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