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 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
 316 {
 317     const char *buf;
 318     const char *file = NULL;
 319     const char *serial;
 320     const char *mediastr = "";
 321     BlockInterfaceType type;
 322     enum { MEDIA_DISK, MEDIA_CDROM } media;
 323     int bus_id, unit_id;
 324     int cyls, heads, secs, translation;
 325     BlockDriver *drv = NULL;
 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
 342     translation = BIOS_ATA_TRANSLATION_AUTO;
 343     media = MEDIA_DISK;
 344
 345     /* Check common options by copying from all_opts to opts, all other options
 346      * are stored in bs_opts. */
 347     id = qemu_opts_id(all_opts);
 348     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
 349     if (error_is_set(&error)) {
 350         qerror_report_err(error);
 351         error_free(error);
 352         return NULL;
 353     }
 354
 355     bs_opts = qdict_new();
 356     qemu_opts_to_qdict(all_opts, bs_opts);
 357     qemu_opts_absorb_qdict(opts, bs_opts, &error);
 358     if (error_is_set(&error)) {
 359         qerror_report_err(error);
 360         error_free(error);
 361         return NULL;
 362     }
 363
 364     if (id) {
 365         qdict_del(bs_opts, "id");
 366     }
 367
 368     /* extract parameters */
 369     bus_id  = qemu_opt_get_number(opts, "bus", 0);
 370     unit_id = qemu_opt_get_number(opts, "unit", -1);
 371     index   = qemu_opt_get_number(opts, "index", -1);
 372
 373     cyls  = qemu_opt_get_number(opts, "cyls", 0);
 374     heads = qemu_opt_get_number(opts, "heads", 0);
 375     secs  = qemu_opt_get_number(opts, "secs", 0);
 376
 377     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 378     ro = qemu_opt_get_bool(opts, "readonly", 0);
 379     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
 380
 381     file = qemu_opt_get(opts, "file");
 382     serial = qemu_opt_get(opts, "serial");
 383
 384     if ((buf = qemu_opt_get(opts, "if")) != NULL) {
 385         for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
 386             ;
 387         if (type == IF_COUNT) {
 388             error_report("unsupported bus type '%s'", buf);
 389             return NULL;
 390         }
 391     } else {
 392         type = block_default_type;
 393     }
 394
 395     max_devs = if_max_devs[type];
 396
 397     if (cyls || heads || secs) {
 398         if (cyls < 1) {
 399             error_report("invalid physical cyls number");
 400             return NULL;
 401         }
 402         if (heads < 1) {
 403             error_report("invalid physical heads number");
 404             return NULL;
 405         }
 406         if (secs < 1) {
 407             error_report("invalid physical secs number");
 408             return NULL;
 409         }
 410     }
 411
 412     if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
 413         if (!cyls) {
 414             error_report("'%s' trans must be used with cyls, heads and secs",
 415                          buf);
 416             return NULL;
 417         }
 418         if (!strcmp(buf, "none"))
 419             translation = BIOS_ATA_TRANSLATION_NONE;
 420         else if (!strcmp(buf, "lba"))
 421             translation = BIOS_ATA_TRANSLATION_LBA;
 422         else if (!strcmp(buf, "auto"))
 423             translation = BIOS_ATA_TRANSLATION_AUTO;
 424         else {
 425             error_report("'%s' invalid translation type", buf);
 426             return NULL;
 427         }
 428     }
 429
 430     if ((buf = qemu_opt_get(opts, "media")) != NULL) {
 431         if (!strcmp(buf, "disk")) {
 432             media = MEDIA_DISK;
 433         } else if (!strcmp(buf, "cdrom")) {
 434             if (cyls || secs || heads) {
 435                 error_report("CHS can't be set with media=%s", buf);
 436                 return NULL;
 437             }
 438             media = MEDIA_CDROM;
 439         } else {
 440             error_report("'%s' invalid media", buf);
 441             return NULL;
 442         }
 443     }
 444
 445     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
 446         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
 447             error_report("invalid discard option");
 448             return NULL;
 449         }
 450     }
 451
 452     bdrv_flags |= BDRV_O_CACHE_WB;
 453     if ((buf = qemu_opt_get(opts, "cache")) != NULL) {
 454         if (bdrv_parse_cache_flags(buf, &bdrv_flags) != 0) {
 455             error_report("invalid cache option");
 456             return NULL;
 457         }
 458     }
 459
 460 #ifdef CONFIG_LINUX_AIO
 461     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 462         if (!strcmp(buf, "native")) {
 463             bdrv_flags |= BDRV_O_NATIVE_AIO;
 464         } else if (!strcmp(buf, "threads")) {
 465             /* this is the default */
 466         } else {
 467            error_report("invalid aio option");
 468            return NULL;
 469         }
 470     }
 471 #endif
 472
 473     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 474         if (is_help_option(buf)) {
 475             error_printf("Supported formats:");
 476             bdrv_iterate_format(bdrv_format_print, NULL);
 477             error_printf("\n");
 478             return NULL;
 479         }
 480         drv = bdrv_find_whitelisted_format(buf, ro);
 481         if (!drv) {
 482             error_report("'%s' invalid format", buf);
 483             return NULL;
 484         }
 485     }
 486
 487     /* disk I/O throttling */
 488     io_limits.bps[BLOCK_IO_LIMIT_TOTAL]  =
 489                            qemu_opt_get_number(opts, "bps", 0);
 490     io_limits.bps[BLOCK_IO_LIMIT_READ]   =
 491                            qemu_opt_get_number(opts, "bps_rd", 0);
 492     io_limits.bps[BLOCK_IO_LIMIT_WRITE]  =
 493                            qemu_opt_get_number(opts, "bps_wr", 0);
 494     io_limits.iops[BLOCK_IO_LIMIT_TOTAL] =
 495                            qemu_opt_get_number(opts, "iops", 0);
 496     io_limits.iops[BLOCK_IO_LIMIT_READ]  =
 497                            qemu_opt_get_number(opts, "iops_rd", 0);
 498     io_limits.iops[BLOCK_IO_LIMIT_WRITE] =
 499                            qemu_opt_get_number(opts, "iops_wr", 0);
 500
 501     if (!do_check_io_limits(&io_limits, &error)) {
 502         error_report("%s", error_get_pretty(error));
 503         error_free(error);
 504         return NULL;
 505     }
 506
 507     if (qemu_opt_get(opts, "boot") != NULL) {
