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