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 "qemu/option.h"
  38 #include "qemu/config-file.h"
  39 #include "qapi/qmp/types.h"
  40 #include "qapi-visit.h"
  41 #include "qapi/qmp-output-visitor.h"
  42 #include "sysemu/sysemu.h"
  43 #include "block/block_int.h"
  44 #include "qmp-commands.h"
  45 #include "trace.h"
  46 #include "sysemu/arch_init.h"
  47
  48 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
  49
  50 static const char *const if_name[IF_COUNT] = {
  51     [IF_NONE] = "none",
  52     [IF_IDE] = "ide",
  53     [IF_SCSI] = "scsi",
  54     [IF_FLOPPY] = "floppy",
  55     [IF_PFLASH] = "pflash",
  56     [IF_MTD] = "mtd",
  57     [IF_SD] = "sd",
  58     [IF_VIRTIO] = "virtio",
  59     [IF_XEN] = "xen",
  60 };
  61
  62 static const int if_max_devs[IF_COUNT] = {
  63     /*
  64      * Do not change these numbers!  They govern how drive option
  65      * index maps to unit and bus.  That mapping is ABI.
  66      *
  67      * All controllers used to imlement if=T drives need to support
  68      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
  69      * Otherwise, some index values map to "impossible" bus, unit
  70      * values.
  71      *
  72      * For instance, if you change [IF_SCSI] to 255, -drive
  73      * if=scsi,index=12 no longer means bus=1,unit=5, but
  74      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
  75      * the drive can't be set up.  Regression.
  76      */
  77     [IF_IDE] = 2,
  78     [IF_SCSI] = 7,
  79 };
  80
  81 /*
  82  * We automatically delete the drive when a device using it gets
  83  * unplugged.  Questionable feature, but we can't just drop it.
  84  * Device models call blockdev_mark_auto_del() to schedule the
  85  * automatic deletion, and generic qdev code calls blockdev_auto_del()
  86  * when deletion is actually safe.
  87  */
  88 void blockdev_mark_auto_del(BlockDriverState *bs)
  89 {
  90     DriveInfo *dinfo = drive_get_by_blockdev(bs);
  91
  92     if (dinfo && !dinfo->enable_auto_del) {
  93         return;
  94     }
  95
  96     if (bs->job) {
  97         block_job_cancel(bs->job);
  98     }
  99     if (dinfo) {
 100         dinfo->auto_del = 1;
 101     }
 102 }
 103
 104 void blockdev_auto_del(BlockDriverState *bs)
 105 {
 106     DriveInfo *dinfo = drive_get_by_blockdev(bs);
 107
 108     if (dinfo && dinfo->auto_del) {
 109         drive_put_ref(dinfo);
 110     }
 111 }
 112
 113 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
 114 {
 115     int max_devs = if_max_devs[type];
 116     return max_devs ? index / max_devs : 0;
 117 }
 118
 119 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
 120 {
 121     int max_devs = if_max_devs[type];
 122     return max_devs ? index % max_devs : index;
 123 }
 124
 125 QemuOpts *drive_def(const char *optstr)
 126 {
 127     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
 128 }
 129
 130 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
 131                     const char *optstr)
 132 {
 133     QemuOpts *opts;
 134     char buf[32];
 135
 136     opts = drive_def(optstr);
 137     if (!opts) {
 138         return NULL;
 139     }
 140     if (type != IF_DEFAULT) {
 141         qemu_opt_set(opts, "if", if_name[type]);
 142     }
 143     if (index >= 0) {
 144         snprintf(buf, sizeof(buf), "%d", index);
 145         qemu_opt_set(opts, "index", buf);
 146     }
 147     if (file)
 148         qemu_opt_set(opts, "file", file);
 149     return opts;
 150 }
 151
 152 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
 153 {
 154     DriveInfo *dinfo;
 155
 156     /* seek interface, bus and unit */
 157
 158     QTAILQ_FOREACH(dinfo, &drives, next) {
 159         if (dinfo->type == type &&
 160             dinfo->bus == bus &&
 161             dinfo->unit == unit)
 162             return dinfo;
 163     }
 164
 165     return NULL;
 166 }
 167
 168 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
 169 {
 170     return drive_get(type,
 171                      drive_index_to_bus_id(type, index),
 172                      drive_index_to_unit_id(type, index));
 173 }
 174
 175 int drive_get_max_bus(BlockInterfaceType type)
 176 {
 177     int max_bus;
 178     DriveInfo *dinfo;
 179
 180     max_bus = -1;
 181     QTAILQ_FOREACH(dinfo, &drives, next) {
 182         if(dinfo->type == type &&
 183            dinfo->bus > max_bus)
 184             max_bus = dinfo->bus;
 185     }
 186     return max_bus;
 187 }
 188
 189 /* Get a block device.  This should only be used for single-drive devices
 190    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
 191    appropriate bus.  */
 192 DriveInfo *drive_get_next(BlockInterfaceType type)
 193 {
 194     static int next_block_unit[IF_COUNT];
 195
 196     return drive_get(type, 0, next_block_unit[type]++);
 197 }
 198
 199 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
 200 {
 201     DriveInfo *dinfo;
 202
 203     QTAILQ_FOREACH(dinfo, &drives, next) {
 204         if (dinfo->bdrv == bs) {
 205             return dinfo;
 206         }
 207     }
 208     return NULL;
 209 }
 210
 211 static void bdrv_format_print(void *opaque, const char *name)
 212 {
 213     error_printf(" %s", name);
 214 }
 215
 216 static void drive_uninit(DriveInfo *dinfo)
 217 {
 218     if (dinfo->opts) {
 219         qemu_opts_del(dinfo->opts);
 220     }
 221
 222     bdrv_unref(dinfo->bdrv);
 223     g_free(dinfo->id);
 224     QTAILQ_REMOVE(&drives, dinfo, next);
 225     g_free(dinfo->serial);
 226     g_free(dinfo);
 227 }
 228
 229 void drive_put_ref(DriveInfo *dinfo)
 230 {
 231     assert(dinfo->refcount);
 232     if (--dinfo->refcount == 0) {
 233         drive_uninit(dinfo);
 234     }
 235 }
 236
 237 void drive_get_ref(DriveInfo *dinfo)
 238 {
 239     dinfo->refcount++;
 240 }
 241
 242 typedef struct {
 243     QEMUBH *bh;
 244     BlockDriverState *bs;
 245 } BDRVPutRefBH;
 246
 247 static void bdrv_put_ref_bh(void *opaque)
 248 {
 249     BDRVPutRefBH *s = opaque;
 250
 251     bdrv_unref(s->bs);
 252     qemu_bh_delete(s->bh);
 253     g_free(s);
 254 }
 255
 256 /*
 257  * Release a BDS reference in a BH
 258  *
 259  * It is not safe to use bdrv_unref() from a callback function when the callers
 260  * still need the BlockDriverState.  In such cases we schedule a BH to release
 261  * the reference.