 508         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 509                 "ignored. Future versions will reject this parameter. Please "
 510                 "update your scripts.\n");
 511     }
 512
 513     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
 514     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 515         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
 516             error_report("werror is not supported by this bus type");
 517             return NULL;
 518         }
 519
 520         on_write_error = parse_block_error_action(buf, 0);
 521         if (on_write_error < 0) {
 522             return NULL;
 523         }
 524     }
 525
 526     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
 527     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 528         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
 529             error_report("rerror is not supported by this bus type");
 530             return NULL;
 531         }
 532
 533         on_read_error = parse_block_error_action(buf, 1);
 534         if (on_read_error < 0) {
 535             return NULL;
 536         }
 537     }
 538
 539     if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
 540         if (type != IF_VIRTIO) {
 541             error_report("addr is not supported by this bus type");
 542             return NULL;
 543         }
 544     }
 545
 546     /* compute bus and unit according index */
 547
 548     if (index != -1) {
 549         if (bus_id != 0 || unit_id != -1) {
 550             error_report("index cannot be used with bus and unit");
 551             return NULL;
 552         }
 553         bus_id = drive_index_to_bus_id(type, index);
 554         unit_id = drive_index_to_unit_id(type, index);
 555     }
 556
 557     /* if user doesn't specify a unit_id,
 558      * try to find the first free
 559      */
 560
 561     if (unit_id == -1) {
 562        unit_id = 0;
 563        while (drive_get(type, bus_id, unit_id) != NULL) {
 564            unit_id++;
 565            if (max_devs && unit_id >= max_devs) {
 566                unit_id -= max_devs;
 567                bus_id++;
 568            }
 569        }
 570     }
 571
 572     /* check unit id */
 573
 574     if (max_devs && unit_id >= max_devs) {
 575         error_report("unit %d too big (max is %d)",
 576                      unit_id, max_devs - 1);
 577         return NULL;
 578     }
 579
 580     /*
 581      * catch multiple definitions
 582      */
 583
 584     if (drive_get(type, bus_id, unit_id) != NULL) {
 585         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 586                      bus_id, unit_id, index);
 587         return NULL;
 588     }
 589
 590     /* init */
 591
 592     dinfo = g_malloc0(sizeof(*dinfo));
 593     if ((buf = qemu_opts_id(opts)) != NULL) {
 594         dinfo->id = g_strdup(buf);
 595     } else {
 596         /* no id supplied -> create one */
 597         dinfo->id = g_malloc0(32);
 598         if (type == IF_IDE || type == IF_SCSI)
 599             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 600         if (max_devs)
 601             snprintf(dinfo->id, 32, "%s%i%s%i",
 602                      if_name[type], bus_id, mediastr, unit_id);
 603         else
 604             snprintf(dinfo->id, 32, "%s%s%i",
 605                      if_name[type], mediastr, unit_id);
 606     }
 607     dinfo->bdrv = bdrv_new(dinfo->id);
 608     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
 609     dinfo->bdrv->read_only = ro;
 610     dinfo->devaddr = devaddr;
 611     dinfo->type = type;
 612     dinfo->bus = bus_id;
 613     dinfo->unit = unit_id;
 614     dinfo->cyls = cyls;
 615     dinfo->heads = heads;
 616     dinfo->secs = secs;
 617     dinfo->trans = translation;
 618     dinfo->opts = all_opts;
 619     dinfo->refcount = 1;
 620     if (serial != NULL) {
 621         dinfo->serial = g_strdup(serial);
 622     }
 623     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 624
 625     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
 626
 627     /* disk I/O throttling */
 628     bdrv_set_io_limits(dinfo->bdrv, &io_limits);
 629
 630     switch(type) {
 631     case IF_IDE:
 632     case IF_SCSI:
 633     case IF_XEN:
 634     case IF_NONE:
 635         dinfo->media_cd = media == MEDIA_CDROM;
 636         break;
 637     case IF_SD:
 638     case IF_FLOPPY:
 639     case IF_PFLASH:
 640     case IF_MTD:
 641         break;
 642     case IF_VIRTIO:
 643     {
 644         /* add virtio block device */
 645         QemuOpts *devopts;
 646         devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
 647         if (arch_type == QEMU_ARCH_S390X) {
 648             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
 649         } else {
 650             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
 651         }
 652         qemu_opt_set(devopts, "drive", dinfo->id);
 653         if (devaddr)
 654             qemu_opt_set(devopts, "addr", devaddr);
 655         break;
 656     }
 657     default:
 658         abort();
 659     }
 660     if (!file || !*file) {
 661         if (qdict_size(bs_opts)) {
 662             file = NULL;
 663         } else {
 664             return dinfo;
 665         }
 666     }
 667     if (snapshot) {
 668         /* always use cache=unsafe with snapshot */
 669         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 670         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 671     }
 672
 673     if (copy_on_read) {
 674         bdrv_flags |= BDRV_O_COPY_ON_READ;
 675     }
 676
 677     if (runstate_check(RUN_STATE_INMIGRATE)) {
 678         bdrv_flags |= BDRV_O_INCOMING;
 679     }
 680
 681     if (media == MEDIA_CDROM) {
 682         /* CDROM is fine for any interface, don't check.  */
 683         ro = 1;
 684     } else if (ro == 1) {
 685         if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
 686             type != IF_NONE && type != IF_PFLASH) {
 687             error_report("readonly not supported by this bus type");
 688             goto err;
 689         }
 690     }
 691
 692     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 693
 694     if (ro && copy_on_read) {
 695         error_report("warning: disabling copy_on_read on readonly drive");
 696     }
 697
 698     ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv);
 699     bs_opts = NULL;
 700
 701     if (ret < 0) {
 702         if (ret == -EMEDIUMTYPE) {
 703             error_report("could not open disk image %s: not in %s format",
