blockdev.c

   1 /*
   2  * QEMU host block devices
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * This work is licensed under the terms of the GNU GPL, version 2 or
   7  * later.  See the COPYING file in the top-level directory.
   8  *
   9  * This file incorporates work covered by the following copyright and
  10  * permission notice:
  11  *
  12  * Copyright (c) 2003-2008 Fabrice Bellard
  13  *
  14  * Permission is hereby granted, free of charge, to any person obtaining a copy
  15  * of this software and associated documentation files (the "Software"), to deal
  16  * in the Software without restriction, including without limitation the rights
  17  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  18  * copies of the Software, and to permit persons to whom the Software is
  19  * furnished to do so, subject to the following conditions:
  20  *
  21  * The above copyright notice and this permission notice shall be included in
  22  * all copies or substantial portions of the Software.
  23  *
  24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  25  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  26  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  27  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  28  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  29  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  30  * THE SOFTWARE.
  31  */
  32
  33 #include "sysemu/blockdev.h"
  34 #include "hw/block/block.h"
  35 #include "block/blockjob.h"
  36 #include "monitor/monitor.h"
  37 #include "qapi/qmp/qerror.h"
  38 #include "qemu/option.h"
  39 #include "qemu/config-file.h"
  40 #include "qapi/qmp/types.h"
  41 #include "qapi-visit.h"
  42 #include "qapi/qmp-output-visitor.h"
  43 #include "sysemu/sysemu.h"
  44 #include "block/block_int.h"
  45 #include "qmp-commands.h"
  46 #include "trace.h"
  47 #include "sysemu/arch_init.h"
  48
  49 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
  50 extern QemuOptsList qemu_common_drive_opts;
  51
  52 static const char *const if_name[IF_COUNT] = {
  53     [IF_NONE] = "none",
  54     [IF_IDE] = "ide",
  55     [IF_SCSI] = "scsi",
  56     [IF_FLOPPY] = "floppy",
  57     [IF_PFLASH] = "pflash",
  58     [IF_MTD] = "mtd",
  59     [IF_SD] = "sd",
  60     [IF_VIRTIO] = "virtio",
  61     [IF_XEN] = "xen",
  62 };
  63
  64 static const int if_max_devs[IF_COUNT] = {
  65     /*
  66      * Do not change these numbers!  They govern how drive option
  67      * index maps to unit and bus.  That mapping is ABI.
  68      *
  69      * All controllers used to imlement if=T drives need to support
  70      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
  71      * Otherwise, some index values map to "impossible" bus, unit
  72      * values.
  73      *
  74      * For instance, if you change [IF_SCSI] to 255, -drive
  75      * if=scsi,index=12 no longer means bus=1,unit=5, but
  76      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
  77      * the drive can't be set up.  Regression.
  78      */
  79     [IF_IDE] = 2,
  80     [IF_SCSI] = 7,
  81 };
  82
  83 /*
  84  * We automatically delete the drive when a device using it gets
  85  * unplugged.  Questionable feature, but we can't just drop it.
  86  * Device models call blockdev_mark_auto_del() to schedule the
  87  * automatic deletion, and generic qdev code calls blockdev_auto_del()
  88  * when deletion is actually safe.
  89  */
  90 void blockdev_mark_auto_del(BlockDriverState *bs)
  91 {
  92     DriveInfo *dinfo = drive_get_by_blockdev(bs);
  93
  94     if (dinfo && !dinfo->enable_auto_del) {
  95         return;
  96     }
  97
  98     if (bs->job) {
  99         block_job_cancel(bs->job);
 100     }
 101     if (dinfo) {
 102         dinfo->auto_del = 1;
 103     }
 104 }
 105
 106 void blockdev_auto_del(BlockDriverState *bs)
 107 {
 108     DriveInfo *dinfo = drive_get_by_blockdev(bs);
 109
 110     if (dinfo && dinfo->auto_del) {
 111         drive_put_ref(dinfo);
 112     }
 113 }
 114
 115 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
 116 {
 117     int max_devs = if_max_devs[type];
 118     return max_devs ? index / max_devs : 0;
 119 }
 120
 121 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
 122 {
 123     int max_devs = if_max_devs[type];
 124     return max_devs ? index % max_devs : index;
 125 }
 126
 127 QemuOpts *drive_def(const char *optstr)
 128 {
 129     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
 130 }
 131
 132 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
 133                     const char *optstr)
 134 {
 135     QemuOpts *opts;
 136     char buf[32];
 137
 138     opts = drive_def(optstr);
 139     if (!opts) {
 140         return NULL;
 141     }
 142     if (type != IF_DEFAULT) {
 143         qemu_opt_set(opts, "if", if_name[type]);
 144     }
 145     if (index >= 0) {
 146         snprintf(buf, sizeof(buf), "%d", index);
 147         qemu_opt_set(opts, "index", buf);
 148     }
 149     if (file)
 150         qemu_opt_set(opts, "file", file);
 151     return opts;
 152 }
 153
 154 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
 155 {
 156     DriveInfo *dinfo;
 157
 158     /* seek interface, bus and unit */
 159
 160     QTAILQ_FOREACH(dinfo, &drives, next) {
 161         if (dinfo->type == type &&
 162             dinfo->bus == bus &&
 163             dinfo->unit == unit)
 164             return dinfo;
 165     }
 166
 167     return NULL;
 168 }
 169
 170 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
 171 {
 172     return drive_get(type,
 173                      drive_index_to_bus_id(type, index),
 174                      drive_index_to_unit_id(type, index));
 175 }
 176
 177 int drive_get_max_bus(BlockInterfaceType type)
 178 {
 179     int max_bus;
 180     DriveInfo *dinfo;
 181
 182     max_bus = -1;
 183     QTAILQ_FOREACH(dinfo, &drives, next) {
 184         if(dinfo->type == type &&
 185            dinfo->bus > max_bus)
 186             max_bus = dinfo->bus;
 187     }
 188     return max_bus;
 189 }
 190
 191 /* Get a block device.  This should only be used for single-drive devices
 192    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
 193    appropriate bus.  */
 194 DriveInfo *drive_get_next(BlockInterfaceType type)
 195 {
 196     static int next_block_unit[IF_COUNT];
 197
 198     return drive_get(type, 0, next_block_unit[type]++);
 199 }
 200
 201 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
 202 {
 203     DriveInfo *dinfo;
 204
 205     QTAILQ_FOREACH(dinfo, &drives, next) {
 206         if (dinfo->bdrv == bs) {
 207             return dinfo;
 208         }
 209     }
 210     return NULL;
 211 }
 212
 213 static void bdrv_format_print(void *opaque, const char *name)
 214 {
 215     error_printf(" %s", name);
 216 }
 217
 218 static void drive_uninit(DriveInfo *dinfo)
 219 {
 220     if (dinfo->opts) {
 221         qemu_opts_del(dinfo->opts);
 222     }
 223
 224     bdrv_unref(dinfo->bdrv);
 225     g_free(dinfo->id);
 226     QTAILQ_REMOVE(&drives, dinfo, next);
 227     g_free(dinfo->serial);
 228     g_free(dinfo);
 229 }
 230
 231 void drive_put_ref(DriveInfo *dinfo)
 232 {
 233     assert(dinfo->refcount);
 234     if (--dinfo->refcount == 0) {
 235         drive_uninit(dinfo);
 236     }
 237 }
 238
 239 void drive_get_ref(DriveInfo *dinfo)
 240 {
 241     dinfo->refcount++;
 242 }
 243
 244 typedef struct {
 245     QEMUBH *bh;
 246     BlockDriverState *bs;
 247 } BDRVPutRefBH;
 248
 249 static void bdrv_put_ref_bh(void *opaque)
 250 {
 251     BDRVPutRefBH *s = opaque;
 252
 253     bdrv_unref(s->bs);
 254     qemu_bh_delete(s->bh);
 255     g_free(s);
 256 }
 257
 258 /*
 259  * Release a BDS reference in a BH
 260  *
 261  * It is not safe to use bdrv_unref() from a callback function when the callers
 262  * still need the BlockDriverState.  In such cases we schedule a BH to release
 263  * the reference.
