blockdev.c

   1 /*
   2  * QEMU host block devices
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * This work is licensed under the terms of the GNU GPL, version 2 or
   7  * later.  See the COPYING file in the top-level directory.
   8  *
   9  * This file incorporates work covered by the following copyright and
  10  * permission notice:
  11  *
  12  * Copyright (c) 2003-2008 Fabrice Bellard
  13  *
  14  * Permission is hereby granted, free of charge, to any person obtaining a copy
  15  * of this software and associated documentation files (the "Software"), to deal
  16  * in the Software without restriction, including without limitation the rights
  17  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  18  * copies of the Software, and to permit persons to whom the Software is
  19  * furnished to do so, subject to the following conditions:
  20  *
  21  * The above copyright notice and this permission notice shall be included in
  22  * all copies or substantial portions of the Software.
  23  *
  24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  25  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  26  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  27  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  28  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  29  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  30  * THE SOFTWARE.
  31  */
  32
  33 #include "sysemu/blockdev.h"
  34 #include "hw/block/block.h"
  35 #include "block/blockjob.h"
  36 #include "monitor/monitor.h"
  37 #include "qapi/qmp/qerror.h"
  38 #include "qemu/option.h"
  39 #include "qemu/config-file.h"
  40 #include "qapi/qmp/types.h"
  41 #include "sysemu/sysemu.h"
  42 #include "block/block_int.h"
  43 #include "qmp-commands.h"
  44 #include "trace.h"
  45 #include "sysemu/arch_init.h"
  46
  47 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
  48 extern QemuOptsList qemu_common_drive_opts;
  49
  50 static const char *const if_name[IF_COUNT] = {
  51     [IF_NONE] = "none",
  52     [IF_IDE] = "ide",
  53     [IF_SCSI] = "scsi",
  54     [IF_FLOPPY] = "floppy",
  55     [IF_PFLASH] = "pflash",
  56     [IF_MTD] = "mtd",
  57     [IF_SD] = "sd",
  58     [IF_VIRTIO] = "virtio",
  59     [IF_XEN] = "xen",
  60 };
  61
  62 static const int if_max_devs[IF_COUNT] = {
  63     /*
  64      * Do not change these numbers!  They govern how drive option
  65      * index maps to unit and bus.  That mapping is ABI.
  66      *
  67      * All controllers used to imlement if=T drives need to support
  68      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
  69      * Otherwise, some index values map to "impossible" bus, unit
  70      * values.
  71      *
  72      * For instance, if you change [IF_SCSI] to 255, -drive
  73      * if=scsi,index=12 no longer means bus=1,unit=5, but
  74      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
  75      * the drive can't be set up.  Regression.
  76      */
  77     [IF_IDE] = 2,
  78     [IF_SCSI] = 7,
  79 };
  80
  81 /*
  82  * We automatically delete the drive when a device using it gets
  83  * unplugged.  Questionable feature, but we can't just drop it.
  84  * Device models call blockdev_mark_auto_del() to schedule the
  85  * automatic deletion, and generic qdev code calls blockdev_auto_del()
  86  * when deletion is actually safe.
  87  */
  88 void blockdev_mark_auto_del(BlockDriverState *bs)
  89 {
  90     DriveInfo *dinfo = drive_get_by_blockdev(bs);
  91
  92     if (bs->job) {
  93         block_job_cancel(bs->job);
  94     }
  95     if (dinfo) {
  96         dinfo->auto_del = 1;
  97     }
  98 }
  99
 100 void blockdev_auto_del(BlockDriverState *bs)
 101 {
 102     DriveInfo *dinfo = drive_get_by_blockdev(bs);
 103
 104     if (dinfo && dinfo->auto_del) {
 105         drive_put_ref(dinfo);
 106     }
 107 }
 108
 109 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
 110 {
 111     int max_devs = if_max_devs[type];
 112     return max_devs ? index / max_devs : 0;
 113 }
 114
 115 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
 116 {
 117     int max_devs = if_max_devs[type];
 118     return max_devs ? index % max_devs : index;
 119 }
 120
 121 QemuOpts *drive_def(const char *optstr)
 122 {
 123     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
 124 }
 125
 126 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
 127                     const char *optstr)
 128 {
 129     QemuOpts *opts;
 130     char buf[32];
 131
 132     opts = drive_def(optstr);
 133     if (!opts) {
 134         return NULL;
 135     }
 136     if (type != IF_DEFAULT) {
 137         qemu_opt_set(opts, "if", if_name[type]);
 138     }
 139     if (index >= 0) {
 140         snprintf(buf, sizeof(buf), "%d", index);
 141         qemu_opt_set(opts, "index", buf);
 142     }
 143     if (file)
 144         qemu_opt_set(opts, "file", file);
 145     return opts;
 146 }
 147
 148 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
 149 {
 150     DriveInfo *dinfo;
 151
 152     /* seek interface, bus and unit */
 153
 154     QTAILQ_FOREACH(dinfo, &drives, next) {
 155         if (dinfo->type == type &&
 156             dinfo->bus == bus &&
 157             dinfo->unit == unit)
 158             return dinfo;
 159     }
 160
 161     return NULL;
 162 }
 163
 164 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
 165 {
 166     return drive_get(type,
 167                      drive_index_to_bus_id(type, index),
 168                      drive_index_to_unit_id(type, index));
 169 }
 170
 171 int drive_get_max_bus(BlockInterfaceType type)
 172 {
 173     int max_bus;
 174     DriveInfo *dinfo;
 175
 176     max_bus = -1;
 177     QTAILQ_FOREACH(dinfo, &drives, next) {
 178         if(dinfo->type == type &&
 179            dinfo->bus > max_bus)
 180             max_bus = dinfo->bus;
 181     }
 182     return max_bus;
 183 }
 184
 185 /* Get a block device.  This should only be used for single-drive devices
 186    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
 187    appropriate bus.  */
 188 DriveInfo *drive_get_next(BlockInterfaceType type)
 189 {
 190     static int next_block_unit[IF_COUNT];
 191
 192     return drive_get(type, 0, next_block_unit[type]++);
 193 }
 194
 195 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
 196 {
 197     DriveInfo *dinfo;
 198
 199     QTAILQ_FOREACH(dinfo, &drives, next) {
 200         if (dinfo->bdrv == bs) {
 201             return dinfo;
 202         }
 203     }
 204     return NULL;
 205 }
 206
 207 static void bdrv_format_print(void *opaque, const char *name)
 208 {
 209     error_printf(" %s", name);
 210 }
 211
 212 static void drive_uninit(DriveInfo *dinfo)
 213 {
 214     qemu_opts_del(dinfo->opts);
 215     bdrv_delete(dinfo->bdrv);
 216     g_free(dinfo->id);
 217     QTAILQ_REMOVE(&drives, dinfo, next);
 218     g_free(dinfo->serial);
 219     g_free(dinfo);
 220 }
 221
 222 void drive_put_ref(DriveInfo *dinfo)
 223 {
 224     assert(dinfo->refcount);
 225     if (--dinfo->refcount == 0) {
 226         drive_uninit(dinfo);
 227     }
 228 }
 229
 230 void drive_get_ref(DriveInfo *dinfo)
 231 {
 232     dinfo->refcount++;
 233 }
 234
 235 typedef struct {
 236     QEMUBH *bh;
 237     DriveInfo *dinfo;
 238 } DrivePutRefBH;
 239
 240 static void drive_put_ref_bh(void *opaque)
 241 {
 242     DrivePutRefBH *s = opaque;
 243
 244     drive_put_ref(s->dinfo);
 245     qemu_bh_delete(s->bh);
 246     g_free(s);
 247 }
 248
 249 /*
 250  * Release a drive reference in a BH
 251  *
 252  * It is not possible to use drive_put_ref() from a callback function when the
 253  * callers still need the drive.  In such cases we schedule a BH to release the
 254  * reference.
