device);
2062     if (!bs) {
2063         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2064         return;
2065     }
2066 
2067     if (!bdrv_is_inserted(bs)) {
2068         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2069         return;
2070     }
2071 
2072     if (!has_format) {
2073         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2074     }
2075     if (format) {
2076         drv = bdrv_find_format(format);
2077         if (!drv) {
2078             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2079             return;
2080         }
2081     }
2082 
2083     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2084         return;
2085     }
2086 
2087     flags = bs->open_flags | BDRV_O_RDWR;
2088 
2089     /* See if we have a backing HD we can use to create our new image
2090      * on top of. */
2091     if (sync == MIRROR_SYNC_MODE_TOP) {
2092         source = bs->backing_hd;
2093         if (!source) {
2094             sync = MIRROR_SYNC_MODE_FULL;
2095         }
2096     }
2097     if (sync == MIRROR_SYNC_MODE_NONE) {
2098         source = bs;
2099     }
2100 
2101     size = bdrv_getlength(bs);
2102     if (size < 0) {
2103         error_setg_errno(errp, -size, "bdrv_getlength failed");
2104         return;
2105     }
2106 
2107     if (mode != NEW_IMAGE_MODE_EXISTING) {
2108         assert(format && drv);
2109         if (source) {
2110             bdrv_img_create(target, format, source->filename,
2111                             source->drv->format_name, NULL,
2112                             size, flags, &local_err, false);
2113         } else {
2114             bdrv_img_create(target, format, NULL, NULL, NULL,
2115                             size, flags, &local_err, false);
2116         }
2117     }
2118 
2119     if (local_err) {
2120         error_propagate(errp, local_err);
2121         return;
2122     }
2123 
2124     target_bs = NULL;
2125     ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2126     if (ret < 0) {
2127         error_propagate(errp, local_err);
2128         return;
2129     }
2130 
2131     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2132                  block_job_cb, bs, &local_err);
2133     if (local_err != NULL) {
2134         bdrv_unref(target_bs);
2135         error_propagate(errp, local_err);
2136         return;
2137     }
2138 }
2139 
2140 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2141 {
2142     return bdrv_named_nodes_list();
2143 }
2144 
2145 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
2146 
2147 void qmp_drive_mirror(const char *device, const char *target,
2148                       bool has_format, const char *format,
2149                       bool has_node_name, const char *node_name,
2150                       bool has_replaces, const char *replaces,
2151                       enum MirrorSyncMode sync,
2152                       bool has_mode, enum NewImageMode mode,
2153                       bool has_speed, int64_t speed,
2154                       bool has_granularity, uint32_t granularity,
2155                       bool has_buf_size, int64_t buf_size,
2156                       bool has_on_source_error, BlockdevOnError on_source_error,
2157                       bool has_on_target_error, BlockdevOnError on_target_error,
2158                       Error **errp)
2159 {
2160     BlockDriverState *bs;
2161     BlockDriverState *source, *target_bs;
2162     BlockDriver *drv = NULL;
2163     Error *local_err = NULL;
2164     QDict *options = NULL;
2165     int flags;
2166     int64_t size;
2167     int ret;
2168 
2169     if (!has_speed) {
2170         speed = 0;
2171     }
2172     if (!has_on_source_error) {
2173         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2174     }
2175     if (!has_on_target_error) {
2176         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2177     }
2178     if (!has_mode) {
2179         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2180     }
2181     if (!has_granularity) {
2182         granularity = 0;
2183     }
2184     if (!has_buf_size) {
2185         buf_size = DEFAULT_MIRROR_BUF_SIZE;
2186     }
2187 
2188     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2189         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2190                   "a value in range [512B, 64MB]");
2191         return;
2192     }
2193     if (granularity & (granularity - 1)) {
2194         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity", "power of 2");
2195         return;
2196     }
2197 
2198     bs = bdrv_find(device);
2199     if (!bs) {
2200         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2201         return;
2202     }
2203 
2204     if (!bdrv_is_inserted(bs)) {
2205         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2206         return;
2207     }
2208 
2209     if (!has_format) {
2210         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2211     }
2212     if (format) {
2213         drv = bdrv_find_format(format);
2214         if (!drv) {
2215             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2216             return;
2217         }
2218     }
2219 