 264  */
 265 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
 266 {
 267     BDRVPutRefBH *s;
 268
 269     s = g_new(BDRVPutRefBH, 1);
 270     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
 271     s->bs = bs;
 272     qemu_bh_schedule(s->bh);
 273 }
 274
 275 static int parse_block_error_action(const char *buf, bool is_read)
 276 {
 277     if (!strcmp(buf, "ignore")) {
 278         return BLOCKDEV_ON_ERROR_IGNORE;
 279     } else if (!is_read && !strcmp(buf, "enospc")) {
 280         return BLOCKDEV_ON_ERROR_ENOSPC;
 281     } else if (!strcmp(buf, "stop")) {
 282         return BLOCKDEV_ON_ERROR_STOP;
 283     } else if (!strcmp(buf, "report")) {
 284         return BLOCKDEV_ON_ERROR_REPORT;
 285     } else {
 286         error_report("'%s' invalid %s error action",
 287                      buf, is_read ? "read" : "write");
 288         return -1;
 289     }
 290 }
 291
 292 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
 293 {
 294     if (throttle_conflicting(cfg)) {
 295         error_setg(errp, "bps/iops/max total values and read/write values"
 296                          " cannot be used at the same time");
 297         return false;
 298     }
 299
 300     if (!throttle_is_valid(cfg)) {
 301         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
 302         return false;
 303     }
 304
 305     return true;
 306 }
 307
 308 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
 309
 310 /* Takes the ownership of bs_opts */
 311 static DriveInfo *blockdev_init(QDict *bs_opts,
 312                                 BlockInterfaceType type,
 313                                 DriveMediaType media)
 314 {
 315     const char *buf;
 316     const char *file = NULL;
 317     const char *serial;
 318     int ro = 0;
 319     int bdrv_flags = 0;
 320     int on_read_error, on_write_error;
 321     DriveInfo *dinfo;
 322     ThrottleConfig cfg;
 323     int snapshot = 0;
 324     bool copy_on_read;
 325     int ret;
 326     Error *error = NULL;
 327     QemuOpts *opts;
 328     const char *id;
 329     bool has_driver_specific_opts;
 330     BlockDriver *drv = NULL;
 331
 332     /* Check common options by copying from bs_opts to opts, all other options
 333      * stay in bs_opts for processing by bdrv_open(). */
 334     id = qdict_get_try_str(bs_opts, "id");
 335     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
 336     if (error_is_set(&error)) {
 337         qerror_report_err(error);
 338         error_free(error);
 339         return NULL;
 340     }
 341
 342     qemu_opts_absorb_qdict(opts, bs_opts, &error);
 343     if (error_is_set(&error)) {
 344         qerror_report_err(error);
 345         error_free(error);
 346         return NULL;
 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     file = qemu_opt_get(opts, "file");
 361     serial = qemu_opt_get(opts, "serial");
 362
 363     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
 364         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
 365             error_report("invalid discard option");
 366             return NULL;
 367         }
 368     }
 369
 370     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
 371         bdrv_flags |= BDRV_O_CACHE_WB;
 372     }
 373     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
 374         bdrv_flags |= BDRV_O_NOCACHE;
 375     }
 376     if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
 377         bdrv_flags |= BDRV_O_NO_FLUSH;
 378     }
 379
 380 #ifdef CONFIG_LINUX_AIO
 381     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 382         if (!strcmp(buf, "native")) {
 383             bdrv_flags |= BDRV_O_NATIVE_AIO;
 384         } else if (!strcmp(buf, "threads")) {
 385             /* this is the default */
 386         } else {
 387            error_report("invalid aio option");
 388            return NULL;
 389         }
 390     }
 391 #endif
 392
 393     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 394         if (is_help_option(buf)) {
 395             error_printf("Supported formats:");
 396             bdrv_iterate_format(bdrv_format_print, NULL);
 397             error_printf("\n");
 398             return NULL;
 399         }
 400
 401         drv = bdrv_find_format(buf);
 402         if (!drv) {
 403             error_report("'%s' invalid format", buf);
 404             return NULL;
 405         }
 406     }
 407
 408     /* disk I/O throttling */
 409     memset(&cfg, 0, sizeof(cfg));
 410     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
 411         qemu_opt_get_number(opts, "throttling.bps-total", 0);
 412     cfg.buckets[THROTTLE_BPS_READ].avg  =
 413         qemu_opt_get_number(opts, "throttling.bps-read", 0);
 414     cfg.buckets[THROTTLE_BPS_WRITE].avg =
 415         qemu_opt_get_number(opts, "throttling.bps-write", 0);
 416     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
 417         qemu_opt_get_number(opts, "throttling.iops-total", 0);
 418     cfg.buckets[THROTTLE_OPS_READ].avg =
 419         qemu_opt_get_number(opts, "throttling.iops-read", 0);
 420     cfg.buckets[THROTTLE_OPS_WRITE].avg =
 421         qemu_opt_get_number(opts, "throttling.iops-write", 0);
 422
 423     cfg.buckets[THROTTLE_BPS_TOTAL].max =
 424         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
 425     cfg.buckets[THROTTLE_BPS_READ].max  =
 426         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
 427     cfg.buckets[THROTTLE_BPS_WRITE].max =
 428         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
 429     cfg.buckets[THROTTLE_OPS_TOTAL].max =
 430         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
 431     cfg.buckets[THROTTLE_OPS_READ].max =
 432         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
 433     cfg.buckets[THROTTLE_OPS_WRITE].max =
 434         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
 435
 436     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
 437
 438     if (!check_throttle_config(&cfg, &error)) {
 439         error_report("%s", error_get_pretty(error));
 440         error_free(error);
 441         return NULL;
 442     }
 443
 444     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
 445     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 446         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
 447             error_report("werror is not supported by this bus type");
 448             return NULL;
 449         }
 450
 451         on_write_error = parse_block_error_action(buf, 0);
 452         if (on_write_error < 0) {
 453             return NULL;
 454         }
 455     }
 456
 457     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
 458     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 459         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
 460             error_report("rerror is not supported by this bus type");
 461             return NULL;
 462         }
 463
 464         on_read_error = parse_block_error_action(buf, 1);
 465         if (on_read_error < 0) {
 466             return NULL;
 467         }
 468     }
 469
 470     /* init */
 471     dinfo = g_malloc0(sizeof(*dinfo));
 472     dinfo->id = g_strdup(qemu_opts_id(opts));
 473     dinfo->bdrv = bdrv_new(dinfo->id);
 474     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
 475     dinfo->bdrv->read_only = ro;
 476     dinfo->type = type;
 477     dinfo->refcount = 1;
 478     if (serial != NULL) {
 479         dinfo->serial = g_strdup(serial);
 480     }
 481     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 482
 483     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
 484
 485     /* disk I/O throttling */
 486     if (throttle_enabled(&cfg)) {
 487         bdrv_io_limits_enable(dinfo->bdrv);
 488         bdrv_set_io_limits(dinfo->bdrv, &cfg);
 489     }
 490
 491     switch(type) {
 492     case IF_IDE:
 493     case IF_SCSI:
 494     case IF_XEN:
 495     case IF_NONE:
 496         dinfo->media_cd = media == MEDIA_CDROM;
 497         break;
 498     case IF_SD:
 499     case IF_FLOPPY:
 500     case IF_PFLASH:
 501     case IF_MTD:
 502     case IF_VIRTIO:
 503         break;
 504     default:
 505         abort();
 506     }
 507     if (!file || !*file) {
 508         if (has_driver_specific_opts) {
 509             file = NULL;
 510         } else {
 511             return dinfo;
 512         }
 513     }
 514     if (snapshot) {
 515         /* always use cache=unsafe with snapshot */
 516         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 517         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 518     }
 519
 520     if (copy_on_read) {
 521         bdrv_flags |= BDRV_O_COPY_ON_READ;
 522     }
 523
 524     if (runstate_check(RUN_STATE_INMIGRATE)) {
 525         bdrv_flags |= BDRV_O_INCOMING;
 526     }
 527
 528     if (media == MEDIA_CDROM) {
 529         /* CDROM is fine for any interface, don't check.  */