 255  */
 256 static void drive_put_ref_bh_schedule(DriveInfo *dinfo)
 257 {
 258     DrivePutRefBH *s;
 259
 260     s = g_new(DrivePutRefBH, 1);
 261     s->bh = qemu_bh_new(drive_put_ref_bh, s);
 262     s->dinfo = dinfo;
 263     qemu_bh_schedule(s->bh);
 264 }
 265
 266 static int parse_block_error_action(const char *buf, bool is_read)
 267 {
 268     if (!strcmp(buf, "ignore")) {
 269         return BLOCKDEV_ON_ERROR_IGNORE;
 270     } else if (!is_read && !strcmp(buf, "enospc")) {
 271         return BLOCKDEV_ON_ERROR_ENOSPC;
 272     } else if (!strcmp(buf, "stop")) {
 273         return BLOCKDEV_ON_ERROR_STOP;
 274     } else if (!strcmp(buf, "report")) {
 275         return BLOCKDEV_ON_ERROR_REPORT;
 276     } else {
 277         error_report("'%s' invalid %s error action",
 278                      buf, is_read ? "read" : "write");
 279         return -1;
 280     }
 281 }
 282
 283 static bool do_check_io_limits(BlockIOLimit *io_limits, Error **errp)
 284 {
 285     bool bps_flag;
 286     bool iops_flag;
 287
 288     assert(io_limits);
 289
 290     bps_flag  = (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] != 0)
 291                  && ((io_limits->bps[BLOCK_IO_LIMIT_READ] != 0)
 292                  || (io_limits->bps[BLOCK_IO_LIMIT_WRITE] != 0));
 293     iops_flag = (io_limits->iops[BLOCK_IO_LIMIT_TOTAL] != 0)
 294                  && ((io_limits->iops[BLOCK_IO_LIMIT_READ] != 0)
 295                  || (io_limits->iops[BLOCK_IO_LIMIT_WRITE] != 0));
 296     if (bps_flag || iops_flag) {
 297         error_setg(errp, "bps(iops) and bps_rd/bps_wr(iops_rd/iops_wr) "
 298                          "cannot be used at the same time");
 299         return false;
 300     }
 301
 302     if (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] < 0 ||
 303         io_limits->bps[BLOCK_IO_LIMIT_WRITE] < 0 ||
 304         io_limits->bps[BLOCK_IO_LIMIT_READ] < 0 ||
 305         io_limits->iops[BLOCK_IO_LIMIT_TOTAL] < 0 ||
 306         io_limits->iops[BLOCK_IO_LIMIT_WRITE] < 0 ||
 307         io_limits->iops[BLOCK_IO_LIMIT_READ] < 0) {
 308         error_setg(errp, "bps and iops values must be 0 or greater");
 309         return false;
 310     }
 311
 312     return true;
 313 }
 314
 315 static DriveInfo *blockdev_init(QemuOpts *all_opts,
 316                                 BlockInterfaceType block_default_type)
 317 {
 318     const char *buf;
 319     const char *file = NULL;
 320     const char *serial;
 321     const char *mediastr = "";
 322     BlockInterfaceType type;
 323     enum { MEDIA_DISK, MEDIA_CDROM } media;
 324     int bus_id, unit_id;
 325     int cyls, heads, secs, translation;
 326     int max_devs;
 327     int index;
 328     int ro = 0;
 329     int bdrv_flags = 0;
 330     int on_read_error, on_write_error;
 331     const char *devaddr;
 332     DriveInfo *dinfo;
 333     BlockIOLimit io_limits;
 334     int snapshot = 0;
 335     bool copy_on_read;
 336     int ret;
 337     Error *error = NULL;
 338     QemuOpts *opts;
 339     QDict *bs_opts;
 340     const char *id;
 341     bool has_driver_specific_opts;
 342     BlockDriver *drv = NULL;
 343
 344     translation = BIOS_ATA_TRANSLATION_AUTO;
 345     media = MEDIA_DISK;
 346
 347     /* Check common options by copying from all_opts to opts, all other options
 348      * are stored in bs_opts. */
 349     id = qemu_opts_id(all_opts);
 350     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
 351     if (error_is_set(&error)) {
 352         qerror_report_err(error);
 353         error_free(error);
 354         return NULL;
 355     }
 356
 357     bs_opts = qdict_new();
 358     qemu_opts_to_qdict(all_opts, bs_opts);
 359     qemu_opts_absorb_qdict(opts, bs_opts, &error);
 360     if (error_is_set(&error)) {
 361         qerror_report_err(error);
 362         error_free(error);
 363         return NULL;
 364     }
 365
 366     if (id) {
 367         qdict_del(bs_opts, "id");
 368     }
 369
 370     has_driver_specific_opts = !!qdict_size(bs_opts);
 371
 372     /* extract parameters */
 373     bus_id  = qemu_opt_get_number(opts, "bus", 0);
 374     unit_id = qemu_opt_get_number(opts, "unit", -1);
 375     index   = qemu_opt_get_number(opts, "index", -1);
 376
 377     cyls  = qemu_opt_get_number(opts, "cyls", 0);
 378     heads = qemu_opt_get_number(opts, "heads", 0);
 379     secs  = qemu_opt_get_number(opts, "secs", 0);
 380
 381     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 382     ro = qemu_opt_get_bool(opts, "read-only", 0);
 383     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
 384
 385     file = qemu_opt_get(opts, "file");
 386     serial = qemu_opt_get(opts, "serial");
 387
 388     if ((buf = qemu_opt_get(opts, "if")) != NULL) {
 389         for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
 390             ;
 391         if (type == IF_COUNT) {
 392             error_report("unsupported bus type '%s'", buf);
 393             return NULL;
 394         }
 395     } else {
 396         type = block_default_type;
 397     }
 398
 399     max_devs = if_max_devs[type];
 400
 401     if (cyls || heads || secs) {
 402         if (cyls < 1) {
 403             error_report("invalid physical cyls number");
 404             return NULL;
 405         }
 406         if (heads < 1) {
 407             error_report("invalid physical heads number");
 408             return NULL;
 409         }
 410         if (secs < 1) {
 411             error_report("invalid physical secs number");
 412             return NULL;
 413         }
 414     }
 415
 416     if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
 417         if (!cyls) {
 418             error_report("'%s' trans must be used with cyls, heads and secs",
 419                          buf);
 420             return NULL;
 421         }
 422         if (!strcmp(buf, "none"))
 423             translation = BIOS_ATA_TRANSLATION_NONE;
 424         else if (!strcmp(buf, "lba"))
 425             translation = BIOS_ATA_TRANSLATION_LBA;
 426         else if (!strcmp(buf, "auto"))
 427             translation = BIOS_ATA_TRANSLATION_AUTO;
 428         else {
 429             error_report("'%s' invalid translation type", buf);
 430             return NULL;
 431         }
 432     }
 433
 434     if ((buf = qemu_opt_get(opts, "media")) != NULL) {
 435         if (!strcmp(buf, "disk")) {
 436             media = MEDIA_DISK;
 437         } else if (!strcmp(buf, "cdrom")) {
 438             if (cyls || secs || heads) {
 439                 error_report("CHS can't be set with media=%s", buf);
 440                 return NULL;
 441             }
 442             media = MEDIA_CDROM;
 443         } else {
 444             error_report("'%s' invalid media", buf);
 445             return NULL;
 446         }
 447     }
 448
 449     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
 450         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
 451             error_report("invalid discard option");
 452             return NULL;
 453         }
 454     }
 455
 456     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
 457         bdrv_flags |= BDRV_O_CACHE_WB;
 458     }