2220     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2221         return;
2222     }
2223 
2224     flags = bs->open_flags | BDRV_O_RDWR;
2225     source = bs->backing_hd;
2226     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2227         sync = MIRROR_SYNC_MODE_FULL;
2228     }
2229     if (sync == MIRROR_SYNC_MODE_NONE) {
2230         source = bs;
2231     }
2232 
2233     size = bdrv_getlength(bs);
2234     if (size < 0) {
2235         error_setg_errno(errp, -size, "bdrv_getlength failed");
2236         return;
2237     }
2238 
2239     if (has_replaces) {
2240         BlockDriverState *to_replace_bs;
2241 
2242         if (!has_node_name) {
2243             error_setg(errp, "a node-name must be provided when replacing a"
2244                              " named node of the graph");
2245             return;
2246         }
2247 
2248         to_replace_bs = check_to_replace_node(replaces, &local_err);
2249 
2250         if (!to_replace_bs) {
2251             error_propagate(errp, local_err);
2252             return;
2253         }
2254 
2255         if (size != bdrv_getlength(to_replace_bs)) {
2256             error_setg(errp, "cannot replace image with a mirror image of "
2257                              "different size");
2258             return;
2259         }
2260     }
2261 
2262     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2263         && mode != NEW_IMAGE_MODE_EXISTING)
2264     {
2265         /* create new image w/o backing file */
2266         assert(format && drv);
2267         bdrv_img_create(target, format,
2268                         NULL, NULL, NULL, size, flags, &local_err, false);
2269     } else {
2270         switch (mode) {
2271         case NEW_IMAGE_MODE_EXISTING:
2272             break;
2273         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2274             /* create new image with backing file */
2275             bdrv_img_create(target, format,
2276                             source->filename,
2277                             source->drv->format_name,
2278                             NULL, size, flags, &local_err, false);
2279             break;
2280         default:
2281             abort();
2282         }
2283     }
2284 
2285     if (local_err) {
2286         error_propagate(errp, local_err);
2287         return;
2288     }
2289 
2290     if (has_node_name) {
2291         options = qdict_new();
2292         qdict_put(options, "node-name", qstring_from_str(node_name));
2293     }
2294 
2295     /* Mirroring takes care of copy-on-write using the source's backing
2296      * file.
2297      */
2298     target_bs = NULL;
2299     ret = bdrv_open(&target_bs, target, NULL, options,
2300                     flags | BDRV_O_NO_BACKING, drv, &local_err);
2301     if (ret < 0) {
2302         error_propagate(errp, local_err);
2303         return;
2304     }
2305 
2306     /* pass the node name to replace to mirror start since it's loose coupling
2307      * and will allow to check whether the node still exist at mirror completion
2308      */
2309     mirror_start(bs, target_bs,
2310                  has_replaces ? replaces : NULL,
2311                  speed, granularity, buf_size, sync,
2312                  on_source_error, on_target_error,
2313                  block_job_cb, bs, &local_err);
2314     if (local_err != NULL) {
2315         bdrv_unref(target_bs);
2316         error_propagate(errp, local_err);
2317         return;
2318     }
2319 }
2320 
2321 static BlockJob *find_block_job(const char *device)
2322 {
2323     BlockDriverState *bs;
2324 
2325     bs = bdrv_find(device);
2326     if (!bs || !bs->job) {
2327         return NULL;
2328     }
2329     return bs->job;
2330 }
2331 
2332 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2333 {
2334     BlockJob *job = find_block_job(device);
2335 
2336     if (!job) {
2337         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2338         return;
2339     }
2340 
2341     block_job_set_speed(job, speed, errp);
2342 }
2343 
2344 void qmp_block_job_cancel(const char *device,
2345                           bool has_force, bool force, Error **errp)
2346 {
2347     BlockJob *job = find_block_job(device);
2348 
2349     if (!has_force) {
2350         force = false;
2351     }
2352 
2353     if (!job) {
2354         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2355         return;
2356     }
2357     if (job->paused && !force) {
2358         error_setg(errp, "The block job for device '%s' is currently paused",
2359                    device);
2360         return;
2361     }
2362 
2363     trace_qmp_block_job_cancel(job);
2364     block_job_cancel(job);
2365 }
2366 
2367 void qmp_block_job_pause(const char *device, Error **errp)
2368 {
2369     BlockJob *job = find_block_job(device);
2370 
2371     if (!job) {
2372         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2373         return;
2374     }
2375 
2376     trace_qmp_block_job_pause(job);
2377     block_job_pause(job);
2378 }
2379 
2380 void qmp_block_job_resume(const char *device, Error **errp)
2381 {
2382     BlockJob *job = find_block_job(device);
2383 
2384     if (!job) {