 530         ro = 1;
 531     }
 532
 533     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 534
 535     if (ro && copy_on_read) {
 536         error_report("warning: disabling copy_on_read on read-only drive");
 537     }
 538
 539     QINCREF(bs_opts);
 540     ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv, &error);
 541
 542     if (ret < 0) {
 543         error_report("could not open disk image %s: %s",
 544                      file ?: dinfo->id, error_get_pretty(error));
 545         goto err;
 546     }
 547
 548     if (bdrv_key_required(dinfo->bdrv))
 549         autostart = 0;
 550
 551     QDECREF(bs_opts);
 552     qemu_opts_del(opts);
 553
 554     return dinfo;
 555
 556 err:
 557     qemu_opts_del(opts);
 558     QDECREF(bs_opts);
 559     bdrv_unref(dinfo->bdrv);
 560     g_free(dinfo->id);
 561     QTAILQ_REMOVE(&drives, dinfo, next);
 562     g_free(dinfo);
 563     return NULL;
 564 }
 565
 566 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
 567 {
 568     const char *value;
 569
 570     value = qemu_opt_get(opts, from);
 571     if (value) {
 572         qemu_opt_set(opts, to, value);
 573         qemu_opt_unset(opts, from);
 574     }
 575 }
 576
 577 QemuOptsList qemu_legacy_drive_opts = {
 578     .name = "drive",
 579     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
 580     .desc = {
 581         {
 582             .name = "bus",
 583             .type = QEMU_OPT_NUMBER,
 584             .help = "bus number",
 585         },{
 586             .name = "unit",
 587             .type = QEMU_OPT_NUMBER,
 588             .help = "unit number (i.e. lun for scsi)",
 589         },{
 590             .name = "index",
 591             .type = QEMU_OPT_NUMBER,
 592             .help = "index number",
 593         },{
 594             .name = "media",
 595             .type = QEMU_OPT_STRING,
 596             .help = "media type (disk, cdrom)",
 597         },{
 598             .name = "if",
 599             .type = QEMU_OPT_STRING,
 600             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
 601         },{
 602             .name = "cyls",
 603             .type = QEMU_OPT_NUMBER,
 604             .help = "number of cylinders (ide disk geometry)",
 605         },{
 606             .name = "heads",
 607             .type = QEMU_OPT_NUMBER,
 608             .help = "number of heads (ide disk geometry)",
 609         },{
 610             .name = "secs",
 611             .type = QEMU_OPT_NUMBER,
 612             .help = "number of sectors (ide disk geometry)",
 613         },{
 614             .name = "trans",
 615             .type = QEMU_OPT_STRING,
 616             .help = "chs translation (auto, lba, none)",
 617         },{
 618             .name = "boot",
 619             .type = QEMU_OPT_BOOL,
 620             .help = "(deprecated, ignored)",
 621         },{
 622             .name = "addr",
 623             .type = QEMU_OPT_STRING,
 624             .help = "pci address (virtio only)",
 625         },
 626         { /* end of list */ }
 627     },
 628 };
 629
 630 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
 631 {
 632     const char *value;
 633     DriveInfo *dinfo = NULL;
 634     QDict *bs_opts;
 635     QemuOpts *legacy_opts;
 636     DriveMediaType media = MEDIA_DISK;
 637     BlockInterfaceType type;
 638     int cyls, heads, secs, translation;
 639     int max_devs, bus_id, unit_id, index;
 640     const char *devaddr;
 641     Error *local_err = NULL;
 642
 643     /* Change legacy command line options into QMP ones */
 644     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
 645     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
 646     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
 647
 648     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
 649     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
 650     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
 651
 652     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
 653     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
 654     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
 655
 656     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
 657     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
 658     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
 659
 660     qemu_opt_rename(all_opts,
 661                     "iops_size", "throttling.iops-size");
 662
 663     qemu_opt_rename(all_opts, "readonly", "read-only");
 664
 665     value = qemu_opt_get(all_opts, "cache");
 666     if (value) {
 667         int flags = 0;
 668
 669         if (bdrv_parse_cache_flags(value, &flags) != 0) {
 670             error_report("invalid cache option");
 671             return NULL;
 672         }
 673
 674         /* Specific options take precedence */
 675         if (!qemu_opt_get(all_opts, "cache.writeback")) {
 676             qemu_opt_set_bool(all_opts, "cache.writeback",
 677                               !!(flags & BDRV_O_CACHE_WB));
 678         }
 679         if (!qemu_opt_get(all_opts, "cache.direct")) {
 680             qemu_opt_set_bool(all_opts, "cache.direct",
 681                               !!(flags & BDRV_O_NOCACHE));
 682         }
 683         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
 684             qemu_opt_set_bool(all_opts, "cache.no-flush",
 685                               !!(flags & BDRV_O_NO_FLUSH));
 686         }
 687         qemu_opt_unset(all_opts, "cache");
 688     }
 689
 690     /* Get a QDict for processing the options */
 691     bs_opts = qdict_new();
 692     qemu_opts_to_qdict(all_opts, bs_opts);
 693
 694     legacy_opts = qemu_opts_create_nofail(&qemu_legacy_drive_opts);
 695     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
 696     if (error_is_set(&local_err)) {
 697         qerror_report_err(local_err);
 698         error_free(local_err);
 699         goto fail;
 700     }
 701
 702     /* Deprecated option boot=[on|off] */
 703     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
 704         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 705                 "ignored. Future versions will reject this parameter. Please "
 706                 "update your scripts.\n");
 707     }
 708
 709     /* Media type */
 710     value = qemu_opt_get(legacy_opts, "media");
 711     if (value) {
 712         if (!strcmp(value, "disk")) {
 713             media = MEDIA_DISK;
 714         } else if (!strcmp(value, "cdrom")) {
 715             media = MEDIA_CDROM;
 716         } else {
 717             error_report("'%s' invalid media", value);
 718             goto fail;
 719         }
 720     }
 721
 722     /* Controller type */
 723     value = qemu_opt_get(legacy_opts, "if");
 724     if (value) {
 725         for (type = 0;
 726              type < IF_COUNT && strcmp(value, if_name[type]);
 727              type++) {
 728         }
 729         if (type == IF_COUNT) {
 730             error_report("unsupported bus type '%s'", value);
 731             goto fail;
 732         }
 733     } else {
 734         type = block_default_type;
 735     }
 736
 737     /* Geometry */
 738     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
 739     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
 740     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
 741
 742     if (cyls || heads || secs) {
 743         if (cyls < 1) {
 744             error_report("invalid physical cyls number");
 745             goto fail;
 746         }
 747         if (heads < 1) {
 748             error_report("invalid physical heads number");
 749             goto fail;
 750         }
 751         if (secs < 1) {
 752             error_report("invalid physical secs number");
 753             goto fail;
 754         }
 755     }
 756
 757     translation = BIOS_ATA_TRANSLATION_AUTO;
 758     value = qemu_opt_get(legacy_opts, "trans");
 759     if (value != NULL) {
 760         if (!cyls) {
 761             error_report("'%s' trans must be used with cyls, heads and secs",
 762                          value);
 763             goto fail;
 764         }
 765         if (!strcmp(value, "none")) {
 766             translation = BIOS_ATA_TRANSLATION_NONE;
 767         } else if (!strcmp(value, "lba")) {
 768             translation = BIOS_ATA_TRANSLATION_LBA;
 769         } else if (!strcmp(value, "auto")) {
 770             translation = BIOS_ATA_TRANSLATION_AUTO;
 771         } else {
 772             error_report("'%s' invalid translation type", value);
 773             goto fail;
 774         }
 775     }
 776
 777     if (media == MEDIA_CDROM) {
 778         if (cyls || secs || heads) {
 779             error_report("CHS can't be set with media=cdrom");
 780             goto fail;
 781         }
 782     }
 783
 784     /* Device address specified by bus/unit or index.
