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_setg_file_open(errp, -ret, 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     int ret;
1066
1067     ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv);
1068     if (ret < 0) {
1069         error_setg_file_open(errp, -ret, filename);
1070         return;
1071     }
1072
1073     if (bdrv_key_required(bs)) {
1074         if (password) {
1075             if (bdrv_set_key(bs, password) < 0) {
1076                 error_set(errp, QERR_INVALID_PASSWORD);
1077             }
1078         } else {
1079             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1080                       bdrv_get_encrypted_filename(bs));
1081         }
1082     } else if (password) {
1083         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1084     }
1085 }
1086
1087 void qmp_change_blockdev(const char *device, const char *filename,
1088                          bool has_format, const char *format, Error **errp)
1089 {
1090     BlockDriverState *bs;
1091     BlockDriver *drv = NULL;
1092     int bdrv_flags;
1093     Error *err = NULL;
1094
1095     bs = bdrv_find(device);
1096     if (!bs) {
1097         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1098         return;
1099     }
1100
1101     if (format) {
1102         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1103         if (!drv) {
1104             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1105             return;
1106         }
1107     }
1108
1109     eject_device(bs, 0, &err);
1110     if (error_is_set(&err)) {
1111         error_propagate(errp, err);
1112         return;
1113     }
1114
1115     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1116     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1117
1118     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1119 }
1120
1121 /* throttling disk I/O limits */
1122 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1123                                int64_t bps_wr, int64_t iops, int64_t iops_rd,
1124                                int64_t iops_wr, Error **errp)
1125 {
1126     BlockIOLimit io_limits;
1127     BlockDriverState *bs;
1128
1129     bs = bdrv_find(device);
1130     if (!bs) {
1131         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1132         return;
1133     }
1134
1135     io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
1136     io_limits.bps[BLOCK_IO_LIMIT_READ]  = bps_rd;
1137     io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
1138     io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
1139     io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
1140     io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
1141
1142     if (!do_check_io_limits(&io_limits, errp)) {
1143         return;
1144     }
1145
1146     bs->io_limits = io_limits;
1147
1148     if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
1149         bdrv_io_limits_enable(bs);
1150     } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
1151         bdrv_io_limits_disable(bs);
1152     } else {
1153         if (bs->block_timer) {
1154             qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
1155         }
1156     }
1157 }
1158
1159 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1160 {
1161     const char *id = qdict_get_str(qdict, "id");
1162     BlockDriverState *bs;
1163
1164     bs = bdrv_find(id);
1165     if (!bs) {
1166         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1167         return -1;
1168     }
1169     if (bdrv_in_use(bs)) {
1170         qerror_report(QERR_DEVICE_IN_USE, id);
1171         return -1;
1172     }
1173
1174     /* quiesce block driver; prevent further io */
1175     bdrv_drain_all();
1176     bdrv_flush(bs);
1177     bdrv_close(bs);
1178
1179     /* if we have a device attached to this BlockDriverState
1180      * then we need to make the drive anonymous until the device
1181      * can be removed.  If this is a drive with no device backing
1182      * then we can just get rid of the block driver state right here.
1183      */
1184     if (bdrv_get_attached_dev(bs)) {
1185         bdrv_make_anon(bs);
1186
1187         /* Further I/O must not pause the guest */
1188         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1189                           BLOCKDEV_ON_ERROR_REPORT);
1190     } else {
1191         drive_uninit(drive_get_by_blockdev(bs));
1192     }
1193
1194     return 0;
1195 }
1196
1197 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1198 {
1199     BlockDriverState *bs;
1200     int ret;
1201
1202     bs = bdrv_find(device);
1203     if (!bs) {
1204         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1205         return;
1206     }
