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
  10 #include "block.h"
  11 #include "blockdev.h"
  12 #include "monitor.h"
  13 #include "qerror.h"
  14 #include "qemu-option.h"
  15 #include "qemu-config.h"
  16 #include "qemu-objects.h"
  17 #include "sysemu.h"
  18 #include "block_int.h"
  19 #include "qmp-commands.h"
  20 #include "trace.h"
  21 #include "arch_init.h"
  22
  23 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
  24
  25 static const char *const if_name[IF_COUNT] = {
  26     [IF_NONE] = "none",
  27     [IF_IDE] = "ide",
  28     [IF_SCSI] = "scsi",
  29     [IF_FLOPPY] = "floppy",
  30     [IF_PFLASH] = "pflash",
  31     [IF_MTD] = "mtd",
  32     [IF_SD] = "sd",
  33     [IF_VIRTIO] = "virtio",
  34     [IF_XEN] = "xen",
  35 };
  36
  37 static const int if_max_devs[IF_COUNT] = {
  38     /*
  39      * Do not change these numbers!  They govern how drive option
  40      * index maps to unit and bus.  That mapping is ABI.
  41      *
  42      * All controllers used to imlement if=T drives need to support
  43      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
  44      * Otherwise, some index values map to "impossible" bus, unit
  45      * values.
  46      *
  47      * For instance, if you change [IF_SCSI] to 255, -drive
  48      * if=scsi,index=12 no longer means bus=1,unit=5, but
  49      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
  50      * the drive can't be set up.  Regression.
  51      */
  52     [IF_IDE] = 2,
  53     [IF_SCSI] = 7,
  54 };
  55
  56 /*
  57  * We automatically delete the drive when a device using it gets
  58  * unplugged.  Questionable feature, but we can't just drop it.
  59  * Device models call blockdev_mark_auto_del() to schedule the
  60  * automatic deletion, and generic qdev code calls blockdev_auto_del()
  61  * when deletion is actually safe.
  62  */
  63 void blockdev_mark_auto_del(BlockDriverState *bs)
  64 {
  65     DriveInfo *dinfo = drive_get_by_blockdev(bs);
  66
  67     if (dinfo) {
  68         dinfo->auto_del = 1;
  69     }
  70 }
  71
  72 void blockdev_auto_del(BlockDriverState *bs)
  73 {
  74     DriveInfo *dinfo = drive_get_by_blockdev(bs);
  75
  76     if (dinfo && dinfo->auto_del) {
  77         drive_put_ref(dinfo);
  78     }
  79 }
  80
  81 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
  82 {
  83     int max_devs = if_max_devs[type];
  84     return max_devs ? index / max_devs : 0;
  85 }
  86
  87 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
  88 {
  89     int max_devs = if_max_devs[type];
  90     return max_devs ? index % max_devs : index;
  91 }
  92
  93 QemuOpts *drive_def(const char *optstr)
  94 {
  95     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
  96 }
  97
  98 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
  99                     const char *optstr)
 100 {
 101     QemuOpts *opts;
 102     char buf[32];
 103
 104     opts = drive_def(optstr);
 105     if (!opts) {
 106         return NULL;
 107     }
 108     if (type != IF_DEFAULT) {
 109         qemu_opt_set(opts, "if", if_name[type]);
 110     }
 111     if (index >= 0) {
 112         snprintf(buf, sizeof(buf), "%d", index);
 113         qemu_opt_set(opts, "index", buf);
 114     }
 115     if (file)
 116         qemu_opt_set(opts, "file", file);
 117     return opts;
 118 }
 119
 120 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
 121 {
 122     DriveInfo *dinfo;
 123
 124     /* seek interface, bus and unit */
 125
 126     QTAILQ_FOREACH(dinfo, &drives, next) {
 127         if (dinfo->type == type &&
 128             dinfo->bus == bus &&
 129             dinfo->unit == unit)
 130             return dinfo;
 131     }
 132
 133     return NULL;
 134 }
 135
 136 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
 137 {
 138     return drive_get(type,
 139                      drive_index_to_bus_id(type, index),
 140                      drive_index_to_unit_id(type, index));
 141 }
 142
 143 int drive_get_max_bus(BlockInterfaceType type)
 144 {
 145     int max_bus;
 146     DriveInfo *dinfo;
 147
 148     max_bus = -1;
 149     QTAILQ_FOREACH(dinfo, &drives, next) {
 150         if(dinfo->type == type &&
 151            dinfo->bus > max_bus)
 152             max_bus = dinfo->bus;
 153     }
