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