 459     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
 460         bdrv_flags |= BDRV_O_NOCACHE;
 461     }
 462     if (qemu_opt_get_bool(opts, "cache.no-flush", true)) {
 463         bdrv_flags |= BDRV_O_NO_FLUSH;
 464     }
 465
 466 #ifdef CONFIG_LINUX_AIO
 467     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 468         if (!strcmp(buf, "native")) {
 469             bdrv_flags |= BDRV_O_NATIVE_AIO;
 470         } else if (!strcmp(buf, "threads")) {
 471             /* this is the default */
 472         } else {
 473            error_report("invalid aio option");
 474            return NULL;
 475         }
 476     }
 477 #endif
 478
 479     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 480         if (is_help_option(buf)) {
 481             error_printf("Supported formats:");
 482             bdrv_iterate_format(bdrv_format_print, NULL);
 483             error_printf("\n");
 484             return NULL;
 485         }
 486
 487         drv = bdrv_find_whitelisted_format(buf, ro);
 488         if (!drv) {
 489             if (!ro && bdrv_find_whitelisted_format(buf, !ro)) {
 490                 error_report("'%s' can be only used as read-only device.", buf);
 491             } else {
 492                 error_report("'%s' invalid format", buf);
 493             }
 494             return NULL;
 495         }
 496     }
 497
 498     /* disk I/O throttling */
 499     io_limits.bps[BLOCK_IO_LIMIT_TOTAL]  =
 500         qemu_opt_get_number(opts, "throttling.bps-total", 0);
 501     io_limits.bps[BLOCK_IO_LIMIT_READ]   =
 502         qemu_opt_get_number(opts, "throttling.bps-read", 0);
 503     io_limits.bps[BLOCK_IO_LIMIT_WRITE]  =
 504         qemu_opt_get_number(opts, "throttling.bps-write", 0);
 505     io_limits.iops[BLOCK_IO_LIMIT_TOTAL] =
 506         qemu_opt_get_number(opts, "throttling.iops-total", 0);
 507     io_limits.iops[BLOCK_IO_LIMIT_READ]  =
 508         qemu_opt_get_number(opts, "throttling.iops-read", 0);
 509     io_limits.iops[BLOCK_IO_LIMIT_WRITE] =
 510         qemu_opt_get_number(opts, "throttling.iops-write", 0);
 511
 512     if (!do_check_io_limits(&io_limits, &error)) {
 513         error_report("%s", error_get_pretty(error));
 514         error_free(error);
 515         return NULL;
 516     }
 517
 518     if (qemu_opt_get(opts, "boot") != NULL) {
 519         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 520                 "ignored. Future versions will reject this parameter. Please "
 521                 "update your scripts.\n");
 522     }
 523
 524     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
 525     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 526         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
 527             error_report("werror is not supported by this bus type");
 528             return NULL;
 529         }
 530
 531         on_write_error = parse_block_error_action(buf, 0);
 532         if (on_write_error < 0) {
 533             return NULL;
 534         }
 535     }
 536
 537     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
 538     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 539         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
 540             error_report("rerror is not supported by this bus type");
 541             return NULL;
 542         }
 543
 544         on_read_error = parse_block_error_action(buf, 1);
 545         if (on_read_error < 0) {
 546             return NULL;
 547         }
 548     }
 549
 550     if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
 551         if (type != IF_VIRTIO) {
 552             error_report("addr is not supported by this bus type");
 553             return NULL;
 554         }
 555     }
 556
 557     /* compute bus and unit according index */
 558
 559     if (index != -1) {
 560         if (bus_id != 0 || unit_id != -1) {
 561             error_report("index cannot be used with bus and unit");
 562             return NULL;
 563         }
 564         bus_id = drive_index_to_bus_id(type, index);
 565         unit_id = drive_index_to_unit_id(type, index);
 566     }
 567
 568     /* if user doesn't specify a unit_id,
 569      * try to find the first free
 570      */
 571
 572     if (unit_id == -1) {
 573        unit_id = 0;
 574        while (drive_get(type, bus_id, unit_id) != NULL) {
 575            unit_id++;
 576            if (max_devs && unit_id >= max_devs) {
 577                unit_id -= max_devs;
 578                bus_id++;
 579            }
 580        }
 581     }
 582
 583     /* check unit id */
 584
 585     if (max_devs && unit_id >= max_devs) {
 586         error_report("unit %d too big (max is %d)",
 587                      unit_id, max_devs - 1);
 588         return NULL;
 589     }
 590
 591     /*
 592      * catch multiple definitions
 593      */
 594
 595     if (drive_get(type, bus_id, unit_id) != NULL) {
 596         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 597                      bus_id, unit_id, index);
 598         return NULL;
 599     }
 600
 601     /* init */
 602
 603     dinfo = g_malloc0(sizeof(*dinfo));
 604     if ((buf = qemu_opts_id(opts)) != NULL) {
 605         dinfo->id = g_strdup(buf);
 606     } else {
 607         /* no id supplied -> create one */
 608         dinfo->id = g_malloc0(32);
 609         if (type == IF_IDE || type == IF_SCSI)
 610             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 611         if (max_devs)
 612             snprintf(dinfo->id, 32, "%s%i%s%i",
 613                      if_name[type], bus_id, mediastr, unit_id);
 614         else
 615             snprintf(dinfo->id, 32, "%s%s%i",
 616                      if_name[type], mediastr, unit_id);
 617     }
 618     dinfo->bdrv = bdrv_new(dinfo->id);
 619     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
 620     dinfo->bdrv->read_only = ro;
 621     dinfo->devaddr = devaddr;
 622     dinfo->type = type;
 623     dinfo->bus = bus_id;
 624     dinfo->unit = unit_id;
 625     dinfo->cyls = cyls;
 626     dinfo->heads = heads;
 627     dinfo->secs = secs;
 628     dinfo->trans = translation;
 629     dinfo->opts = all_opts;
 630     dinfo->refcount = 1;
 631     if (serial != NULL) {
 632         dinfo->serial = g_strdup(serial);
 633     }
 634     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 635
 636     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
 637
 638     /* disk I/O throttling */
 639     bdrv_set_io_limits(dinfo->bdrv, &io_limits);
 640
 641     switch(type) {
 642     case IF_IDE:
 643     case IF_SCSI:
 644     case IF_XEN:
 645     case IF_NONE:
 646         dinfo->media_cd = media == MEDIA_CDROM;
 647         break;
 648     case IF_SD:
 649     case IF_FLOPPY:
 650     case IF_PFLASH:
 651     case IF_MTD:
 652         break;
 653     case IF_VIRTIO:
 654     {
 655         /* add virtio block device */
 656         QemuOpts *devopts;
 657         devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
 658         if (arch_type == QEMU_ARCH_S390X) {
 659             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
 660         } else {
 661             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