 704                          file ?: dinfo->id, drv->format_name);
 705         } else {
 706             error_report("could not open disk image %s: %s",
 707                          file ?: dinfo->id, strerror(-ret));
 708         }
 709         goto err;
 710     }
 711
 712     if (bdrv_key_required(dinfo->bdrv))
 713         autostart = 0;
 714
 715     qemu_opts_del(opts);
 716
 717     return dinfo;
 718
 719 err:
 720     qemu_opts_del(opts);
 721     QDECREF(bs_opts);
 722     bdrv_delete(dinfo->bdrv);
 723     g_free(dinfo->id);
 724     QTAILQ_REMOVE(&drives, dinfo, next);
 725     g_free(dinfo);
 726     return NULL;
 727 }
 728
 729 void do_commit(Monitor *mon, const QDict *qdict)
 730 {
 731     const char *device = qdict_get_str(qdict, "device");
 732     BlockDriverState *bs;
 733     int ret;
 734
 735     if (!strcmp(device, "all")) {
 736         ret = bdrv_commit_all();
 737     } else {
 738         bs = bdrv_find(device);
 739         if (!bs) {
 740             monitor_printf(mon, "Device '%s' not found\n", device);
 741             return;
 742         }
 743         ret = bdrv_commit(bs);
 744     }
 745     if (ret < 0) {
 746         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
 747                        strerror(-ret));
 748     }
 749 }
 750
 751 static void blockdev_do_action(int kind, void *data, Error **errp)
 752 {
 753     TransactionAction action;
 754     TransactionActionList list;
 755
 756     action.kind = kind;
 757     action.data = data;
 758     list.value = &action;
 759     list.next = NULL;
 760     qmp_transaction(&list, errp);
 761 }
 762
 763 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
 764                                 bool has_format, const char *format,
 765                                 bool has_mode, enum NewImageMode mode,
 766                                 Error **errp)
 767 {
 768     BlockdevSnapshot snapshot = {
 769         .device = (char *) device,
 770         .snapshot_file = (char *) snapshot_file,
 771         .has_format = has_format,
 772         .format = (char *) format,
 773         .has_mode = has_mode,
 774         .mode = mode,
 775     };
 776     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
 777                        &snapshot, errp);
 778 }
 779
 780
 781 /* New and old BlockDriverState structs for group snapshots */
 782
 783 typedef struct BlkTransactionStates BlkTransactionStates;
 784
 785 /* Only prepare() may fail. In a single transaction, only one of commit() or
 786    abort() will be called, clean() will always be called if it present. */
 787 typedef struct BdrvActionOps {
 788     /* Size of state struct, in bytes. */
 789     size_t instance_size;
 790     /* Prepare the work, must NOT be NULL. */
 791     void (*prepare)(BlkTransactionStates *common, Error **errp);
 792     /* Commit the changes, must NOT be NULL. */
 793     void (*commit)(BlkTransactionStates *common);
 794     /* Abort the changes on fail, can be NULL. */
 795     void (*abort)(BlkTransactionStates *common);
 796     /* Clean up resource in the end, can be NULL. */
 797     void (*clean)(BlkTransactionStates *common);
 798 } BdrvActionOps;
 799
 800 /*
 801  * This structure must be arranged as first member in child type, assuming
 802  * that compiler will also arrange it to the same address with parent instance.
 803  * Later it will be used in free().
 804  */
 805 struct BlkTransactionStates {
 806     TransactionAction *action;
 807     const BdrvActionOps *ops;
 808     QSIMPLEQ_ENTRY(BlkTransactionStates) entry;
 809 };
 810
 811 /* external snapshot private data */
 812 typedef struct ExternalSnapshotStates {
 813     BlkTransactionStates common;
 814     BlockDriverState *old_bs;
 815     BlockDriverState *new_bs;
 816 } ExternalSnapshotStates;
 817
 818 static void external_snapshot_prepare(BlkTransactionStates *common,
 819                                       Error **errp)
 820 {
 821     BlockDriver *proto_drv;
 822     BlockDriver *drv;
 823     int flags, ret;
 824     Error *local_err = NULL;
 825     const char *device;
 826     const char *new_image_file;
 827     const char *format = "qcow2";
 828     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
 829     ExternalSnapshotStates *states =
 830                              DO_UPCAST(ExternalSnapshotStates, common, common);
 831     TransactionAction *action = common->action;
 832
 833     /* get parameters */
 834     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
 835
 836     device = action->blockdev_snapshot_sync->device;
 837     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
 838     if (action->blockdev_snapshot_sync->has_format) {
 839         format = action->blockdev_snapshot_sync->format;
 840     }
 841     if (action->blockdev_snapshot_sync->has_mode) {
 842         mode = action->blockdev_snapshot_sync->mode;
 843     }
 844
 845     /* start processing */
 846     drv = bdrv_find_format(format);
 847     if (!drv) {
 848         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 849         return;
 850     }
 851
 852     states->old_bs = bdrv_find(device);
 853     if (!states->old_bs) {
 854         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 855         return;
 856     }
 857
 858     if (!bdrv_is_inserted(states->old_bs)) {
 859         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
 860         return;
 861     }
 862
 863     if (bdrv_in_use(states->old_bs)) {
 864         error_set(errp, QERR_DEVICE_IN_USE, device);
 865         return;
 866     }
 867
 868     if (!bdrv_is_read_only(states->old_bs)) {
 869         if (bdrv_flush(states->old_bs)) {
 870             error_set(errp, QERR_IO_ERROR);
 871             return;
 872         }
 873     }
 874
 875     flags = states->old_bs->open_flags;
 876
 877     proto_drv = bdrv_find_protocol(new_image_file);
 878     if (!proto_drv) {
 879         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 880         return;
 881     }
 882
 883     /* create new image w/backing file */
 884     if (mode != NEW_IMAGE_MODE_EXISTING) {
 885         bdrv_img_create(new_image_file, format,
 886                         states->old_bs->filename,
 887                         states->old_bs->drv->format_name,
 888                         NULL, -1, flags, &local_err, false);
 889         if (error_is_set(&local_err)) {
 890             error_propagate(errp, local_err);
 891             return;
 892         }
 893     }
 894