 785      * If none was specified, try to find the first free one. */
 786     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
 787     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
 788     index   = qemu_opt_get_number(legacy_opts, "index", -1);
 789
 790     max_devs = if_max_devs[type];
 791
 792     if (index != -1) {
 793         if (bus_id != 0 || unit_id != -1) {
 794             error_report("index cannot be used with bus and unit");
 795             goto fail;
 796         }
 797         bus_id = drive_index_to_bus_id(type, index);
 798         unit_id = drive_index_to_unit_id(type, index);
 799     }
 800
 801     if (unit_id == -1) {
 802        unit_id = 0;
 803        while (drive_get(type, bus_id, unit_id) != NULL) {
 804            unit_id++;
 805            if (max_devs && unit_id >= max_devs) {
 806                unit_id -= max_devs;
 807                bus_id++;
 808            }
 809        }
 810     }
 811
 812     if (max_devs && unit_id >= max_devs) {
 813         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
 814         goto fail;
 815     }
 816
 817     if (drive_get(type, bus_id, unit_id) != NULL) {
 818         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 819                      bus_id, unit_id, index);
 820         goto fail;
 821     }
 822
 823     /* no id supplied -> create one */
 824     if (qemu_opts_id(all_opts) == NULL) {
 825         char *new_id;
 826         const char *mediastr = "";
 827         if (type == IF_IDE || type == IF_SCSI) {
 828             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 829         }
 830         if (max_devs) {
 831             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
 832                                      mediastr, unit_id);
 833         } else {
 834             new_id = g_strdup_printf("%s%s%i", if_name[type],
 835                                      mediastr, unit_id);
 836         }
 837         qdict_put(bs_opts, "id", qstring_from_str(new_id));
 838         g_free(new_id);
 839     }
 840
 841     /* Add virtio block device */
 842     devaddr = qemu_opt_get(legacy_opts, "addr");
 843     if (devaddr && type != IF_VIRTIO) {
 844         error_report("addr is not supported by this bus type");
 845         goto fail;
 846     }
 847
 848     if (type == IF_VIRTIO) {
 849         QemuOpts *devopts;
 850         devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
 851         if (arch_type == QEMU_ARCH_S390X) {
 852             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
 853         } else {
 854             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
 855         }
 856         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
 857         if (devaddr) {
 858             qemu_opt_set(devopts, "addr", devaddr);
 859         }
 860     }
 861
 862     /* Actual block device init: Functionality shared with blockdev-add */
 863     dinfo = blockdev_init(bs_opts, type, media);
 864     if (dinfo == NULL) {
 865         goto fail;
 866     }
 867
 868     /* Set legacy DriveInfo fields */
 869     dinfo->enable_auto_del = true;
 870     dinfo->opts = all_opts;
 871
 872     dinfo->cyls = cyls;
 873     dinfo->heads = heads;
 874     dinfo->secs = secs;
 875     dinfo->trans = translation;
 876
 877     dinfo->bus = bus_id;
 878     dinfo->unit = unit_id;
 879     dinfo->devaddr = devaddr;
 880
 881 fail:
 882     qemu_opts_del(legacy_opts);
 883     return dinfo;
 884 }
 885
 886 void do_commit(Monitor *mon, const QDict *qdict)
 887 {
 888     const char *device = qdict_get_str(qdict, "device");
 889     BlockDriverState *bs;
 890     int ret;
 891
 892     if (!strcmp(device, "all")) {
 893         ret = bdrv_commit_all();
 894     } else {
 895         bs = bdrv_find(device);
 896         if (!bs) {
 897             monitor_printf(mon, "Device '%s' not found\n", device);
 898             return;
 899         }
 900         ret = bdrv_commit(bs);
 901     }
 902     if (ret < 0) {
 903         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
 904                        strerror(-ret));
 905     }
 906 }
 907
 908 static void blockdev_do_action(int kind, void *data, Error **errp)
 909 {
 910     TransactionAction action;
 911     TransactionActionList list;
 912
 913     action.kind = kind;
 914     action.data = data;
 915     list.value = &action;
 916     list.next = NULL;
 917     qmp_transaction(&list, errp);
 918 }
 919
 920 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
 921                                 bool has_format, const char *format,
 922                                 bool has_mode, enum NewImageMode mode,
 923                                 Error **errp)
 924 {
 925     BlockdevSnapshot snapshot = {
 926         .device = (char *) device,
 927         .snapshot_file = (char *) snapshot_file,
 928         .has_format = has_format,
 929         .format = (char *) format,
 930         .has_mode = has_mode,
 931         .mode = mode,
 932     };
 933     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
 934                        &snapshot, errp);
 935 }
 936
 937 void qmp_blockdev_snapshot_internal_sync(const char *device,
 938                                          const char *name,
 939                                          Error **errp)
 940 {
 941     BlockdevSnapshotInternal snapshot = {
 942         .device = (char *) device,
 943         .name = (char *) name
 944     };
 945
 946     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
 947                        &snapshot, errp);
 948 }
 949
 950 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
 951                                                          bool has_id,
 952                                                          const char *id,
 953                                                          bool has_name,
 954                                                          const char *name,
 955                                                          Error **errp)
 956 {
 957     BlockDriverState *bs = bdrv_find(device);
 958     QEMUSnapshotInfo sn;
 959     Error *local_err = NULL;
 960     SnapshotInfo *info = NULL;
 961     int ret;
 962
 963     if (!bs) {
 964         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 965         return NULL;
 966     }
 967
 968     if (!has_id) {
 969         id = NULL;
 970     }
 971
 972     if (!has_name) {
 973         name = NULL;
 974     }
 975
 976     if (!id && !name) {
 977         error_setg(errp, "Name or id must be provided");
 978         return NULL;
 979     }
 980
 981     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
 982     if (error_is_set(&local_err)) {
 983         error_propagate(errp, local_err);
 984         return NULL;
 985     }
 986     if (!ret) {
 987         error_setg(errp,
 988                    "Snapshot with id '%s' and name '%s' does not exist on "
 989                    "device '%s'",
 990                    STR_OR_NULL(id), STR_OR_NULL(name), device);
 991         return NULL;
 992     }
 993
 994     bdrv_snapshot_delete(bs, id, name, &local_err);
 995     if (error_is_set(&local_err)) {
 996         error_propagate(errp, local_err);
 997         return NULL;
 998     }
 999
1000     info = g_malloc0(sizeof(SnapshotInfo));
1001     info->id = g_strdup(sn.id_str);
1002     info->name = g_strdup(sn.name);
1003     info->date_nsec = sn.date_nsec;
1004     info->date_sec = sn.date_sec;
1005     info->vm_state_size = sn.vm_state_size;
1006     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1007     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1008
1009     return info;
1010 }
1011
1012 /* New and old BlockDriverState structs for group snapshots */
1013
1014 typedef struct BlkTransactionState BlkTransactionState;
1015
1016 /* Only prepare() may fail. In a single transaction, only one of commit() or
1017    abort() will be called, clean() will always be called if it present. */
1018 typedef struct BdrvActionOps {
1019     /* Size of state struct, in bytes. */
1020     size_t instance_size;
1021     /* Prepare the work, must NOT be NULL. */
1022     void (*prepare)(BlkTransactionState *common, Error **errp);
1023     /* Commit the changes, can be NULL. */
1024     void (*commit)(BlkTransactionState *common);
1025     /* Abort the changes on fail, can be NULL. */
1026     void (*abort)(BlkTransactionState *common);
1027     /* Clean up resource in the end, can be NULL. */
1028     void (*clean)(BlkTransactionState *common);
1029 } BdrvActionOps;
1030
1031 /*
1032  * This structure must be arranged as first member in child type, assuming
1033  * that compiler will also arrange it to the same address with parent instance.
1034  * Later it will be used in free().