 662         }
 663         qemu_opt_set(devopts, "drive", dinfo->id);
 664         if (devaddr)
 665             qemu_opt_set(devopts, "addr", devaddr);
 666         break;
 667     }
 668     default:
 669         abort();
 670     }
 671     if (!file || !*file) {
 672         if (has_driver_specific_opts) {
 673             file = NULL;
 674         } else {
 675             return dinfo;
 676         }
 677     }
 678     if (snapshot) {
 679         /* always use cache=unsafe with snapshot */
 680         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 681         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 682     }
 683
 684     if (copy_on_read) {
 685         bdrv_flags |= BDRV_O_COPY_ON_READ;
 686     }
 687
 688     if (runstate_check(RUN_STATE_INMIGRATE)) {
 689         bdrv_flags |= BDRV_O_INCOMING;
 690     }
 691
 692     if (media == MEDIA_CDROM) {
 693         /* CDROM is fine for any interface, don't check.  */
 694         ro = 1;
 695     } else if (ro == 1) {
 696         if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
 697             type != IF_NONE && type != IF_PFLASH) {
 698             error_report("read-only not supported by this bus type");
 699             goto err;
 700         }
 701     }
 702
 703     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 704
 705     if (ro && copy_on_read) {
 706         error_report("warning: disabling copy_on_read on read-only drive");
 707     }
 708
 709     QINCREF(bs_opts);
 710     ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv);
 711
 712     if (ret < 0) {
 713         if (ret == -EMEDIUMTYPE) {
 714             error_report("could not open disk image %s: not in %s format",
 715                          file ?: dinfo->id, drv ? drv->format_name :
 716                          qdict_get_str(bs_opts, "driver"));
 717         } else {
 718             error_report("could not open disk image %s: %s",
 719                          file ?: dinfo->id, strerror(-ret));
 720         }
 721         goto err;
 722     }
 723
 724     if (bdrv_key_required(dinfo->bdrv))
 725         autostart = 0;
 726
 727     QDECREF(bs_opts);
 728     qemu_opts_del(opts);
 729
 730     return dinfo;
 731
 732 err:
 733     qemu_opts_del(opts);
 734     QDECREF(bs_opts);
 735     bdrv_delete(dinfo->bdrv);
 736     g_free(dinfo->id);
 737     QTAILQ_REMOVE(&drives, dinfo, next);
 738     g_free(dinfo);
 739     return NULL;
 740 }
 741
 742 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
 743 {
 744     const char *value;
 745
 746     value = qemu_opt_get(opts, from);
 747     if (value) {
 748         qemu_opt_set(opts, to, value);
 749         qemu_opt_unset(opts, from);
 750     }
 751 }
 752
 753 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
 754 {
 755     const char *value;
 756
 757     /* Change legacy command line options into QMP ones */
 758     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
 759     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
 760     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
 761
 762     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
 763     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
 764     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
 765
 766     qemu_opt_rename(all_opts, "readonly", "read-only");
 767
 768     value = qemu_opt_get(all_opts, "cache");
 769     if (value) {
 770         int flags = 0;
 771
 772         if (bdrv_parse_cache_flags(value, &flags) != 0) {
 773             error_report("invalid cache option");
 774             return NULL;
 775         }
 776
 777         /* Specific options take precedence */
 778         if (!qemu_opt_get(all_opts, "cache.writeback")) {
 779             qemu_opt_set_bool(all_opts, "cache.writeback",
 780                               !!(flags & BDRV_O_CACHE_WB));
 781         }
 782         if (!qemu_opt_get(all_opts, "cache.direct")) {
 783             qemu_opt_set_bool(all_opts, "cache.direct",
 784                               !!(flags & BDRV_O_NOCACHE));
 785         }
 786         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
 787             qemu_opt_set_bool(all_opts, "cache.no-flush",
 788                               !!(flags & BDRV_O_NO_FLUSH));
 789         }
 790         qemu_opt_unset(all_opts, "cache");
 791     }
 792
 793     return blockdev_init(all_opts, block_default_type);
 794 }
 795
 796 void do_commit(Monitor *mon, const QDict *qdict)
 797 {
 798     const char *device = qdict_get_str(qdict, "device");
 799     BlockDriverState *bs;
 800     int ret;
 801
 802     if (!strcmp(device, "all")) {
 803         ret = bdrv_commit_all();
 804     } else {
 805         bs = bdrv_find(device);
 806         if (!bs) {
 807             monitor_printf(mon, "Device '%s' not found\n", device);
 808             return;
 809         }
 810         ret = bdrv_commit(bs);
 811     }
 812     if (ret < 0) {
 813         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
 814                        strerror(-ret));
 815     }
 816 }
 817
 818 static void blockdev_do_action(int kind, void *data, Error **errp)
 819 {
 820     TransactionAction action;
 821     TransactionActionList list;
 822
 823     action.kind = kind;
 824     action.data = data;
 825     list.value = &action;
 826     list.next = NULL;
 827     qmp_transaction(&list, errp);
 828 }
 829
 830 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
 831                                 bool has_format, const char *format,
 832                                 bool has_mode, enum NewImageMode mode,
 833                                 Error **errp)
 834 {
 835     BlockdevSnapshot snapshot = {
 836         .device = (char *) device,
 837         .snapshot_file = (char *) snapshot_file,
 838         .has_format = has_format,
 839         .format = (char *) format,
 840         .has_mode = has_mode,
 841         .mode = mode,
 842     };
 843     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
 844                        &snapshot, errp);
 845 }
 846
 847
 848 /* New and old BlockDriverState structs for group snapshots */
 849
 850 typedef struct BlkTransactionState BlkTransactionState;
 851
 852 /* Only prepare() may fail. In a single transaction, only one of commit() or
 853    abort() will be called, clean() will always be called if it present. */
 854 typedef struct BdrvActionOps {
 855     /* Size of state struct, in bytes. */
 856     size_t instance_size;
 857     /* Prepare the work, must NOT be NULL. */
 858     void (*prepare)(BlkTransactionState *common, Error **errp);
 859     /* Commit the changes, can be NULL. */
 860     void (*commit)(BlkTransactionState *common);
 861     /* Abort the changes on fail, can be NULL. */
 862     void (*abort)(BlkTransactionState *common);
 863     /* Clean up resource in the end, can be NULL. */
 864     void (*clean)(BlkTransactionState *common);
 865 } BdrvActionOps;
 866
 867 /*
 868  * This structure must be arranged as first member in child type, assuming
 869  * that compiler will also arrange it to the same address with parent instance.
 870  * Later it will be used in free().