 262  */
 263 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
 264 {
 265     BDRVPutRefBH *s;
 266
 267     s = g_new(BDRVPutRefBH, 1);
 268     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
 269     s->bs = bs;
 270     qemu_bh_schedule(s->bh);
 271 }
 272
 273 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
 274 {
 275     if (!strcmp(buf, "ignore")) {
 276         return BLOCKDEV_ON_ERROR_IGNORE;
 277     } else if (!is_read && !strcmp(buf, "enospc")) {
 278         return BLOCKDEV_ON_ERROR_ENOSPC;
 279     } else if (!strcmp(buf, "stop")) {
 280         return BLOCKDEV_ON_ERROR_STOP;
 281     } else if (!strcmp(buf, "report")) {
 282         return BLOCKDEV_ON_ERROR_REPORT;
 283     } else {
 284         error_setg(errp, "'%s' invalid %s error action",
 285                    buf, is_read ? "read" : "write");
 286         return -1;
 287     }
 288 }
 289
 290 static inline int parse_enum_option(const char *lookup[], const char *buf,
 291                                     int max, int def, Error **errp)
 292 {
 293     int i;
 294
 295     if (!buf) {
 296         return def;
 297     }
 298
 299     for (i = 0; i < max; i++) {
 300         if (!strcmp(buf, lookup[i])) {
 301             return i;
 302         }
 303     }
 304
 305     error_setg(errp, "invalid parameter value: %s", buf);
 306     return def;
 307 }
 308
 309 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
 310 {
 311     if (throttle_conflicting(cfg)) {
 312         error_setg(errp, "bps/iops/max total values and read/write values"
 313                          " cannot be used at the same time");
 314         return false;
 315     }
 316
 317     if (!throttle_is_valid(cfg)) {
 318         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
 319         return false;
 320     }
 321
 322     return true;
 323 }
 324
 325 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
 326
 327 /* Takes the ownership of bs_opts */
 328 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
 329                                 Error **errp)
 330 {
 331     const char *buf;
 332     const char *serial;
 333     int ro = 0;
 334     int bdrv_flags = 0;
 335     int on_read_error, on_write_error;
 336     DriveInfo *dinfo;
 337     ThrottleConfig cfg;
 338     int snapshot = 0;
 339     bool copy_on_read;
 340     int ret;
 341     Error *error = NULL;
 342     QemuOpts *opts;
 343     const char *id;
 344     bool has_driver_specific_opts;
 345     BlockdevDetectZeroesOptions detect_zeroes;
 346     BlockDriver *drv = NULL;
 347
 348     /* Check common options by copying from bs_opts to opts, all other options
 349      * stay in bs_opts for processing by bdrv_open(). */
 350     id = qdict_get_try_str(bs_opts, "id");
 351     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
 352     if (error) {
 353         error_propagate(errp, error);
 354         return NULL;
 355     }
 356
 357     qemu_opts_absorb_qdict(opts, bs_opts, &error);
 358     if (error) {
 359         error_propagate(errp, error);
 360         goto early_err;
 361     }
 362
 363     if (id) {
 364         qdict_del(bs_opts, "id");
 365     }
 366
 367     has_driver_specific_opts = !!qdict_size(bs_opts);
 368
 369     /* extract parameters */
 370     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 371     ro = qemu_opt_get_bool(opts, "read-only", 0);
 372     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
 373
 374     serial = qemu_opt_get(opts, "serial");
 375
 376     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
 377         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
 378             error_setg(errp, "invalid discard option");
 379             goto early_err;
 380         }
 381     }
 382
 383     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
 384         bdrv_flags |= BDRV_O_CACHE_WB;
 385     }
 386     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
 387         bdrv_flags |= BDRV_O_NOCACHE;
 388     }
 389     if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
 390         bdrv_flags |= BDRV_O_NO_FLUSH;
 391     }
 392
 393 #ifdef CONFIG_LINUX_AIO
 394     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 395         if (!strcmp(buf, "native")) {
 396             bdrv_flags |= BDRV_O_NATIVE_AIO;
 397         } else if (!strcmp(buf, "threads")) {
 398             /* this is the default */
 399         } else {
 400            error_setg(errp, "invalid aio option");
 401            goto early_err;
 402         }
 403     }
 404 #endif
 405
 406     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 407         if (is_help_option(buf)) {
 408             error_printf("Supported formats:");
 409             bdrv_iterate_format(bdrv_format_print, NULL);
 410             error_printf("\n");
 411             goto early_err;
 412         }
 413
 414         drv = bdrv_find_format(buf);
 415         if (!drv) {
 416             error_setg(errp, "'%s' invalid format", buf);
 417             goto early_err;
 418         }
 419     }
 420
 421     /* disk I/O throttling */
 422     memset(&cfg, 0, sizeof(cfg));
 423     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
 424         qemu_opt_get_number(opts, "throttling.bps-total", 0);
 425     cfg.buckets[THROTTLE_BPS_READ].avg  =
 426         qemu_opt_get_number(opts, "throttling.bps-read", 0);
 427     cfg.buckets[THROTTLE_BPS_WRITE].avg =
 428         qemu_opt_get_number(opts, "throttling.bps-write", 0);
 429     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
 430         qemu_opt_get_number(opts, "throttling.iops-total", 0);
 431     cfg.buckets[THROTTLE_OPS_READ].avg =
 432         qemu_opt_get_number(opts, "throttling.iops-read", 0);
 433     cfg.buckets[THROTTLE_OPS_WRITE].avg =
 434         qemu_opt_get_number(opts, "throttling.iops-write", 0);
 435
 436     cfg.buckets[THROTTLE_BPS_TOTAL].max =
 437         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
 438     cfg.buckets[THROTTLE_BPS_READ].max  =
 439         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
 440     cfg.buckets[THROTTLE_BPS_WRITE].max =
 441         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
 442     cfg.buckets[THROTTLE_OPS_TOTAL].max =
 443         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
 444     cfg.buckets[THROTTLE_OPS_READ].max =
 445         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
 446     cfg.buckets[THROTTLE_OPS_WRITE].max =
 447         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
 448
 449     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
 450
 451     if (!check_throttle_config(&cfg, &error)) {
 452         error_propagate(errp, error);
 453         goto early_err;
 454     }
 455
 456     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
 457     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 458         on_write_error = parse_block_error_action(buf, 0, &error);
 459         if (error) {
 460             error_propagate(errp, error);
 461             goto early_err;
 462         }
 463     }
 464
 465     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
 466     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 467         on_read_error = parse_block_error_action(buf, 1, &error);
 468         if (error) {
 469             error_propagate(errp, error);
 470             goto early_err;
 471         }
 472     }
 473
 474     detect_zeroes =
 475         parse_enum_option(BlockdevDetectZeroesOptions_lookup,
 476                           qemu_opt_get(opts, "detect-zeroes"),
 477                           BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
 478                           BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
 479                           &error);
 480     if (error) {
 481         error_propagate(errp, error);
 482         goto early_err;
 483     }
 484
 485     if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
 486         !(bdrv_flags & BDRV_O_UNMAP)) {
 487         error_setg(errp, "setting detect-zeroes to unmap is not allowed "
 488                          "without setting discard operation to unmap");
 489         goto early_err;
 490     }
 491
 492     /* init */
 493     dinfo = g_malloc0(sizeof(*dinfo));
 494     dinfo->id = g_strdup(qemu_opts_id(opts));
 495     dinfo->bdrv = bdrv_new(dinfo->id, &error);
 496     if (error) {
 497         error_propagate(errp, error);
 498         goto bdrv_new_err;
 499     }
 500     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
 501     dinfo->bdrv->read_only = ro;
 502     dinfo->bdrv->detect_zeroes = detect_zeroes;
 503     dinfo->refcount = 1;
 504     if (serial != NULL) {
 505         dinfo->serial = g_strdup(serial);
 506     }
 507     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 508
 509     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
 510
 511     /* disk I/O throttling */
 512     if (throttle_enabled(&cfg)) {
 513         bdrv_io_limits_enable(dinfo->bdrv);
 514         bdrv_set_io_limits(dinfo->bdrv, &cfg);
 515     }
 516
 517     if (!file || !*file) {
 518         if (has_driver_specific_opts) {
 519             file = NULL;
 520         } else {
 521             QDECREF(bs_opts);
 522             qemu_opts_del(opts);
 523             return dinfo;
 524         }
 525     }
 526     if (snapshot) {
 527         /* always use cache=unsafe with snapshot */
 528         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 529         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 530     }
 531
 532     if (copy_on_read) {
 533         bdrv_flags |= BDRV_O_COPY_ON_READ;
 534     }
 535
 536     if (runstate_check(RUN_STATE_INMIGRATE)) {
 537         bdrv_flags |= BDRV_O_INCOMING;
 538     }
 539
 540     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 541
 542     QINCREF(bs_opts);
 543     ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
 544
 545     if (ret < 0) {
 546         error_setg(errp, "could not open disk image %s: %s",
 547                    file ?: dinfo->id, error_get_pretty(error));
 548         error_free(error);
 549         goto err;
 550     }
 551
 552     if (bdrv_key_required(dinfo->bdrv))
