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 "sysemu/sysemu.h"
  42 #include "block/block_int.h"
  43 #include "qmp-commands.h"
  44 #include "trace.h"
  45 #include "sysemu/arch_init.h"
  46
  47 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
  48 extern QemuOptsList qemu_common_drive_opts;
  49
  50 static const char *const if_name[IF_COUNT] = {
  51     [IF_NONE] = "none",
  52     [IF_IDE] = "ide",
  53     [IF_SCSI] = "scsi",
  54     [IF_FLOPPY] = "floppy",
  55     [IF_PFLASH] = "pflash",
  56     [IF_MTD] = "mtd",
  57     [IF_SD] = "sd",
  58     [IF_VIRTIO] = "virtio",
  59     [IF_XEN] = "xen",
  60 };
  61
  62 static const int if_max_devs[IF_COUNT] = {
  63     /*
  64      * Do not change these numbers!  They govern how drive option
  65      * index maps to unit and bus.  That mapping is ABI.
  66      *
  67      * All controllers used to imlement if=T drives need to support
  68      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
  69      * Otherwise, some index values map to "impossible" bus, unit
  70      * values.
  71      *
  72      * For instance, if you change [IF_SCSI] to 255, -drive
  73      * if=scsi,index=12 no longer means bus=1,unit=5, but
  74      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
  75      * the drive can't be set up.  Regression.
  76      */
  77     [IF_IDE] = 2,
  78     [IF_SCSI] = 7,
  79 };
  80
  81 /*
  82  * We automatically delete the drive when a device using it gets
  83  * unplugged.  Questionable feature, but we can't just drop it.
  84  * Device models call blockdev_mark_auto_del() to schedule the
  85  * automatic deletion, and generic qdev code calls blockdev_auto_del()
  86  * when deletion is actually safe.
  87  */
  88 void blockdev_mark_auto_del(BlockDriverState *bs)
  89 {
  90     DriveInfo *dinfo = drive_get_by_blockdev(bs);
  91
  92     if (bs->job) {
  93         block_job_cancel(bs->job);
  94     }
  95     if (dinfo) {
  96         dinfo->auto_del = 1;
  97     }
  98 }
  99
 100 void blockdev_auto_del(BlockDriverState *bs)
 101 {
 102     DriveInfo *dinfo = drive_get_by_blockdev(bs);
 103
 104     if (dinfo && dinfo->auto_del) {
 105         drive_put_ref(dinfo);
 106     }
 107 }
 108
 109 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
 110 {
 111     int max_devs = if_max_devs[type];
 112     return max_devs ? index / max_devs : 0;
 113 }
 114
 115 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
 116 {
 117     int max_devs = if_max_devs[type];
 118     return max_devs ? index % max_devs : index;
 119 }
 120
 121 QemuOpts *drive_def(const char *optstr)
 122 {
 123     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
 124 }
 125
 126 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
 127                     const char *optstr)
 128 {
 129     QemuOpts *opts;
 130     char buf[32];
 131
 132     opts = drive_def(optstr);
 133     if (!opts) {
 134         return NULL;
 135     }
 136     if (type != IF_DEFAULT) {
 137         qemu_opt_set(opts, "if", if_name[type]);
 138     }
 139     if (index >= 0) {
 140         snprintf(buf, sizeof(buf), "%d", index);
 141         qemu_opt_set(opts, "index", buf);
 142     }
 143     if (file)
 144         qemu_opt_set(opts, "file", file);
 145     return opts;
 146 }
 147
 148 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
 149 {
 150     DriveInfo *dinfo;
 151
 152     /* seek interface, bus and unit */
 153
 154     QTAILQ_FOREACH(dinfo, &drives, next) {
 155         if (dinfo->type == type &&
 156             dinfo->bus == bus &&
 157             dinfo->unit == unit)
 158             return dinfo;
 159     }
 160
 161     return NULL;
 162 }
 163
 164 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
 165 {
 166     return drive_get(type,
 167                      drive_index_to_bus_id(type, index),
 168                      drive_index_to_unit_id(type, index));
 169 }
 170
 171 int drive_get_max_bus(BlockInterfaceType type)
 172 {
 173     int max_bus;
 174     DriveInfo *dinfo;
 175
 176     max_bus = -1;
 177     QTAILQ_FOREACH(dinfo, &drives, next) {
 178         if(dinfo->type == type &&
 179            dinfo->bus > max_bus)
 180             max_bus = dinfo->bus;
 181     }
 182     return max_bus;
 183 }
 184
 185 /* Get a block device.  This should only be used for single-drive devices
 186    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
 187    appropriate bus.  */
 188 DriveInfo *drive_get_next(BlockInterfaceType type)
 189 {
 190     static int next_block_unit[IF_COUNT];
 191
 192     return drive_get(type, 0, next_block_unit[type]++);
 193 }
 194
 195 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
 196 {
 197     DriveInfo *dinfo;
 198
 199     QTAILQ_FOREACH(dinfo, &drives, next) {
 200         if (dinfo->bdrv == bs) {
 201             return dinfo;
 202         }
 203     }
 204     return NULL;
 205 }
 206
 207 static void bdrv_format_print(void *opaque, const char *name)
 208 {
 209     error_printf(" %s", name);
 210 }
 211
 212 static void drive_uninit(DriveInfo *dinfo)
 213 {
 214     qemu_opts_del(dinfo->opts);
 215     bdrv_delete(dinfo->bdrv);
 216     g_free(dinfo->id);
 217     QTAILQ_REMOVE(&drives, dinfo, next);
 218     g_free(dinfo->serial);
 219     g_free(dinfo);
 220 }
 221
 222 void drive_put_ref(DriveInfo *dinfo)
 223 {
 224     assert(dinfo->refcount);
 225     if (--dinfo->refcount == 0) {
 226         drive_uninit(dinfo);
 227     }
 228 }
 229
 230 void drive_get_ref(DriveInfo *dinfo)
 231 {
 232     dinfo->refcount++;
 233 }
 234
 235 typedef struct {
 236     QEMUBH *bh;
 237     DriveInfo *dinfo;
 238 } DrivePutRefBH;
 239
 240 static void drive_put_ref_bh(void *opaque)
 241 {
 242     DrivePutRefBH *s = opaque;
 243
 244     drive_put_ref(s->dinfo);
 245     qemu_bh_delete(s->bh);
 246     g_free(s);
 247 }
 248
 249 /*
 250  * Release a drive reference in a BH
 251  *
 252  * It is not possible to use drive_put_ref() from a callback function when the
 253  * callers still need the drive.  In such cases we schedule a BH to release the
 254  * reference.
