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