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