 871  */
 872 struct BlkTransactionState {
 873     TransactionAction *action;
 874     const BdrvActionOps *ops;
 875     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
 876 };
 877
 878 /* external snapshot private data */
 879 typedef struct ExternalSnapshotState {
 880     BlkTransactionState common;
 881     BlockDriverState *old_bs;
 882     BlockDriverState *new_bs;
 883 } ExternalSnapshotState;
 884
 885 static void external_snapshot_prepare(BlkTransactionState *common,
 886                                       Error **errp)
 887 {
 888     BlockDriver *drv;
 889     int flags, ret;
 890     Error *local_err = NULL;
 891     const char *device;
 892     const char *new_image_file;
 893     const char *format = "qcow2";
 894     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
 895     ExternalSnapshotState *state =
 896                              DO_UPCAST(ExternalSnapshotState, common, common);
 897     TransactionAction *action = common->action;
 898
 899     /* get parameters */
 900     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
 901
 902     device = action->blockdev_snapshot_sync->device;
 903     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
 904     if (action->blockdev_snapshot_sync->has_format) {
 905         format = action->blockdev_snapshot_sync->format;
 906     }
 907     if (action->blockdev_snapshot_sync->has_mode) {
 908         mode = action->blockdev_snapshot_sync->mode;
 909     }
 910
 911     /* start processing */
 912     drv = bdrv_find_format(format);
 913     if (!drv) {
 914         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 915         return;
 916     }
 917
 918     state->old_bs = bdrv_find(device);
 919     if (!state->old_bs) {
 920         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 921         return;
 922     }
 923
 924     if (!bdrv_is_inserted(state->old_bs)) {
 925         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
 926         return;
 927     }
 928
 929     if (bdrv_in_use(state->old_bs)) {
 930         error_set(errp, QERR_DEVICE_IN_USE, device);
 931         return;
 932     }
 933
 934     if (!bdrv_is_read_only(state->old_bs)) {
 935         if (bdrv_flush(state->old_bs)) {
 936             error_set(errp, QERR_IO_ERROR);
 937             return;
 938         }
 939     }
 940
 941     flags = state->old_bs->open_flags;
 942
 943     /* create new image w/backing file */
 944     if (mode != NEW_IMAGE_MODE_EXISTING) {
 945         bdrv_img_create(new_image_file, format,
 946                         state->old_bs->filename,
 947                         state->old_bs->drv->format_name,
 948                         NULL, -1, flags, &local_err, false);
 949         if (error_is_set(&local_err)) {
 950             error_propagate(errp, local_err);
 951             return;
 952         }
 953     }
 954
 955     /* We will manually add the backing_hd field to the bs later */
 956     state->new_bs = bdrv_new("");
 957     /* TODO Inherit bs->options or only take explicit options with an
 958      * extended QMP command? */
 959     ret = bdrv_open(state->new_bs, new_image_file, NULL,
 960                     flags | BDRV_O_NO_BACKING, drv);
 961     if (ret != 0) {
 962         error_setg_file_open(errp, -ret, new_image_file);
 963     }
 964 }
 965
 966 static void external_snapshot_commit(BlkTransactionState *common)
 967 {
 968     ExternalSnapshotState *state =
 969                              DO_UPCAST(ExternalSnapshotState, common, common);
 970
 971     /* This removes our old bs and adds the new bs */
 972     bdrv_append(state->new_bs, state->old_bs);
 973     /* We don't need (or want) to use the transactional
 974      * bdrv_reopen_multiple() across all the entries at once, because we
 975      * don't want to abort all of them if one of them fails the reopen */
 976     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
 977                 NULL);
 978 }
 979
 980 static void external_snapshot_abort(BlkTransactionState *common)
 981 {
 982     ExternalSnapshotState *state =
 983                              DO_UPCAST(ExternalSnapshotState, common, common);
 984     if (state->new_bs) {
 985         bdrv_delete(state->new_bs);
 986     }
 987 }
 988
 989 typedef struct DriveBackupState {
 990     BlkTransactionState common;
 991     BlockDriverState *bs;
 992     BlockJob *job;
 993 } DriveBackupState;
 994
 995 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
 996 {
 997     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
 998     DriveBackup *backup;
 999     Error *local_err = NULL;
1000
1001     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1002     backup = common->action->drive_backup;
1003
1004     qmp_drive_backup(backup->device, backup->target,
1005                      backup->has_format, backup->format,
1006                      backup->sync,
1007                      backup->has_mode, backup->mode,
1008                      backup->has_speed, backup->speed,
1009                      backup->has_on_source_error, backup->on_source_error,
1010                      backup->has_on_target_error, backup->on_target_error,
1011                      &local_err);
1012     if (error_is_set(&local_err)) {
1013         error_propagate(errp, local_err);
1014         state->bs = NULL;
1015         state->job = NULL;
1016         return;
1017     }
1018
1019     state->bs = bdrv_find(backup->device);
1020     state->job = state->bs->job;
1021 }
1022
1023 static void drive_backup_abort(BlkTransactionState *common)
1024 {
1025     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1026     BlockDriverState *bs = state->bs;
1027
1028     /* Only cancel if it's the job we started */
1029     if (bs && bs->job && bs->job == state->job) {
1030         block_job_cancel_sync(bs->job);
1031     }
1032 }
1033
1034 static void abort_prepare(BlkTransactionState *common, Error **errp)
1035 {
1036     error_setg(errp, "Transaction aborted using Abort action");
1037 }
1038
1039 static void abort_commit(BlkTransactionState *common)
1040 {
1041     g_assert_not_reached(); /* this action never succeeds */
1042 }
1043
1044 static const BdrvActionOps actions[] = {
1045     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1046         .instance_size = sizeof(ExternalSnapshotState),
1047         .prepare  = external_snapshot_prepare,
1048         .commit   = external_snapshot_commit,
1049         .abort = external_snapshot_abort,
1050     },
1051     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1052         .instance_size = sizeof(DriveBackupState),
1053         .prepare = drive_backup_prepare,
1054         .abort = drive_backup_abort,
1055     },
1056     [TRANSACTION_ACTION_KIND_ABORT] = {
1057         .instance_size = sizeof(BlkTransactionState),
1058         .prepare = abort_prepare,
1059         .commit = abort_commit,
1060     },
1061 };
1062
1063 /*
1064  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1065  *  then we do not pivot any of the devices in the group, and abandon the
1066  *  snapshots
1067  */
1068 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1069 {
1070     TransactionActionList *dev_entry = dev_list;
1071     BlkTransactionState *state, *next;
1072     Error *local_err = NULL;
1073
1074     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1075     QSIMPLEQ_INIT(&snap_bdrv_states);
1076
1077     /* drain all i/o before any snapshots */
1078     bdrv_drain_all();
1079
1080     /* We don't do anything in this loop that commits us to the snapshot */
1081     while (NULL != dev_entry) {
1082         TransactionAction *dev_info = NULL;
1083         const BdrvActionOps *ops;
1084
1085         dev_info = dev_entry->value;
1086         dev_entry = dev_entry->next;
1087
1088         assert(dev_info->kind < ARRAY_SIZE(actions));
1089
1090         ops = &actions[dev_info->kind];
1091         state = g_malloc0(ops->instance_size);
1092         state->ops = ops;
1093         state->action = dev_info;
1094         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1095
1096         state->ops->prepare(state, &local_err);
1097         if (error_is_set(&local_err)) {
1098             error_propagate(errp, local_err);
1099             goto delete_and_fail;
1100         }
1101     }
1102
1103     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1104         if (state->ops->commit) {
1105             state->ops->commit(state);
1106         }
1107     }
1108
1109     /* success */
1110     goto exit;
1111
1112 delete_and_fail:
1113     /*
1114     * failure, and it is all-or-none; abandon each new bs, and keep using
1115     * the original bs for all images
1116     */
1117     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1118         if (state->ops->abort) {
1119             state->ops->abort(state);
1120         }
1121     }
1122 exit:
1123     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1124         if (state->ops->clean) {
1125             state->ops->clean(state);
1126         }
1127         g_free(state);
1128     }
1129 }
1130
1131
1132 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1133 {
1134     if (bdrv_in_use(bs)) {
1135         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1136         return;
1137     }
1138     if (!bdrv_dev_has_removable_media(bs)) {
1139         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1140         return;
1141     }
1142
1143     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1144         bdrv_dev_eject_request(bs, force);
1145         if (!force) {
1146             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1147             return;
1148         }
1149     }
1150
1151     bdrv_close(bs);
1152 }
1153
1154 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1155 {
1156     BlockDriverState *bs;
1157
1158     bs = bdrv_find(device);
1159     if (!bs) {
1160         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1161         return;
1162     }
1163
1164     eject_device(bs, force, errp);
1165 }
1166
1167 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1168 {