 154     return max_bus;
 155 }
 156
 157 /* Get a block device.  This should only be used for single-drive devices
 158    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
 159    appropriate bus.  */
 160 DriveInfo *drive_get_next(BlockInterfaceType type)
 161 {
 162     static int next_block_unit[IF_COUNT];
 163
 164     return drive_get(type, 0, next_block_unit[type]++);
 165 }
 166
 167 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
 168 {
 169     DriveInfo *dinfo;
 170
 171     QTAILQ_FOREACH(dinfo, &drives, next) {
 172         if (dinfo->bdrv == bs) {
 173             return dinfo;
 174         }
 175     }
 176     return NULL;
 177 }
 178
 179 static void bdrv_format_print(void *opaque, const char *name)
 180 {
 181     error_printf(" %s", name);
 182 }
 183
 184 static void drive_uninit(DriveInfo *dinfo)
 185 {
 186     qemu_opts_del(dinfo->opts);
 187     bdrv_delete(dinfo->bdrv);
 188     g_free(dinfo->id);
 189     QTAILQ_REMOVE(&drives, dinfo, next);
 190     g_free(dinfo);
 191 }
 192
 193 void drive_put_ref(DriveInfo *dinfo)
 194 {
 195     assert(dinfo->refcount);
 196     if (--dinfo->refcount == 0) {
 197         drive_uninit(dinfo);
 198     }
 199 }
 200
 201 void drive_get_ref(DriveInfo *dinfo)
 202 {
 203     dinfo->refcount++;
 204 }
 205
 206 typedef struct {
 207     QEMUBH *bh;
 208     DriveInfo *dinfo;
 209 } DrivePutRefBH;
 210
 211 static void drive_put_ref_bh(void *opaque)
 212 {
 213     DrivePutRefBH *s = opaque;
 214
 215     drive_put_ref(s->dinfo);
 216     qemu_bh_delete(s->bh);
 217     g_free(s);
 218 }
 219
 220 /*
 221  * Release a drive reference in a BH
 222  *
 223  * It is not possible to use drive_put_ref() from a callback function when the
 224  * callers still need the drive.  In such cases we schedule a BH to release the
 225  * reference.
 226  */
 227 static void drive_put_ref_bh_schedule(DriveInfo *dinfo)
 228 {
 229     DrivePutRefBH *s;
 230
 231     s = g_new(DrivePutRefBH, 1);
 232     s->bh = qemu_bh_new(drive_put_ref_bh, s);
 233     s->dinfo = dinfo;
 234     qemu_bh_schedule(s->bh);
 235 }
 236
 237 static int parse_block_error_action(const char *buf, int is_read)
 238 {
 239     if (!strcmp(buf, "ignore")) {
 240         return BLOCK_ERR_IGNORE;
 241     } else if (!is_read && !strcmp(buf, "enospc")) {
 242         return BLOCK_ERR_STOP_ENOSPC;
 243     } else if (!strcmp(buf, "stop")) {
 244         return BLOCK_ERR_STOP_ANY;
 245     } else if (!strcmp(buf, "report")) {
 246         return BLOCK_ERR_REPORT;
 247     } else {
 248         error_report("'%s' invalid %s error action",
 249                      buf, is_read ? "read" : "write");
 250         return -1;
 251     }
 252 }
 253
 254 static bool do_check_io_limits(BlockIOLimit *io_limits)
 255 {
 256     bool bps_flag;
 257     bool iops_flag;
 258
 259     assert(io_limits);
 260
 261     bps_flag  = (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] != 0)
 262                  && ((io_limits->bps[BLOCK_IO_LIMIT_READ] != 0)
 263                  || (io_limits->bps[BLOCK_IO_LIMIT_WRITE] != 0));
 264     iops_flag = (io_limits->iops[BLOCK_IO_LIMIT_TOTAL] != 0)
 265                  && ((io_limits->iops[BLOCK_IO_LIMIT_READ] != 0)
 266                  || (io_limits->iops[BLOCK_IO_LIMIT_WRITE] != 0));
 267     if (bps_flag || iops_flag) {
 268         return false;
 269     }
 270
 271     return true;
 272 }
 273
 274 DriveInfo *drive_init(QemuOpts *opts, int default_to_scsi)
 275 {
 276     const char *buf;
 277     const char *file = NULL;
 278     char devname[128];
 279     const char *serial;
 280     const char *mediastr = "";
 281     BlockInterfaceType type;
 282     enum { MEDIA_DISK, MEDIA_CDROM } media;
 283     int bus_id, unit_id;
 284     int cyls, heads, secs, translation;
 285     BlockDriver *drv = NULL;
 286     int max_devs;
 287     int index;
 288     int ro = 0;
 289     int bdrv_flags = 0;
 290     int on_read_error, on_write_error;
 291     const char *devaddr;
 292     DriveInfo *dinfo;
 293     BlockIOLimit io_limits;
 294     int snapshot = 0;
 295     bool copy_on_read;
 296     int ret;
 297
 298     translation = BIOS_ATA_TRANSLATION_AUTO;
 299     media = MEDIA_DISK;
 300
 301     /* extract parameters */
 302     bus_id  = qemu_opt_get_number(opts, "bus", 0);
