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