1035  */
1036 struct BlkTransactionState {
1037     TransactionAction *action;
1038     const BdrvActionOps *ops;
1039     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1040 };
1041
1042 /* internal snapshot private data */
1043 typedef struct InternalSnapshotState {
1044     BlkTransactionState common;
1045     BlockDriverState *bs;
1046     QEMUSnapshotInfo sn;
1047 } InternalSnapshotState;
1048
1049 static void internal_snapshot_prepare(BlkTransactionState *common,
1050                                       Error **errp)
1051 {
1052     const char *device;
1053     const char *name;
1054     BlockDriverState *bs;
1055     QEMUSnapshotInfo old_sn, *sn;
1056     bool ret;
1057     qemu_timeval tv;
1058     BlockdevSnapshotInternal *internal;
1059     InternalSnapshotState *state;
1060     int ret1;
1061
1062     g_assert(common->action->kind ==
1063              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1064     internal = common->action->blockdev_snapshot_internal_sync;
1065     state = DO_UPCAST(InternalSnapshotState, common, common);
1066
1067     /* 1. parse input */
1068     device = internal->device;
1069     name = internal->name;
1070
1071     /* 2. check for validation */
1072     bs = bdrv_find(device);
1073     if (!bs) {
1074         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1075         return;
1076     }
1077
1078     if (!bdrv_is_inserted(bs)) {
1079         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1080         return;
1081     }
1082
1083     if (bdrv_is_read_only(bs)) {
1084         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1085         return;
1086     }
1087
1088     if (!bdrv_can_snapshot(bs)) {
1089         error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1090                   bs->drv->format_name, device, "internal snapshot");
1091         return;
1092     }
1093
1094     if (!strlen(name)) {
1095         error_setg(errp, "Name is empty");
1096         return;
1097     }
1098
1099     /* check whether a snapshot with name exist */
1100     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn, errp);
1101     if (error_is_set(errp)) {
1102         return;
1103     } else if (ret) {
1104         error_setg(errp,
1105                    "Snapshot with name '%s' already exists on device '%s'",
1106                    name, device);
1107         return;
1108     }
1109
1110     /* 3. take the snapshot */
1111     sn = &state->sn;
1112     pstrcpy(sn->name, sizeof(sn->name), name);
1113     qemu_gettimeofday(&tv);
1114     sn->date_sec = tv.tv_sec;
1115     sn->date_nsec = tv.tv_usec * 1000;
1116     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1117
1118     ret1 = bdrv_snapshot_create(bs, sn);
1119     if (ret1 < 0) {
1120         error_setg_errno(errp, -ret1,
1121                          "Failed to create snapshot '%s' on device '%s'",
1122                          name, device);
1123         return;
1124     }
1125
1126     /* 4. succeed, mark a snapshot is created */
1127     state->bs = bs;
1128 }
1129
1130 static void internal_snapshot_abort(BlkTransactionState *common)
1131 {
1132     InternalSnapshotState *state =
1133                              DO_UPCAST(InternalSnapshotState, common, common);
1134     BlockDriverState *bs = state->bs;
1135     QEMUSnapshotInfo *sn = &state->sn;
1136     Error *local_error = NULL;
1137
1138     if (!bs) {
1139         return;
1140     }
1141
1142     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1143         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1144                      "device '%s' in abort: %s",
1145                      sn->id_str,
1146                      sn->name,
1147                      bdrv_get_device_name(bs),
1148                      error_get_pretty(local_error));
1149         error_free(local_error);
1150     }
1151 }
1152
1153 /* external snapshot private data */
1154 typedef struct ExternalSnapshotState {
1155     BlkTransactionState common;
1156     BlockDriverState *old_bs;
1157     BlockDriverState *new_bs;
1158 } ExternalSnapshotState;
1159
1160 static void external_snapshot_prepare(BlkTransactionState *common,
1161                                       Error **errp)
1162 {
1163     BlockDriver *drv;
1164     int flags, ret;
1165     Error *local_err = NULL;
1166     const char *device;
1167     const char *new_image_file;
1168     const char *format = "qcow2";
1169     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1170     ExternalSnapshotState *state =
1171                              DO_UPCAST(ExternalSnapshotState, common, common);
1172     TransactionAction *action = common->action;
1173
1174     /* get parameters */
1175     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1176
1177     device = action->blockdev_snapshot_sync->device;
1178     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1179     if (action->blockdev_snapshot_sync->has_format) {
1180         format = action->blockdev_snapshot_sync->format;
1181     }
1182     if (action->blockdev_snapshot_sync->has_mode) {
1183         mode = action->blockdev_snapshot_sync->mode;
1184     }
1185
1186     /* start processing */
1187     drv = bdrv_find_format(format);
1188     if (!drv) {
1189         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1190         return;
1191     }
1192
1193     state->old_bs = bdrv_find(device);
1194     if (!state->old_bs) {
1195         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1196         return;
1197     }
1198
1199     if (!bdrv_is_inserted(state->old_bs)) {
1200         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1201         return;
1202     }
1203
1204     if (bdrv_in_use(state->old_bs)) {
1205         error_set(errp, QERR_DEVICE_IN_USE, device);
1206         return;
1207     }
1208
1209     if (!bdrv_is_read_only(state->old_bs)) {
1210         if (bdrv_flush(state->old_bs)) {
1211             error_set(errp, QERR_IO_ERROR);
1212             return;
1213         }
1214     }
1215
1216     if (bdrv_check_ext_snapshot(state->old_bs) != EXT_SNAPSHOT_ALLOWED) {
1217         error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1218         return;
1219     }
1220
1221     flags = state->old_bs->open_flags;
1222
1223     /* create new image w/backing file */
1224     if (mode != NEW_IMAGE_MODE_EXISTING) {
1225         bdrv_img_create(new_image_file, format,
1226                         state->old_bs->filename,
1227                         state->old_bs->drv->format_name,
1228                         NULL, -1, flags, &local_err, false);
1229         if (error_is_set(&local_err)) {
1230             error_propagate(errp, local_err);
1231             return;
1232         }
1233     }
1234
1235     /* We will manually add the backing_hd field to the bs later */
1236     state->new_bs = bdrv_new("");
1237     /* TODO Inherit bs->options or only take explicit options with an
1238      * extended QMP command? */
1239     ret = bdrv_open(state->new_bs, new_image_file, NULL,
1240                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1241     if (ret != 0) {
1242         error_propagate(errp, local_err);
1243     }
1244 }
1245
1246 static void external_snapshot_commit(BlkTransactionState *common)
1247 {
1248     ExternalSnapshotState *state =
1249                              DO_UPCAST(ExternalSnapshotState, common, common);
1250
1251     /* This removes our old bs and adds the new bs */
1252     bdrv_append(state->new_bs, state->old_bs);
1253     /* We don't need (or want) to use the transactional
1254      * bdrv_reopen_multiple() across all the entries at once, because we
1255      * don't want to abort all of them if one of them fails the reopen */
1256     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1257                 NULL);
1258 }
1259
1260 static void external_snapshot_abort(BlkTransactionState *common)
1261 {
1262     ExternalSnapshotState *state =
1263                              DO_UPCAST(ExternalSnapshotState, common, common);
1264     if (state->new_bs) {
1265         bdrv_unref(state->new_bs);
1266     }
1267 }
1268
1269 typedef struct DriveBackupState {
1270     BlkTransactionState common;
1271     BlockDriverState *bs;
1272     BlockJob *job;
1273 } DriveBackupState;
1274
1275 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1276 {
1277     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1278     DriveBackup *backup;
1279     Error *local_err = NULL;
1280
1281     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1282     backup = common->action->drive_backup;
1283
1284     qmp_drive_backup(backup->device, backup->target,
1285                      backup->has_format, backup->format,
1286                      backup->sync,
1287                      backup->has_mode, backup->mode,
1288                      backup->has_speed, backup->speed,
1289                      backup->has_on_source_error, backup->on_source_error,
1290                      backup->has_on_target_error, backup->on_target_error,
1291                      &local_err);
1292     if (error_is_set(&local_err)) {
1293         error_propagate(errp, local_err);
1294         state->bs = NULL;
1295         state->job = NULL;
1296         return;
1297     }
1298
1299     state->bs = bdrv_find(backup->device);
1300     state->job = state->bs->job;
1301 }
1302
1303 static void drive_backup_abort(BlkTransactionState *common)
1304 {
1305     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1306     BlockDriverState *bs = state->bs;
1307
1308     /* Only cancel if it's the job we started */
1309     if (bs && bs->job && bs->job == state->job) {
1310         block_job_cancel_sync(bs->job);
1311     }
1312 }
1313
1314 static void abort_prepare(BlkTransactionState *common, Error **errp)
1315 {
1316     error_setg(errp, "Transaction aborted using Abort action");
1317 }
1318
1319 static void abort_commit(BlkTransactionState *common)
1320 {
1321     g_assert_not_reached(); /* this action never succeeds */
1322 }
1323
1324 static const BdrvActionOps actions[] = {
1325     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1326         .instance_size = sizeof(ExternalSnapshotState),
1327         .prepare  = external_snapshot_prepare,
1328         .commit   = external_snapshot_commit,
1329         .abort = external_snapshot_abort,
1330     },
1331     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1332         .instance_size = sizeof(DriveBackupState),
1333         .prepare = drive_backup_prepare,
1334         .abort = drive_backup_abort,
1335     },
1336     [TRANSACTION_ACTION_KIND_ABORT] = {
1337         .instance_size = sizeof(BlkTransactionState),
1338         .prepare = abort_prepare,
1339         .commit = abort_commit,
1340     },
1341     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1342         .instance_size = sizeof(InternalSnapshotState),
1343         .prepare  = internal_snapshot_prepare,
1344         .abort = internal_snapshot_abort,
1345     },
1346 };
1347
1348 /*
1349  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1350  *  then we do not pivot any of the devices in the group, and abandon the