 303     unit_id = qemu_opt_get_number(opts, "unit", -1);
 304     index   = qemu_opt_get_number(opts, "index", -1);
 305
 306     cyls  = qemu_opt_get_number(opts, "cyls", 0);
 307     heads = qemu_opt_get_number(opts, "heads", 0);
 308     secs  = qemu_opt_get_number(opts, "secs", 0);
 309
 310     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 311     ro = qemu_opt_get_bool(opts, "readonly", 0);
 312     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
 313
 314     file = qemu_opt_get(opts, "file");
 315     serial = qemu_opt_get(opts, "serial");
 316
 317     if ((buf = qemu_opt_get(opts, "if")) != NULL) {
 318         pstrcpy(devname, sizeof(devname), buf);
 319         for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
 320             ;
 321         if (type == IF_COUNT) {
 322             error_report("unsupported bus type '%s'", buf);
 323             return NULL;
 324         }
 325     } else {
 326         type = default_to_scsi ? IF_SCSI : IF_IDE;
 327         pstrcpy(devname, sizeof(devname), if_name[type]);
 328     }
 329
 330     max_devs = if_max_devs[type];
 331
 332     if (cyls || heads || secs) {
 333         if (cyls < 1 || (type == IF_IDE && cyls > 16383)) {
 334             error_report("invalid physical cyls number");
 335             return NULL;
 336         }
 337         if (heads < 1 || (type == IF_IDE && heads > 16)) {
 338             error_report("invalid physical heads number");
 339             return NULL;
 340         }
 341         if (secs < 1 || (type == IF_IDE && secs > 63)) {
 342             error_report("invalid physical secs number");
 343             return NULL;
 344         }
 345     }
 346
 347     if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
 348         if (!cyls) {
 349             error_report("'%s' trans must be used with cyls, heads and secs",
 350                          buf);
 351             return NULL;
 352         }
 353         if (!strcmp(buf, "none"))
 354             translation = BIOS_ATA_TRANSLATION_NONE;
 355         else if (!strcmp(buf, "lba"))
 356             translation = BIOS_ATA_TRANSLATION_LBA;
 357         else if (!strcmp(buf, "auto"))
 358             translation = BIOS_ATA_TRANSLATION_AUTO;
 359         else {
 360             error_report("'%s' invalid translation type", buf);
 361             return NULL;
 362         }
 363     }
 364
 365     if ((buf = qemu_opt_get(opts, "media")) != NULL) {
 366         if (!strcmp(buf, "disk")) {
 367             media = MEDIA_DISK;
 368         } else if (!strcmp(buf, "cdrom")) {
 369             if (cyls || secs || heads) {
 370                 error_report("CHS can't be set with media=%s", buf);
 371                 return NULL;
 372             }
 373             media = MEDIA_CDROM;
 374         } else {
 375             error_report("'%s' invalid media", buf);
 376             return NULL;
 377         }
 378     }
 379
 380     if ((buf = qemu_opt_get(opts, "cache")) != NULL) {
 381         if (bdrv_parse_cache_flags(buf, &bdrv_flags) != 0) {
 382             error_report("invalid cache option");
 383             return NULL;
 384         }
 385     }
 386
 387 #ifdef CONFIG_LINUX_AIO
 388     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 389         if (!strcmp(buf, "native")) {
 390             bdrv_flags |= BDRV_O_NATIVE_AIO;
 391         } else if (!strcmp(buf, "threads")) {
 392             /* this is the default */
 393         } else {
 394            error_report("invalid aio option");
 395            return NULL;
 396         }
 397     }
 398 #endif
 399
 400     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 401        if (strcmp(buf, "?") == 0) {
 402            error_printf("Supported formats:");
 403            bdrv_iterate_format(bdrv_format_print, NULL);
 404            error_printf("\n");
 405            return NULL;
 406         }
 407         drv = bdrv_find_whitelisted_format(buf);
 408         if (!drv) {
 409             error_report("'%s' invalid format", buf);
 410             return NULL;
 411         }
 412     }
 413
 414     /* disk I/O throttling */
 415     io_limits.bps[BLOCK_IO_LIMIT_TOTAL]  =
 416                            qemu_opt_get_number(opts, "bps", 0);
 417     io_limits.bps[BLOCK_IO_LIMIT_READ]   =