 255  */
 256 static void drive_put_ref_bh_schedule(DriveInfo *dinfo)
 257 {
 258     DrivePutRefBH *s;
 259
 260     s = g_new(DrivePutRefBH, 1);
 261     s->bh = qemu_bh_new(drive_put_ref_bh, s);
 262     s->dinfo = dinfo;
 263     qemu_bh_schedule(s->bh);
 264 }
 265
 266 static int parse_block_error_action(const char *buf, bool is_read)
 267 {
 268     if (!strcmp(buf, "ignore")) {
 269         return BLOCKDEV_ON_ERROR_IGNORE;
 270     } else if (!is_read && !strcmp(buf, "enospc")) {
 271         return BLOCKDEV_ON_ERROR_ENOSPC;
 272     } else if (!strcmp(buf, "stop")) {
 273         return BLOCKDEV_ON_ERROR_STOP;
 274     } else if (!strcmp(buf, "report")) {
 275         return BLOCKDEV_ON_ERROR_REPORT;
 276     } else {
 277         error_report("'%s' invalid %s error action",
 278                      buf, is_read ? "read" : "write");
 279         return -1;
 280     }
 281 }
 282
 283 static bool do_check_io_limits(BlockIOLimit *io_limits, Error **errp)
 284 {
 285     bool bps_flag;
 286     bool iops_flag;
 287
 288     assert(io_limits);
 289
 290     bps_flag  = (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] != 0)
 291                  && ((io_limits->bps[BLOCK_IO_LIMIT_READ] != 0)
 292                  || (io_limits->bps[BLOCK_IO_LIMIT_WRITE] != 0));
 293     iops_flag = (io_limits->iops[BLOCK_IO_LIMIT_TOTAL] != 0)
 294                  && ((io_limits->iops[BLOCK_IO_LIMIT_READ] != 0)
 295                  || (io_limits->iops[BLOCK_IO_LIMIT_WRITE] != 0));
 296     if (bps_flag || iops_flag) {
 297         error_setg(errp, "bps(iops) and bps_rd/bps_wr(iops_rd/iops_wr) "
 298                          "cannot be used at the same time");
 299         return false;
 300     }
 301
 302     if (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] < 0 ||
 303         io_limits->bps[BLOCK_IO_LIMIT_WRITE] < 0 ||
 304         io_limits->bps[BLOCK_IO_LIMIT_READ] < 0 ||
 305         io_limits->iops[BLOCK_IO_LIMIT_TOTAL] < 0 ||
 306         io_limits->iops[BLOCK_IO_LIMIT_WRITE] < 0 ||
 307         io_limits->iops[BLOCK_IO_LIMIT_READ] < 0) {
 308         error_setg(errp, "bps and iops values must be 0 or greater");
 309         return false;
 310     }
 311
 312     return true;
 313 }
 314
 315 static DriveInfo *blockdev_init(QemuOpts *all_opts,
 316                                 BlockInterfaceType block_default_type)
 317 {
 318     const char *buf;
 319     const char *file = NULL;
 320     const char *serial;
 321     const char *mediastr = "";
 322     BlockInterfaceType type;
 323     enum { MEDIA_DISK, MEDIA_CDROM } media;
 324     int bus_id, unit_id;
 325     int cyls, heads, secs, translation;
 326     int max_devs;
 327     int index;
 328     int ro = 0;
 329     int bdrv_flags = 0;
 330     int on_read_error, on_write_error;
 331     const char *devaddr;
 332     DriveInfo *dinfo;
 333     BlockIOLimit io_limits;
 334     int snapshot = 0;
 335     bool copy_on_read;
 336     int ret;
 337     Error *error = NULL;
 338     QemuOpts *opts;
 339     QDict *bs_opts;
 340     const char *id;
 341     bool has_driver_specific_opts;
 342
 343     translation = BIOS_ATA_TRANSLATION_AUTO;
 344     media = MEDIA_DISK;
 345
 346     /* Check common options by copying from all_opts to opts, all other options
 347      * are stored in bs_opts. */
 348     id = qemu_opts_id(all_opts);
 349     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
 350     if (error_is_set(&error)) {
 351         qerror_report_err(error);
 352         error_free(error);
 353         return NULL;
 354     }
 355
 356     bs_opts = qdict_new();
 357     qemu_opts_to_qdict(all_opts, bs_opts);
 358     qemu_opts_absorb_qdict(opts, bs_opts, &error);
 359     if (error_is_set(&error)) {
 360         qerror_report_err(error);
 361         error_free(error);
 362         return NULL;
 363     }
 364
 365     if (id) {
 366         qdict_del(bs_opts, "id");
 367     }
 368
 369     has_driver_specific_opts = !!qdict_size(bs_opts);
 370
 371     /* extract parameters */
 372     bus_id  = qemu_opt_get_number(opts, "bus", 0);
 373     unit_id = qemu_opt_get_number(opts, "unit", -1);
 374     index   = qemu_opt_get_number(opts, "index", -1);
 375
 376     cyls  = qemu_opt_get_number(opts, "cyls", 0);
 377     heads = qemu_opt_get_number(opts, "heads", 0);
 378     secs  = qemu_opt_get_number(opts, "secs", 0);
 379
 380     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 381     ro = qemu_opt_get_bool(opts, "read-only", 0);
 382     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
 383
 384     file = qemu_opt_get(opts, "file");
 385     serial = qemu_opt_get(opts, "serial");
 386
 387     if ((buf = qemu_opt_get(opts, "if")) != NULL) {
 388         for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
 389             ;
 390         if (type == IF_COUNT) {
 391             error_report("unsupported bus type '%s'", buf);
 392             return NULL;
 393         }
 394     } else {
 395         type = block_default_type;
 396     }
 397
 398     max_devs = if_max_devs[type];
 399
 400     if (cyls || heads || secs) {
 401         if (cyls < 1) {
 402             error_report("invalid physical cyls number");
 403             return NULL;
 404         }
 405         if (heads < 1) {
 406             error_report("invalid physical heads number");
 407             return NULL;
 408         }
 409         if (secs < 1) {
 410             error_report("invalid physical secs number");
 411             return NULL;
 412         }
 413     }
 414
 415     if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
 416         if (!cyls) {
 417             error_report("'%s' trans must be used with cyls, heads and secs",
 418                          buf);
 419             return NULL;
 420         }
 421         if (!strcmp(buf, "none"))
 422             translation = BIOS_ATA_TRANSLATION_NONE;
 423         else if (!strcmp(buf, "lba"))
 424             translation = BIOS_ATA_TRANSLATION_LBA;
 425         else if (!strcmp(buf, "auto"))
 426             translation = BIOS_ATA_TRANSLATION_AUTO;
 427         else {
 428             error_report("'%s' invalid translation type", buf);
 429             return NULL;
 430         }
 431     }
 432
 433     if ((buf = qemu_opt_get(opts, "media")) != NULL) {
 434         if (!strcmp(buf, "disk")) {
 435             media = MEDIA_DISK;
 436         } else if (!strcmp(buf, "cdrom")) {
 437             if (cyls || secs || heads) {
 438                 error_report("CHS can't be set with media=%s", buf);
 439                 return NULL;
 440             }
 441             media = MEDIA_CDROM;
 442         } else {
 443             error_report("'%s' invalid media", buf);
 444             return NULL;
 445         }
 446     }
 447
 448     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
 449         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
 450             error_report("invalid discard option");
 451             return NULL;
 452         }
 453     }
 454
 455     bdrv_flags = 0;
 456     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
 457         bdrv_flags |= BDRV_O_CACHE_WB;
 458     }
 459     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
 460         bdrv_flags |= BDRV_O_NOCACHE;
 461     }
 462     if (qemu_opt_get_bool(opts, "cache.no-flush", true)) {
 463         bdrv_flags |= BDRV_O_NO_FLUSH;
 464     }
 465
 466 #ifdef CONFIG_LINUX_AIO
 467     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 468         if (!strcmp(buf, "native")) {
 469             bdrv_flags |= BDRV_O_NATIVE_AIO;
 470         } else if (!strcmp(buf, "threads")) {
 471             /* this is the default */
 472         } else {
 473            error_report("invalid aio option");
 474            return NULL;
 475         }
 476     }
 477 #endif
 478
 479     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 480         if (is_help_option(buf)) {
 481             error_printf("Supported formats:");
 482             bdrv_iterate_format(bdrv_format_print, NULL);
 483             error_printf("\n");
 484             return NULL;
 485         }
 486
 487         qdict_put(bs_opts, "driver", qstring_from_str(buf));
 488     }
 489
 490     /* disk I/O throttling */