2385         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2386         return;
2387     }
2388 
2389     trace_qmp_block_job_resume(job);
2390     block_job_resume(job);
2391 }
2392 
2393 void qmp_block_job_complete(const char *device, Error **errp)
2394 {
2395     BlockJob *job = find_block_job(device);
2396 
2397     if (!job) {
2398         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2399         return;
2400     }
2401 
2402     trace_qmp_block_job_complete(job);
2403     block_job_complete(job, errp);
2404 }
2405 
2406 void qmp_change_backing_file(const char *device,
2407                              const char *image_node_name,
2408                              const char *backing_file,
2409                              Error **errp)
2410 {
2411     BlockDriverState *bs = NULL;
2412     BlockDriverState *image_bs = NULL;
2413     Error *local_err = NULL;
2414     bool ro;
2415     int open_flags;
2416     int ret;
2417 
2418     /* find the top layer BDS of the chain */
2419     bs = bdrv_find(device);
2420     if (!bs) {
2421         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2422         return;
2423     }
2424 
2425     image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
2426     if (local_err) {
2427         error_propagate(errp, local_err);
2428         return;
2429     }
2430 
2431     if (!image_bs) {
2432         error_setg(errp, "image file not found");
2433         return;
2434     }
2435 
2436     if (bdrv_find_base(image_bs) == image_bs) {
2437         error_setg(errp, "not allowing backing file change on an image "
2438                          "without a backing file");
2439         return;
2440     }
2441 
2442     /* even though we are not necessarily operating on bs, we need it to
2443      * determine if block ops are currently prohibited on the chain */
2444     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
2445         return;
2446     }
2447 
2448     /* final sanity check */
2449     if (!bdrv_chain_contains(bs, image_bs)) {
2450         error_setg(errp, "'%s' and image file are not in the same chain",
2451                    device);
2452         return;
2453     }
2454 
2455     /* if not r/w, reopen to make r/w */
2456     open_flags = image_bs->open_flags;
2457     ro = bdrv_is_read_only(image_bs);
2458 
2459     if (ro) {
2460         bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
2461         if (local_err) {
2462             error_propagate(errp, local_err);
2463             return;
2464         }
2465     }
2466 
2467     ret = bdrv_change_backing_file(image_bs, backing_file,
2468                                image_bs->drv ? image_bs->drv->format_name : "");
2469 
2470     if (ret < 0) {
2471         error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
2472                          backing_file);
2473         /* don't exit here, so we can try to restore open flags if
2474          * appropriate */
2475     }
2476 
2477     if (ro) {
2478         bdrv_reopen(image_bs, open_flags, &local_err);
2479         if (local_err) {
2480             error_propagate(errp, local_err); /* will preserve prior errp */
2481         }
2482     }
2483 }
2484 
2485 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2486 {
2487     QmpOutputVisitor *ov = qmp_output_visitor_new();
2488     DriveInfo *dinfo;
2489     QObject *obj;
2490     QDict *qdict;
2491     Error *local_err = NULL;
2492 
2493     /* Require an ID in the top level */
2494     if (!options->has_id) {
2495         error_setg(errp, "Block device needs an ID");
2496         goto fail;
2497     }
2498 
2499     /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
2500      * cache.direct=false instead of silently switching to aio=threads, except
2501      * when called from drive_new().
2502      *
2503      * For now, simply forbidding the combination for all drivers will do. */
2504     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2505         bool direct = options->has_cache &&
2506                       options->cache->has_direct &&
2507                       options->cache->direct;
2508         if (!direct) {
2509             error_setg(errp, "aio=native requires cache.direct=true");
2510             goto fail;
2511         }
2512     }
2513 
2514     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2515                                &options, NULL, &local_err);
2516     if (local_err) {
2517         error_propagate(errp, local_err);
2518         goto fail;
2519     }
2520 
2521     obj = qmp_output_get_qobject(ov);
2522     qdict = qobject_to_qdict(obj);
2523 
2524     qdict_flatten(qdict);
2525 
2526     dinfo = blockdev_init(NULL, qdict, &local_err);
2527     if (local_err) {
2528         error_propagate(errp, local_err);
2529         goto fail;
2530     }
2531 
2532     if (bdrv_key_required(dinfo->bdrv)) {
2533         drive_del(dinfo);
2534         error_setg(errp, "blockdev-add doesn't support encrypted devices");
2535         goto fail;
2536     }
2537 
2538 fail:
2539     qmp_output_visitor_cleanup(ov);
2540 }
2541 
2542 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2543 {
2544     BlockJobInfoList **prev = opaque;
2545     BlockJob *job = bs->job;
2546 