 418                            qemu_opt_get_number(opts, "bps_rd", 0);
 419     io_limits.bps[BLOCK_IO_LIMIT_WRITE]  =
 420                            qemu_opt_get_number(opts, "bps_wr", 0);
 421     io_limits.iops[BLOCK_IO_LIMIT_TOTAL] =
 422                            qemu_opt_get_number(opts, "iops", 0);
 423     io_limits.iops[BLOCK_IO_LIMIT_READ]  =
 424                            qemu_opt_get_number(opts, "iops_rd", 0);
 425     io_limits.iops[BLOCK_IO_LIMIT_WRITE] =
 426                            qemu_opt_get_number(opts, "iops_wr", 0);
 427
 428     if (!do_check_io_limits(&io_limits)) {
 429         error_report("bps(iops) and bps_rd/bps_wr(iops_rd/iops_wr) "
 430                      "cannot be used at the same time");
 431         return NULL;
 432     }
 433
 434     if (qemu_opt_get(opts, "boot") != NULL) {
 435         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 436                 "ignored. Future versions will reject this parameter. Please "
 437                 "update your scripts.\n");
 438     }
 439
 440     on_write_error = BLOCK_ERR_STOP_ENOSPC;
 441     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 442         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
 443             error_report("werror is not supported by this bus type");
 444             return NULL;
 445         }
 446
 447         on_write_error = parse_block_error_action(buf, 0);
 448         if (on_write_error < 0) {
 449             return NULL;
 450         }
 451     }
 452
 453     on_read_error = BLOCK_ERR_REPORT;
 454     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 455         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
 456             error_report("rerror is not supported by this bus type");
 457             return NULL;
 458         }
 459
 460         on_read_error = parse_block_error_action(buf, 1);
 461         if (on_read_error < 0) {
 462             return NULL;
 463         }
 464     }
 465
 466     if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
 467         if (type != IF_VIRTIO) {
 468             error_report("addr is not supported by this bus type");
 469             return NULL;
 470         }
 471     }
 472
 473     /* compute bus and unit according index */
 474
 475     if (index != -1) {
 476         if (bus_id != 0 || unit_id != -1) {
 477             error_report("index cannot be used with bus and unit");
 478             return NULL;
 479         }
 480         bus_id = drive_index_to_bus_id(type, index);
 481         unit_id = drive_index_to_unit_id(type, index);
 482     }
 483
 484     /* if user doesn't specify a unit_id,
 485      * try to find the first free
 486      */
 487
 488     if (unit_id == -1) {
 489        unit_id = 0;
 490        while (drive_get(type, bus_id, unit_id) != NULL) {
 491            unit_id++;
 492            if (max_devs && unit_id >= max_devs) {
 493                unit_id -= max_devs;
 494                bus_id++;
 495            }
 496        }
 497     }
 498
 499     /* check unit id */
 500
 501     if (max_devs && unit_id >= max_devs) {
 502         error_report("unit %d too big (max is %d)",
 503                      unit_id, max_devs - 1);
 504         return NULL;
 505     }
 506
 507     /*
 508      * catch multiple definitions
 509      */
 510
 511     if (drive_get(type, bus_id, unit_id) != NULL) {
 512         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 513                      bus_id, unit_id, index);
 514         return NULL;
 515     }
 516
 517     /* init */
 518
 519     dinfo = g_malloc0(sizeof(*dinfo));
 520     if ((buf = qemu_opts_id(opts)) != NULL) {
 521         dinfo->id = g_strdup(buf);
 522     } else {
 523         /* no id supplied -> create one */
 524         dinfo->id = g_malloc0(32);
 525         if (type == IF_IDE || type == IF_SCSI)
 526             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 527         if (max_devs)
 528             snprintf(dinfo->id, 32, "%s%i%s%i",
 529                      devname, bus_id, mediastr, unit_id);
 530         else
 531             snprintf(dinfo->id, 32, "%s%s%i",
 532                      devname, mediastr, unit_id);
 533     }
 534     dinfo->bdrv = bdrv_new(dinfo->id);
 535     dinfo->devaddr = devaddr;
 536     dinfo->type = type;