 553         autostart = 0;
 554
 555     QDECREF(bs_opts);
 556     qemu_opts_del(opts);
 557
 558     return dinfo;
 559
 560 err:
 561     bdrv_unref(dinfo->bdrv);
 562     QTAILQ_REMOVE(&drives, dinfo, next);
 563 bdrv_new_err:
 564     g_free(dinfo->id);
 565     g_free(dinfo);
 566 early_err:
 567     QDECREF(bs_opts);
 568     qemu_opts_del(opts);
 569     return NULL;
 570 }
 571
 572 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
 573 {
 574     const char *value;
 575
 576     value = qemu_opt_get(opts, from);
 577     if (value) {
 578         qemu_opt_set(opts, to, value);
 579         qemu_opt_unset(opts, from);
 580     }
 581 }
 582
 583 QemuOptsList qemu_legacy_drive_opts = {
 584     .name = "drive",
 585     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
 586     .desc = {
 587         {
 588             .name = "bus",
 589             .type = QEMU_OPT_NUMBER,
 590             .help = "bus number",
 591         },{
 592             .name = "unit",
 593             .type = QEMU_OPT_NUMBER,
 594             .help = "unit number (i.e. lun for scsi)",
 595         },{
 596             .name = "index",
 597             .type = QEMU_OPT_NUMBER,
 598             .help = "index number",
 599         },{
 600             .name = "media",
 601             .type = QEMU_OPT_STRING,
 602             .help = "media type (disk, cdrom)",
 603         },{
 604             .name = "if",
 605             .type = QEMU_OPT_STRING,
 606             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
 607         },{
 608             .name = "cyls",
 609             .type = QEMU_OPT_NUMBER,
 610             .help = "number of cylinders (ide disk geometry)",
 611         },{
 612             .name = "heads",
 613             .type = QEMU_OPT_NUMBER,
 614             .help = "number of heads (ide disk geometry)",
 615         },{
 616             .name = "secs",
 617             .type = QEMU_OPT_NUMBER,
 618             .help = "number of sectors (ide disk geometry)",
 619         },{
 620             .name = "trans",
 621             .type = QEMU_OPT_STRING,
 622             .help = "chs translation (auto, lba, none)",
 623         },{
 624             .name = "boot",
 625             .type = QEMU_OPT_BOOL,
 626             .help = "(deprecated, ignored)",
 627         },{
 628             .name = "addr",
 629             .type = QEMU_OPT_STRING,
 630             .help = "pci address (virtio only)",
 631         },{
 632             .name = "file",
 633             .type = QEMU_OPT_STRING,
 634             .help = "file name",
 635         },
 636
 637         /* Options that are passed on, but have special semantics with -drive */
 638         {
 639             .name = "read-only",
 640             .type = QEMU_OPT_BOOL,
 641             .help = "open drive file as read-only",
 642         },{
 643             .name = "rerror",
 644             .type = QEMU_OPT_STRING,
 645             .help = "read error action",
 646         },{
 647             .name = "werror",
 648             .type = QEMU_OPT_STRING,
 649             .help = "write error action",
 650         },{
 651             .name = "copy-on-read",
 652             .type = QEMU_OPT_BOOL,
 653             .help = "copy read data from backing file into image file",
 654         },
 655
 656         { /* end of list */ }
 657     },
 658 };
 659
 660 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
 661 {
 662     const char *value;
 663     DriveInfo *dinfo = NULL;
 664     QDict *bs_opts;
 665     QemuOpts *legacy_opts;
 666     DriveMediaType media = MEDIA_DISK;
 667     BlockInterfaceType type;
 668     int cyls, heads, secs, translation;
 669     int max_devs, bus_id, unit_id, index;
 670     const char *devaddr;
 671     const char *werror, *rerror;
 672     bool read_only = false;
 673     bool copy_on_read;
 674     const char *filename;
 675     Error *local_err = NULL;
 676
 677     /* Change legacy command line options into QMP ones */
 678     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
 679     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
 680     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
 681
 682     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
 683     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
 684     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
 685
 686     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
 687     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
 688     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
 689
 690     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
 691     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
 692     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
 693
 694     qemu_opt_rename(all_opts,
 695                     "iops_size", "throttling.iops-size");
 696
 697     qemu_opt_rename(all_opts, "readonly", "read-only");
 698
 699     value = qemu_opt_get(all_opts, "cache");
 700     if (value) {
 701         int flags = 0;
 702
 703         if (bdrv_parse_cache_flags(value, &flags) != 0) {
 704             error_report("invalid cache option");
 705             return NULL;
 706         }
 707
 708         /* Specific options take precedence */
 709         if (!qemu_opt_get(all_opts, "cache.writeback")) {
 710             qemu_opt_set_bool(all_opts, "cache.writeback",
 711                               !!(flags & BDRV_O_CACHE_WB));
 712         }
 713         if (!qemu_opt_get(all_opts, "cache.direct")) {
 714             qemu_opt_set_bool(all_opts, "cache.direct",
 715                               !!(flags & BDRV_O_NOCACHE));
 716         }
 717         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
 718             qemu_opt_set_bool(all_opts, "cache.no-flush",
 719                               !!(flags & BDRV_O_NO_FLUSH));
 720         }
 721         qemu_opt_unset(all_opts, "cache");
 722     }
 723
 724     /* Get a QDict for processing the options */
 725     bs_opts = qdict_new();
 726     qemu_opts_to_qdict(all_opts, bs_opts);
 727
 728     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
 729                                    &error_abort);
 730     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
 731     if (local_err) {
 732         error_report("%s", error_get_pretty(local_err));
 733         error_free(local_err);
 734         goto fail;
 735     }
 736
 737     /* Deprecated option boot=[on|off] */
 738     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
 739         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 740                 "ignored. Future versions will reject this parameter. Please "
 741                 "update your scripts.\n");
 742     }
 743
 744     /* Media type */
 745     value = qemu_opt_get(legacy_opts, "media");
 746     if (value) {
 747         if (!strcmp(value, "disk")) {
 748             media = MEDIA_DISK;
 749         } else if (!strcmp(value, "cdrom")) {
 750             media = MEDIA_CDROM;
 751             read_only = true;
 752         } else {
 753             error_report("'%s' invalid media", value);
 754             goto fail;
 755         }
 756     }
 757
 758     /* copy-on-read is disabled with a warning for read-only devices */
 759     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
 760     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
 761
 762     if (read_only && copy_on_read) {
 763         error_report("warning: disabling copy-on-read on read-only drive");
 764         copy_on_read = false;
 765     }
 766
 767     qdict_put(bs_opts, "read-only",
 768               qstring_from_str(read_only ? "on" : "off"));
 769     qdict_put(bs_opts, "copy-on-read",
 770               qstring_from_str(copy_on_read ? "on" :"off"));
 771
 772     /* Controller type */
 773     value = qemu_opt_get(legacy_opts, "if");
 774     if (value) {
 775         for (type = 0;
 776              type < IF_COUNT && strcmp(value, if_name[type]);
 777              type++) {
 778         }
 779         if (type == IF_COUNT) {
 780             error_report("unsupported bus type '%s'", value);
 781             goto fail;
 782         }
 783     } else {
 784         type = block_default_type;
 785     }
 786
 787     /* Geometry */
 788     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
 789     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
 790     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
 791
 792     if (cyls || heads || secs) {
 793         if (cyls < 1) {
 794             error_report("invalid physical cyls number");
 795             goto fail;
 796         }
 797         if (heads < 1) {
 798             error_report("invalid physical heads number");
 799             goto fail;
 800         }
 801         if (secs < 1) {
 802             error_report("invalid physical secs number");
 803             goto fail;
 804         }
 805     }
 806
 807     translation = BIOS_ATA_TRANSLATION_AUTO;
 808     value = qemu_opt_get(legacy_opts, "trans");
 809     if (value != NULL) {
 810         if (!cyls) {
 811             error_report("'%s' trans must be used with cyls, heads and secs",
 812                          value);
 813             goto fail;
 814         }
 815         if (!strcmp(value, "none")) {
 816             translation = BIOS_ATA_TRANSLATION_NONE;
 817         } else if (!strcmp(value, "lba")) {
 818             translation = BIOS_ATA_TRANSLATION_LBA;
 819         } else if (!strcmp(value, "large")) {
 820             translation = BIOS_ATA_TRANSLATION_LARGE;
 821         } else if (!strcmp(value, "rechs")) {
 822             translation = BIOS_ATA_TRANSLATION_RECHS;
 823         } else if (!strcmp(value, "auto")) {
 824             translation = BIOS_ATA_TRANSLATION_AUTO;
 825         } else {
 826             error_report("'%s' invalid translation type", value);
 827             goto fail;
 828         }
 829     }
 830
 831     if (media == MEDIA_CDROM) {
 832         if (cyls || secs || heads) {
 833             error_report("CHS can't be set with media=cdrom");
 834             goto fail;
 835         }
 836     }
 837
 838     /* Device address specified by bus/unit or index.
