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