 895     /* We will manually add the backing_hd field to the bs later */
 896     states->new_bs = bdrv_new("");
 897     /* TODO Inherit bs->options or only take explicit options with an
 898      * extended QMP command? */
 899     ret = bdrv_open(states->new_bs, new_image_file, NULL,
 900                     flags | BDRV_O_NO_BACKING, drv);
 901     if (ret != 0) {
 902         error_set(errp, QERR_OPEN_FILE_FAILED, new_image_file);
 903     }
 904 }
 905
 906 static void external_snapshot_commit(BlkTransactionStates *common)
 907 {
 908     ExternalSnapshotStates *states =
 909                              DO_UPCAST(ExternalSnapshotStates, common, common);
 910
 911     /* This removes our old bs from the bdrv_states, and adds the new bs */
 912     bdrv_append(states->new_bs, states->old_bs);
 913     /* We don't need (or want) to use the transactional
 914      * bdrv_reopen_multiple() across all the entries at once, because we
 915      * don't want to abort all of them if one of them fails the reopen */
 916     bdrv_reopen(states->new_bs, states->new_bs->open_flags & ~BDRV_O_RDWR,
 917                 NULL);
 918 }
 919
 920 static void external_snapshot_abort(BlkTransactionStates *common)
 921 {
 922     ExternalSnapshotStates *states =
 923                              DO_UPCAST(ExternalSnapshotStates, common, common);
 924     if (states->new_bs) {
 925         bdrv_delete(states->new_bs);
 926     }
 927 }
 928
 929 static const BdrvActionOps actions[] = {
 930     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
 931         .instance_size = sizeof(ExternalSnapshotStates),
 932         .prepare  = external_snapshot_prepare,
 933         .commit   = external_snapshot_commit,
 934         .abort = external_snapshot_abort,
 935     },
 936 };
 937
 938 /*
 939  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
 940  *  then we do not pivot any of the devices in the group, and abandon the
 941  *  snapshots
 942  */
 943 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
 944 {
 945     TransactionActionList *dev_entry = dev_list;
 946     BlkTransactionStates *states, *next;
 947     Error *local_err = NULL;
 948
 949     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionStates) snap_bdrv_states;
 950     QSIMPLEQ_INIT(&snap_bdrv_states);
 951
 952     /* drain all i/o before any snapshots */
 953     bdrv_drain_all();
 954
 955     /* We don't do anything in this loop that commits us to the snapshot */
 956     while (NULL != dev_entry) {
 957         TransactionAction *dev_info = NULL;
 958         const BdrvActionOps *ops;
 959
 960         dev_info = dev_entry->value;
 961         dev_entry = dev_entry->next;
 962
 963         assert(dev_info->kind < ARRAY_SIZE(actions));
 964
 965         ops = &actions[dev_info->kind];
 966         states = g_malloc0(ops->instance_size);
 967         states->ops = ops;
 968         states->action = dev_info;
 969         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, states, entry);
 970
 971         states->ops->prepare(states, &local_err);
 972         if (error_is_set(&local_err)) {
 973             error_propagate(errp, local_err);
 974             goto delete_and_fail;
 975         }
 976     }
 977
 978     QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
 979         states->ops->commit(states);
 980     }
 981
 982     /* success */
 983     goto exit;
 984
 985 delete_and_fail:
 986     /*
 987     * failure, and it is all-or-none; abandon each new bs, and keep using
 988     * the original bs for all images
 989     */
 990     QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
 991         if (states->ops->abort) {
 992             states->ops->abort(states);
 993         }
 994     }
 995 exit:
 996     QSIMPLEQ_FOREACH_SAFE(states, &snap_bdrv_states, entry, next) {
 997         if (states->ops->clean) {
 998             states->ops->clean(states);
 999         }
1000         g_free(states);
1001     }
1002 }
1003
1004
1005 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1006 {
1007     if (bdrv_in_use(bs)) {
1008         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1009         return;
1010     }
1011     if (!bdrv_dev_has_removable_media(bs)) {
1012         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1013         return;
1014     }
1015
1016     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1017         bdrv_dev_eject_request(bs, force);
1018         if (!force) {
1019             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1020             return;
1021         }
1022     }
1023
1024     bdrv_close(bs);
1025 }
1026
1027 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1028 {
1029     BlockDriverState *bs;
1030
1031     bs = bdrv_find(device);
1032     if (!bs) {
1033         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1034         return;
1035     }
1036
1037     eject_device(bs, force, errp);
1038 }
1039
1040 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1041 {
1042     BlockDriverState *bs;
1043     int err;
1044
1045     bs = bdrv_find(device);
1046     if (!bs) {
1047         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1048         return;
1049     }
1050
1051     err = bdrv_set_key(bs, password);
1052     if (err == -EINVAL) {
1053         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1054         return;
1055     } else if (err < 0) {
1056         error_set(errp, QERR_INVALID_PASSWORD);
1057         return;
1058     }
1059 }
1060
1061 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1062                                     int bdrv_flags, BlockDriver *drv,
1063                                     const char *password, Error **errp)
1064 {
1065     if (bdrv_open(bs, filename, NULL, bdrv_flags, drv) < 0) {
1066         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
1067         return;
1068     }
1069
1070     if (bdrv_key_required(bs)) {
1071         if (password) {
1072             if (bdrv_set_key(bs, password) < 0) {
1073                 error_set(errp, QERR_INVALID_PASSWORD);
1074             }
1075         } else {
1076             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1077                       bdrv_get_encrypted_filename(bs));
1078         }
1079     } else if (password) {
1080         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1081     }
1082 }
1083
1084 void qmp_change_blockdev(const char *device, const char *filename,
1085                          bool has_format, const char *format, Error **errp)
1086 {
1087     BlockDriverState *bs;