 537     dinfo->bus = bus_id;
 538     dinfo->unit = unit_id;
 539     dinfo->opts = opts;
 540     dinfo->refcount = 1;
 541     if (serial)
 542         strncpy(dinfo->serial, serial, sizeof(dinfo->serial) - 1);
 543     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 544
 545     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
 546
 547     /* disk I/O throttling */
 548     bdrv_set_io_limits(dinfo->bdrv, &io_limits);
 549
 550     switch(type) {
 551     case IF_IDE:
 552     case IF_SCSI:
 553     case IF_XEN:
 554     case IF_NONE:
 555         switch(media) {
 556         case MEDIA_DISK:
 557             if (cyls != 0) {
 558                 bdrv_set_geometry_hint(dinfo->bdrv, cyls, heads, secs);
 559                 bdrv_set_translation_hint(dinfo->bdrv, translation);
 560             }
 561             break;
 562         case MEDIA_CDROM:
 563             dinfo->media_cd = 1;
 564             break;
 565         }
 566         break;
 567     case IF_SD:
 568     case IF_FLOPPY:
 569     case IF_PFLASH:
 570     case IF_MTD:
 571         break;
 572     case IF_VIRTIO:
 573         /* add virtio block device */
 574         opts = qemu_opts_create(qemu_find_opts("device"), NULL, 0);
 575         if (arch_type == QEMU_ARCH_S390X) {
 576             qemu_opt_set(opts, "driver", "virtio-blk-s390");
 577         } else {
 578             qemu_opt_set(opts, "driver", "virtio-blk-pci");
 579         }
 580         qemu_opt_set(opts, "drive", dinfo->id);
 581         if (devaddr)
 582             qemu_opt_set(opts, "addr", devaddr);
 583         break;
 584     default:
 585         abort();
 586     }
 587     if (!file || !*file) {
 588         return dinfo;
 589     }
 590     if (snapshot) {
 591         /* always use cache=unsafe with snapshot */
 592         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 593         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 594     }
 595
 596     if (copy_on_read) {
 597         bdrv_flags |= BDRV_O_COPY_ON_READ;
 598     }
 599
 600     if (media == MEDIA_CDROM) {
 601         /* CDROM is fine for any interface, don't check.  */
 602         ro = 1;
 603     } else if (ro == 1) {
 604         if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
 605             type != IF_NONE && type != IF_PFLASH) {
 606             error_report("readonly not supported by this bus type");
 607             goto err;
 608         }
 609     }
 610
 611     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 612
 613     ret = bdrv_open(dinfo->bdrv, file, bdrv_flags, drv);
 614     if (ret < 0) {
 615         error_report("could not open disk image %s: %s",
 616                      file, strerror(-ret));
 617         goto err;
 618     }
 619
 620     if (bdrv_key_required(dinfo->bdrv))
 621         autostart = 0;
 622     return dinfo;
 623
 624 err:
 625     bdrv_delete(dinfo->bdrv);
 626     g_free(dinfo->id);
 627     QTAILQ_REMOVE(&drives, dinfo, next);
 628     g_free(dinfo);
 629     return NULL;
 630 }
 631
 632 void do_commit(Monitor *mon, const QDict *qdict)
 633 {
 634     const char *device = qdict_get_str(qdict, "device");
 635     BlockDriverState *bs;
 636
 637     if (!strcmp(device, "all")) {
 638         bdrv_commit_all();
 639     } else {
 640         int ret;
 641
 642         bs = bdrv_find(device);
 643         if (!bs) {
 644             qerror_report(QERR_DEVICE_NOT_FOUND, device);
 645             return;
 646         }
 647         ret = bdrv_commit(bs);
 648         if (ret == -EBUSY) {
 649             qerror_report(QERR_DEVICE_IN_USE, device);
 650             return;
 651         }
 652     }
 653 }
 654
 655 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
 656                                 bool has_format, const char *format,
 657                                 Error **errp)
 658 {
 659     BlockDriverState *bs;
 660     BlockDriver *drv, *old_drv, *proto_drv;
 661     int ret = 0;
 662     int flags;
 663     char old_filename[1024];
 664
 665     bs = bdrv_find(device);
 666     if (!bs) {
 667         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 668         return;
 669     }
 670     if (bdrv_in_use(bs)) {
 671         error_set(errp, QERR_DEVICE_IN_USE, device);
 672         return;
 673     }
 674