 491     io_limits.bps[BLOCK_IO_LIMIT_TOTAL]  =
 492         qemu_opt_get_number(opts, "throttling.bps-total", 0);
 493     io_limits.bps[BLOCK_IO_LIMIT_READ]   =
 494         qemu_opt_get_number(opts, "throttling.bps-read", 0);
 495     io_limits.bps[BLOCK_IO_LIMIT_WRITE]  =
 496         qemu_opt_get_number(opts, "throttling.bps-write", 0);
 497     io_limits.iops[BLOCK_IO_LIMIT_TOTAL] =
 498         qemu_opt_get_number(opts, "throttling.iops-total", 0);
 499     io_limits.iops[BLOCK_IO_LIMIT_READ]  =
 500         qemu_opt_get_number(opts, "throttling.iops-read", 0);
 501     io_limits.iops[BLOCK_IO_LIMIT_WRITE] =
 502         qemu_opt_get_number(opts, "throttling.iops-write", 0);
 503
 504     if (!do_check_io_limits(&io_limits, &error)) {
 505         error_report("%s", error_get_pretty(error));
 506         error_free(error);
 507         return NULL;
 508     }
 509
 510     if (qemu_opt_get(opts, "boot") != NULL) {
 511         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 512                 "ignored. Future versions will reject this parameter. Please "
 513                 "update your scripts.\n");
 514     }
 515
 516     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
 517     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 518         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
 519             error_report("werror is not supported by this bus type");
 520             return NULL;
 521         }
 522
 523         on_write_error = parse_block_error_action(buf, 0);
 524         if (on_write_error < 0) {
 525             return NULL;
 526         }
 527     }
 528
 529     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
 530     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 531         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
 532             error_report("rerror is not supported by this bus type");
 533             return NULL;
 534         }
 535
 536         on_read_error = parse_block_error_action(buf, 1);
 537         if (on_read_error < 0) {
 538             return NULL;
 539         }
 540     }
 541
 542     if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
 543         if (type != IF_VIRTIO) {
 544             error_report("addr is not supported by this bus type");
 545             return NULL;
 546         }
 547     }
 548
 549     /* compute bus and unit according index */
 550
 551     if (index != -1) {
 552         if (bus_id != 0 || unit_id != -1) {
 553             error_report("index cannot be used with bus and unit");
 554             return NULL;
 555         }
 556         bus_id = drive_index_to_bus_id(type, index);
 557         unit_id = drive_index_to_unit_id(type, index);
 558     }
 559
 560     /* if user doesn't specify a unit_id,
 561      * try to find the first free
 562      */
 563
 564     if (unit_id == -1) {
 565        unit_id = 0;
 566        while (drive_get(type, bus_id, unit_id) != NULL) {
 567            unit_id++;
 568            if (max_devs && unit_id >= max_devs) {
 569                unit_id -= max_devs;
 570                bus_id++;
 571            }
 572        }
 573     }
 574
 575     /* check unit id */
 576
 577     if (max_devs && unit_id >= max_devs) {
 578         error_report("unit %d too big (max is %d)",
 579                      unit_id, max_devs - 1);
 580         return NULL;
 581     }
 582
 583     /*
 584      * catch multiple definitions
 585      */
 586
 587     if (drive_get(type, bus_id, unit_id) != NULL) {
 588         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 589                      bus_id, unit_id, index);
 590         return NULL;
 591     }
 592
 593     /* init */
 594
 595     dinfo = g_malloc0(sizeof(*dinfo));
 596     if ((buf = qemu_opts_id(opts)) != NULL) {
 597         dinfo->id = g_strdup(buf);
 598     } else {
 599         /* no id supplied -> create one */
 600         dinfo->id = g_malloc0(32);
 601         if (type == IF_IDE || type == IF_SCSI)
 602             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 603         if (max_devs)
 604             snprintf(dinfo->id, 32, "%s%i%s%i",
 605                      if_name[type], bus_id, mediastr, unit_id);
 606         else
 607             snprintf(dinfo->id, 32, "%s%s%i",
 608                      if_name[type], mediastr, unit_id);
 609     }
 610     dinfo->bdrv = bdrv_new(dinfo->id);
 611     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
 612     dinfo->bdrv->read_only = ro;
 613     dinfo->devaddr = devaddr;
 614     dinfo->type = type;
 615     dinfo->bus = bus_id;
 616     dinfo->unit = unit_id;
 617     dinfo->cyls = cyls;
 618     dinfo->heads = heads;
 619     dinfo->secs = secs;
 620     dinfo->trans = translation;
 621     dinfo->opts = all_opts;
 622     dinfo->refcount = 1;
 623     if (serial != NULL) {
 624         dinfo->serial = g_strdup(serial);
 625     }
 626     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 627
 628     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
 629
 630     /* disk I/O throttling */
 631     bdrv_set_io_limits(dinfo->bdrv, &io_limits);
 632
 633     switch(type) {
 634     case IF_IDE:
 635     case IF_SCSI:
 636     case IF_XEN:
 637     case IF_NONE:
 638         dinfo->media_cd = media == MEDIA_CDROM;
 639         break;
 640     case IF_SD:
 641     case IF_FLOPPY:
 642     case IF_PFLASH:
 643     case IF_MTD:
 644         break;
 645     case IF_VIRTIO:
 646     {
 647         /* add virtio block device */
 648         QemuOpts *devopts;
 649         devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
 650         if (arch_type == QEMU_ARCH_S390X) {
 651             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
 652         } else {
 653             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
 654         }
 655         qemu_opt_set(devopts, "drive", dinfo->id);
 656         if (devaddr)
 657             qemu_opt_set(devopts, "addr", devaddr);
 658         break;
 659     }
 660     default:
 661         abort();
 662     }
 663     if (!file || !*file) {
 664         if (has_driver_specific_opts) {
 665             file = NULL;
 666         } else {
 667             return dinfo;
 668         }
 669     }
 670     if (snapshot) {
 671         /* always use cache=unsafe with snapshot */
 672         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 673         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 674     }
 675
 676     if (copy_on_read) {
 677         bdrv_flags |= BDRV_O_COPY_ON_READ;
 678     }
 679
 680     if (runstate_check(RUN_STATE_INMIGRATE)) {
 681         bdrv_flags |= BDRV_O_INCOMING;
 682     }
 683
 684     if (media == MEDIA_CDROM) {
 685         /* CDROM is fine for any interface, don't check.  */
 686         ro = 1;
 687     } else if (ro == 1) {
 688         if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
 689             type != IF_NONE && type != IF_PFLASH) {
 690             error_report("read-only not supported by this bus type");
 691             goto err;
 692         }
 693     }
 694
 695     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 696
 697     if (ro && copy_on_read) {
 698         error_report("warning: disabling copy_on_read on read-only drive");
 699     }
 700
 701     QINCREF(bs_opts);
 702     ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, NULL);
 703
 704     if (ret < 0) {
 705         if (ret == -EMEDIUMTYPE) {
 706             error_report("could not open disk image %s: not in %s format",
 707                          file ?: dinfo->id, qdict_get_str(bs_opts, "driver"));
 708         } else {
 709             error_report("could not open disk image %s: %s",
 710                          file ?: dinfo->id, strerror(-ret));
 711         }
 712         goto err;
 713     }
 714
 715     if (bdrv_key_required(dinfo->bdrv))
 716         autostart = 0;
 717
 718     QDECREF(bs_opts);
 719     qemu_opts_del(opts);
 720
 721     return dinfo;
 722
 723 err:
 724     qemu_opts_del(opts);
 725     QDECREF(bs_opts);
 726     bdrv_delete(dinfo->bdrv);
 727     g_free(dinfo->id);
 728     QTAILQ_REMOVE(&drives, dinfo, next);
 729     g_free(dinfo);
 730     return NULL;
 731 }
 732
 733 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