1207
1208     if (size < 0) {
1209         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1210         return;
1211     }
1212
1213     /* complete all in-flight operations before resizing the device */
1214     bdrv_drain_all();
1215
1216     ret = bdrv_truncate(bs, size);
1217     switch (ret) {
1218     case 0:
1219         break;
1220     case -ENOMEDIUM:
1221         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1222         break;
1223     case -ENOTSUP:
1224         error_set(errp, QERR_UNSUPPORTED);
1225         break;
1226     case -EACCES:
1227         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1228         break;
1229     case -EBUSY:
1230         error_set(errp, QERR_DEVICE_IN_USE, device);
1231         break;
1232     default:
1233         error_setg_errno(errp, -ret, "Could not resize");
1234         break;
1235     }
1236 }
1237
1238 static void block_job_cb(void *opaque, int ret)
1239 {
1240     BlockDriverState *bs = opaque;
1241     QObject *obj;
1242
1243     trace_block_job_cb(bs, bs->job, ret);
1244
1245     assert(bs->job);
1246     obj = qobject_from_block_job(bs->job);
1247     if (ret < 0) {
1248         QDict *dict = qobject_to_qdict(obj);
1249         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1250     }
1251
1252     if (block_job_is_cancelled(bs->job)) {
1253         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1254     } else {
1255         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1256     }
1257     qobject_decref(obj);
1258
1259     drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1260 }
1261
1262 void qmp_block_stream(const char *device, bool has_base,
1263                       const char *base, bool has_speed, int64_t speed,
1264                       bool has_on_error, BlockdevOnError on_error,
1265                       Error **errp)
1266 {
1267     BlockDriverState *bs;
1268     BlockDriverState *base_bs = NULL;
1269     Error *local_err = NULL;
1270
1271     if (!has_on_error) {
1272         on_error = BLOCKDEV_ON_ERROR_REPORT;
1273     }
1274
1275     bs = bdrv_find(device);
1276     if (!bs) {
1277         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1278         return;
1279     }
1280
1281     if (base) {
1282         base_bs = bdrv_find_backing_image(bs, base);
1283         if (base_bs == NULL) {
1284             error_set(errp, QERR_BASE_NOT_FOUND, base);
1285             return;
1286         }
1287     }
1288
1289     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1290                  on_error, block_job_cb, bs, &local_err);
1291     if (error_is_set(&local_err)) {
1292         error_propagate(errp, local_err);
1293         return;
1294     }
1295
1296     /* Grab a reference so hotplug does not delete the BlockDriverState from
1297      * underneath us.
1298      */
1299     drive_get_ref(drive_get_by_blockdev(bs));
1300
1301     trace_qmp_block_stream(bs, bs->job);
1302 }
1303
1304 void qmp_block_commit(const char *device,
1305                       bool has_base, const char *base, const char *top,
1306                       bool has_speed, int64_t speed,
1307                       Error **errp)
1308 {
1309     BlockDriverState *bs;
1310     BlockDriverState *base_bs, *top_bs;
1311     Error *local_err = NULL;
1312     /* This will be part of the QMP command, if/when the
1313      * BlockdevOnError change for blkmirror makes it in
1314      */
1315     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1316
1317     /* drain all i/o before commits */
1318     bdrv_drain_all();
1319
1320     bs = bdrv_find(device);
1321     if (!bs) {
1322         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1323         return;
1324     }
1325
1326     /* default top_bs is the active layer */
1327     top_bs = bs;
1328
1329     if (top) {
1330         if (strcmp(bs->filename, top) != 0) {
1331             top_bs = bdrv_find_backing_image(bs, top);
1332         }
1333     }
1334
1335     if (top_bs == NULL) {
1336         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1337         return;
1338     }
1339
1340     if (has_base && base) {
1341         base_bs = bdrv_find_backing_image(top_bs, base);
1342     } else {
1343         base_bs = bdrv_find_base(top_bs);
1344     }
1345
1346     if (base_bs == NULL) {
1347         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1348         return;
1349     }
1350
1351     commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1352                 &local_err);
1353     if (local_err != NULL) {
1354         error_propagate(errp, local_err);
1355         return;
1356     }
1357     /* Grab a reference so hotplug does not delete the BlockDriverState from
1358      * underneath us.
