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