 734 {
 735     const char *value;
 736
 737     value = qemu_opt_get(opts, from);
 738     if (value) {
 739         qemu_opt_set(opts, to, value);
 740         qemu_opt_unset(opts, from);
 741     }
 742 }
 743
 744 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
 745 {
 746     const char *value;
 747
 748     /* Change legacy command line options into QMP ones */
 749     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
 750     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
 751     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
 752
 753     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
 754     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
 755     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
 756
 757     qemu_opt_rename(all_opts, "readonly", "read-only");
 758
 759     value = qemu_opt_get(all_opts, "cache");
 760     if (value) {
 761         int flags = 0;
 762
 763         if (bdrv_parse_cache_flags(value, &flags) != 0) {
 764             error_report("invalid cache option");
 765             return NULL;
 766         }
 767
 768         /* Specific options take precedence */
 769         if (!qemu_opt_get(all_opts, "cache.writeback")) {
 770             qemu_opt_set_bool(all_opts, "cache.writeback",
 771                               !!(flags & BDRV_O_CACHE_WB));
 772         }
 773         if (!qemu_opt_get(all_opts, "cache.direct")) {
 774             qemu_opt_set_bool(all_opts, "cache.direct",
 775                               !!(flags & BDRV_O_NOCACHE));
 776         }
 777         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
 778             qemu_opt_set_bool(all_opts, "cache.no-flush",
 779                               !!(flags & BDRV_O_NO_FLUSH));
 780         }
 781         qemu_opt_unset(all_opts, "cache");
 782     }
 783
 784     return blockdev_init(all_opts, block_default_type);
 785 }
 786
 787 void do_commit(Monitor *mon, const QDict *qdict)
 788 {
 789     const char *device = qdict_get_str(qdict, "device");
 790     BlockDriverState *bs;
 791     int ret;
 792
 793     if (!strcmp(device, "all")) {
 794         ret = bdrv_commit_all();
 795     } else {
 796         bs = bdrv_find(device);
 797         if (!bs) {
 798             monitor_printf(mon, "Device '%s' not found\n", device);
 799             return;
 800         }
 801         ret = bdrv_commit(bs);
 802     }
 803     if (ret < 0) {
 804         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
 805                        strerror(-ret));
 806     }
 807 }
 808
 809 static void blockdev_do_action(int kind, void *data, Error **errp)
 810 {
 811     TransactionAction action;
 812     TransactionActionList list;
 813
 814     action.kind = kind;
 815     action.data = data;
 816     list.value = &action;
 817     list.next = NULL;
 818     qmp_transaction(&list, errp);
 819 }
 820
 821 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
 822                                 bool has_format, const char *format,
 823                                 bool has_mode, enum NewImageMode mode,
 824                                 Error **errp)
 825 {
 826     BlockdevSnapshot snapshot = {
 827         .device = (char *) device,
 828         .snapshot_file = (char *) snapshot_file,
 829         .has_format = has_format,
 830         .format = (char *) format,
 831         .has_mode = has_mode,
 832         .mode = mode,
 833     };
 834     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
 835                        &snapshot, errp);
 836 }
 837
 838
 839 /* New and old BlockDriverState structs for group snapshots */
 840
 841 typedef struct BlkTransactionState BlkTransactionState;
 842
 843 /* Only prepare() may fail. In a single transaction, only one of commit() or
 844    abort() will be called, clean() will always be called if it present. */
 845 typedef struct BdrvActionOps {
 846     /* Size of state struct, in bytes. */
 847     size_t instance_size;
 848     /* Prepare the work, must NOT be NULL. */
 849     void (*prepare)(BlkTransactionState *common, Error **errp);
 850     /* Commit the changes, can be NULL. */
 851     void (*commit)(BlkTransactionState *common);
 852     /* Abort the changes on fail, can be NULL. */
 853     void (*abort)(BlkTransactionState *common);
 854     /* Clean up resource in the end, can be NULL. */
 855     void (*clean)(BlkTransactionState *common);
 856 } BdrvActionOps;
 857
 858 /*
 859  * This structure must be arranged as first member in child type, assuming
 860  * that compiler will also arrange it to the same address with parent instance.
 861  * Later it will be used in free().
 862  */
 863 struct BlkTransactionState {
 864     TransactionAction *action;
 865     const BdrvActionOps *ops;
 866     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
 867 };
 868
 869 /* external snapshot private data */
 870 typedef struct ExternalSnapshotState {
 871     BlkTransactionState common;
 872     BlockDriverState *old_bs;
 873     BlockDriverState *new_bs;
 874 } ExternalSnapshotState;
 875
 876 static void external_snapshot_prepare(BlkTransactionState *common,
 877                                       Error **errp)
 878 {
 879     BlockDriver *drv;
 880     int flags, ret;
 881     Error *local_err = NULL;
 882     const char *device;
 883     const char *new_image_file;
 884     const char *format = "qcow2";
 885     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
 886     ExternalSnapshotState *state =
 887                              DO_UPCAST(ExternalSnapshotState, common, common);
 888     TransactionAction *action = common->action;
 889
 890     /* get parameters */
 891     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
 892
 893     device = action->blockdev_snapshot_sync->device;
 894     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
 895     if (action->blockdev_snapshot_sync->has_format) {
 896         format = action->blockdev_snapshot_sync->format;
 897     }
 898     if (action->blockdev_snapshot_sync->has_mode) {
 899         mode = action->blockdev_snapshot_sync->mode;
 900     }
 901
 902     /* start processing */
 903     drv = bdrv_find_format(format);
 904     if (!drv) {
 905         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 906         return;
 907     }
 908
 909     state->old_bs = bdrv_find(device);
 910     if (!state->old_bs) {
 911         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 912         return;
 913     }
 914
 915     if (!bdrv_is_inserted(state->old_bs)) {
 916         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
 917         return;
 918     }
 919
 920     if (bdrv_in_use(state->old_bs)) {
 921         error_set(errp, QERR_DEVICE_IN_USE, device);
 922         return;
 923     }
 924
 925     if (!bdrv_is_read_only(state->old_bs)) {
 926         if (bdrv_flush(state->old_bs)) {
 927             error_set(errp, QERR_IO_ERROR);
 928             return;
 929         }
 930     }
 931
 932     flags = state->old_bs->open_flags;
 933
 934     /* create new image w/backing file */
 935     if (mode != NEW_IMAGE_MODE_EXISTING) {
 936         bdrv_img_create(new_image_file, format,
 937                         state->old_bs->filename,
 938                         state->old_bs->drv->format_name,
 939                         NULL, -1, flags, &local_err, false);
 940         if (error_is_set(&local_err)) {
 941             error_propagate(errp, local_err);
 942             return;
 943         }
 944     }
 945
 946     /* We will manually add the backing_hd field to the bs later */
 947     state->new_bs = bdrv_new("");
 948     /* TODO Inherit bs->options or only take explicit options with an
 949      * extended QMP command? */
 950     ret = bdrv_open(state->new_bs, new_image_file, NULL,
 951                     flags | BDRV_O_NO_BACKING, drv);
 952     if (ret != 0) {
 953         error_setg_file_open(errp, -ret, new_image_file);
 954     }
 955 }
 956
 957 static void external_snapshot_commit(BlkTransactionState *common)
 958 {
 959     ExternalSnapshotState *state =
 960                              DO_UPCAST(ExternalSnapshotState, common, common);
 961
 962     /* This removes our old bs and adds the new bs */
