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