 675     pstrcpy(old_filename, sizeof(old_filename), bs->filename);
 676
 677     old_drv = bs->drv;
 678     flags = bs->open_flags;
 679
 680     if (!has_format) {
 681         format = "qcow2";
 682     }
 683
 684     drv = bdrv_find_format(format);
 685     if (!drv) {
 686         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 687         return;
 688     }
 689
 690     proto_drv = bdrv_find_protocol(snapshot_file);
 691     if (!proto_drv) {
 692         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 693         return;
 694     }
 695
 696     ret = bdrv_img_create(snapshot_file, format, bs->filename,
 697                           bs->drv->format_name, NULL, -1, flags);
 698     if (ret) {
 699         error_set(errp, QERR_UNDEFINED_ERROR);
 700         return;
 701     }
 702
 703     bdrv_drain_all();
 704     bdrv_flush(bs);
 705
 706     bdrv_close(bs);
 707     ret = bdrv_open(bs, snapshot_file, flags, drv);
 708     /*
 709      * If reopening the image file we just created fails, fall back
 710      * and try to re-open the original image. If that fails too, we
 711      * are in serious trouble.
 712      */
 713     if (ret != 0) {
 714         ret = bdrv_open(bs, old_filename, flags, old_drv);
 715         if (ret != 0) {
 716             error_set(errp, QERR_OPEN_FILE_FAILED, old_filename);
 717         } else {
 718             error_set(errp, QERR_OPEN_FILE_FAILED, snapshot_file);
 719         }
 720     }
 721 }
 722
 723
 724 /* New and old BlockDriverState structs for group snapshots */
 725 typedef struct BlkGroupSnapshotStates {
 726     BlockDriverState *old_bs;
 727     BlockDriverState *new_bs;
 728     QSIMPLEQ_ENTRY(BlkGroupSnapshotStates) entry;
 729 } BlkGroupSnapshotStates;
 730
 731 /*
 732  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
 733  *  then we do not pivot any of the devices in the group, and abandon the
 734  *  snapshots
 735  */
 736 void qmp_blockdev_group_snapshot_sync(SnapshotDevList *dev_list,
 737                                       Error **errp)
 738 {
 739     int ret = 0;
 740     SnapshotDevList *dev_entry = dev_list;
 741     SnapshotDev *dev_info = NULL;
 742     BlkGroupSnapshotStates *states;
 743     BlockDriver *proto_drv;
 744     BlockDriver *drv;
 745     int flags;
 746     const char *format;
 747     const char *snapshot_file;
 748
 749     QSIMPLEQ_HEAD(snap_bdrv_states, BlkGroupSnapshotStates) snap_bdrv_states;
 750     QSIMPLEQ_INIT(&snap_bdrv_states);
 751
 752     /* drain all i/o before any snapshots */
 753     bdrv_drain_all();
 754
 755     /* We don't do anything in this loop that commits us to the snapshot */
 756     while (NULL != dev_entry) {
 757         dev_info = dev_entry->value;
 758         dev_entry = dev_entry->next;
 759
 760         states = g_malloc0(sizeof(BlkGroupSnapshotStates));
 761         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, states, entry);
 762
 763         states->old_bs = bdrv_find(dev_info->device);
 764
 765         if (!states->old_bs) {
 766             error_set(errp, QERR_DEVICE_NOT_FOUND, dev_info->device);
 767             goto delete_and_fail;
 768         }
 769
 770         if (bdrv_in_use(states->old_bs)) {
 771             error_set(errp, QERR_DEVICE_IN_USE, dev_info->device);
 772             goto delete_and_fail;
 773         }
 774
 775         if (!bdrv_is_read_only(states->old_bs) &&
 776              bdrv_is_inserted(states->old_bs)) {
 777
 778             if (bdrv_flush(states->old_bs)) {
 779                 error_set(errp, QERR_IO_ERROR);
 780                 goto delete_and_fail;
 781             }
 782         }
 783
 784         snapshot_file = dev_info->snapshot_file;
 785
 786         flags = states->old_bs->open_flags;
 787
 788         if (!dev_info->has_format) {
 789             format = "qcow2";
 790         } else {
 791             format = dev_info->format;
 792         }
 793
 794         drv = bdrv_find_format(format);
 795         if (!drv) {
 796             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 797             goto delete_and_fail;
 798         }
 799
 800         proto_drv = bdrv_find_protocol(snapshot_file);
 801         if (!proto_drv) {
 802             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 803             goto delete_and_fail;