 839      * If none was specified, try to find the first free one. */
 840     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
 841     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
 842     index   = qemu_opt_get_number(legacy_opts, "index", -1);
 843
 844     max_devs = if_max_devs[type];
 845
 846     if (index != -1) {
 847         if (bus_id != 0 || unit_id != -1) {
 848             error_report("index cannot be used with bus and unit");
 849             goto fail;
 850         }
 851         bus_id = drive_index_to_bus_id(type, index);
 852         unit_id = drive_index_to_unit_id(type, index);
 853     }
 854
 855     if (unit_id == -1) {
 856        unit_id = 0;
 857        while (drive_get(type, bus_id, unit_id) != NULL) {
 858            unit_id++;
 859            if (max_devs && unit_id >= max_devs) {
 860                unit_id -= max_devs;
 861                bus_id++;
 862            }
 863        }
 864     }
 865
 866     if (max_devs && unit_id >= max_devs) {
 867         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
 868         goto fail;
 869     }
 870
 871     if (drive_get(type, bus_id, unit_id) != NULL) {
 872         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 873                      bus_id, unit_id, index);
 874         goto fail;
 875     }
 876
 877     /* no id supplied -> create one */
 878     if (qemu_opts_id(all_opts) == NULL) {
 879         char *new_id;
 880         const char *mediastr = "";
 881         if (type == IF_IDE || type == IF_SCSI) {
 882             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 883         }
 884         if (max_devs) {
 885             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
 886                                      mediastr, unit_id);
 887         } else {
 888             new_id = g_strdup_printf("%s%s%i", if_name[type],
 889                                      mediastr, unit_id);
 890         }
 891         qdict_put(bs_opts, "id", qstring_from_str(new_id));
 892         g_free(new_id);
 893     }
 894
 895     /* Add virtio block device */
 896     devaddr = qemu_opt_get(legacy_opts, "addr");
 897     if (devaddr && type != IF_VIRTIO) {
 898         error_report("addr is not supported by this bus type");
 899         goto fail;
 900     }
 901
 902     if (type == IF_VIRTIO) {
 903         QemuOpts *devopts;
 904         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
 905                                    &error_abort);
 906         if (arch_type == QEMU_ARCH_S390X) {
 907             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
 908         } else {
 909             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
 910         }
 911         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
 912         if (devaddr) {
 913             qemu_opt_set(devopts, "addr", devaddr);
 914         }
 915     }
 916
 917     filename = qemu_opt_get(legacy_opts, "file");
 918
 919     /* Check werror/rerror compatibility with if=... */
 920     werror = qemu_opt_get(legacy_opts, "werror");
 921     if (werror != NULL) {
 922         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
 923             type != IF_NONE) {
 924             error_report("werror is not supported by this bus type");
 925             goto fail;
 926         }
 927         qdict_put(bs_opts, "werror", qstring_from_str(werror));
 928     }
 929
 930     rerror = qemu_opt_get(legacy_opts, "rerror");
 931     if (rerror != NULL) {
 932         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
 933             type != IF_NONE) {
 934             error_report("rerror is not supported by this bus type");
 935             goto fail;
 936         }
 937         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
 938     }
 939
 940     /* Actual block device init: Functionality shared with blockdev-add */
 941     dinfo = blockdev_init(filename, bs_opts, &local_err);
 942     if (dinfo == NULL) {
 943         if (local_err) {
 944             error_report("%s", error_get_pretty(local_err));
 945             error_free(local_err);
 946         }
 947         goto fail;
 948     } else {
 949         assert(!local_err);
 950     }
 951
 952     /* Set legacy DriveInfo fields */
 953     dinfo->enable_auto_del = true;
 954     dinfo->opts = all_opts;
 955
 956     dinfo->cyls = cyls;
 957     dinfo->heads = heads;
 958     dinfo->secs = secs;
 959     dinfo->trans = translation;
 960
 961     dinfo->type = type;
 962     dinfo->bus = bus_id;
 963     dinfo->unit = unit_id;
 964     dinfo->devaddr = devaddr;
 965
 966     switch(type) {
 967     case IF_IDE:
 968     case IF_SCSI:
 969     case IF_XEN:
 970     case IF_NONE:
 971         dinfo->media_cd = media == MEDIA_CDROM;
 972         break;
 973     default:
 974         break;
 975     }
 976
 977 fail:
 978     qemu_opts_del(legacy_opts);
 979     return dinfo;
 980 }
 981
 982 void do_commit(Monitor *mon, const QDict *qdict)
 983 {
 984     const char *device = qdict_get_str(qdict, "device");
 985     BlockDriverState *bs;
 986     int ret;
 987
 988     if (!strcmp(device, "all")) {
 989         ret = bdrv_commit_all();
 990     } else {
 991         bs = bdrv_find(device);
 992         if (!bs) {
 993             monitor_printf(mon, "Device '%s' not found\n", device);
 994             return;
 995         }
 996         ret = bdrv_commit(bs);
 997     }
 998     if (ret < 0) {
 999         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1000                        strerror(-ret));
1001     }
1002 }
1003
1004 static void blockdev_do_action(int kind, void *data, Error **errp)
1005 {
1006     TransactionAction action;
1007     TransactionActionList list;
1008
1009     action.kind = kind;
1010     action.data = data;
1011     list.value = &action;
1012     list.next = NULL;
1013     qmp_transaction(&list, errp);
1014 }
1015
1016 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1017                                 bool has_node_name, const char *node_name,
1018                                 const char *snapshot_file,
1019                                 bool has_snapshot_node_name,
1020                                 const char *snapshot_node_name,
1021                                 bool has_format, const char *format,
1022                                 bool has_mode, NewImageMode mode, Error **errp)
1023 {
1024     BlockdevSnapshot snapshot = {
1025         .has_device = has_device,
1026         .device = (char *) device,
1027         .has_node_name = has_node_name,
1028         .node_name = (char *) node_name,
1029         .snapshot_file = (char *) snapshot_file,
1030         .has_snapshot_node_name = has_snapshot_node_name,
1031         .snapshot_node_name = (char *) snapshot_node_name,
1032         .has_format = has_format,
1033         .format = (char *) format,
1034         .has_mode = has_mode,
1035         .mode = mode,
1036     };
1037     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1038                        &snapshot, errp);
1039 }
1040
1041 void qmp_blockdev_snapshot_internal_sync(const char *device,
1042                                          const char *name,
1043                                          Error **errp)
1044 {
1045     BlockdevSnapshotInternal snapshot = {
1046         .device = (char *) device,
1047         .name = (char *) name
1048     };
1049
1050     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1051                        &snapshot, errp);
1052 }
1053
1054 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1055                                                          bool has_id,
1056                                                          const char *id,
1057                                                          bool has_name,
1058                                                          const char *name,
1059                                                          Error **errp)
1060 {
1061     BlockDriverState *bs = bdrv_find(device);
1062     QEMUSnapshotInfo sn;
1063     Error *local_err = NULL;
1064     SnapshotInfo *info = NULL;
1065     int ret;
1066
1067     if (!bs) {
1068         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1069         return NULL;
1070     }
1071
1072     if (!has_id) {
1073         id = NULL;
1074     }
1075
1076     if (!has_name) {
1077         name = NULL;
1078     }
1079
1080     if (!id && !name) {
1081         error_setg(errp, "Name or id must be provided");
1082         return NULL;
1083     }
1084
1085     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1086     if (local_err) {
1087         error_propagate(errp, local_err);
1088         return NULL;
1089     }
1090     if (!ret) {
1091         error_setg(errp,
1092                    "Snapshot with id '%s' and name '%s' does not exist on "
1093                    "device '%s'",
1094                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1095         return NULL;
1096     }
1097
1098     bdrv_snapshot_delete(bs, id, name, &local_err);
1099     if (local_err) {
1100         error_propagate(errp, local_err);
1101         return NULL;
1102     }
1103
1104     info = g_malloc0(sizeof(SnapshotInfo));
1105     info->id = g_strdup(sn.id_str);
1106     info->name = g_strdup(sn.name);
1107     info->date_nsec = sn.date_nsec;
1108     info->date_sec = sn.date_sec;
1109     info->vm_state_size = sn.vm_state_size;
1110     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1111     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1112
1113     return info;
1114 }
1115
1116 /* New and old BlockDriverState structs for group snapshots */
1117
1118 typedef struct BlkTransactionState BlkTransactionState;
1119
1120 /* Only prepare() may fail. In a single transaction, only one of commit() or
1121    abort() will be called, clean() will always be called if it present. */
1122 typedef struct BdrvActionOps {
1123     /* Size of state struct, in bytes. */
1124     size_t instance_size;
1125     /* Prepare the work, must NOT be NULL. */
1126     void (*prepare)(BlkTransactionState *common, Error **errp);
1127     /* Commit the changes, can be NULL. */
1128     void (*commit)(BlkTransactionState *common);
1129     /* Abort the changes on fail, can be NULL. */
1130     void (*abort)(BlkTransactionState *common);
1131     /* Clean up resource in the end, can be NULL. */
1132     void (*clean)(BlkTransactionState *common);
1133 } BdrvActionOps;
1134
1135 /*
1136  * This structure must be arranged as first member in child type, assuming
1137  * that compiler will also arrange it to the same address with parent instance.
1138  * Later it will be used in free().