 963     bdrv_append(state->new_bs, state->old_bs);
 964     /* We don't need (or want) to use the transactional
 965      * bdrv_reopen_multiple() across all the entries at once, because we
 966      * don't want to abort all of them if one of them fails the reopen */
 967     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
 968                 NULL);
 969 }
 970
 971 static void external_snapshot_abort(BlkTransactionState *common)
 972 {
 973     ExternalSnapshotState *state =
 974                              DO_UPCAST(ExternalSnapshotState, common, common);
 975     if (state->new_bs) {
 976         bdrv_delete(state->new_bs);
 977     }
 978 }
 979
 980 typedef struct DriveBackupState {
 981     BlkTransactionState common;
 982     BlockDriverState *bs;
 983     BlockJob *job;
 984 } DriveBackupState;
 985
 986 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
 987 {
 988     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
 989     DriveBackup *backup;
 990     Error *local_err = NULL;
 991
 992     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
 993     backup = common->action->drive_backup;
 994
 995     qmp_drive_backup(backup->device, backup->target,
 996                      backup->has_format, backup->format,
 997                      backup->sync,
 998                      backup->has_mode, backup->mode,
 999                      backup->has_speed, backup->speed,
1000                      backup->has_on_source_error, backup->on_source_error,
1001                      backup->has_on_target_error, backup->on_target_error,
1002                      &local_err);
1003     if (error_is_set(&local_err)) {
1004         error_propagate(errp, local_err);
1005         state->bs = NULL;
1006         state->job = NULL;
1007         return;
1008     }
1009
1010     state->bs = bdrv_find(backup->device);
1011     state->job = state->bs->job;
1012 }
1013
1014 static void drive_backup_abort(BlkTransactionState *common)
1015 {
1016     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1017     BlockDriverState *bs = state->bs;
1018
1019     /* Only cancel if it's the job we started */
1020     if (bs && bs->job && bs->job == state->job) {
1021         block_job_cancel_sync(bs->job);
1022     }
1023 }
1024
1025 static void abort_prepare(BlkTransactionState *common, Error **errp)
1026 {
1027     error_setg(errp, "Transaction aborted using Abort action");
1028 }
1029
1030 static void abort_commit(BlkTransactionState *common)
1031 {
1032     assert(false); /* this action never succeeds */
1033 }
1034
1035 static const BdrvActionOps actions[] = {
1036     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1037         .instance_size = sizeof(ExternalSnapshotState),
1038         .prepare  = external_snapshot_prepare,
1039         .commit   = external_snapshot_commit,
1040         .abort = external_snapshot_abort,
1041     },
1042     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1043         .instance_size = sizeof(DriveBackupState),
1044         .prepare = drive_backup_prepare,
1045         .abort = drive_backup_abort,
1046     },
1047     [TRANSACTION_ACTION_KIND_ABORT] = {
1048         .instance_size = sizeof(BlkTransactionState),
1049         .prepare = abort_prepare,
1050         .commit = abort_commit,
1051     },
1052 };
1053
1054 /*
1055  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1056  *  then we do not pivot any of the devices in the group, and abandon the
1057  *  snapshots
1058  */
1059 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1060 {
1061     TransactionActionList *dev_entry = dev_list;
1062     BlkTransactionState *state, *next;
1063     Error *local_err = NULL;
1064
1065     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1066     QSIMPLEQ_INIT(&snap_bdrv_states);
1067
1068     /* drain all i/o before any snapshots */
1069     bdrv_drain_all();
1070
1071     /* We don't do anything in this loop that commits us to the snapshot */
1072     while (NULL != dev_entry) {
1073         TransactionAction *dev_info = NULL;
1074         const BdrvActionOps *ops;
1075
1076         dev_info = dev_entry->value;
1077         dev_entry = dev_entry->next;
1078
1079         assert(dev_info->kind < ARRAY_SIZE(actions));
1080
1081         ops = &actions[dev_info->kind];
1082         state = g_malloc0(ops->instance_size);
1083         state->ops = ops;
1084         state->action = dev_info;
1085         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1086
1087         state->ops->prepare(state, &local_err);
1088         if (error_is_set(&local_err)) {
1089             error_propagate(errp, local_err);
1090             goto delete_and_fail;
1091         }
1092     }
1093
1094     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1095         if (state->ops->commit) {
1096             state->ops->commit(state);
1097         }
1098     }
1099
1100     /* success */
1101     goto exit;
1102
1103 delete_and_fail:
1104     /*
1105     * failure, and it is all-or-none; abandon each new bs, and keep using
1106     * the original bs for all images
1107     */
1108     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1109         if (state->ops->abort) {
1110             state->ops->abort(state);
1111         }
1112     }
1113 exit:
1114     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1115         if (state->ops->clean) {
1116             state->ops->clean(state);
1117         }
1118         g_free(state);
1119     }
1120 }
1121
1122
1123 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1124 {
1125     if (bdrv_in_use(bs)) {
1126         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1127         return;
1128     }
1129     if (!bdrv_dev_has_removable_media(bs)) {
1130         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1131         return;
1132     }
1133
1134     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1135         bdrv_dev_eject_request(bs, force);
1136         if (!force) {
1137             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1138             return;
1139         }
1140     }
1141
1142     bdrv_close(bs);
1143 }
1144
1145 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1146 {
1147     BlockDriverState *bs;
1148
1149     bs = bdrv_find(device);
1150     if (!bs) {
1151         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1152         return;
1153     }
1154
1155     eject_device(bs, force, errp);
1156 }
1157
1158 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1159 {
1160     BlockDriverState *bs;
1161     int err;
1162
1163     bs = bdrv_find(device);
1164     if (!bs) {
1165         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1166         return;
1167     }
1168
1169     err = bdrv_set_key(bs, password);
1170     if (err == -EINVAL) {
1171         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1172         return;
1173     } else if (err < 0) {
1174         error_set(errp, QERR_INVALID_PASSWORD);
1175         return;
1176     }
1177 }
1178
1179 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1180                                     int bdrv_flags, BlockDriver *drv,
1181                                     const char *password, Error **errp)
1182 {
1183     int ret;
1184
1185     ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv);
1186     if (ret < 0) {
1187         error_setg_file_open(errp, -ret, filename);
1188         return;
1189     }
1190
1191     if (bdrv_key_required(bs)) {
1192         if (password) {
1193             if (bdrv_set_key(bs, password) < 0) {
1194                 error_set(errp, QERR_INVALID_PASSWORD);
1195             }
1196         } else {
1197             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1198                       bdrv_get_encrypted_filename(bs));
1199         }
1200     } else if (password) {
1201         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1202     }
1203 }
1204
1205 void qmp_change_blockdev(const char *device, const char *filename,
1206                          bool has_format, const char *format, Error **errp)
1207 {
1208     BlockDriverState *bs;
1209     BlockDriver *drv = NULL;
1210     int bdrv_flags;
1211     Error *err = NULL;
1212
1213     bs = bdrv_find(device);
1214     if (!bs) {