1088     BlockDriver *drv = NULL;
1089     int bdrv_flags;
1090     Error *err = NULL;
1091
1092     bs = bdrv_find(device);
1093     if (!bs) {
1094         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1095         return;
1096     }
1097
1098     if (format) {
1099         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1100         if (!drv) {
1101             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1102             return;
1103         }
1104     }
1105
1106     eject_device(bs, 0, &err);
1107     if (error_is_set(&err)) {
1108         error_propagate(errp, err);
1109         return;
1110     }
1111
1112     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1113     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1114
1115     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1116 }
1117
1118 /* throttling disk I/O limits */
1119 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1120                                int64_t bps_wr, int64_t iops, int64_t iops_rd,
1121                                int64_t iops_wr, Error **errp)
1122 {
1123     BlockIOLimit io_limits;
1124     BlockDriverState *bs;
1125
1126     bs = bdrv_find(device);
1127     if (!bs) {
1128         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1129         return;
1130     }
1131
1132     io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
1133     io_limits.bps[BLOCK_IO_LIMIT_READ]  = bps_rd;
1134     io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
1135     io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
1136     io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
1137     io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
1138
1139     if (!do_check_io_limits(&io_limits, errp)) {
1140         return;
1141     }
1142
1143     bs->io_limits = io_limits;
1144
1145     if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
1146         bdrv_io_limits_enable(bs);
1147     } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
1148         bdrv_io_limits_disable(bs);
1149     } else {
1150         if (bs->block_timer) {
1151             qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
1152         }
1153     }
1154 }
1155
1156 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1157 {
1158     const char *id = qdict_get_str(qdict, "id");
1159     BlockDriverState *bs;
1160
1161     bs = bdrv_find(id);
1162     if (!bs) {
1163         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1164         return -1;
1165     }
1166     if (bdrv_in_use(bs)) {
1167         qerror_report(QERR_DEVICE_IN_USE, id);
1168         return -1;
1169     }
1170
1171     /* quiesce block driver; prevent further io */
1172     bdrv_drain_all();
1173     bdrv_flush(bs);
1174     bdrv_close(bs);
1175
1176     /* if we have a device attached to this BlockDriverState
1177      * then we need to make the drive anonymous until the device
1178      * can be removed.  If this is a drive with no device backing
1179      * then we can just get rid of the block driver state right here.
1180      */
1181     if (bdrv_get_attached_dev(bs)) {
1182         bdrv_make_anon(bs);
1183
1184         /* Further I/O must not pause the guest */
1185         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1186                           BLOCKDEV_ON_ERROR_REPORT);
1187     } else {
1188         drive_uninit(drive_get_by_blockdev(bs));
1189     }
1190
1191     return 0;
1192 }
1193
1194 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1195 {
1196     BlockDriverState *bs;
1197     int ret;
1198
1199     bs = bdrv_find(device);
1200     if (!bs) {
1201         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1202         return;
1203     }
1204
1205     if (size < 0) {
1206         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1207         return;
1208     }
1209
1210     /* complete all in-flight operations before resizing the device */
1211     bdrv_drain_all();
1212
1213     ret = bdrv_truncate(bs, size);
1214     switch (ret) {
1215     case 0:
1216         break;
1217     case -ENOMEDIUM:
1218         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1219         break;
1220     case -ENOTSUP:
1221         error_set(errp, QERR_UNSUPPORTED);
1222         break;
1223     case -EACCES:
1224         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1225         break;
1226     case -EBUSY:
1227         error_set(errp, QERR_DEVICE_IN_USE, device);
1228         break;
1229     default:
1230         error_setg_errno(errp, -ret, "Could not resize");
1231         break;
1232     }
1233 }
1234
1235 static void block_job_cb(void *opaque, int ret)
1236 {
1237     BlockDriverState *bs = opaque;
1238     QObject *obj;
1239
1240     trace_block_job_cb(bs, bs->job, ret);
1241
1242     assert(bs->job);
1243     obj = qobject_from_block_job(bs->job);
1244     if (ret < 0) {
1245         QDict *dict = qobject_to_qdict(obj);
1246         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1247     }
1248
1249     if (block_job_is_cancelled(bs->job)) {
1250         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1251     } else {
1252         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1253     }
1254     qobject_decref(obj);
1255
1256     drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1257 }
1258
1259 void qmp_block_stream(const char *device, bool has_base,
1260                       const char *base, bool has_speed, int64_t speed,
1261                       bool has_on_error, BlockdevOnError on_error,
1262                       Error **errp)
1263 {
1264     BlockDriverState *bs;
1265     BlockDriverState *base_bs = NULL;
1266     Error *local_err = NULL;
1267
1268     if (!has_on_error) {
1269         on_error = BLOCKDEV_ON_ERROR_REPORT;
1270     }
1271
1272     bs = bdrv_find(device);
1273     if (!bs) {
1274         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1275         return;
1276     }
1277
1278     if (base) {
1279         base_bs = bdrv_find_backing_image(bs, base);
1280         if (base_bs == NULL) {
1281             error_set(errp, QERR_BASE_NOT_FOUND, base);
1282             return;
1283         }
1284     }
1285
1286     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1287                  on_error, block_job_cb, bs, &local_err);
1288     if (error_is_set(&local_err)) {
1289         error_propagate(errp, local_err);
1290         return;
1291     }
1292
1293     /* Grab a reference so hotplug does not delete the BlockDriverState from
1294      * underneath us.