1359      */
1360     drive_get_ref(drive_get_by_blockdev(bs));
1361 }
1362
1363 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1364
1365 void qmp_drive_mirror(const char *device, const char *target,
1366                       bool has_format, const char *format,
1367                       enum MirrorSyncMode sync,
1368                       bool has_mode, enum NewImageMode mode,
1369                       bool has_speed, int64_t speed,
1370                       bool has_granularity, uint32_t granularity,
1371                       bool has_buf_size, int64_t buf_size,
1372                       bool has_on_source_error, BlockdevOnError on_source_error,
1373                       bool has_on_target_error, BlockdevOnError on_target_error,
1374                       Error **errp)
1375 {
1376     BlockDriverState *bs;
1377     BlockDriverState *source, *target_bs;
1378     BlockDriver *proto_drv;
1379     BlockDriver *drv = NULL;
1380     Error *local_err = NULL;
1381     int flags;
1382     uint64_t size;
1383     int ret;
1384
1385     if (!has_speed) {
1386         speed = 0;
1387     }
1388     if (!has_on_source_error) {
1389         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1390     }
1391     if (!has_on_target_error) {
1392         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1393     }
1394     if (!has_mode) {
1395         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1396     }
1397     if (!has_granularity) {
1398         granularity = 0;
1399     }
1400     if (!has_buf_size) {
1401         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1402     }
1403
1404     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1405         error_set(errp, QERR_INVALID_PARAMETER, device);
1406         return;
1407     }
1408     if (granularity & (granularity - 1)) {
1409         error_set(errp, QERR_INVALID_PARAMETER, device);
1410         return;
1411     }
1412
1413     bs = bdrv_find(device);
1414     if (!bs) {
1415         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1416         return;
1417     }
1418
1419     if (!bdrv_is_inserted(bs)) {
1420         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1421         return;
1422     }
1423
1424     if (!has_format) {
1425         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1426     }
1427     if (format) {
1428         drv = bdrv_find_format(format);
1429         if (!drv) {
1430             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1431             return;
1432         }
1433     }
1434
1435     if (bdrv_in_use(bs)) {
1436         error_set(errp, QERR_DEVICE_IN_USE, device);
1437         return;
1438     }
1439
1440     flags = bs->open_flags | BDRV_O_RDWR;
1441     source = bs->backing_hd;
1442     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1443         sync = MIRROR_SYNC_MODE_FULL;
1444     }
1445
1446     proto_drv = bdrv_find_protocol(target);
1447     if (!proto_drv) {
1448         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1449         return;
1450     }
1451
1452     bdrv_get_geometry(bs, &size);
1453     size *= 512;
1454     if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1455         /* create new image w/o backing file */
1456         assert(format && drv);
1457         bdrv_img_create(target, format,
1458                         NULL, NULL, NULL, size, flags, &local_err, false);
1459     } else {
1460         switch (mode) {
1461         case NEW_IMAGE_MODE_EXISTING:
1462             ret = 0;
1463             break;
1464         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1465             /* create new image with backing file */
1466             bdrv_img_create(target, format,
1467                             source->filename,
1468                             source->drv->format_name,
1469                             NULL, size, flags, &local_err, false);
1470             break;
1471         default:
1472             abort();
1473         }
1474     }
1475
1476     if (error_is_set(&local_err)) {
1477         error_propagate(errp, local_err);
1478         return;
1479     }
1480
1481     /* Mirroring takes care of copy-on-write using the source's backing
1482      * file.
1483      */
1484     target_bs = bdrv_new("");
1485     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv);
1486
1487     if (ret < 0) {
1488         bdrv_delete(target_bs);
1489         error_setg_file_open(errp, -ret, target);
1490         return;
1491     }
1492
1493     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1494                  on_source_error, on_target_error,
1495                  block_job_cb, bs, &local_err);
1496     if (local_err != NULL) {
1497         bdrv_delete(target_bs);
1498         error_propagate(errp, local_err);
1499         return;
1500     }
1501
1502     /* Grab a reference so hotplug does not delete the BlockDriverState from
1503      * underneath us.
1504      */
1505     drive_get_ref(drive_get_by_blockdev(bs));
1506 }
1507
1508 static BlockJob *find_block_job(const char *device)
1509 {
1510     BlockDriverState *bs;
1511
1512     bs = bdrv_find(device);
1513     if (!bs || !bs->job) {
1514         return NULL;
1515     }
1516     return bs->job;
1517 }
1518
1519 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1520 {
1521     BlockJob *job = find_block_job(device);
1522
1523     if (!job) {
1524         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1525         return;
1526     }
1527
1528     block_job_set_speed(job, speed, errp);
1529 }
1530
1531 void qmp_block_job_cancel(const char *device,