1215         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1216         return;
1217     }
1218
1219     if (format) {
1220         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1221         if (!drv) {
1222             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1223             return;
1224         }
1225     }
1226
1227     eject_device(bs, 0, &err);
1228     if (error_is_set(&err)) {
1229         error_propagate(errp, err);
1230         return;
1231     }
1232
1233     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1234     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1235
1236     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1237 }
1238
1239 /* throttling disk I/O limits */
1240 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1241                                int64_t bps_wr, int64_t iops, int64_t iops_rd,
1242                                int64_t iops_wr, Error **errp)
1243 {
1244     BlockIOLimit io_limits;
1245     BlockDriverState *bs;
1246
1247     bs = bdrv_find(device);
1248     if (!bs) {
1249         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1250         return;
1251     }
1252
1253     io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
1254     io_limits.bps[BLOCK_IO_LIMIT_READ]  = bps_rd;
1255     io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
1256     io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
1257     io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
1258     io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
1259
1260     if (!do_check_io_limits(&io_limits, errp)) {
1261         return;
1262     }
1263
1264     bs->io_limits = io_limits;
1265
1266     if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
1267         bdrv_io_limits_enable(bs);
1268     } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
1269         bdrv_io_limits_disable(bs);
1270     } else {
1271         if (bs->block_timer) {
1272             qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
1273         }
1274     }
1275 }
1276
1277 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1278 {
1279     const char *id = qdict_get_str(qdict, "id");
1280     BlockDriverState *bs;
1281
1282     bs = bdrv_find(id);
1283     if (!bs) {
1284         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1285         return -1;
1286     }
1287     if (bdrv_in_use(bs)) {
1288         qerror_report(QERR_DEVICE_IN_USE, id);
1289         return -1;
1290     }
1291
1292     /* quiesce block driver; prevent further io */
1293     bdrv_drain_all();
1294     bdrv_flush(bs);
1295     bdrv_close(bs);
1296
1297     /* if we have a device attached to this BlockDriverState
1298      * then we need to make the drive anonymous until the device
1299      * can be removed.  If this is a drive with no device backing
1300      * then we can just get rid of the block driver state right here.
1301      */
1302     if (bdrv_get_attached_dev(bs)) {
1303         bdrv_make_anon(bs);
1304
1305         /* Further I/O must not pause the guest */
1306         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1307                           BLOCKDEV_ON_ERROR_REPORT);
1308     } else {
1309         drive_uninit(drive_get_by_blockdev(bs));
1310     }
1311
1312     return 0;
1313 }
1314
1315 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1316 {
1317     BlockDriverState *bs;
1318     int ret;
1319
1320     bs = bdrv_find(device);
1321     if (!bs) {
1322         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1323         return;
1324     }
1325
1326     if (size < 0) {
1327         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1328         return;
1329     }
1330
1331     /* complete all in-flight operations before resizing the device */
1332     bdrv_drain_all();
1333
1334     ret = bdrv_truncate(bs, size);
1335     switch (ret) {
1336     case 0:
1337         break;
1338     case -ENOMEDIUM:
1339         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1340         break;
1341     case -ENOTSUP:
1342         error_set(errp, QERR_UNSUPPORTED);
1343         break;
1344     case -EACCES:
1345         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1346         break;
1347     case -EBUSY:
1348         error_set(errp, QERR_DEVICE_IN_USE, device);
1349         break;
1350     default:
1351         error_setg_errno(errp, -ret, "Could not resize");
1352         break;
1353     }
1354 }
1355
1356 static void block_job_cb(void *opaque, int ret)
1357 {
1358     BlockDriverState *bs = opaque;
1359     QObject *obj;
1360
1361     trace_block_job_cb(bs, bs->job, ret);
1362
1363     assert(bs->job);
1364     obj = qobject_from_block_job(bs->job);
1365     if (ret < 0) {
1366         QDict *dict = qobject_to_qdict(obj);
1367         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1368     }
1369
1370     if (block_job_is_cancelled(bs->job)) {
1371         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1372     } else {
1373         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1374     }
1375     qobject_decref(obj);
1376
1377     drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1378 }
1379
1380 void qmp_block_stream(const char *device, bool has_base,
1381                       const char *base, bool has_speed, int64_t speed,
1382                       bool has_on_error, BlockdevOnError on_error,
1383                       Error **errp)
1384 {
1385     BlockDriverState *bs;
1386     BlockDriverState *base_bs = NULL;
1387     Error *local_err = NULL;
1388
1389     if (!has_on_error) {
1390         on_error = BLOCKDEV_ON_ERROR_REPORT;
1391     }
1392
1393     bs = bdrv_find(device);
1394     if (!bs) {
1395         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1396         return;
1397     }
1398
1399     if (base) {
1400         base_bs = bdrv_find_backing_image(bs, base);
1401         if (base_bs == NULL) {
1402             error_set(errp, QERR_BASE_NOT_FOUND, base);
1403             return;
1404         }
1405     }
1406
1407     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1408                  on_error, block_job_cb, bs, &local_err);
1409     if (error_is_set(&local_err)) {
1410         error_propagate(errp, local_err);
1411         return;
1412     }
1413
1414     /* Grab a reference so hotplug does not delete the BlockDriverState from
1415      * underneath us.
1416      */
1417     drive_get_ref(drive_get_by_blockdev(bs));
1418
1419     trace_qmp_block_stream(bs, bs->job);
1420 }
1421
1422 void qmp_block_commit(const char *device,
1423                       bool has_base, const char *base, const char *top,
1424                       bool has_speed, int64_t speed,
1425                       Error **errp)
1426 {
1427     BlockDriverState *bs;
1428     BlockDriverState *base_bs, *top_bs;
1429     Error *local_err = NULL;
1430     /* This will be part of the QMP command, if/when the
1431      * BlockdevOnError change for blkmirror makes it in
1432      */
1433     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1434
1435     /* drain all i/o before commits */
1436     bdrv_drain_all();
1437
1438     bs = bdrv_find(device);
1439     if (!bs) {
1440         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1441         return;
1442     }
1443
1444     /* default top_bs is the active layer */
1445     top_bs = bs;
1446
1447     if (top) {
1448         if (strcmp(bs->filename, top) != 0) {
1449             top_bs = bdrv_find_backing_image(bs, top);
1450         }
1451     }
1452
1453     if (top_bs == NULL) {
1454         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1455         return;
1456     }
1457
1458     if (has_base && base) {
1459         base_bs = bdrv_find_backing_image(top_bs, base);
1460     } else {
1461         base_bs = bdrv_find_base(top_bs);
1462     }
1463
1464     if (base_bs == NULL) {
1465         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1466         return;
1467     }
1468
1469     commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1470                 &local_err);
1471     if (local_err != NULL) {
1472         error_propagate(errp, local_err);
1473         return;
1474     }
1475     /* Grab a reference so hotplug does not delete the BlockDriverState from
1476      * underneath us.