1139  */
1140 struct BlkTransactionState {
1141     TransactionAction *action;
1142     const BdrvActionOps *ops;
1143     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1144 };
1145
1146 /* internal snapshot private data */
1147 typedef struct InternalSnapshotState {
1148     BlkTransactionState common;
1149     BlockDriverState *bs;
1150     QEMUSnapshotInfo sn;
1151 } InternalSnapshotState;
1152
1153 static void internal_snapshot_prepare(BlkTransactionState *common,
1154                                       Error **errp)
1155 {
1156     Error *local_err = NULL;
1157     const char *device;
1158     const char *name;
1159     BlockDriverState *bs;
1160     QEMUSnapshotInfo old_sn, *sn;
1161     bool ret;
1162     qemu_timeval tv;
1163     BlockdevSnapshotInternal *internal;
1164     InternalSnapshotState *state;
1165     int ret1;
1166
1167     g_assert(common->action->kind ==
1168              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1169     internal = common->action->blockdev_snapshot_internal_sync;
1170     state = DO_UPCAST(InternalSnapshotState, common, common);
1171
1172     /* 1. parse input */
1173     device = internal->device;
1174     name = internal->name;
1175
1176     /* 2. check for validation */
1177     bs = bdrv_find(device);
1178     if (!bs) {
1179         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1180         return;
1181     }
1182
1183     if (!bdrv_is_inserted(bs)) {
1184         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1185         return;
1186     }
1187
1188     if (bdrv_is_read_only(bs)) {
1189         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1190         return;
1191     }
1192
1193     if (!bdrv_can_snapshot(bs)) {
1194         error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1195                   bs->drv->format_name, device, "internal snapshot");
1196         return;
1197     }
1198
1199     if (!strlen(name)) {
1200         error_setg(errp, "Name is empty");
1201         return;
1202     }
1203
1204     /* check whether a snapshot with name exist */
1205     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1206                                             &local_err);
1207     if (local_err) {
1208         error_propagate(errp, local_err);
1209         return;
1210     } else if (ret) {
1211         error_setg(errp,
1212                    "Snapshot with name '%s' already exists on device '%s'",
1213                    name, device);
1214         return;
1215     }
1216
1217     /* 3. take the snapshot */
1218     sn = &state->sn;
1219     pstrcpy(sn->name, sizeof(sn->name), name);
1220     qemu_gettimeofday(&tv);
1221     sn->date_sec = tv.tv_sec;
1222     sn->date_nsec = tv.tv_usec * 1000;
1223     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1224
1225     ret1 = bdrv_snapshot_create(bs, sn);
1226     if (ret1 < 0) {
1227         error_setg_errno(errp, -ret1,
1228                          "Failed to create snapshot '%s' on device '%s'",
1229                          name, device);
1230         return;
1231     }
1232
1233     /* 4. succeed, mark a snapshot is created */
1234     state->bs = bs;
1235 }
1236
1237 static void internal_snapshot_abort(BlkTransactionState *common)
1238 {
1239     InternalSnapshotState *state =
1240                              DO_UPCAST(InternalSnapshotState, common, common);
1241     BlockDriverState *bs = state->bs;
1242     QEMUSnapshotInfo *sn = &state->sn;
1243     Error *local_error = NULL;
1244
1245     if (!bs) {
1246         return;
1247     }
1248
1249     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1250         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1251                      "device '%s' in abort: %s",
1252                      sn->id_str,
1253                      sn->name,
1254                      bdrv_get_device_name(bs),
1255                      error_get_pretty(local_error));
1256         error_free(local_error);
1257     }
1258 }
1259
1260 /* external snapshot private data */
1261 typedef struct ExternalSnapshotState {
1262     BlkTransactionState common;
1263     BlockDriverState *old_bs;
1264     BlockDriverState *new_bs;
1265 } ExternalSnapshotState;
1266
1267 static void external_snapshot_prepare(BlkTransactionState *common,
1268                                       Error **errp)
1269 {
1270     BlockDriver *drv;
1271     int flags, ret;
1272     QDict *options = NULL;
1273     Error *local_err = NULL;
1274     bool has_device = false;
1275     const char *device;
1276     bool has_node_name = false;
1277     const char *node_name;
1278     bool has_snapshot_node_name = false;
1279     const char *snapshot_node_name;
1280     const char *new_image_file;
1281     const char *format = "qcow2";
1282     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1283     ExternalSnapshotState *state =
1284                              DO_UPCAST(ExternalSnapshotState, common, common);
1285     TransactionAction *action = common->action;
1286
1287     /* get parameters */
1288     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1289
1290     has_device = action->blockdev_snapshot_sync->has_device;
1291     device = action->blockdev_snapshot_sync->device;
1292     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1293     node_name = action->blockdev_snapshot_sync->node_name;
1294     has_snapshot_node_name =
1295         action->blockdev_snapshot_sync->has_snapshot_node_name;
1296     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1297
1298     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1299     if (action->blockdev_snapshot_sync->has_format) {
1300         format = action->blockdev_snapshot_sync->format;
1301     }
1302     if (action->blockdev_snapshot_sync->has_mode) {
1303         mode = action->blockdev_snapshot_sync->mode;
1304     }
1305
1306     /* start processing */
1307     drv = bdrv_find_format(format);
1308     if (!drv) {
1309         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1310         return;
1311     }
1312
1313     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1314                                    has_node_name ? node_name : NULL,
1315                                    &local_err);
1316     if (local_err) {
1317         error_propagate(errp, local_err);
1318         return;
1319     }
1320
1321     if (has_node_name && !has_snapshot_node_name) {
1322         error_setg(errp, "New snapshot node name missing");
1323         return;
1324     }
1325
1326     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1327         error_setg(errp, "New snapshot node name already existing");
1328         return;
1329     }
1330
1331     if (!bdrv_is_inserted(state->old_bs)) {
1332         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1333         return;
1334     }
1335
1336     if (bdrv_op_is_blocked(state->old_bs,
1337                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1338         return;
1339     }
1340
1341     if (!bdrv_is_read_only(state->old_bs)) {
1342         if (bdrv_flush(state->old_bs)) {
1343             error_set(errp, QERR_IO_ERROR);
1344             return;
1345         }
1346     }
1347
1348     if (!bdrv_is_first_non_filter(state->old_bs)) {
1349         error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1350         return;
1351     }
1352
1353     flags = state->old_bs->open_flags;
1354
1355     /* create new image w/backing file */
1356     if (mode != NEW_IMAGE_MODE_EXISTING) {
1357         bdrv_img_create(new_image_file, format,
1358                         state->old_bs->filename,
1359                         state->old_bs->drv->format_name,
1360                         NULL, -1, flags, &local_err, false);
1361         if (local_err) {
1362             error_propagate(errp, local_err);
1363             return;
1364         }
1365     }
1366
1367     if (has_snapshot_node_name) {
1368         options = qdict_new();
1369         qdict_put(options, "node-name",
1370                   qstring_from_str(snapshot_node_name));
1371     }
1372
1373     /* TODO Inherit bs->options or only take explicit options with an
1374      * extended QMP command? */
1375     assert(state->new_bs == NULL);
1376     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1377                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1378     /* We will manually add the backing_hd field to the bs later */
1379     if (ret != 0) {
1380         error_propagate(errp, local_err);
1381     }
1382 }
1383
1384 static void external_snapshot_commit(BlkTransactionState *common)
1385 {
1386     ExternalSnapshotState *state =
1387                              DO_UPCAST(ExternalSnapshotState, common, common);
1388
1389     /* This removes our old bs and adds the new bs */
1390     bdrv_append(state->new_bs, state->old_bs);
1391     /* We don't need (or want) to use the transactional
1392      * bdrv_reopen_multiple() across all the entries at once, because we
1393      * don't want to abort all of them if one of them fails the reopen */
1394     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1395                 NULL);
1396 }
1397
1398 static void external_snapshot_abort(BlkTransactionState *common)
1399 {
1400     ExternalSnapshotState *state =
1401                              DO_UPCAST(ExternalSnapshotState, common, common);
1402     if (state->new_bs) {
1403         bdrv_unref(state->new_bs);
1404     }
1405 }
1406
1407 typedef struct DriveBackupState {
1408     BlkTransactionState common;
1409     BlockDriverState *bs;
1410     BlockJob *job;
1411 } DriveBackupState;
1412
1413 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1414 {
1415     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1416     DriveBackup *backup;
1417     Error *local_err = NULL;
1418
1419     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1420     backup = common->action->drive_backup;
1421
1422     qmp_drive_backup(backup->device, backup->target,
1423                      backup->has_format, backup->format,
1424                      backup->sync,
1425                      backup->has_mode, backup->mode,
1426                      backup->has_speed, backup->speed,
1427                      backup->has_on_source_error, backup->on_source_error,
1428                      backup->has_on_target_error, backup->on_target_error,
1429                      &local_err);
1430     if (local_err) {
1431         error_propagate(errp, local_err);
1432         state->bs = NULL;
1433         state->job = NULL;
1434         return;
1435     }
1436
1437     state->bs = bdrv_find(backup->device);
1438     state->job = state->bs->job;
1439 }
1440
1441 static void drive_backup_abort(BlkTransactionState *common)
1442 {
1443     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1444     BlockDriverState *bs = state->bs;
1445
1446     /* Only cancel if it's the job we started */
1447     if (bs && bs->job && bs->job == state->job) {
1448         block_job_cancel_sync(bs->job);
1449     }
1450 }
1451
1452 static void abort_prepare(BlkTransactionState *common, Error **errp)