1169     BlockDriverState *bs;
1170     int err;
1171
1172     bs = bdrv_find(device);
1173     if (!bs) {
1174         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1175         return;
1176     }
1177
1178     err = bdrv_set_key(bs, password);
1179     if (err == -EINVAL) {
1180         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1181         return;
1182     } else if (err < 0) {
1183         error_set(errp, QERR_INVALID_PASSWORD);
1184         return;
1185     }
1186 }
1187
1188 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1189                                     int bdrv_flags, BlockDriver *drv,
1190                                     const char *password, Error **errp)
1191 {
1192     int ret;
1193
1194     ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv);
1195     if (ret < 0) {
1196         error_setg_file_open(errp, -ret, filename);
1197         return;
1198     }
1199
1200     if (bdrv_key_required(bs)) {
1201         if (password) {
1202             if (bdrv_set_key(bs, password) < 0) {
1203                 error_set(errp, QERR_INVALID_PASSWORD);
1204             }
1205         } else {
1206             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1207                       bdrv_get_encrypted_filename(bs));
1208         }
1209     } else if (password) {
1210         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1211     }
1212 }
1213
1214 void qmp_change_blockdev(const char *device, const char *filename,
1215                          bool has_format, const char *format, Error **errp)
1216 {
1217     BlockDriverState *bs;
1218     BlockDriver *drv = NULL;
1219     int bdrv_flags;
1220     Error *err = NULL;
1221
1222     bs = bdrv_find(device);
1223     if (!bs) {
1224         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1225         return;
1226     }
1227
1228     if (format) {
1229         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1230         if (!drv) {
1231             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1232             return;
1233         }
1234     }
1235
1236     eject_device(bs, 0, &err);
1237     if (error_is_set(&err)) {
1238         error_propagate(errp, err);
1239         return;
1240     }
1241
1242     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1243     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1244
1245     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1246 }
1247
1248 /* throttling disk I/O limits */
1249 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1250                                int64_t bps_wr, int64_t iops, int64_t iops_rd,
1251                                int64_t iops_wr, Error **errp)
1252 {
1253     BlockIOLimit io_limits;
1254     BlockDriverState *bs;
1255
1256     bs = bdrv_find(device);
1257     if (!bs) {
1258         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1259         return;
1260     }
1261
1262     io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
1263     io_limits.bps[BLOCK_IO_LIMIT_READ]  = bps_rd;
1264     io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
1265     io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
1266     io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
1267     io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
1268
1269     if (!do_check_io_limits(&io_limits, errp)) {
1270         return;
1271     }
1272
1273     bs->io_limits = io_limits;
1274
1275     if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
1276         bdrv_io_limits_enable(bs);
1277     } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
1278         bdrv_io_limits_disable(bs);
1279     } else {
1280         if (bs->block_timer) {
1281             timer_mod(bs->block_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL));
1282         }
1283     }
1284 }
1285
1286 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1287 {
1288     const char *id = qdict_get_str(qdict, "id");
1289     BlockDriverState *bs;
1290
1291     bs = bdrv_find(id);
1292     if (!bs) {
1293         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1294         return -1;
1295     }
1296     if (bdrv_in_use(bs)) {
1297         qerror_report(QERR_DEVICE_IN_USE, id);
1298         return -1;
1299     }
1300
1301     /* quiesce block driver; prevent further io */
1302     bdrv_drain_all();
1303     bdrv_flush(bs);
1304     bdrv_close(bs);
1305
1306     /* if we have a device attached to this BlockDriverState
1307      * then we need to make the drive anonymous until the device
1308      * can be removed.  If this is a drive with no device backing
1309      * then we can just get rid of the block driver state right here.
1310      */
1311     if (bdrv_get_attached_dev(bs)) {
1312         bdrv_make_anon(bs);
1313
1314         /* Further I/O must not pause the guest */
1315         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1316                           BLOCKDEV_ON_ERROR_REPORT);
1317     } else {
1318         drive_uninit(drive_get_by_blockdev(bs));
1319     }
1320
1321     return 0;
1322 }
1323
1324 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1325 {
1326     BlockDriverState *bs;
1327     int ret;
1328
1329     bs = bdrv_find(device);
1330     if (!bs) {
1331         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1332         return;
1333     }
1334
1335     if (size < 0) {
1336         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1337         return;
1338     }
1339
1340     /* complete all in-flight operations before resizing the device */
1341     bdrv_drain_all();
1342
1343     ret = bdrv_truncate(bs, size);
1344     switch (ret) {
1345     case 0:
1346         break;
1347     case -ENOMEDIUM:
1348         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1349         break;
1350     case -ENOTSUP:
1351         error_set(errp, QERR_UNSUPPORTED);
1352         break;
1353     case -EACCES:
1354         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1355         break;
1356     case -EBUSY:
1357         error_set(errp, QERR_DEVICE_IN_USE, device);
1358         break;
1359     default:
1360         error_setg_errno(errp, -ret, "Could not resize");
1361         break;
1362     }
1363 }
1364
1365 static void block_job_cb(void *opaque, int ret)
1366 {
1367     BlockDriverState *bs = opaque;
1368     QObject *obj;
1369
1370     trace_block_job_cb(bs, bs->job, ret);
1371
1372     assert(bs->job);
1373     obj = qobject_from_block_job(bs->job);
1374     if (ret < 0) {
1375         QDict *dict = qobject_to_qdict(obj);
1376         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1377     }
1378
1379     if (block_job_is_cancelled(bs->job)) {
1380         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1381     } else {
1382         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1383     }
1384     qobject_decref(obj);
1385
1386     drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1387 }
1388
1389 void qmp_block_stream(const char *device, bool has_base,
1390                       const char *base, bool has_speed, int64_t speed,
1391                       bool has_on_error, BlockdevOnError on_error,
1392                       Error **errp)
1393 {
1394     BlockDriverState *bs;
1395     BlockDriverState *base_bs = NULL;
1396     Error *local_err = NULL;
1397
1398     if (!has_on_error) {
1399         on_error = BLOCKDEV_ON_ERROR_REPORT;
1400     }
1401
1402     bs = bdrv_find(device);
1403     if (!bs) {
1404         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1405         return;
1406     }
1407
1408     if (base) {
1409         base_bs = bdrv_find_backing_image(bs, base);
1410         if (base_bs == NULL) {
1411             error_set(errp, QERR_BASE_NOT_FOUND, base);
1412             return;
1413         }
1414     }
1415
1416     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1417                  on_error, block_job_cb, bs, &local_err);
1418     if (error_is_set(&local_err)) {
1419         error_propagate(errp, local_err);
1420         return;
1421     }
1422
1423     /* Grab a reference so hotplug does not delete the BlockDriverState from
1424      * underneath us.
1425      */
1426     drive_get_ref(drive_get_by_blockdev(bs));
1427
1428     trace_qmp_block_stream(bs, bs->job);
1429 }
1430
1431 void qmp_block_commit(const char *device,
1432                       bool has_base, const char *base, const char *top,
1433                       bool has_speed, int64_t speed,
1434                       Error **errp)
1435 {
1436     BlockDriverState *bs;
1437     BlockDriverState *base_bs, *top_bs;
1438     Error *local_err = NULL;
1439     /* This will be part of the QMP command, if/when the
1440      * BlockdevOnError change for blkmirror makes it in
1441      */
1442     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1443
1444     /* drain all i/o before commits */
1445     bdrv_drain_all();
1446
1447     bs = bdrv_find(device);
1448     if (!bs) {
1449         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1450         return;
1451     }
1452
1453     /* default top_bs is the active layer */
1454     top_bs = bs;
1455
1456     if (top) {
1457         if (strcmp(bs->filename, top) != 0) {
1458             top_bs = bdrv_find_backing_image(bs, top);
1459         }
1460     }
1461
1462     if (top_bs == NULL) {
1463         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1464         return;
1465     }
1466
1467     if (has_base && base) {
1468         base_bs = bdrv_find_backing_image(top_bs, base);
1469     } else {
1470         base_bs = bdrv_find_base(top_bs);
1471     }
1472
1473     if (base_bs == NULL) {
1474         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1475         return;
1476     }
1477
1478     commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1479                 &local_err);
1480     if (local_err != NULL) {
1481         error_propagate(errp, local_err);
1482         return;
1483     }
1484     /* Grab a reference so hotplug does not delete the BlockDriverState from
1485      * underneath us.
