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