1453 {
1454     error_setg(errp, "Transaction aborted using Abort action");
1455 }
1456
1457 static void abort_commit(BlkTransactionState *common)
1458 {
1459     g_assert_not_reached(); /* this action never succeeds */
1460 }
1461
1462 static const BdrvActionOps actions[] = {
1463     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1464         .instance_size = sizeof(ExternalSnapshotState),
1465         .prepare  = external_snapshot_prepare,
1466         .commit   = external_snapshot_commit,
1467         .abort = external_snapshot_abort,
1468     },
1469     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1470         .instance_size = sizeof(DriveBackupState),
1471         .prepare = drive_backup_prepare,
1472         .abort = drive_backup_abort,
1473     },
1474     [TRANSACTION_ACTION_KIND_ABORT] = {
1475         .instance_size = sizeof(BlkTransactionState),
1476         .prepare = abort_prepare,
1477         .commit = abort_commit,
1478     },
1479     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1480         .instance_size = sizeof(InternalSnapshotState),
1481         .prepare  = internal_snapshot_prepare,
1482         .abort = internal_snapshot_abort,
1483     },
1484 };
1485
1486 /*
1487  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1488  *  then we do not pivot any of the devices in the group, and abandon the
1489  *  snapshots
1490  */
1491 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1492 {
1493     TransactionActionList *dev_entry = dev_list;
1494     BlkTransactionState *state, *next;
1495     Error *local_err = NULL;
1496
1497     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1498     QSIMPLEQ_INIT(&snap_bdrv_states);
1499
1500     /* drain all i/o before any snapshots */
1501     bdrv_drain_all();
1502
1503     /* We don't do anything in this loop that commits us to the snapshot */
1504     while (NULL != dev_entry) {
1505         TransactionAction *dev_info = NULL;
1506         const BdrvActionOps *ops;
1507
1508         dev_info = dev_entry->value;
1509         dev_entry = dev_entry->next;
1510
1511         assert(dev_info->kind < ARRAY_SIZE(actions));
1512
1513         ops = &actions[dev_info->kind];
1514         assert(ops->instance_size > 0);
1515
1516         state = g_malloc0(ops->instance_size);
1517         state->ops = ops;
1518         state->action = dev_info;
1519         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1520
1521         state->ops->prepare(state, &local_err);
1522         if (local_err) {
1523             error_propagate(errp, local_err);
1524             goto delete_and_fail;
1525         }
1526     }
1527
1528     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1529         if (state->ops->commit) {
1530             state->ops->commit(state);
1531         }
1532     }
1533
1534     /* success */
1535     goto exit;
1536
1537 delete_and_fail:
1538     /*
1539     * failure, and it is all-or-none; abandon each new bs, and keep using
1540     * the original bs for all images
1541     */
1542     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1543         if (state->ops->abort) {
1544             state->ops->abort(state);
1545         }
1546     }
1547 exit:
1548     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1549         if (state->ops->clean) {
1550             state->ops->clean(state);
1551         }
1552         g_free(state);
1553     }
1554 }
1555
1556
1557 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1558 {
1559     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1560         return;
1561     }
1562     if (!bdrv_dev_has_removable_media(bs)) {
1563         error_setg(errp, "Device '%s' is not removable",
1564                    bdrv_get_device_name(bs));
1565         return;
1566     }
1567
1568     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1569         bdrv_dev_eject_request(bs, force);
1570         if (!force) {
1571             error_setg(errp, "Device '%s' is locked",
1572                        bdrv_get_device_name(bs));
1573             return;
1574         }
1575     }
1576
1577     bdrv_close(bs);
1578 }
1579
1580 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1581 {
1582     BlockDriverState *bs;
1583
1584     bs = bdrv_find(device);
1585     if (!bs) {
1586         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1587         return;
1588     }
1589
1590     eject_device(bs, force, errp);
1591 }
1592
1593 void qmp_block_passwd(bool has_device, const char *device,
1594                       bool has_node_name, const char *node_name,
1595                       const char *password, Error **errp)
1596 {
1597     Error *local_err = NULL;
1598     BlockDriverState *bs;
1599     int err;
1600
1601     bs = bdrv_lookup_bs(has_device ? device : NULL,
1602                         has_node_name ? node_name : NULL,
1603                         &local_err);
1604     if (local_err) {
1605         error_propagate(errp, local_err);
1606         return;
1607     }
1608
1609     err = bdrv_set_key(bs, password);
1610     if (err == -EINVAL) {
1611         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1612         return;
1613     } else if (err < 0) {
1614         error_set(errp, QERR_INVALID_PASSWORD);
1615         return;
1616     }
1617 }
1618
1619 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1620                                     int bdrv_flags, BlockDriver *drv,
1621                                     const char *password, Error **errp)
1622 {
1623     Error *local_err = NULL;
1624     int ret;
1625
1626     ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1627     if (ret < 0) {
1628         error_propagate(errp, local_err);
1629         return;
1630     }
1631
1632     if (bdrv_key_required(bs)) {
1633         if (password) {
1634             if (bdrv_set_key(bs, password) < 0) {
1635                 error_set(errp, QERR_INVALID_PASSWORD);
1636             }
1637         } else {
1638             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1639                       bdrv_get_encrypted_filename(bs));
1640         }
1641     } else if (password) {
1642         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1643     }
1644 }
1645
1646 void qmp_change_blockdev(const char *device, const char *filename,
1647                          const char *format, Error **errp)
1648 {
1649     BlockDriverState *bs;
1650     BlockDriver *drv = NULL;
1651     int bdrv_flags;
1652     Error *err = NULL;
1653
1654     bs = bdrv_find(device);
1655     if (!bs) {
1656         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1657         return;
1658     }
1659
1660     if (format) {
1661         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1662         if (!drv) {
1663             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1664             return;
1665         }
1666     }
1667
1668     eject_device(bs, 0, &err);
1669     if (err) {
1670         error_propagate(errp, err);
1671         return;
1672     }
1673
1674     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1675     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1676
1677     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1678 }
1679
1680 /* throttling disk I/O limits */
1681 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1682                                int64_t bps_wr,
1683                                int64_t iops,
1684                                int64_t iops_rd,
1685                                int64_t iops_wr,
1686                                bool has_bps_max,
1687                                int64_t bps_max,
1688                                bool has_bps_rd_max,
1689                                int64_t bps_rd_max,
1690                                bool has_bps_wr_max,
1691                                int64_t bps_wr_max,
1692                                bool has_iops_max,
1693                                int64_t iops_max,
1694                                bool has_iops_rd_max,
1695                                int64_t iops_rd_max,
1696                                bool has_iops_wr_max,
1697                                int64_t iops_wr_max,
1698                                bool has_iops_size,
1699                                int64_t iops_size, Error **errp)
1700 {
1701     ThrottleConfig cfg;
1702     BlockDriverState *bs;
1703
1704     bs = bdrv_find(device);
1705     if (!bs) {
1706         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1707         return;
1708     }
1709
1710     memset(&cfg, 0, sizeof(cfg));
1711     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1712     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1713     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1714
1715     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1716     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1717     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1718
1719     if (has_bps_max) {
1720         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1721     }
1722     if (has_bps_rd_max) {
1723         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1724     }
1725     if (has_bps_wr_max) {
1726         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1727     }
1728     if (has_iops_max) {
1729         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1730     }
1731     if (has_iops_rd_max) {
1732         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1733     }
1734     if (has_iops_wr_max) {
1735         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1736     }
1737
1738     if (has_iops_size) {
1739         cfg.op_size = iops_size;
1740     }
1741
1742     if (!check_throttle_config(&cfg, errp)) {
1743         return;
1744     }
1745
1746     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1747         bdrv_io_limits_enable(bs);
1748     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1749         bdrv_io_limits_disable(bs);
1750     }
1751
1752     if (bs->io_limits_enabled) {
1753         bdrv_set_io_limits(bs, &cfg);
1754     }
1755 }
1756
1757 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1758 {
1759     const char *id = qdict_get_str(qdict, "id");
1760     BlockDriverState *bs;
1761     Error *local_err = NULL;
1762
1763     bs = bdrv_find(id);
1764     if (!bs) {
1765         error_report("Device '%s' not found", id);
1766         return -1;
1767     }
1768     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1769         error_report("%s", error_get_pretty(local_err));
1770         error_free(local_err);
1771         return -1;
1772     }
1773
1774     /* quiesce block driver; prevent further io */
1775     bdrv_drain_all();
1776     bdrv_flush(bs);
1777     bdrv_close(bs);
1778
1779     /* if we have a device attached to this BlockDriverState
1780      * then we need to make the drive anonymous until the device
1781      * can be removed.  If this is a drive with no device backing
1782      * then we can just get rid of the block driver state right here.