1351  *  snapshots
1352  */
1353 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1354 {
1355     TransactionActionList *dev_entry = dev_list;
1356     BlkTransactionState *state, *next;
1357     Error *local_err = NULL;
1358
1359     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1360     QSIMPLEQ_INIT(&snap_bdrv_states);
1361
1362     /* drain all i/o before any snapshots */
1363     bdrv_drain_all();
1364
1365     /* We don't do anything in this loop that commits us to the snapshot */
1366     while (NULL != dev_entry) {
1367         TransactionAction *dev_info = NULL;
1368         const BdrvActionOps *ops;
1369
1370         dev_info = dev_entry->value;
1371         dev_entry = dev_entry->next;
1372
1373         assert(dev_info->kind < ARRAY_SIZE(actions));
1374
1375         ops = &actions[dev_info->kind];
1376         assert(ops->instance_size > 0);
1377
1378         state = g_malloc0(ops->instance_size);
1379         state->ops = ops;
1380         state->action = dev_info;
1381         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1382
1383         state->ops->prepare(state, &local_err);
1384         if (error_is_set(&local_err)) {
1385             error_propagate(errp, local_err);
1386             goto delete_and_fail;
1387         }
1388     }
1389
1390     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1391         if (state->ops->commit) {
1392             state->ops->commit(state);
1393         }
1394     }
1395
1396     /* success */
1397     goto exit;
1398
1399 delete_and_fail:
1400     /*
1401     * failure, and it is all-or-none; abandon each new bs, and keep using
1402     * the original bs for all images
1403     */
1404     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1405         if (state->ops->abort) {
1406             state->ops->abort(state);
1407         }
1408     }
1409 exit:
1410     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1411         if (state->ops->clean) {
1412             state->ops->clean(state);
1413         }
1414         g_free(state);
1415     }
1416 }
1417
1418
1419 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1420 {
1421     if (bdrv_in_use(bs)) {
1422         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1423         return;
1424     }
1425     if (!bdrv_dev_has_removable_media(bs)) {
1426         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1427         return;
1428     }
1429
1430     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1431         bdrv_dev_eject_request(bs, force);
1432         if (!force) {
1433             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1434             return;
1435         }
1436     }
1437
1438     bdrv_close(bs);
1439 }
1440
1441 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1442 {
1443     BlockDriverState *bs;
1444
1445     bs = bdrv_find(device);
1446     if (!bs) {
1447         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1448         return;
1449     }
1450
1451     eject_device(bs, force, errp);
1452 }
1453
1454 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1455 {
1456     BlockDriverState *bs;
1457     int err;
1458
1459     bs = bdrv_find(device);
1460     if (!bs) {
1461         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1462         return;
1463     }
1464
1465     err = bdrv_set_key(bs, password);
1466     if (err == -EINVAL) {
1467         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1468         return;
1469     } else if (err < 0) {
1470         error_set(errp, QERR_INVALID_PASSWORD);
1471         return;
1472     }
1473 }
1474
1475 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1476                                     int bdrv_flags, BlockDriver *drv,
1477                                     const char *password, Error **errp)
1478 {
1479     Error *local_err = NULL;
1480     int ret;
1481
1482     ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv, &local_err);
1483     if (ret < 0) {
1484         error_propagate(errp, local_err);
1485         return;
1486     }
1487
1488     if (bdrv_key_required(bs)) {
1489         if (password) {
1490             if (bdrv_set_key(bs, password) < 0) {
1491                 error_set(errp, QERR_INVALID_PASSWORD);
1492             }
1493         } else {
1494             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1495                       bdrv_get_encrypted_filename(bs));
1496         }
1497     } else if (password) {
1498         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1499     }
1500 }
1501
1502 void qmp_change_blockdev(const char *device, const char *filename,
1503                          bool has_format, const char *format, Error **errp)
1504 {
1505     BlockDriverState *bs;
1506     BlockDriver *drv = NULL;
1507     int bdrv_flags;
1508     Error *err = NULL;
1509
1510     bs = bdrv_find(device);
1511     if (!bs) {
1512         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1513         return;
1514     }
1515
1516     if (format) {
1517         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1518         if (!drv) {
1519             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1520             return;
1521         }
1522     }
1523
1524     eject_device(bs, 0, &err);
1525     if (error_is_set(&err)) {
1526         error_propagate(errp, err);
1527         return;
1528     }
1529
1530     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1531     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1532
1533     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1534 }
1535
1536 /* throttling disk I/O limits */
1537 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1538                                int64_t bps_wr,
1539                                int64_t iops,
1540                                int64_t iops_rd,
1541                                int64_t iops_wr,
1542                                bool has_bps_max,
1543                                int64_t bps_max,
1544                                bool has_bps_rd_max,
1545                                int64_t bps_rd_max,
1546                                bool has_bps_wr_max,
1547                                int64_t bps_wr_max,
1548                                bool has_iops_max,
1549                                int64_t iops_max,
1550                                bool has_iops_rd_max,
1551                                int64_t iops_rd_max,
1552                                bool has_iops_wr_max,
1553                                int64_t iops_wr_max,
1554                                bool has_iops_size,
1555                                int64_t iops_size, Error **errp)
1556 {
1557     ThrottleConfig cfg;
1558     BlockDriverState *bs;
1559
1560     bs = bdrv_find(device);
1561     if (!bs) {
1562         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1563         return;
1564     }
1565
1566     memset(&cfg, 0, sizeof(cfg));
1567     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1568     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1569     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1570
1571     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1572     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1573     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1574
1575     if (has_bps_max) {
1576         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1577     }
1578     if (has_bps_rd_max) {
1579         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1580     }
1581     if (has_bps_wr_max) {
1582         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1583     }
1584     if (has_iops_max) {
1585         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1586     }
1587     if (has_iops_rd_max) {
1588         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1589     }
1590     if (has_iops_wr_max) {
1591         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1592     }
1593
1594     if (has_iops_size) {
1595         cfg.op_size = iops_size;
1596     }
1597
1598     if (!check_throttle_config(&cfg, errp)) {
1599         return;
1600     }
1601
1602     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1603         bdrv_io_limits_enable(bs);
1604     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1605         bdrv_io_limits_disable(bs);
1606     }
1607
1608     if (bs->io_limits_enabled) {
1609         bdrv_set_io_limits(bs, &cfg);
1610     }
1611 }
1612
1613 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1614 {
1615     const char *id = qdict_get_str(qdict, "id");
1616     BlockDriverState *bs;
1617
1618     bs = bdrv_find(id);
1619     if (!bs) {
1620         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1621         return -1;
1622     }
1623     if (bdrv_in_use(bs)) {
1624         qerror_report(QERR_DEVICE_IN_USE, id);
1625         return -1;
1626     }
1627
1628     /* quiesce block driver; prevent further io */
1629     bdrv_drain_all();
1630     bdrv_flush(bs);
1631     bdrv_close(bs);
1632
1633     /* if we have a device attached to this BlockDriverState
1634      * then we need to make the drive anonymous until the device
1635      * can be removed.  If this is a drive with no device backing
1636      * then we can just get rid of the block driver state right here.
1637      */
1638     if (bdrv_get_attached_dev(bs)) {
1639         bdrv_make_anon(bs);
1640
1641         /* Further I/O must not pause the guest */
1642         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1643                           BLOCKDEV_ON_ERROR_REPORT);
1644     } else {
1645         drive_uninit(drive_get_by_blockdev(bs));
1646     }
1647
1648     return 0;
1649 }
1650
1651 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1652 {
1653     BlockDriverState *bs;
1654     int ret;
1655
1656     bs = bdrv_find(device);
1657     if (!bs) {
1658         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1659         return;
1660     }
1661
1662     if (size < 0) {
1663         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1664         return;
1665     }
1666
1667     /* complete all in-flight operations before resizing the device */
1668     bdrv_drain_all();
1669
1670     ret = bdrv_truncate(bs, size);
1671     switch (ret) {
1672     case 0:
1673         break;
1674     case -ENOMEDIUM:
1675         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1676         break;
1677     case -ENOTSUP:
1678         error_set(errp, QERR_UNSUPPORTED);
1679         break;
1680     case -EACCES:
1681         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1682         break;
1683     case -EBUSY:
1684         error_set(errp, QERR_DEVICE_IN_USE, device);
1685         break;
1686     default:
1687         error_setg_errno(errp, -ret, "Could not resize");
1688         break;
1689     }
1690 }
1691
1692 static void block_job_cb(void *opaque, int ret)
1693 {
1694     BlockDriverState *bs = opaque;
1695     QObject *obj;
1696