1477      */
1478     drive_get_ref(drive_get_by_blockdev(bs));
1479 }
1480
1481 void qmp_drive_backup(const char *device, const char *target,
1482                       bool has_format, const char *format,
1483                       enum MirrorSyncMode sync,
1484                       bool has_mode, enum NewImageMode mode,
1485                       bool has_speed, int64_t speed,
1486                       bool has_on_source_error, BlockdevOnError on_source_error,
1487                       bool has_on_target_error, BlockdevOnError on_target_error,
1488                       Error **errp)
1489 {
1490     BlockDriverState *bs;
1491     BlockDriverState *target_bs;
1492     BlockDriverState *source = NULL;
1493     BlockDriver *drv = NULL;
1494     Error *local_err = NULL;
1495     int flags;
1496     int64_t size;
1497     int ret;
1498
1499     if (!has_speed) {
1500         speed = 0;
1501     }
1502     if (!has_on_source_error) {
1503         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1504     }
1505     if (!has_on_target_error) {
1506         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1507     }
1508     if (!has_mode) {
1509         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1510     }
1511
1512     bs = bdrv_find(device);
1513     if (!bs) {
1514         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1515         return;
1516     }
1517
1518     if (!bdrv_is_inserted(bs)) {
1519         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1520         return;
1521     }
1522
1523     if (!has_format) {
1524         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1525     }
1526     if (format) {
1527         drv = bdrv_find_format(format);
1528         if (!drv) {
1529             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1530             return;
1531         }
1532     }
1533
1534     if (bdrv_in_use(bs)) {
1535         error_set(errp, QERR_DEVICE_IN_USE, device);
1536         return;
1537     }
1538
1539     flags = bs->open_flags | BDRV_O_RDWR;
1540
1541     /* See if we have a backing HD we can use to create our new image
1542      * on top of. */
1543     if (sync == MIRROR_SYNC_MODE_TOP) {
1544         source = bs->backing_hd;
1545         if (!source) {
1546             sync = MIRROR_SYNC_MODE_FULL;
1547         }
1548     }
1549     if (sync == MIRROR_SYNC_MODE_NONE) {
1550         source = bs;
1551     }
1552
1553     size = bdrv_getlength(bs);
1554     if (size < 0) {
1555         error_setg_errno(errp, -size, "bdrv_getlength failed");
1556         return;
1557     }
1558
1559     if (mode != NEW_IMAGE_MODE_EXISTING) {
1560         assert(format && drv);
1561         if (source) {
1562             bdrv_img_create(target, format, source->filename,
1563                             source->drv->format_name, NULL,
1564                             size, flags, &local_err, false);
1565         } else {
1566             bdrv_img_create(target, format, NULL, NULL, NULL,
1567                             size, flags, &local_err, false);
1568         }
1569     }
1570
1571     if (error_is_set(&local_err)) {
1572         error_propagate(errp, local_err);
1573         return;
1574     }
1575
1576     target_bs = bdrv_new("");
1577     ret = bdrv_open(target_bs, target, NULL, flags, drv);
1578     if (ret < 0) {
1579         bdrv_delete(target_bs);
1580         error_setg_file_open(errp, -ret, target);
1581         return;
1582     }
1583
1584     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1585                  block_job_cb, bs, &local_err);
1586     if (local_err != NULL) {
1587         bdrv_delete(target_bs);
1588         error_propagate(errp, local_err);
1589         return;
1590     }
1591
1592     /* Grab a reference so hotplug does not delete the BlockDriverState from
1593      * underneath us.
1594      */
1595     drive_get_ref(drive_get_by_blockdev(bs));
1596 }
1597
1598 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1599
1600 void qmp_drive_mirror(const char *device, const char *target,
1601                       bool has_format, const char *format,
1602                       enum MirrorSyncMode sync,
1603                       bool has_mode, enum NewImageMode mode,
1604                       bool has_speed, int64_t speed,
1605                       bool has_granularity, uint32_t granularity,
1606                       bool has_buf_size, int64_t buf_size,
1607                       bool has_on_source_error, BlockdevOnError on_source_error,
1608                       bool has_on_target_error, BlockdevOnError on_target_error,
1609                       Error **errp)
1610 {
1611     BlockDriverState *bs;
1612     BlockDriverState *source, *target_bs;
1613     BlockDriver *drv = NULL;
1614     Error *local_err = NULL;
1615     int flags;
1616     int64_t size;
1617     int ret;
1618
1619     if (!has_speed) {
1620         speed = 0;
1621     }
1622     if (!has_on_source_error) {
1623         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1624     }
1625     if (!has_on_target_error) {
1626         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1627     }
1628     if (!has_mode) {
1629         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1630     }
1631     if (!has_granularity) {
1632         granularity = 0;
1633     }
1634     if (!has_buf_size) {
1635         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1636     }
1637
1638     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1639         error_set(errp, QERR_INVALID_PARAMETER, device);
1640         return;
1641     }
1642     if (granularity & (granularity - 1)) {
1643         error_set(errp, QERR_INVALID_PARAMETER, device);
1644         return;
1645     }
1646
1647     bs = bdrv_find(device);
1648     if (!bs) {
1649         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1650         return;
1651     }
1652
1653     if (!bdrv_is_inserted(bs)) {
1654         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1655         return;
1656     }
1657
1658     if (!has_format) {
1659         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1660     }
1661     if (format) {
1662         drv = bdrv_find_format(format);
1663         if (!drv) {
1664             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1665             return;
1666         }
1667     }
1668
1669     if (bdrv_in_use(bs)) {
1670         error_set(errp, QERR_DEVICE_IN_USE, device);
1671         return;
1672     }
1673
1674     flags = bs->open_flags | BDRV_O_RDWR;
1675     source = bs->backing_hd;
1676     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1677         sync = MIRROR_SYNC_MODE_FULL;
1678     }
1679
1680     size = bdrv_getlength(bs);
1681     if (size < 0) {
1682         error_setg_errno(errp, -size, "bdrv_getlength failed");
1683         return;
1684     }
1685
1686     if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1687         /* create new image w/o backing file */
1688         assert(format && drv);
1689         bdrv_img_create(target, format,
1690                         NULL, NULL, NULL, size, flags, &local_err, false);
1691     } else {
1692         switch (mode) {
1693         case NEW_IMAGE_MODE_EXISTING:
1694             ret = 0;
1695             break;
1696         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1697             /* create new image with backing file */
1698             bdrv_img_create(target, format,
1699                             source->filename,
1700                             source->drv->format_name,
1701                             NULL, size, flags, &local_err, false);
1702             break;
1703         default:
1704             abort();
1705         }
1706     }
1707
1708     if (error_is_set(&local_err)) {
1709         error_propagate(errp, local_err);
1710         return;
1711     }
1712
1713     /* Mirroring takes care of copy-on-write using the source's backing
1714      * file.
1715      */
1716     target_bs = bdrv_new("");
1717     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv);
1718     if (ret < 0) {
1719         bdrv_delete(target_bs);
1720         error_setg_file_open(errp, -ret, target);
1721         return;
1722     }
1723
1724     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1725                  on_source_error, on_target_error,
1726                  block_job_cb, bs, &local_err);
1727     if (local_err != NULL) {
1728         bdrv_delete(target_bs);
1729         error_propagate(errp, local_err);
1730         return;
1731     }
1732
1733     /* Grab a reference so hotplug does not delete the BlockDriverState from
1734      * underneath us.
