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