2547     if (job) {
2548         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2549         elem->value = block_job_query(bs->job);
2550         (*prev)->next = elem;
2551         *prev = elem;
2552     }
2553 }
2554 
2555 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2556 {
2557     /* Dummy is a fake list element for holding the head pointer */
2558     BlockJobInfoList dummy = {};
2559     BlockJobInfoList *prev = &dummy;
2560     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2561     return dummy.next;
2562 }
2563 
2564 QemuOptsList qemu_common_drive_opts = {
2565     .name = "drive",
2566     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2567     .desc = {
2568         {
2569             .name = "snapshot",
2570             .type = QEMU_OPT_BOOL,
2571             .help = "enable/disable snapshot mode",
2572         },{
2573             .name = "discard",
2574             .type = QEMU_OPT_STRING,
2575             .help = "discard operation (ignore/off, unmap/on)",
2576         },{
2577             .name = "cache.writeback",
2578             .type = QEMU_OPT_BOOL,
2579             .help = "enables writeback mode for any caches",
2580         },{
2581             .name = "cache.direct",
2582             .type = QEMU_OPT_BOOL,
2583             .help = "enables use of O_DIRECT (bypass the host page cache)",
2584         },{
2585             .name = "cache.no-flush",
2586             .type = QEMU_OPT_BOOL,
2587             .help = "ignore any flush requests for the device",
2588         },{
2589             .name = "aio",
2590             .type = QEMU_OPT_STRING,
2591             .help = "host AIO implementation (threads, native)",
2592         },{
2593             .name = "format",
2594             .type = QEMU_OPT_STRING,
2595             .help = "disk format (raw, qcow2, ...)",
2596         },{
2597             .name = "rerror",
2598             .type = QEMU_OPT_STRING,
2599             .help = "read error action",
2600         },{
2601             .name = "werror",
2602             .type = QEMU_OPT_STRING,
2603             .help = "write error action",
2604         },{
2605             .name = "read-only",
2606             .type = QEMU_OPT_BOOL,
2607             .help = "open drive file as read-only",
2608         },{
2609             .name = "throttling.iops-total",
2610             .type = QEMU_OPT_NUMBER,
2611             .help = "limit total I/O operations per second",
2612         },{
2613             .name = "throttling.iops-read",
2614             .type = QEMU_OPT_NUMBER,
2615             .help = "limit read operations per second",
2616         },{
2617             .name = "throttling.iops-write",
2618             .type = QEMU_OPT_NUMBER,
2619             .help = "limit write operations per second",
2620         },{
2621             .name = "throttling.bps-total",
2622             .type = QEMU_OPT_NUMBER,
2623             .help = "limit total bytes per second",
2624         },{
2625             .name = "throttling.bps-read",
2626             .type = QEMU_OPT_NUMBER,
2627             .help = "limit read bytes per second",
2628         },{
2629             .name = "throttling.bps-write",
2630             .type = QEMU_OPT_NUMBER,
2631             .help = "limit write bytes per second",
2632         },{
2633             .name = "throttling.iops-total-max",
2634             .type = QEMU_OPT_NUMBER,
2635             .help = "I/O operations burst",
2636         },{
2637             .name = "throttling.iops-read-max",
2638             .type = QEMU_OPT_NUMBER,
2639             .help = "I/O operations read burst",
2640         },{
2641             .name = "throttling.iops-write-max",
2642             .type = QEMU_OPT_NUMBER,
2643             .help = "I/O operations write burst",
2644         },{
2645             .name = "throttling.bps-total-max",
2646             .type = QEMU_OPT_NUMBER,
2647             .help = "total bytes burst",
2648         },{
2649             .name = "throttling.bps-read-max",
2650             .type = QEMU_OPT_NUMBER,
2651             .help = "total bytes read burst",
2652         },{
2653             .name = "throttling.bps-write-max",
2654             .type = QEMU_OPT_NUMBER,
2655             .help = "total bytes write burst",
2656         },{
2657             .name = "throttling.iops-size",
2658             .type = QEMU_OPT_NUMBER,
2659             .help = "when limiting by iops max size of an I/O in bytes",
2660         },{
2661             .name = "copy-on-read",
2662             .type = QEMU_OPT_BOOL,
2663             .help = "copy read data from backing file into image file",
2664         },{
2665             .name = "detect-zeroes",
2666             .type = QEMU_OPT_STRING,
2667             .help = "try to optimize zero writes (off, on, unmap)",
2668         },
2669         { /* end of list */ }
2670     },
2671 };
2672 
2673 QemuOptsList qemu_drive_opts = {
2674     .name = "drive",
2675     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2676     .desc = {
2677         /*
2678          * no elements => accept any params
2679          * validation will happen later
2680          */
2681         { /* end of list */ }
2682     },
2683 };
2684