1295      */
1296     drive_get_ref(drive_get_by_blockdev(bs));
1297
1298     trace_qmp_block_stream(bs, bs->job);
1299 }
1300
1301 void qmp_block_commit(const char *device,
1302                       bool has_base, const char *base, const char *top,
1303                       bool has_speed, int64_t speed,
1304                       Error **errp)
1305 {
1306     BlockDriverState *bs;
1307     BlockDriverState *base_bs, *top_bs;
1308     Error *local_err = NULL;
1309     /* This will be part of the QMP command, if/when the
1310      * BlockdevOnError change for blkmirror makes it in
1311      */
1312     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1313
1314     /* drain all i/o before commits */
1315     bdrv_drain_all();
1316
1317     bs = bdrv_find(device);
1318     if (!bs) {
1319         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1320         return;
1321     }
1322
1323     /* default top_bs is the active layer */
1324     top_bs = bs;
1325
1326     if (top) {
1327         if (strcmp(bs->filename, top) != 0) {
1328             top_bs = bdrv_find_backing_image(bs, top);
1329         }
1330     }
1331
1332     if (top_bs == NULL) {
1333         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1334         return;
1335     }
1336
1337     if (has_base && base) {
1338         base_bs = bdrv_find_backing_image(top_bs, base);
1339     } else {
1340         base_bs = bdrv_find_base(top_bs);
1341     }
1342
1343     if (base_bs == NULL) {
1344         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1345         return;
1346     }
1347
1348     commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1349                 &local_err);
1350     if (local_err != NULL) {
1351         error_propagate(errp, local_err);
1352         return;
1353     }
1354     /* Grab a reference so hotplug does not delete the BlockDriverState from
1355      * underneath us.
1356      */
1357     drive_get_ref(drive_get_by_blockdev(bs));
1358 }
1359
1360 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1361
1362 void qmp_drive_mirror(const char *device, const char *target,
1363                       bool has_format, const char *format,
1364                       enum MirrorSyncMode sync,
1365                       bool has_mode, enum NewImageMode mode,
1366                       bool has_speed, int64_t speed,
1367                       bool has_granularity, uint32_t granularity,
1368                       bool has_buf_size, int64_t buf_size,
1369                       bool has_on_source_error, BlockdevOnError on_source_error,
1370                       bool has_on_target_error, BlockdevOnError on_target_error,
1371                       Error **errp)
1372 {
1373     BlockDriverState *bs;
1374     BlockDriverState *source, *target_bs;
1375     BlockDriver *proto_drv;
1376     BlockDriver *drv = NULL;
1377     Error *local_err = NULL;
1378     int flags;
1379     uint64_t size;
1380     int ret;
1381
1382     if (!has_speed) {
1383         speed = 0;
1384     }
1385     if (!has_on_source_error) {
1386         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1387     }
1388     if (!has_on_target_error) {
1389         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1390     }
1391     if (!has_mode) {
1392         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1393     }
1394     if (!has_granularity) {
1395         granularity = 0;
1396     }
1397     if (!has_buf_size) {
1398         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1399     }
1400
1401     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1402         error_set(errp, QERR_INVALID_PARAMETER, device);
1403         return;
1404     }
1405     if (granularity & (granularity - 1)) {
1406         error_set(errp, QERR_INVALID_PARAMETER, device);
1407         return;
1408     }
1409
1410     bs = bdrv_find(device);
1411     if (!bs) {
1412         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1413         return;
1414     }
1415
1416     if (!bdrv_is_inserted(bs)) {
1417         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1418         return;
1419     }
1420
1421     if (!has_format) {
1422         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1423     }
1424     if (format) {
1425         drv = bdrv_find_format(format);
1426         if (!drv) {
1427             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1428             return;
1429         }
1430     }
1431
1432     if (bdrv_in_use(bs)) {
1433         error_set(errp, QERR_DEVICE_IN_USE, device);
1434         return;
1435     }
1436
1437     flags = bs->open_flags | BDRV_O_RDWR;
1438     source = bs->backing_hd;
1439     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1440         sync = MIRROR_SYNC_MODE_FULL;
1441     }
1442
1443     proto_drv = bdrv_find_protocol(target);
1444     if (!proto_drv) {
1445         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1446         return;
1447     }
1448
1449     bdrv_get_geometry(bs, &size);
1450     size *= 512;
1451     if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1452         /* create new image w/o backing file */
1453         assert(format && drv);
1454         bdrv_img_create(target, format,
1455                         NULL, NULL, NULL, size, flags, &local_err, false);
1456     } else {
1457         switch (mode) {
1458         case NEW_IMAGE_MODE_EXISTING:
1459             ret = 0;
1460             break;
1461         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1462             /* create new image with backing file */
1463             bdrv_img_create(target, format,
1464                             source->filename,
1465                             source->drv->format_name,
1466                             NULL, size, flags, &local_err, false);
1467             break;
1468         default:
1469             abort();
1470         }
1471     }
1472
1473     if (error_is_set(&local_err)) {
1474         error_propagate(errp, local_err);
1475         return;
1476     }
1477
1478     /* Mirroring takes care of copy-on-write using the source's backing
1479      * file.