1697     trace_block_job_cb(bs, bs->job, ret);
1698
1699     assert(bs->job);
1700     obj = qobject_from_block_job(bs->job);
1701     if (ret < 0) {
1702         QDict *dict = qobject_to_qdict(obj);
1703         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1704     }
1705
1706     if (block_job_is_cancelled(bs->job)) {
1707         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1708     } else {
1709         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1710     }
1711     qobject_decref(obj);
1712
1713     bdrv_put_ref_bh_schedule(bs);
1714 }
1715
1716 void qmp_block_stream(const char *device, bool has_base,
1717                       const char *base, bool has_speed, int64_t speed,
1718                       bool has_on_error, BlockdevOnError on_error,
1719                       Error **errp)
1720 {
1721     BlockDriverState *bs;
1722     BlockDriverState *base_bs = NULL;
1723     Error *local_err = NULL;
1724
1725     if (!has_on_error) {
1726         on_error = BLOCKDEV_ON_ERROR_REPORT;
1727     }
1728
1729     bs = bdrv_find(device);
1730     if (!bs) {
1731         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1732         return;
1733     }
1734
1735     if (base) {
1736         base_bs = bdrv_find_backing_image(bs, base);
1737         if (base_bs == NULL) {
1738             error_set(errp, QERR_BASE_NOT_FOUND, base);
1739             return;
1740         }
1741     }
1742
1743     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1744                  on_error, block_job_cb, bs, &local_err);
1745     if (error_is_set(&local_err)) {
1746         error_propagate(errp, local_err);
1747         return;
1748     }
1749
1750     trace_qmp_block_stream(bs, bs->job);
1751 }
1752
1753 void qmp_block_commit(const char *device,
1754                       bool has_base, const char *base, const char *top,
1755                       bool has_speed, int64_t speed,
1756                       Error **errp)
1757 {
1758     BlockDriverState *bs;
1759     BlockDriverState *base_bs, *top_bs;
1760     Error *local_err = NULL;
1761     /* This will be part of the QMP command, if/when the
1762      * BlockdevOnError change for blkmirror makes it in
1763      */
1764     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1765
1766     /* drain all i/o before commits */
1767     bdrv_drain_all();
1768
1769     bs = bdrv_find(device);
1770     if (!bs) {
1771         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1772         return;
1773     }
1774
1775     /* default top_bs is the active layer */
1776     top_bs = bs;
1777
1778     if (top) {
1779         if (strcmp(bs->filename, top) != 0) {
1780             top_bs = bdrv_find_backing_image(bs, top);
1781         }
1782     }
1783
1784     if (top_bs == NULL) {
1785         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1786         return;
1787     }
1788
1789     if (has_base && base) {
1790         base_bs = bdrv_find_backing_image(top_bs, base);
1791     } else {
1792         base_bs = bdrv_find_base(top_bs);
1793     }
1794
1795     if (base_bs == NULL) {
1796         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1797         return;
1798     }
1799
1800     commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1801                 &local_err);
1802     if (local_err != NULL) {
1803         error_propagate(errp, local_err);
1804         return;
1805     }
1806 }
1807
1808 void qmp_drive_backup(const char *device, const char *target,
1809                       bool has_format, const char *format,
1810                       enum MirrorSyncMode sync,
1811                       bool has_mode, enum NewImageMode mode,
1812                       bool has_speed, int64_t speed,
1813                       bool has_on_source_error, BlockdevOnError on_source_error,
1814                       bool has_on_target_error, BlockdevOnError on_target_error,
1815                       Error **errp)
1816 {
1817     BlockDriverState *bs;
1818     BlockDriverState *target_bs;
1819     BlockDriverState *source = NULL;
1820     BlockDriver *drv = NULL;
1821     Error *local_err = NULL;
1822     int flags;
1823     int64_t size;
1824     int ret;
1825
1826     if (!has_speed) {
1827         speed = 0;
1828     }
1829     if (!has_on_source_error) {
1830         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1831     }
1832     if (!has_on_target_error) {
1833         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1834     }
1835     if (!has_mode) {
1836         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1837     }
1838
1839     bs = bdrv_find(device);
1840     if (!bs) {
1841         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1842         return;
1843     }
1844
1845     if (!bdrv_is_inserted(bs)) {
1846         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1847         return;
1848     }
1849
1850     if (!has_format) {
1851         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1852     }
1853     if (format) {
1854         drv = bdrv_find_format(format);
1855         if (!drv) {
1856             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1857             return;
1858         }
1859     }
1860
1861     if (bdrv_in_use(bs)) {
1862         error_set(errp, QERR_DEVICE_IN_USE, device);
1863         return;
1864     }
1865
1866     flags = bs->open_flags | BDRV_O_RDWR;
1867
1868     /* See if we have a backing HD we can use to create our new image
1869      * on top of. */
1870     if (sync == MIRROR_SYNC_MODE_TOP) {
1871         source = bs->backing_hd;
1872         if (!source) {
1873             sync = MIRROR_SYNC_MODE_FULL;
1874         }
1875     }
1876     if (sync == MIRROR_SYNC_MODE_NONE) {
1877         source = bs;
1878     }
1879
1880     size = bdrv_getlength(bs);
1881     if (size < 0) {
1882         error_setg_errno(errp, -size, "bdrv_getlength failed");
1883         return;
1884     }
1885
1886     if (mode != NEW_IMAGE_MODE_EXISTING) {
1887         assert(format && drv);
1888         if (source) {
1889             bdrv_img_create(target, format, source->filename,
1890                             source->drv->format_name, NULL,
1891                             size, flags, &local_err, false);
1892         } else {
1893             bdrv_img_create(target, format, NULL, NULL, NULL,
1894                             size, flags, &local_err, false);
1895         }
1896     }
1897
1898     if (error_is_set(&local_err)) {
1899         error_propagate(errp, local_err);
1900         return;
1901     }
1902
1903     target_bs = bdrv_new("");
1904     ret = bdrv_open(target_bs, target, NULL, flags, drv, &local_err);
1905     if (ret < 0) {
1906         bdrv_unref(target_bs);
1907         error_propagate(errp, local_err);
1908         return;
1909     }
1910
1911     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1912                  block_job_cb, bs, &local_err);
1913     if (local_err != NULL) {
1914         bdrv_unref(target_bs);
1915         error_propagate(errp, local_err);
1916         return;
1917     }
1918 }
1919
1920 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1921
1922 void qmp_drive_mirror(const char *device, const char *target,
1923                       bool has_format, const char *format,
1924                       enum MirrorSyncMode sync,
1925                       bool has_mode, enum NewImageMode mode,
1926                       bool has_speed, int64_t speed,
1927                       bool has_granularity, uint32_t granularity,
1928                       bool has_buf_size, int64_t buf_size,
1929                       bool has_on_source_error, BlockdevOnError on_source_error,
1930                       bool has_on_target_error, BlockdevOnError on_target_error,
1931                       Error **errp)
1932 {
1933     BlockDriverState *bs;
1934     BlockDriverState *source, *target_bs;
1935     BlockDriver *drv = NULL;
1936     Error *local_err = NULL;
1937     int flags;
1938     int64_t size;
1939     int ret;
1940
1941     if (!has_speed) {
1942         speed = 0;
1943     }
1944     if (!has_on_source_error) {
1945         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1946     }
1947     if (!has_on_target_error) {
1948         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1949     }
1950     if (!has_mode) {
1951         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1952     }
1953     if (!has_granularity) {
1954         granularity = 0;
1955     }
1956     if (!has_buf_size) {
1957         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1958     }
1959
1960     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1961         error_set(errp, QERR_INVALID_PARAMETER, device);
1962         return;
1963     }
1964     if (granularity & (granularity - 1)) {
1965         error_set(errp, QERR_INVALID_PARAMETER, device);
1966         return;
1967     }
1968
1969     bs = bdrv_find(device);
1970     if (!bs) {
1971         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1972         return;
1973     }
1974
1975     if (!bdrv_is_inserted(bs)) {
1976         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1977         return;
1978     }
1979
1980     if (!has_format) {
1981         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1982     }
1983     if (format) {
1984         drv = bdrv_find_format(format);
1985         if (!drv) {
1986             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1987             return;
1988         }
1989     }
1990
1991     if (bdrv_in_use(bs)) {
1992         error_set(errp, QERR_DEVICE_IN_USE, device);
1993         return;
1994     }
1995
1996     flags = bs->open_flags | BDRV_O_RDWR;
1997     source = bs->backing_hd;
1998     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1999         sync = MIRROR_SYNC_MODE_FULL;
2000     }
2001
2002     size = bdrv_getlength(bs);
2003     if (size < 0) {
2004         error_setg_errno(errp, -size, "bdrv_getlength failed");
2005         return;
2006     }
2007
2008     if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
2009         /* create new image w/o backing file */
2010         assert(format && drv);
2011         bdrv_img_create(target, format,
2012                         NULL, NULL, NULL, size, flags, &local_err, false);
2013     } else {
2014         switch (mode) {
2015         case NEW_IMAGE_MODE_EXISTING:
2016             break;
2017         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2018             /* create new image with backing file */
2019             bdrv_img_create(target, format,
2020                             source->filename,
2021                             source->drv->format_name,
2022                             NULL, size, flags, &local_err, false);
2023             break;
2024         default:
2025             abort();
2026         }
2027     }
2028
2029     if (error_is_set(&local_err)) {
2030         error_propagate(errp, local_err);
2031         return;
2032     }
2033
2034     /* Mirroring takes care of copy-on-write using the source's backing
2035      * file.