1486      */
1487     drive_get_ref(drive_get_by_blockdev(bs));
1488 }
1489
1490 void qmp_drive_backup(const char *device, const char *target,
1491                       bool has_format, const char *format,
1492                       enum MirrorSyncMode sync,
1493                       bool has_mode, enum NewImageMode mode,
1494                       bool has_speed, int64_t speed,
1495                       bool has_on_source_error, BlockdevOnError on_source_error,
1496                       bool has_on_target_error, BlockdevOnError on_target_error,
1497                       Error **errp)
1498 {
1499     BlockDriverState *bs;
1500     BlockDriverState *target_bs;
1501     BlockDriverState *source = NULL;
1502     BlockDriver *drv = NULL;
1503     Error *local_err = NULL;
1504     int flags;
1505     int64_t size;
1506     int ret;
1507
1508     if (!has_speed) {
1509         speed = 0;
1510     }
1511     if (!has_on_source_error) {
1512         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1513     }
1514     if (!has_on_target_error) {
1515         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1516     }
1517     if (!has_mode) {
1518         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1519     }
1520
1521     bs = bdrv_find(device);
1522     if (!bs) {
1523         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1524         return;
1525     }
1526
1527     if (!bdrv_is_inserted(bs)) {
1528         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1529         return;
1530     }
1531
1532     if (!has_format) {
1533         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1534     }
1535     if (format) {
1536         drv = bdrv_find_format(format);
1537         if (!drv) {
1538             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1539             return;
1540         }
1541     }
1542
1543     if (bdrv_in_use(bs)) {
1544         error_set(errp, QERR_DEVICE_IN_USE, device);
1545         return;
1546     }
1547
1548     flags = bs->open_flags | BDRV_O_RDWR;
1549
1550     /* See if we have a backing HD we can use to create our new image
1551      * on top of. */
1552     if (sync == MIRROR_SYNC_MODE_TOP) {
1553         source = bs->backing_hd;
1554         if (!source) {
1555             sync = MIRROR_SYNC_MODE_FULL;
1556         }
1557     }
1558     if (sync == MIRROR_SYNC_MODE_NONE) {
1559         source = bs;
1560     }
1561
1562     size = bdrv_getlength(bs);
1563     if (size < 0) {
1564         error_setg_errno(errp, -size, "bdrv_getlength failed");
1565         return;
1566     }
1567
1568     if (mode != NEW_IMAGE_MODE_EXISTING) {
1569         assert(format && drv);
1570         if (source) {
1571             bdrv_img_create(target, format, source->filename,
1572                             source->drv->format_name, NULL,
1573                             size, flags, &local_err, false);
1574         } else {
1575             bdrv_img_create(target, format, NULL, NULL, NULL,
1576                             size, flags, &local_err, false);
1577         }
1578     }
1579
1580     if (error_is_set(&local_err)) {
1581         error_propagate(errp, local_err);
1582         return;
1583     }
1584
1585     target_bs = bdrv_new("");
1586     ret = bdrv_open(target_bs, target, NULL, flags, drv);
1587     if (ret < 0) {
1588         bdrv_delete(target_bs);
1589         error_setg_file_open(errp, -ret, target);
1590         return;
1591     }
1592
1593     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1594                  block_job_cb, bs, &local_err);
1595     if (local_err != NULL) {
1596         bdrv_delete(target_bs);
1597         error_propagate(errp, local_err);
1598         return;
1599     }
1600
1601     /* Grab a reference so hotplug does not delete the BlockDriverState from
1602      * underneath us.
1603      */
1604     drive_get_ref(drive_get_by_blockdev(bs));
1605 }
1606
1607 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1608
1609 void qmp_drive_mirror(const char *device, const char *target,
1610                       bool has_format, const char *format,
1611                       enum MirrorSyncMode sync,
1612                       bool has_mode, enum NewImageMode mode,
1613                       bool has_speed, int64_t speed,
1614                       bool has_granularity, uint32_t granularity,
1615                       bool has_buf_size, int64_t buf_size,
1616                       bool has_on_source_error, BlockdevOnError on_source_error,
1617                       bool has_on_target_error, BlockdevOnError on_target_error,
1618                       Error **errp)
1619 {
1620     BlockDriverState *bs;
1621     BlockDriverState *source, *target_bs;
1622     BlockDriver *drv = NULL;
1623     Error *local_err = NULL;
1624     int flags;
1625     int64_t size;
1626     int ret;
1627
1628     if (!has_speed) {
1629         speed = 0;
1630     }
1631     if (!has_on_source_error) {
1632         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1633     }
1634     if (!has_on_target_error) {
1635         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1636     }
1637     if (!has_mode) {
1638         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1639     }
1640     if (!has_granularity) {
1641         granularity = 0;
1642     }
1643     if (!has_buf_size) {
1644         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1645     }
1646
1647     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1648         error_set(errp, QERR_INVALID_PARAMETER, device);
1649         return;
1650     }
1651     if (granularity & (granularity - 1)) {
1652         error_set(errp, QERR_INVALID_PARAMETER, device);
1653         return;
1654     }
1655
1656     bs = bdrv_find(device);
1657     if (!bs) {
1658         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1659         return;
1660     }
1661
1662     if (!bdrv_is_inserted(bs)) {
1663         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1664         return;
1665     }
1666
1667     if (!has_format) {
1668         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1669     }
1670     if (format) {
1671         drv = bdrv_find_format(format);
1672         if (!drv) {
1673             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1674             return;
1675         }
1676     }
1677
1678     if (bdrv_in_use(bs)) {
1679         error_set(errp, QERR_DEVICE_IN_USE, device);
1680         return;
1681     }
1682
1683     flags = bs->open_flags | BDRV_O_RDWR;
1684     source = bs->backing_hd;
1685     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1686         sync = MIRROR_SYNC_MODE_FULL;
1687     }
1688
1689     size = bdrv_getlength(bs);
1690     if (size < 0) {
1691         error_setg_errno(errp, -size, "bdrv_getlength failed");
1692         return;
1693     }
1694
1695     if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1696         /* create new image w/o backing file */
1697         assert(format && drv);
1698         bdrv_img_create(target, format,
1699                         NULL, NULL, NULL, size, flags, &local_err, false);
1700     } else {
1701         switch (mode) {
1702         case NEW_IMAGE_MODE_EXISTING:
1703             ret = 0;
1704             break;
1705         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1706             /* create new image with backing file */
1707             bdrv_img_create(target, format,
1708                             source->filename,
1709                             source->drv->format_name,
1710                             NULL, size, flags, &local_err, false);
1711             break;
1712         default:
1713             abort();
1714         }
1715     }
1716
1717     if (error_is_set(&local_err)) {
1718         error_propagate(errp, local_err);
1719         return;
1720     }
1721
1722     /* Mirroring takes care of copy-on-write using the source's backing
1723      * file.
1724      */
1725     target_bs = bdrv_new("");
1726     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv);
1727     if (ret < 0) {
1728         bdrv_delete(target_bs);
1729         error_setg_file_open(errp, -ret, target);
1730         return;
1731     }
1732
1733     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1734                  on_source_error, on_target_error,
1735                  block_job_cb, bs, &local_err);
1736     if (local_err != NULL) {
1737         bdrv_delete(target_bs);
1738         error_propagate(errp, local_err);
1739         return;
1740     }
1741
1742     /* Grab a reference so hotplug does not delete the BlockDriverState from
1743      * underneath us.