1480      */
1481     target_bs = bdrv_new("");
1482     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv);
1483
1484     if (ret < 0) {
1485         bdrv_delete(target_bs);
1486         error_set(errp, QERR_OPEN_FILE_FAILED, target);
1487         return;
1488     }
1489
1490     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1491                  on_source_error, on_target_error,
1492                  block_job_cb, bs, &local_err);
1493     if (local_err != NULL) {
1494         bdrv_delete(target_bs);
1495         error_propagate(errp, local_err);
1496         return;
1497     }
1498
1499     /* Grab a reference so hotplug does not delete the BlockDriverState from
1500      * underneath us.
1501      */
1502     drive_get_ref(drive_get_by_blockdev(bs));
1503 }
1504
1505 static BlockJob *find_block_job(const char *device)
1506 {
1507     BlockDriverState *bs;
1508
1509     bs = bdrv_find(device);
1510     if (!bs || !bs->job) {
1511         return NULL;
1512     }
1513     return bs->job;
1514 }
1515
1516 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1517 {
1518     BlockJob *job = find_block_job(device);
1519
1520     if (!job) {
1521         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1522         return;
1523     }
1524
1525     block_job_set_speed(job, speed, errp);
1526 }
1527
1528 void qmp_block_job_cancel(const char *device,
1529                           bool has_force, bool force, Error **errp)
1530 {
1531     BlockJob *job = find_block_job(device);
1532
1533     if (!has_force) {
1534         force = false;
1535     }
1536
1537     if (!job) {
1538         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1539         return;
1540     }
1541     if (job->paused && !force) {
1542         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1543         return;
1544     }
1545
1546     trace_qmp_block_job_cancel(job);
1547     block_job_cancel(job);
1548 }
1549
1550 void qmp_block_job_pause(const char *device, Error **errp)
1551 {
1552     BlockJob *job = find_block_job(device);
1553
1554     if (!job) {
1555         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1556         return;
1557     }
1558
1559     trace_qmp_block_job_pause(job);
1560     block_job_pause(job);
1561 }
1562
1563 void qmp_block_job_resume(const char *device, Error **errp)
1564 {
1565     BlockJob *job = find_block_job(device);
1566
1567     if (!job) {
1568         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1569         return;
1570     }
1571
1572     trace_qmp_block_job_resume(job);
1573     block_job_resume(job);
1574 }
1575
1576 void qmp_block_job_complete(const char *device, Error **errp)
1577 {
1578     BlockJob *job = find_block_job(device);
1579
1580     if (!job) {
1581         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1582         return;
1583     }
1584
1585     trace_qmp_block_job_complete(job);
1586     block_job_complete(job, errp);
1587 }
1588
1589 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1590 {
1591     BlockJobInfoList **prev = opaque;
1592     BlockJob *job = bs->job;
1593
1594     if (job) {
1595         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1596         elem->value = block_job_query(bs->job);
1597         (*prev)->next = elem;
1598         *prev = elem;
1599     }
1600 }
1601
1602 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1603 {
1604     /* Dummy is a fake list element for holding the head pointer */
1605     BlockJobInfoList dummy = {};
1606     BlockJobInfoList *prev = &dummy;
1607     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1608     return dummy.next;
1609 }
1610
1611 QemuOptsList qemu_common_drive_opts = {
1612     .name = "drive",
1613     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
1614     .desc = {
1615         {
1616             .name = "bus",
1617             .type = QEMU_OPT_NUMBER,
1618             .help = "bus number",
1619         },{
1620             .name = "unit",
1621             .type = QEMU_OPT_NUMBER,
1622             .help = "unit number (i.e. lun for scsi)",
1623         },{
1624             .name = "if",
1625             .type = QEMU_OPT_STRING,
1626             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1627         },{
1628             .name = "index",
1629             .type = QEMU_OPT_NUMBER,
1630             .help = "index number",
1631         },{
1632             .name = "cyls",
1633             .type = QEMU_OPT_NUMBER,
1634             .help = "number of cylinders (ide disk geometry)",
1635         },{
1636             .name = "heads",
1637             .type = QEMU_OPT_NUMBER,
1638             .help = "number of heads (ide disk geometry)",
1639         },{
1640             .name = "secs",
1641             .type = QEMU_OPT_NUMBER,
1642             .help = "number of sectors (ide disk geometry)",
1643         },{
1644             .name = "trans",
1645             .type = QEMU_OPT_STRING,
1646             .help = "chs translation (auto, lba. none)",
1647         },{
1648             .name = "media",
1649             .type = QEMU_OPT_STRING,
1650             .help = "media type (disk, cdrom)",
1651         },{
1652             .name = "snapshot",
1653             .type = QEMU_OPT_BOOL,
1654             .help = "enable/disable snapshot mode",
1655         },{
1656             .name = "file",
1657             .type = QEMU_OPT_STRING,
1658             .help = "disk image",
1659         },{
1660             .name = "discard",
1661             .type = QEMU_OPT_STRING,
1662             .help = "discard operation (ignore/off, unmap/on)",
1663         },{
1664             .name = "cache",
1665             .type = QEMU_OPT_STRING,
1666             .help = "host cache usage (none, writeback, writethrough, "
1667                     "directsync, unsafe)",
1668         },{
1669             .name = "aio",
1670             .type = QEMU_OPT_STRING,
1671             .help = "host AIO implementation (threads, native)",
1672         },{
1673             .name = "format",
1674             .type = QEMU_OPT_STRING,
1675             .help = "disk format (raw, qcow2, ...)",
1676         },{
1677             .name = "serial",
1678             .type = QEMU_OPT_STRING,
1679             .help = "disk serial number",
1680         },{
1681             .name = "rerror",