1532                           bool has_force, bool force, Error **errp)
1533 {
1534     BlockJob *job = find_block_job(device);
1535
1536     if (!has_force) {
1537         force = false;
1538     }
1539
1540     if (!job) {
1541         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1542         return;
1543     }
1544     if (job->paused && !force) {
1545         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1546         return;
1547     }
1548
1549     trace_qmp_block_job_cancel(job);
1550     block_job_cancel(job);
1551 }
1552
1553 void qmp_block_job_pause(const char *device, Error **errp)
1554 {
1555     BlockJob *job = find_block_job(device);
1556
1557     if (!job) {
1558         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1559         return;
1560     }
1561
1562     trace_qmp_block_job_pause(job);
1563     block_job_pause(job);
1564 }
1565
1566 void qmp_block_job_resume(const char *device, Error **errp)
1567 {
1568     BlockJob *job = find_block_job(device);
1569
1570     if (!job) {
1571         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1572         return;
1573     }
1574
1575     trace_qmp_block_job_resume(job);
1576     block_job_resume(job);
1577 }
1578
1579 void qmp_block_job_complete(const char *device, Error **errp)
1580 {
1581     BlockJob *job = find_block_job(device);
1582
1583     if (!job) {
1584         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1585         return;
1586     }
1587
1588     trace_qmp_block_job_complete(job);
1589     block_job_complete(job, errp);
1590 }
1591
1592 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1593 {
1594     BlockJobInfoList **prev = opaque;
1595     BlockJob *job = bs->job;
1596
1597     if (job) {
1598         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1599         elem->value = block_job_query(bs->job);
1600         (*prev)->next = elem;
1601         *prev = elem;
1602     }
1603 }
1604
1605 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1606 {
1607     /* Dummy is a fake list element for holding the head pointer */
1608     BlockJobInfoList dummy = {};
1609     BlockJobInfoList *prev = &dummy;
1610     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1611     return dummy.next;
1612 }
1613
1614 QemuOptsList qemu_common_drive_opts = {
1615     .name = "drive",
1616     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
1617     .desc = {
1618         {
1619             .name = "bus",
1620             .type = QEMU_OPT_NUMBER,
1621             .help = "bus number",
1622         },{
1623             .name = "unit",
1624             .type = QEMU_OPT_NUMBER,
1625             .help = "unit number (i.e. lun for scsi)",
1626         },{
1627             .name = "if",
1628             .type = QEMU_OPT_STRING,
1629             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1630         },{
1631             .name = "index",
1632             .type = QEMU_OPT_NUMBER,
1633             .help = "index number",
1634         },{
1635             .name = "cyls",
1636             .type = QEMU_OPT_NUMBER,
1637             .help = "number of cylinders (ide disk geometry)",
1638         },{
1639             .name = "heads",
1640             .type = QEMU_OPT_NUMBER,
1641             .help = "number of heads (ide disk geometry)",
1642         },{
1643             .name = "secs",
1644             .type = QEMU_OPT_NUMBER,
1645             .help = "number of sectors (ide disk geometry)",
1646         },{
1647             .name = "trans",
1648             .type = QEMU_OPT_STRING,
1649             .help = "chs translation (auto, lba. none)",
1650         },{
1651             .name = "media",
1652             .type = QEMU_OPT_STRING,
1653             .help = "media type (disk, cdrom)",
1654         },{
1655             .name = "snapshot",
1656             .type = QEMU_OPT_BOOL,
1657             .help = "enable/disable snapshot mode",
1658         },{
1659             .name = "file",
1660             .type = QEMU_OPT_STRING,
1661             .help = "disk image",
1662         },{
1663             .name = "discard",
1664             .type = QEMU_OPT_STRING,
1665             .help = "discard operation (ignore/off, unmap/on)",
1666         },{
1667             .name = "cache",
1668             .type = QEMU_OPT_STRING,
1669             .help = "host cache usage (none, writeback, writethrough, "
1670                     "directsync, unsafe)",
1671         },{
1672             .name = "aio",
1673             .type = QEMU_OPT_STRING,
1674             .help = "host AIO implementation (threads, native)",
1675         },{
1676             .name = "format",
1677             .type = QEMU_OPT_STRING,
1678             .help = "disk format (raw, qcow2, ...)",
1679         },{
1680             .name = "serial",
1681             .type = QEMU_OPT_STRING,
1682             .help = "disk serial number",
1683         },{
1684             .name = "rerror",
1685             .type = QEMU_OPT_STRING,
1686             .help = "read error action",
1687         },{
1688             .name = "werror",
1689             .type = QEMU_OPT_STRING,
1690             .help = "write error action",
1691         },{
1692             .name = "addr",
1693             .type = QEMU_OPT_STRING,
1694             .help = "pci address (virtio only)",
1695         },{
1696             .name = "readonly",
1697             .type = QEMU_OPT_BOOL,
1698             .help = "open drive file as read-only",
1699         },{
1700             .name = "iops",
1701             .type = QEMU_OPT_NUMBER,
1702             .help = "limit total I/O operations per second",
1703         },{