1744      */
1745     drive_get_ref(drive_get_by_blockdev(bs));
1746 }
1747
1748 static BlockJob *find_block_job(const char *device)
1749 {
1750     BlockDriverState *bs;
1751
1752     bs = bdrv_find(device);
1753     if (!bs || !bs->job) {
1754         return NULL;
1755     }
1756     return bs->job;
1757 }
1758
1759 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1760 {
1761     BlockJob *job = find_block_job(device);
1762
1763     if (!job) {
1764         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1765         return;
1766     }
1767
1768     block_job_set_speed(job, speed, errp);
1769 }
1770
1771 void qmp_block_job_cancel(const char *device,
1772                           bool has_force, bool force, Error **errp)
1773 {
1774     BlockJob *job = find_block_job(device);
1775
1776     if (!has_force) {
1777         force = false;
1778     }
1779
1780     if (!job) {
1781         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1782         return;
1783     }
1784     if (job->paused && !force) {
1785         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1786         return;
1787     }
1788
1789     trace_qmp_block_job_cancel(job);
1790     block_job_cancel(job);
1791 }
1792
1793 void qmp_block_job_pause(const char *device, Error **errp)
1794 {
1795     BlockJob *job = find_block_job(device);
1796
1797     if (!job) {
1798         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1799         return;
1800     }
1801
1802     trace_qmp_block_job_pause(job);
1803     block_job_pause(job);
1804 }
1805
1806 void qmp_block_job_resume(const char *device, Error **errp)
1807 {
1808     BlockJob *job = find_block_job(device);
1809
1810     if (!job) {
1811         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1812         return;
1813     }
1814
1815     trace_qmp_block_job_resume(job);
1816     block_job_resume(job);
1817 }
1818
1819 void qmp_block_job_complete(const char *device, Error **errp)
1820 {
1821     BlockJob *job = find_block_job(device);
1822
1823     if (!job) {
1824         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1825         return;
1826     }
1827
1828     trace_qmp_block_job_complete(job);
1829     block_job_complete(job, errp);
1830 }
1831
1832 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1833 {
1834     BlockJobInfoList **prev = opaque;
1835     BlockJob *job = bs->job;
1836
1837     if (job) {
1838         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1839         elem->value = block_job_query(bs->job);
1840         (*prev)->next = elem;
1841         *prev = elem;
1842     }
1843 }
1844
1845 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1846 {
1847     /* Dummy is a fake list element for holding the head pointer */
1848     BlockJobInfoList dummy = {};
1849     BlockJobInfoList *prev = &dummy;
1850     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1851     return dummy.next;
1852 }
1853
1854 QemuOptsList qemu_common_drive_opts = {
1855     .name = "drive",
1856     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
1857     .desc = {
1858         {
1859             .name = "bus",
1860             .type = QEMU_OPT_NUMBER,
1861             .help = "bus number",
1862         },{
1863             .name = "unit",
1864             .type = QEMU_OPT_NUMBER,
1865             .help = "unit number (i.e. lun for scsi)",
1866         },{
1867             .name = "if",
1868             .type = QEMU_OPT_STRING,
1869             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1870         },{
1871             .name = "index",
1872             .type = QEMU_OPT_NUMBER,
1873             .help = "index number",
1874         },{
1875             .name = "cyls",
1876             .type = QEMU_OPT_NUMBER,
1877             .help = "number of cylinders (ide disk geometry)",
1878         },{
1879             .name = "heads",
1880             .type = QEMU_OPT_NUMBER,
1881             .help = "number of heads (ide disk geometry)",
1882         },{
1883             .name = "secs",
1884             .type = QEMU_OPT_NUMBER,
1885             .help = "number of sectors (ide disk geometry)",
1886         },{
1887             .name = "trans",
1888             .type = QEMU_OPT_STRING,
1889             .help = "chs translation (auto, lba. none)",
1890         },{
1891             .name = "media",
1892             .type = QEMU_OPT_STRING,
1893             .help = "media type (disk, cdrom)",
1894         },{
1895             .name = "snapshot",
1896             .type = QEMU_OPT_BOOL,
1897             .help = "enable/disable snapshot mode",
1898         },{
1899             .name = "file",
1900             .type = QEMU_OPT_STRING,
1901             .help = "disk image",
1902         },{
1903             .name = "discard",
1904             .type = QEMU_OPT_STRING,
1905             .help = "discard operation (ignore/off, unmap/on)",
1906         },{
1907             .name = "cache.writeback",
1908             .type = QEMU_OPT_BOOL,
1909             .help = "enables writeback mode for any caches",
1910         },{
1911             .name = "cache.direct",
1912             .type = QEMU_OPT_BOOL,
1913             .help = "enables use of O_DIRECT (bypass the host page cache)",
1914         },{
1915             .name = "cache.no-flush",
1916             .type = QEMU_OPT_BOOL,
1917             .help = "ignore any flush requests for the device",
1918         },{
1919             .name = "aio",
1920             .type = QEMU_OPT_STRING,
1921             .help = "host AIO implementation (threads, native)",
1922         },{
1923             .name = "format",
1924             .type = QEMU_OPT_STRING,
1925             .help = "disk format (raw, qcow2, ...)",
1926         },{
1927             .name = "serial",
1928             .type = QEMU_OPT_STRING,
1929             .help = "disk serial number",
1930         },{
1931             .name = "rerror",
1932             .type = QEMU_OPT_STRING,
1933             .help = "read error action",
1934         },{
1935             .name = "werror",
1936             .type = QEMU_OPT_STRING,
1937             .help = "write error action",
1938         },{
1939             .name = "addr",
1940             .type = QEMU_OPT_STRING,
1941             .help = "pci address (virtio only)",
1942         },{
1943             .name = "read-only",
1944             .type = QEMU_OPT_BOOL,
1945             .help = "open drive file as read-only",
1946         },{
1947             .name = "throttling.iops-total",
1948             .type = QEMU_OPT_NUMBER,
1949             .help = "limit total I/O operations per second",
1950         },{
1951             .name = "throttling.iops-read",
1952             .type = QEMU_OPT_NUMBER,
1953             .help = "limit read operations per second",
1954         },{
1955             .name = "throttling.iops-write",
1956             .type = QEMU_OPT_NUMBER,
1957             .help = "limit write operations per second",
1958         },{
1959             .name = "throttling.bps-total",
1960             .type = QEMU_OPT_NUMBER,
1961             .help = "limit total bytes per second",
1962         },{
1963             .name = "throttling.bps-read",
1964             .type = QEMU_OPT_NUMBER,
1965             .help = "limit read bytes per second",
1966         },{
1967             .name = "throttling.bps-write",
1968             .type = QEMU_OPT_NUMBER,
1969             .help = "limit write bytes per second",
1970         },{
1971             .name = "copy-on-read",
1972             .type = QEMU_OPT_BOOL,
1973             .help = "copy read data from backing file into image file",
1974         },{
1975             .name = "boot",
1976             .type = QEMU_OPT_BOOL,
1977             .help = "(deprecated, ignored)",
1978         },
1979         { /* end of list */ }
1980     },
1981 };
1982
1983 QemuOptsList qemu_drive_opts = {
1984     .name = "drive",
1985     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1986     .desc = {
1987         /*
1988          * no elements => accept any params
1989          * validation will happen later
1990          */
1991         { /* end of list */ }
1992     },
1993 };