2036      */
2037     target_bs = bdrv_new("");
2038     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv,
2039                     &local_err);
2040     if (ret < 0) {
2041         bdrv_unref(target_bs);
2042         error_propagate(errp, local_err);
2043         return;
2044     }
2045
2046     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2047                  on_source_error, on_target_error,
2048                  block_job_cb, bs, &local_err);
2049     if (local_err != NULL) {
2050         bdrv_unref(target_bs);
2051         error_propagate(errp, local_err);
2052         return;
2053     }
2054 }
2055
2056 static BlockJob *find_block_job(const char *device)
2057 {
2058     BlockDriverState *bs;
2059
2060     bs = bdrv_find(device);
2061     if (!bs || !bs->job) {
2062         return NULL;
2063     }
2064     return bs->job;
2065 }
2066
2067 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2068 {
2069     BlockJob *job = find_block_job(device);
2070
2071     if (!job) {
2072         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2073         return;
2074     }
2075
2076     block_job_set_speed(job, speed, errp);
2077 }
2078
2079 void qmp_block_job_cancel(const char *device,
2080                           bool has_force, bool force, Error **errp)
2081 {
2082     BlockJob *job = find_block_job(device);
2083
2084     if (!has_force) {
2085         force = false;
2086     }
2087
2088     if (!job) {
2089         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2090         return;
2091     }
2092     if (job->paused && !force) {
2093         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
2094         return;
2095     }
2096
2097     trace_qmp_block_job_cancel(job);
2098     block_job_cancel(job);
2099 }
2100
2101 void qmp_block_job_pause(const char *device, Error **errp)
2102 {
2103     BlockJob *job = find_block_job(device);
2104
2105     if (!job) {
2106         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2107         return;
2108     }
2109
2110     trace_qmp_block_job_pause(job);
2111     block_job_pause(job);
2112 }
2113
2114 void qmp_block_job_resume(const char *device, Error **errp)
2115 {
2116     BlockJob *job = find_block_job(device);
2117
2118     if (!job) {
2119         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2120         return;
2121     }
2122
2123     trace_qmp_block_job_resume(job);
2124     block_job_resume(job);
2125 }
2126
2127 void qmp_block_job_complete(const char *device, Error **errp)
2128 {
2129     BlockJob *job = find_block_job(device);
2130
2131     if (!job) {
2132         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2133         return;
2134     }
2135
2136     trace_qmp_block_job_complete(job);
2137     block_job_complete(job, errp);
2138 }
2139
2140 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2141 {
2142     QmpOutputVisitor *ov = qmp_output_visitor_new();
2143     QObject *obj;
2144     QDict *qdict;
2145     DriveInfo *dinfo;
2146     Error *local_err = NULL;
2147
2148     /* Require an ID in the top level */
2149     if (!options->has_id) {
2150         error_setg(errp, "Block device needs an ID");
2151         goto fail;
2152     }
2153
2154     /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2155      * cache.direct=false instead of silently switching to aio=threads, except
2156      * if called from drive_init.
2157      *
2158      * For now, simply forbidding the combination for all drivers will do. */
2159     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2160         bool direct = options->cache->has_direct && options->cache->direct;
2161         if (!options->has_cache && !direct) {
2162             error_setg(errp, "aio=native requires cache.direct=true");
2163             goto fail;
2164         }
2165     }
2166
2167     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2168                                &options, NULL, &local_err);
2169     if (error_is_set(&local_err)) {
2170         error_propagate(errp, local_err);
2171         goto fail;
2172     }
2173
2174     obj = qmp_output_get_qobject(ov);
2175     qdict = qobject_to_qdict(obj);
2176
2177     qdict_flatten(qdict);
2178
2179     dinfo = blockdev_init(qdict, IF_NONE, MEDIA_DISK);
2180     if (!dinfo) {
2181         error_setg(errp, "Could not open image");
2182         goto fail;
2183     }
2184
2185 fail:
2186     qmp_output_visitor_cleanup(ov);
2187 }
2188
2189 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2190 {
2191     BlockJobInfoList **prev = opaque;
2192     BlockJob *job = bs->job;
2193
2194     if (job) {
2195         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2196         elem->value = block_job_query(bs->job);
2197         (*prev)->next = elem;
2198         *prev = elem;
2199     }
2200 }
2201
2202 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2203 {
2204     /* Dummy is a fake list element for holding the head pointer */
2205     BlockJobInfoList dummy = {};
2206     BlockJobInfoList *prev = &dummy;
2207     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2208     return dummy.next;
2209 }
2210
2211 QemuOptsList qemu_common_drive_opts = {
2212     .name = "drive",
2213     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2214     .desc = {
2215         {
2216             .name = "snapshot",
2217             .type = QEMU_OPT_BOOL,
2218             .help = "enable/disable snapshot mode",
2219         },{
2220             .name = "file",
2221             .type = QEMU_OPT_STRING,
2222             .help = "disk image",
2223         },{
2224             .name = "discard",
2225             .type = QEMU_OPT_STRING,
2226             .help = "discard operation (ignore/off, unmap/on)",
2227         },{
2228             .name = "cache.writeback",
2229             .type = QEMU_OPT_BOOL,
2230             .help = "enables writeback mode for any caches",
2231         },{
2232             .name = "cache.direct",
2233             .type = QEMU_OPT_BOOL,
2234             .help = "enables use of O_DIRECT (bypass the host page cache)",
2235         },{
2236             .name = "cache.no-flush",
2237             .type = QEMU_OPT_BOOL,
2238             .help = "ignore any flush requests for the device",
2239         },{
2240             .name = "aio",
2241             .type = QEMU_OPT_STRING,
2242             .help = "host AIO implementation (threads, native)",
2243         },{
2244             .name = "format",
2245             .type = QEMU_OPT_STRING,
2246             .help = "disk format (raw, qcow2, ...)",
2247         },{
2248             .name = "serial",
2249             .type = QEMU_OPT_STRING,
2250             .help = "disk serial number",
2251         },{
2252             .name = "rerror",
2253             .type = QEMU_OPT_STRING,
2254             .help = "read error action",
2255         },{
2256             .name = "werror",
2257             .type = QEMU_OPT_STRING,
2258             .help = "write error action",
2259         },{
2260             .name = "read-only",
2261             .type = QEMU_OPT_BOOL,
2262             .help = "open drive file as read-only",
2263         },{
2264             .name = "throttling.iops-total",
2265             .type = QEMU_OPT_NUMBER,
2266             .help = "limit total I/O operations per second",
2267         },{
2268             .name = "throttling.iops-read",
2269             .type = QEMU_OPT_NUMBER,
2270             .help = "limit read operations per second",
2271         },{
2272             .name = "throttling.iops-write",
2273             .type = QEMU_OPT_NUMBER,
2274             .help = "limit write operations per second",
2275         },{
2276             .name = "throttling.bps-total",
2277             .type = QEMU_OPT_NUMBER,
2278             .help = "limit total bytes per second",
2279         },{
2280             .name = "throttling.bps-read",
2281             .type = QEMU_OPT_NUMBER,
2282             .help = "limit read bytes per second",
2283         },{
2284             .name = "throttling.bps-write",
2285             .type = QEMU_OPT_NUMBER,
2286             .help = "limit write bytes per second",
2287         },{
2288             .name = "throttling.iops-total-max",
2289             .type = QEMU_OPT_NUMBER,
2290             .help = "I/O operations burst",
2291         },{
2292             .name = "throttling.iops-read-max",
2293             .type = QEMU_OPT_NUMBER,
2294             .help = "I/O operations read burst",
2295         },{
2296             .name = "throttling.iops-write-max",
2297             .type = QEMU_OPT_NUMBER,
2298             .help = "I/O operations write burst",
2299         },{
2300             .name = "throttling.bps-total-max",
2301             .type = QEMU_OPT_NUMBER,
2302             .help = "total bytes burst",
2303         },{
2304             .name = "throttling.bps-read-max",
2305             .type = QEMU_OPT_NUMBER,
2306             .help = "total bytes read burst",
2307         },{
2308             .name = "throttling.bps-write-max",
2309             .type = QEMU_OPT_NUMBER,
2310             .help = "total bytes write burst",
2311         },{
2312             .name = "throttling.iops-size",
2313             .type = QEMU_OPT_NUMBER,
2314             .help = "when limiting by iops max size of an I/O in bytes",
2315         },{
2316             .name = "copy-on-read",
2317             .type = QEMU_OPT_BOOL,
2318             .help = "copy read data from backing file into image file",
2319         },
2320         { /* end of list */ }
2321     },
2322 };
2323
2324 QemuOptsList qemu_drive_opts = {
2325     .name = "drive",
2326     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2327     .desc = {
2328         /*
2329          * no elements => accept any params
2330          * validation will happen later
2331          */
2332         { /* end of list */ }
2333     },
2334 };