1682             .type = QEMU_OPT_STRING,
1683             .help = "read error action",
1684         },{
1685             .name = "werror",
1686             .type = QEMU_OPT_STRING,
1687             .help = "write error action",
1688         },{
1689             .name = "addr",
1690             .type = QEMU_OPT_STRING,
1691             .help = "pci address (virtio only)",
1692         },{
1693             .name = "readonly",
1694             .type = QEMU_OPT_BOOL,
1695             .help = "open drive file as read-only",
1696         },{
1697             .name = "iops",
1698             .type = QEMU_OPT_NUMBER,
1699             .help = "limit total I/O operations per second",
1700         },{
1701             .name = "iops_rd",
1702             .type = QEMU_OPT_NUMBER,
1703             .help = "limit read operations per second",
1704         },{
1705             .name = "iops_wr",
1706             .type = QEMU_OPT_NUMBER,
1707             .help = "limit write operations per second",
1708         },{
1709             .name = "bps",
1710             .type = QEMU_OPT_NUMBER,
1711             .help = "limit total bytes per second",
1712         },{
1713             .name = "bps_rd",
1714             .type = QEMU_OPT_NUMBER,
1715             .help = "limit read bytes per second",
1716         },{
1717             .name = "bps_wr",
1718             .type = QEMU_OPT_NUMBER,
1719             .help = "limit write bytes per second",
1720         },{
1721             .name = "copy-on-read",
1722             .type = QEMU_OPT_BOOL,
1723             .help = "copy read data from backing file into image file",
1724         },{
1725             .name = "boot",
1726             .type = QEMU_OPT_BOOL,
1727             .help = "(deprecated, ignored)",
1728         },
1729         { /* end of list */ }
1730     },
1731 };
1732
1733 QemuOptsList qemu_drive_opts = {
1734     .name = "drive",
1735     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1736     .desc = {
1737         {
1738             .name = "bus",
1739             .type = QEMU_OPT_NUMBER,
1740             .help = "bus number",
1741         },{
1742             .name = "unit",
1743             .type = QEMU_OPT_NUMBER,
1744             .help = "unit number (i.e. lun for scsi)",
1745         },{
1746             .name = "if",
1747             .type = QEMU_OPT_STRING,
1748             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1749         },{
1750             .name = "index",
1751             .type = QEMU_OPT_NUMBER,
1752             .help = "index number",
1753         },{
1754             .name = "cyls",
1755             .type = QEMU_OPT_NUMBER,
1756             .help = "number of cylinders (ide disk geometry)",
1757         },{
1758             .name = "heads",
1759             .type = QEMU_OPT_NUMBER,
1760             .help = "number of heads (ide disk geometry)",
1761         },{
1762             .name = "secs",
1763             .type = QEMU_OPT_NUMBER,
1764             .help = "number of sectors (ide disk geometry)",
1765         },{
1766             .name = "trans",
1767             .type = QEMU_OPT_STRING,
1768             .help = "chs translation (auto, lba. none)",
1769         },{
1770             .name = "media",
1771             .type = QEMU_OPT_STRING,
1772             .help = "media type (disk, cdrom)",
1773         },{
1774             .name = "snapshot",
1775             .type = QEMU_OPT_BOOL,
1776             .help = "enable/disable snapshot mode",
1777         },{
1778             .name = "file",
1779             .type = QEMU_OPT_STRING,
1780             .help = "disk image",
1781         },{
1782             .name = "discard",
1783             .type = QEMU_OPT_STRING,
1784             .help = "discard operation (ignore/off, unmap/on)",
1785         },{
1786             .name = "cache",
1787             .type = QEMU_OPT_STRING,
1788             .help = "host cache usage (none, writeback, writethrough, "
1789                     "directsync, unsafe)",
1790         },{
1791             .name = "aio",
1792             .type = QEMU_OPT_STRING,
1793             .help = "host AIO implementation (threads, native)",
1794         },{
1795             .name = "format",
1796             .type = QEMU_OPT_STRING,
1797             .help = "disk format (raw, qcow2, ...)",
1798         },{
1799             .name = "serial",
1800             .type = QEMU_OPT_STRING,
1801             .help = "disk serial number",
1802         },{
1803             .name = "rerror",
1804             .type = QEMU_OPT_STRING,
1805             .help = "read error action",
1806         },{
1807             .name = "werror",
1808             .type = QEMU_OPT_STRING,
1809             .help = "write error action",
1810         },{
1811             .name = "addr",
1812             .type = QEMU_OPT_STRING,
1813             .help = "pci address (virtio only)",
1814         },{
1815             .name = "readonly",
1816             .type = QEMU_OPT_BOOL,
1817             .help = "open drive file as read-only",
1818         },{
1819             .name = "iops",
1820             .type = QEMU_OPT_NUMBER,
1821             .help = "limit total I/O operations per second",
1822         },{
1823             .name = "iops_rd",
1824             .type = QEMU_OPT_NUMBER,
1825             .help = "limit read operations per second",
1826         },{
1827             .name = "iops_wr",
1828             .type = QEMU_OPT_NUMBER,
1829             .help = "limit write operations per second",
1830         },{
1831             .name = "bps",
1832             .type = QEMU_OPT_NUMBER,
1833             .help = "limit total bytes per second",
1834         },{
1835             .name = "bps_rd",
1836             .type = QEMU_OPT_NUMBER,
1837             .help = "limit read bytes per second",
1838         },{
1839             .name = "bps_wr",
1840             .type = QEMU_OPT_NUMBER,
1841             .help = "limit write bytes per second",
1842         },{
1843             .name = "copy-on-read",
1844             .type = QEMU_OPT_BOOL,
1845             .help = "copy read data from backing file into image file",
1846         },{
1847             .name = "boot",
1848             .type = QEMU_OPT_BOOL,
1849             .help = "(deprecated, ignored)",
1850         },
1851         { /* end of list */ }
1852     },
1853 };