1783      */
1784     if (bdrv_get_attached_dev(bs)) {
1785         bdrv_make_anon(bs);
1786
1787         /* Further I/O must not pause the guest */
1788         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1789                           BLOCKDEV_ON_ERROR_REPORT);
1790     } else {
1791         drive_uninit(drive_get_by_blockdev(bs));
1792     }
1793
1794     return 0;
1795 }
1796
1797 void qmp_block_resize(bool has_device, const char *device,
1798                       bool has_node_name, const char *node_name,
1799                       int64_t size, Error **errp)
1800 {
1801     Error *local_err = NULL;
1802     BlockDriverState *bs;
1803     int ret;
1804
1805     bs = bdrv_lookup_bs(has_device ? device : NULL,
1806                         has_node_name ? node_name : NULL,
1807                         &local_err);
1808     if (local_err) {
1809         error_propagate(errp, local_err);
1810         return;
1811     }
1812
1813     if (!bdrv_is_first_non_filter(bs)) {
1814         error_set(errp, QERR_FEATURE_DISABLED, "resize");
1815         return;
1816     }
1817
1818     if (size < 0) {
1819         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1820         return;
1821     }
1822
1823     /* complete all in-flight operations before resizing the device */
1824     bdrv_drain_all();
1825
1826     ret = bdrv_truncate(bs, size);
1827     switch (ret) {
1828     case 0:
1829         break;
1830     case -ENOMEDIUM:
1831         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1832         break;
1833     case -ENOTSUP:
1834         error_set(errp, QERR_UNSUPPORTED);
1835         break;
1836     case -EACCES:
1837         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1838         break;
1839     case -EBUSY:
1840         error_set(errp, QERR_DEVICE_IN_USE, device);
1841         break;
1842     default:
1843         error_setg_errno(errp, -ret, "Could not resize");
1844         break;
1845     }
1846 }
1847
1848 static void block_job_cb(void *opaque, int ret)
1849 {
1850     BlockDriverState *bs = opaque;
1851     QObject *obj;
1852
1853     trace_block_job_cb(bs, bs->job, ret);
1854
1855     assert(bs->job);
1856     obj = qobject_from_block_job(bs->job);
1857     if (ret < 0) {
1858         QDict *dict = qobject_to_qdict(obj);
1859         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1860     }
1861
1862     if (block_job_is_cancelled(bs->job)) {
1863         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1864     } else {
1865         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1866     }
1867     qobject_decref(obj);
1868
1869     bdrv_put_ref_bh_schedule(bs);
1870 }
1871
1872 void qmp_block_stream(const char *device, bool has_base,
1873                       const char *base, bool has_speed, int64_t speed,
1874                       bool has_on_error, BlockdevOnError on_error,
1875                       Error **errp)
1876 {
1877     BlockDriverState *bs;
1878     BlockDriverState *base_bs = NULL;
1879     Error *local_err = NULL;
1880
1881     if (!has_on_error) {
1882         on_error = BLOCKDEV_ON_ERROR_REPORT;
1883     }
1884
1885     bs = bdrv_find(device);
1886     if (!bs) {
1887         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1888         return;
1889     }
1890
1891     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
1892         return;
1893     }
1894
1895     if (base) {
1896         base_bs = bdrv_find_backing_image(bs, base);
1897         if (base_bs == NULL) {
1898             error_set(errp, QERR_BASE_NOT_FOUND, base);
1899             return;
1900         }
1901     }
1902
1903     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1904                  on_error, block_job_cb, bs, &local_err);
1905     if (local_err) {
1906         error_propagate(errp, local_err);
1907         return;
1908     }
1909
1910     trace_qmp_block_stream(bs, bs->job);
1911 }
1912
1913 void qmp_block_commit(const char *device,
1914                       bool has_base, const char *base, const char *top,
1915                       bool has_speed, int64_t speed,
1916                       Error **errp)
1917 {
1918     BlockDriverState *bs;
1919     BlockDriverState *base_bs, *top_bs;
1920     Error *local_err = NULL;
1921     /* This will be part of the QMP command, if/when the
1922      * BlockdevOnError change for blkmirror makes it in
1923      */
1924     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1925
1926     if (!has_speed) {
1927         speed = 0;
1928     }
1929
1930     /* drain all i/o before commits */
1931     bdrv_drain_all();
1932
1933     bs = bdrv_find(device);
1934     if (!bs) {
1935         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1936         return;
1937     }
1938
1939     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, errp)) {
1940         return;
1941     }
1942
1943     /* default top_bs is the active layer */
1944     top_bs = bs;
1945
1946     if (top) {
1947         if (strcmp(bs->filename, top) != 0) {
1948             top_bs = bdrv_find_backing_image(bs, top);
1949         }
1950     }
1951
1952     if (top_bs == NULL) {
1953         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1954         return;
1955     }
1956
1957     if (has_base && base) {
1958         base_bs = bdrv_find_backing_image(top_bs, base);
1959     } else {
1960         base_bs = bdrv_find_base(top_bs);
1961     }
1962
1963     if (base_bs == NULL) {
1964         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1965         return;
1966     }
1967
1968     if (top_bs == bs) {
1969         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1970                             bs, &local_err);
1971     } else {
1972         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1973                     &local_err);
1974     }
1975     if (local_err != NULL) {
1976         error_propagate(errp, local_err);
1977         return;
1978     }
1979 }
1980
1981 void qmp_drive_backup(const char *device, const char *target,
1982                       bool has_format, const char *format,
1983                       enum MirrorSyncMode sync,
1984                       bool has_mode, enum NewImageMode mode,
1985                       bool has_speed, int64_t speed,
1986                       bool has_on_source_error, BlockdevOnError on_source_error,
1987                       bool has_on_target_error, BlockdevOnError on_target_error,
1988                       Error **errp)
1989 {
1990     BlockDriverState *bs;
1991     BlockDriverState *target_bs;
1992     BlockDriverState *source = NULL;
1993     BlockDriver *drv = NULL;
1994     Error *local_err = NULL;
1995     int flags;
1996     int64_t size;
1997     int ret;
1998
1999     if (!has_speed) {
2000         speed = 0;
2001     }
2002     if (!has_on_source_error) {
2003         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2004     }
2005     if (!has_on_target_error) {
2006         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2007     }
2008     if (!has_mode) {
2009         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2010     }
2011
2012     bs = bdrv_find(device);
2013     if (!bs) {
2014         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2015         return;
2016     }
2017
2018     if (!bdrv_is_inserted(bs)) {
2019         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2020         return;
2021     }
2022
2023     if (!has_format) {
2024         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2025     }
2026     if (format) {
2027         drv = bdrv_find_format(format);
2028         if (!drv) {
2029             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2030             return;
2031         }
2032     }
2033
2034     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2035         return;
2036     }
2037
2038     flags = bs->open_flags | BDRV_O_RDWR;
2039
2040     /* See if we have a backing HD we can use to create our new image
2041      * on top of. */
2042     if (sync == MIRROR_SYNC_MODE_TOP) {
2043         source = bs->backing_hd;
2044         if (!source) {
2045             sync = MIRROR_SYNC_MODE_FULL;
2046         }
2047     }
2048     if (sync == MIRROR_SYNC_MODE_NONE) {
2049         source = bs;
2050     }
2051
2052     size = bdrv_getlength(bs);
2053     if (size < 0) {
2054         error_setg_errno(errp, -size, "bdrv_getlength failed");
2055         return;
2056     }
2057
2058     if (mode != NEW_IMAGE_MODE_EXISTING) {
2059         assert(format && drv);
2060         if (source) {
2061             bdrv_img_create(target, format, source->filename,
2062                             source->drv->format_name, NULL,
2063                             size, flags, &local_err, false);
2064         } else {
2065             bdrv_img_create(target, format, NULL, NULL, NULL,
2066                             size, flags, &local_err, false);
2067         }
2068     }
2069
2070     if (local_err) {
2071         error_propagate(errp, local_err);
2072         return;
2073     }
2074
2075     target_bs = NULL;
2076     ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2077     if (ret < 0) {
2078         error_propagate(errp, local_err);
2079         return;
2080     }
2081
2082     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2083                  block_job_cb, bs, &local_err);
2084     if (local_err != NULL) {
2085         bdrv_unref(target_bs);
2086         error_propagate(errp, local_err);
2087         return;
2088     }
2089 }
2090
2091 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2092 {
2093     return bdrv_named_nodes_list();
2094 }
2095
2096 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
2097
2098 void qmp_drive_mirror(const char *device, const char *target,
2099                       bool has_format, const char *format,
2100                       enum MirrorSyncMode sync,
2101                       bool has_mode, enum NewImageMode mode,
2102                       bool has_speed, int64_t speed,
2103                       bool has_granularity, uint32_t granularity,
2104                       bool has_buf_size, int64_t buf_size,
2105                       bool has_on_source_error, BlockdevOnError on_source_error,
2106                       bool has_on_target_error, BlockdevOnError on_target_error,
2107                       Error **errp)
2108 {
2109     BlockDriverState *bs;
2110     BlockDriverState *source, *target_bs;
2111     BlockDriver *drv = NULL;
2112     Error *local_err = NULL;
2113     int flags;
2114     int64_t size;
2115     int ret;
2116
2117     if (!has_speed) {
2118         speed = 0;
2119     }
2120     if (!has_on_source_error) {
2121         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2122     }
2123     if (!has_on_target_error) {
2124         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2125     }
2126     if (!has_mode) {
2127         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2128     }
2129     if (!has_granularity) {
2130         granularity = 0;
2131     }
2132     if (!has_buf_size) {
2133         buf_size = DEFAULT_MIRROR_BUF_SIZE;
2134     }
2135
2136     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2137         error_set(errp, QERR_INVALID_PARAMETER, device);
2138         return;
2139     }
2140     if (granularity & (granularity - 1)) {
2141         error_set(errp, QERR_INVALID_PARAMETER, device);
2142         return;
2143     }
2144
2145     bs = bdrv_find(device);
2146     if (!bs) {
2147         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2148         return;
2149     }
2150
2151     if (!bdrv_is_inserted(bs)) {
2152         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2153         return;
2154     }
2155
2156     if (!has_format) {
2157         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2158     }
2159     if (format) {
2160         drv = bdrv_find_format(format);
2161         if (!drv) {
2162             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2163             return;
2164         }
2165     }
2166
2167     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2168         return;
2169     }
2170
2171     flags = bs->open_flags | BDRV_O_RDWR;
2172     source = bs->backing_hd;
2173     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2174         sync = MIRROR_SYNC_MODE_FULL;
2175     }
2176     if (sync == MIRROR_SYNC_MODE_NONE) {
2177         source = bs;
2178     }
2179
2180     size = bdrv_getlength(bs);
2181     if (size < 0) {
2182         error_setg_errno(errp, -size, "bdrv_getlength failed");
2183         return;
2184     }
2185
2186     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2187         && mode != NEW_IMAGE_MODE_EXISTING)
2188     {
2189         /* create new image w/o backing file */
2190         assert(format && drv);
2191         bdrv_img_create(target, format,
2192                         NULL, NULL, NULL, size, flags, &local_err, false);
2193     } else {
2194         switch (mode) {
2195         case NEW_IMAGE_MODE_EXISTING:
2196             break;
2197         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2198             /* create new image with backing file */
2199             bdrv_img_create(target, format,
2200                             source->filename,
2201                             source->drv->format_name,
2202                             NULL, size, flags, &local_err, false);
2203             break;
2204         default:
2205             abort();
2206         }
2207     }
2208
2209     if (local_err) {
2210         error_propagate(errp, local_err);
2211         return;
2212     }
2213
2214     /* Mirroring takes care of copy-on-write using the source's backing
2215      * file.