1735      */
1736     drive_get_ref(drive_get_by_blockdev(bs));
1737 }
1738
1739 static BlockJob *find_block_job(const char *device)
1740 {
1741     BlockDriverState *bs;
1742
1743     bs = bdrv_find(device);
1744     if (!bs || !bs->job) {
1745         return NULL;
1746     }
1747     return bs->job;
1748 }
1749
1750 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1751 {
1752     BlockJob *job = find_block_job(device);
1753
1754     if (!job) {
1755         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1756         return;
1757     }
1758
1759     block_job_set_speed(job, speed, errp);
1760 }
1761
1762 void qmp_block_job_cancel(const char *device,
1763                           bool has_force, bool force, Error **errp)
1764 {
1765     BlockJob *job = find_block_job(device);
1766
1767     if (!has_force) {
1768         force = false;
1769     }
1770
1771     if (!job) {
1772         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1773         return;
1774     }
1775     if (job->paused && !force) {
1776         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1777         return;
1778     }
1779
1780     trace_qmp_block_job_cancel(job);
1781     block_job_cancel(job);
1782 }
1783
1784 void qmp_block_job_pause(const char *device, Error **errp)
1785 {
1786     BlockJob *job = find_block_job(device);
1787
1788     if (!job) {
1789         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1790         return;
1791     }
1792
1793     trace_qmp_block_job_pause(job);
1794     block_job_pause(job);
1795 }
1796
1797 void qmp_block_job_resume(const char *device, Error **errp)
1798 {
1799     BlockJob *job = find_block_job(device);
1800
1801     if (!job) {
1802         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1803         return;
1804     }
1805
1806     trace_qmp_block_job_resume(job);
1807     block_job_resume(job);
1808 }
1809
1810 void qmp_block_job_complete(const char *device, Error **errp)
1811 {
1812     BlockJob *job = find_block_job(device);
1813
1814     if (!job) {
1815         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1816         return;
1817     }
1818
1819     trace_qmp_block_job_complete(job);
1820     block_job_complete(job, errp);
1821 }
1822
1823 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1824 {
1825     BlockJobInfoList **prev = opaque;
1826     BlockJob *job = bs->job;
1827
1828     if (job) {
1829         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1830         elem->value = block_job_query(bs->job);
1831         (*prev)->next = elem;
1832         *prev = elem;
1833     }
1834 }
1835
1836 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1837 {
1838     /* Dummy is a fake list element for holding the head pointer */
1839     BlockJobInfoList dummy = {};
1840     BlockJobInfoList *prev = &dummy;
1841     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1842     return dummy.next;
1843 }
1844
1845 QemuOptsList qemu_common_drive_opts = {
1846     .name = "drive",
1847     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
1848     .desc = {
1849         {
1850             .name = "bus",
1851             .type = QEMU_OPT_NUMBER,
1852             .help = "bus number",
1853         },{
1854             .name = "unit",
1855             .type = QEMU_OPT_NUMBER,
1856             .help = "unit number (i.e. lun for scsi)",
1857         },{
1858             .name = "if",
1859             .type = QEMU_OPT_STRING,
1860             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1861         },{
1862             .name = "index",
1863             .type = QEMU_OPT_NUMBER,
1864             .help = "index number",
1865         },{
1866             .name = "cyls",
1867             .type = QEMU_OPT_NUMBER,
1868             .help = "number of cylinders (ide disk geometry)",
1869         },{
1870             .name = "heads",
1871             .type = QEMU_OPT_NUMBER,
1872             .help = "number of heads (ide disk geometry)",
1873         },{
1874             .name = "secs",
1875             .type = QEMU_OPT_NUMBER,
1876             .help = "number of sectors (ide disk geometry)",
1877         },{
1878             .name = "trans",
1879             .type = QEMU_OPT_STRING,
1880             .help = "chs translation (auto, lba. none)",
1881         },{
1882             .name = "media",
1883             .type = QEMU_OPT_STRING,
1884             .help = "media type (disk, cdrom)",
1885         },{
1886             .name = "snapshot",
1887             .type = QEMU_OPT_BOOL,
1888             .help = "enable/disable snapshot mode",
1889         },{
1890             .name = "file",
1891             .type = QEMU_OPT_STRING,
1892             .help = "disk image",
1893         },{
1894             .name = "discard",
1895             .type = QEMU_OPT_STRING,
1896             .help = "discard operation (ignore/off, unmap/on)",
1897         },{
1898             .name = "cache.writeback",
1899             .type = QEMU_OPT_BOOL,
1900             .help = "enables writeback mode for any caches",
1901         },{
1902             .name = "cache.direct",
1903             .type = QEMU_OPT_BOOL,
1904             .help = "enables use of O_DIRECT (bypass the host page cache)",
1905         },{
1906             .name = "cache.no-flush",
1907             .type = QEMU_OPT_BOOL,
1908             .help = "ignore any flush requests for the device",
1909         },{
1910             .name = "aio",
1911             .type = QEMU_OPT_STRING,
1912             .help = "host AIO implementation (threads, native)",
1913         },{
1914             .name = "format",
1915             .type = QEMU_OPT_STRING,
1916             .help = "disk format (raw, qcow2, ...)",
1917         },{
1918             .name = "serial",
1919             .type = QEMU_OPT_STRING,
1920             .help = "disk serial number",
1921         },{
1922             .name = "rerror",
1923             .type = QEMU_OPT_STRING,
1924             .help = "read error action",
1925         },{
1926             .name = "werror",
1927             .type = QEMU_OPT_STRING,
1928             .help = "write error action",
1929         },{
1930             .name = "addr",
1931             .type = QEMU_OPT_STRING,
1932             .help = "pci address (virtio only)",
1933         },{
1934             .name = "read-only",
1935             .type = QEMU_OPT_BOOL,
1936             .help = "open drive file as read-only",
1937         },{
1938             .name = "throttling.iops-total",
1939             .type = QEMU_OPT_NUMBER,
1940             .help = "limit total I/O operations per second",
1941         },{
1942             .name = "throttling.iops-read",
1943             .type = QEMU_OPT_NUMBER,
1944             .help = "limit read operations per second",
1945         },{
1946             .name = "throttling.iops-write",
1947             .type = QEMU_OPT_NUMBER,
1948             .help = "limit write operations per second",
1949         },{
1950             .name = "throttling.bps-total",
1951             .type = QEMU_OPT_NUMBER,
1952             .help = "limit total bytes per second",
1953         },{
1954             .name = "throttling.bps-read",
1955             .type = QEMU_OPT_NUMBER,
1956             .help = "limit read bytes per second",
1957         },{
1958             .name = "throttling.bps-write",
1959             .type = QEMU_OPT_NUMBER,
1960             .help = "limit write bytes per second",
1961         },{
1962             .name = "copy-on-read",
1963             .type = QEMU_OPT_BOOL,
1964             .help = "copy read data from backing file into image file",
1965         },{
1966             .name = "boot",
1967             .type = QEMU_OPT_BOOL,
1968             .help = "(deprecated, ignored)",
1969         },
1970         { /* end of list */ }
1971     },
1972 };
1973
1974 QemuOptsList qemu_drive_opts = {
1975     .name = "drive",
1976     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1977     .desc = {
1978         /*
1979          * no elements => accept any params
1980          * validation will happen later
1981          */
1982         { /* end of list */ }
1983     },
1984 };