1704             .name = "iops_rd",
1705             .type = QEMU_OPT_NUMBER,
1706             .help = "limit read operations per second",
1707         },{
1708             .name = "iops_wr",
1709             .type = QEMU_OPT_NUMBER,
1710             .help = "limit write operations per second",
1711         },{
1712             .name = "bps",
1713             .type = QEMU_OPT_NUMBER,
1714             .help = "limit total bytes per second",
1715         },{
1716             .name = "bps_rd",
1717             .type = QEMU_OPT_NUMBER,
1718             .help = "limit read bytes per second",
1719         },{
1720             .name = "bps_wr",
1721             .type = QEMU_OPT_NUMBER,
1722             .help = "limit write bytes per second",
1723         },{
1724             .name = "copy-on-read",
1725             .type = QEMU_OPT_BOOL,
1726             .help = "copy read data from backing file into image file",
1727         },{
1728             .name = "boot",
1729             .type = QEMU_OPT_BOOL,
1730             .help = "(deprecated, ignored)",
1731         },
1732         { /* end of list */ }
1733     },
1734 };
1735
1736 QemuOptsList qemu_drive_opts = {
1737     .name = "drive",
1738     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1739     .desc = {
1740         {
1741             .name = "bus",
1742             .type = QEMU_OPT_NUMBER,
1743             .help = "bus number",
1744         },{
1745             .name = "unit",
1746             .type = QEMU_OPT_NUMBER,
1747             .help = "unit number (i.e. lun for scsi)",
1748         },{
1749             .name = "if",
1750             .type = QEMU_OPT_STRING,
1751             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1752         },{
1753             .name = "index",
1754             .type = QEMU_OPT_NUMBER,
1755             .help = "index number",
1756         },{
1757             .name = "cyls",
1758             .type = QEMU_OPT_NUMBER,
1759             .help = "number of cylinders (ide disk geometry)",
1760         },{
1761             .name = "heads",
1762             .type = QEMU_OPT_NUMBER,
1763             .help = "number of heads (ide disk geometry)",
1764         },{
1765             .name = "secs",
1766             .type = QEMU_OPT_NUMBER,
1767             .help = "number of sectors (ide disk geometry)",
1768         },{
1769             .name = "trans",
1770             .type = QEMU_OPT_STRING,
1771             .help = "chs translation (auto, lba. none)",
1772         },{
1773             .name = "media",
1774             .type = QEMU_OPT_STRING,
1775             .help = "media type (disk, cdrom)",
1776         },{
1777             .name = "snapshot",
1778             .type = QEMU_OPT_BOOL,
1779             .help = "enable/disable snapshot mode",
1780         },{
1781             .name = "file",
1782             .type = QEMU_OPT_STRING,
1783             .help = "disk image",
1784         },{
1785             .name = "discard",
1786             .type = QEMU_OPT_STRING,
1787             .help = "discard operation (ignore/off, unmap/on)",
1788         },{
1789             .name = "cache",
1790             .type = QEMU_OPT_STRING,
1791             .help = "host cache usage (none, writeback, writethrough, "
1792                     "directsync, unsafe)",
1793         },{
1794             .name = "aio",
1795             .type = QEMU_OPT_STRING,
1796             .help = "host AIO implementation (threads, native)",
1797         },{
1798             .name = "format",
1799             .type = QEMU_OPT_STRING,
1800             .help = "disk format (raw, qcow2, ...)",
1801         },{
1802             .name = "serial",
1803             .type = QEMU_OPT_STRING,
1804             .help = "disk serial number",
1805         },{
1806             .name = "rerror",
1807             .type = QEMU_OPT_STRING,
1808             .help = "read error action",
1809         },{
1810             .name = "werror",
1811             .type = QEMU_OPT_STRING,
1812             .help = "write error action",
1813         },{
1814             .name = "addr",
1815             .type = QEMU_OPT_STRING,
1816             .help = "pci address (virtio only)",
1817         },{
1818             .name = "readonly",
1819             .type = QEMU_OPT_BOOL,
1820             .help = "open drive file as read-only",
1821         },{
1822             .name = "iops",
1823             .type = QEMU_OPT_NUMBER,
1824             .help = "limit total I/O operations per second",
1825         },{
1826             .name = "iops_rd",
1827             .type = QEMU_OPT_NUMBER,
1828             .help = "limit read operations per second",
1829         },{
1830             .name = "iops_wr",
1831             .type = QEMU_OPT_NUMBER,
1832             .help = "limit write operations per second",
1833         },{
1834             .name = "bps",
1835             .type = QEMU_OPT_NUMBER,
1836             .help = "limit total bytes per second",
1837         },{
1838             .name = "bps_rd",
1839             .type = QEMU_OPT_NUMBER,
1840             .help = "limit read bytes per second",
1841         },{
1842             .name = "bps_wr",
1843             .type = QEMU_OPT_NUMBER,
1844             .help = "limit write bytes per second",
1845         },{
1846             .name = "copy-on-read",
1847             .type = QEMU_OPT_BOOL,
1848             .help = "copy read data from backing file into image file",
1849         },{
1850             .name = "boot",
1851             .type = QEMU_OPT_BOOL,
1852             .help = "(deprecated, ignored)",
1853         },
1854         { /* end of list */ }
1855     },
1856 };