2216      */
2217     target_bs = NULL;
2218     ret = bdrv_open(&target_bs, target, NULL, NULL, flags | BDRV_O_NO_BACKING,
2219                     drv, &local_err);
2220     if (ret < 0) {
2221         error_propagate(errp, local_err);
2222         return;
2223     }
2224
2225     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2226                  on_source_error, on_target_error,
2227                  block_job_cb, bs, &local_err);
2228     if (local_err != NULL) {
2229         bdrv_unref(target_bs);
2230         error_propagate(errp, local_err);
2231         return;
2232     }
2233 }
2234
2235 static BlockJob *find_block_job(const char *device)
2236 {
2237     BlockDriverState *bs;
2238
2239     bs = bdrv_find(device);
2240     if (!bs || !bs->job) {
2241         return NULL;
2242     }
2243     return bs->job;
2244 }
2245
2246 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2247 {
2248     BlockJob *job = find_block_job(device);
2249
2250     if (!job) {
2251         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2252         return;
2253     }
2254
2255     block_job_set_speed(job, speed, errp);
2256 }
2257
2258 void qmp_block_job_cancel(const char *device,
2259                           bool has_force, bool force, Error **errp)
2260 {
2261     BlockJob *job = find_block_job(device);
2262
2263     if (!has_force) {
2264         force = false;
2265     }
2266
2267     if (!job) {
2268         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2269         return;
2270     }
2271     if (job->paused && !force) {
2272         error_setg(errp, "The block job for device '%s' is currently paused",
2273                    device);
2274         return;
2275     }
2276
2277     trace_qmp_block_job_cancel(job);
2278     block_job_cancel(job);
2279 }
2280
2281 void qmp_block_job_pause(const char *device, Error **errp)
2282 {
2283     BlockJob *job = find_block_job(device);
2284
2285     if (!job) {
2286         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2287         return;
2288     }
2289
2290     trace_qmp_block_job_pause(job);
2291     block_job_pause(job);
2292 }
2293
2294 void qmp_block_job_resume(const char *device, Error **errp)
2295 {
2296     BlockJob *job = find_block_job(device);
2297
2298     if (!job) {
2299         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2300         return;
2301     }
2302
2303     trace_qmp_block_job_resume(job);
2304     block_job_resume(job);
2305 }
2306
2307 void qmp_block_job_complete(const char *device, Error **errp)
2308 {
2309     BlockJob *job = find_block_job(device);
2310
2311     if (!job) {
2312         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2313         return;
2314     }
2315
2316     trace_qmp_block_job_complete(job);
2317     block_job_complete(job, errp);
2318 }
2319
2320 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2321 {
2322     QmpOutputVisitor *ov = qmp_output_visitor_new();
2323     DriveInfo *dinfo;
2324     QObject *obj;
2325     QDict *qdict;
2326     Error *local_err = NULL;
2327
2328     /* Require an ID in the top level */
2329     if (!options->has_id) {
2330         error_setg(errp, "Block device needs an ID");
2331         goto fail;
2332     }
2333
2334     /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2335      * cache.direct=false instead of silently switching to aio=threads, except
2336      * if called from drive_init.
2337      *
2338      * For now, simply forbidding the combination for all drivers will do. */
2339     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2340         bool direct = options->has_cache &&
2341                       options->cache->has_direct &&
2342                       options->cache->direct;
2343         if (!direct) {
2344             error_setg(errp, "aio=native requires cache.direct=true");
2345             goto fail;
2346         }
2347     }
2348
2349     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2350                                &options, NULL, &local_err);
2351     if (local_err) {
2352         error_propagate(errp, local_err);
2353         goto fail;
2354     }
2355
2356     obj = qmp_output_get_qobject(ov);
2357     qdict = qobject_to_qdict(obj);
2358
2359     qdict_flatten(qdict);
2360
2361     dinfo = blockdev_init(NULL, qdict, &local_err);
2362     if (local_err) {
2363         error_propagate(errp, local_err);
2364         goto fail;
2365     }
2366
2367     if (bdrv_key_required(dinfo->bdrv)) {
2368         drive_uninit(dinfo);
2369         error_setg(errp, "blockdev-add doesn't support encrypted devices");
2370         goto fail;
2371     }
2372
2373 fail:
2374     qmp_output_visitor_cleanup(ov);
2375 }
2376
2377 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2378 {
2379     BlockJobInfoList **prev = opaque;
2380     BlockJob *job = bs->job;
2381
2382     if (job) {
2383         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2384         elem->value = block_job_query(bs->job);
2385         (*prev)->next = elem;
2386         *prev = elem;
2387     }
2388 }
2389
2390 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2391 {
2392     /* Dummy is a fake list element for holding the head pointer */
2393     BlockJobInfoList dummy = {};
2394     BlockJobInfoList *prev = &dummy;
2395     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2396     return dummy.next;
2397 }
2398
2399 QemuOptsList qemu_common_drive_opts = {
2400     .name = "drive",
2401     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2402     .desc = {
2403         {
2404             .name = "snapshot",
2405             .type = QEMU_OPT_BOOL,
2406             .help = "enable/disable snapshot mode",
2407         },{
2408             .name = "discard",
2409             .type = QEMU_OPT_STRING,
2410             .help = "discard operation (ignore/off, unmap/on)",
2411         },{
2412             .name = "cache.writeback",
2413             .type = QEMU_OPT_BOOL,
2414             .help = "enables writeback mode for any caches",
2415         },{
2416             .name = "cache.direct",
2417             .type = QEMU_OPT_BOOL,
2418             .help = "enables use of O_DIRECT (bypass the host page cache)",
2419         },{
2420             .name = "cache.no-flush",
2421             .type = QEMU_OPT_BOOL,
2422             .help = "ignore any flush requests for the device",
2423         },{
2424             .name = "aio",
2425             .type = QEMU_OPT_STRING,
2426             .help = "host AIO implementation (threads, native)",
2427         },{
2428             .name = "format",
2429             .type = QEMU_OPT_STRING,
2430             .help = "disk format (raw, qcow2, ...)",
2431         },{
2432             .name = "serial",
2433             .type = QEMU_OPT_STRING,
2434             .help = "disk serial number",
2435         },{
2436             .name = "rerror",
2437             .type = QEMU_OPT_STRING,
2438             .help = "read error action",
2439         },{
2440             .name = "werror",
2441             .type = QEMU_OPT_STRING,
2442             .help = "write error action",
2443         },{
2444             .name = "read-only",
2445             .type = QEMU_OPT_BOOL,
2446             .help = "open drive file as read-only",
2447         },{
2448             .name = "throttling.iops-total",
2449             .type = QEMU_OPT_NUMBER,
2450             .help = "limit total I/O operations per second",
2451         },{
2452             .name = "throttling.iops-read",
2453             .type = QEMU_OPT_NUMBER,
2454             .help = "limit read operations per second",
2455         },{
2456             .name = "throttling.iops-write",
2457             .type = QEMU_OPT_NUMBER,
2458             .help = "limit write operations per second",
2459         },{
2460             .name = "throttling.bps-total",
2461             .type = QEMU_OPT_NUMBER,
2462             .help = "limit total bytes per second",
2463         },{
2464             .name = "throttling.bps-read",
2465             .type = QEMU_OPT_NUMBER,
2466             .help = "limit read bytes per second",
2467         },{
2468             .name = "throttling.bps-write",
2469             .type = QEMU_OPT_NUMBER,
2470             .help = "limit write bytes per second",
2471         },{
2472             .name = "throttling.iops-total-max",
2473             .type = QEMU_OPT_NUMBER,
2474             .help = "I/O operations burst",
2475         },{
2476             .name = "throttling.iops-read-max",
2477             .type = QEMU_OPT_NUMBER,
2478             .help = "I/O operations read burst",
2479         },{
2480             .name = "throttling.iops-write-max",
2481             .type = QEMU_OPT_NUMBER,
2482             .help = "I/O operations write burst",
2483         },{
2484             .name = "throttling.bps-total-max",
2485             .type = QEMU_OPT_NUMBER,
2486             .help = "total bytes burst",
2487         },{
2488             .name = "throttling.bps-read-max",
2489             .type = QEMU_OPT_NUMBER,
2490             .help = "total bytes read burst",
2491         },{
2492             .name = "throttling.bps-write-max",
2493             .type = QEMU_OPT_NUMBER,
2494             .help = "total bytes write burst",
2495         },{
2496             .name = "throttling.iops-size",
2497             .type = QEMU_OPT_NUMBER,
2498             .help = "when limiting by iops max size of an I/O in bytes",
2499         },{
2500             .name = "copy-on-read",
2501             .type = QEMU_OPT_BOOL,
2502             .help = "copy read data from backing file into image file",
2503         },{
2504             .name = "detect-zeroes",
2505             .type = QEMU_OPT_STRING,
2506             .help = "try to optimize zero writes (off, on, unmap)",
2507         },
2508         { /* end of list */ }
2509     },
2510 };
2511
2512 QemuOptsList qemu_drive_opts = {
2513     .name = "drive",
2514     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2515     .desc = {
2516         /*
2517          * no elements => accept any params
2518          * validation will happen later
2519          */
2520         { /* end of list */ }
2521     },
2522 };