 804         }
 805
 806         /* create new image w/backing file */
 807         ret = bdrv_img_create(snapshot_file, format,
 808                               states->old_bs->filename,
 809                               drv->format_name, NULL, -1, flags);
 810         if (ret) {
 811             error_set(errp, QERR_OPEN_FILE_FAILED, snapshot_file);
 812             goto delete_and_fail;
 813         }
 814
 815         /* We will manually add the backing_hd field to the bs later */
 816         states->new_bs = bdrv_new("");
 817         ret = bdrv_open(states->new_bs, snapshot_file,
 818                         flags | BDRV_O_NO_BACKING, drv);
 819         if (ret != 0) {
 820             error_set(errp, QERR_OPEN_FILE_FAILED, snapshot_file);
 821             goto delete_and_fail;
 822         }
 823     }
 824
 825
 826     /* Now we are going to do the actual pivot.  Everything up to this point
 827      * is reversible, but we are committed at this point */
 828     QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
 829         /* This removes our old bs from the bdrv_states, and adds the new bs */
 830         bdrv_append(states->new_bs, states->old_bs);
 831     }
 832
 833     /* success */
 834     goto exit;
 835
 836 delete_and_fail:
 837     /*
 838     * failure, and it is all-or-none; abandon each new bs, and keep using
 839     * the original bs for all images
 840     */
 841     QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
 842         if (states->new_bs) {
 843              bdrv_delete(states->new_bs);
 844         }
 845     }
 846 exit:
 847     QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
 848         g_free(states);
 849     }
 850     return;
 851 }
 852
 853
 854 static void eject_device(BlockDriverState *bs, int force, Error **errp)
 855 {
 856     if (bdrv_in_use(bs)) {
 857         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
 858         return;
 859     }
 860     if (!bdrv_dev_has_removable_media(bs)) {
 861         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
 862         return;
 863     }
 864
 865     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
 866         bdrv_dev_eject_request(bs, force);
 867         if (!force) {
 868             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
 869             return;
 870         }
 871     }
 872
 873     bdrv_close(bs);
 874 }
 875
 876 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
 877 {
 878     BlockDriverState *bs;
 879
 880     bs = bdrv_find(device);
 881     if (!bs) {
 882         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 883         return;
 884     }
 885
 886     eject_device(bs, force, errp);
 887 }
 888
 889 void qmp_block_passwd(const char *device, const char *password, Error **errp)
 890 {
 891     BlockDriverState *bs;
 892     int err;
 893
 894     bs = bdrv_find(device);
 895     if (!bs) {
 896         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 897         return;
 898     }
 899
 900     err = bdrv_set_key(bs, password);
 901     if (err == -EINVAL) {
 902         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
 903         return;
 904     } else if (err < 0) {
 905         error_set(errp, QERR_INVALID_PASSWORD);
 906         return;
 907     }
 908 }
 909
 910 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
 911                                     int bdrv_flags, BlockDriver *drv,
 912                                     const char *password, Error **errp)
 913 {
 914     if (bdrv_open(bs, filename, bdrv_flags, drv) < 0) {
 915         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
 916         return;
 917     }
 918
 919     if (bdrv_key_required(bs)) {
 920         if (password) {
 921             if (bdrv_set_key(bs, password) < 0) {
 922                 error_set(errp, QERR_INVALID_PASSWORD);
 923             }
 924         } else {
 925             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
 926                       bdrv_get_encrypted_filename(bs));
 927         }
 928     } else if (password) {
 929         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
 930     }
 931 }
 932
 933 void qmp_change_blockdev(const char *device, const char *filename,
 934                          bool has_format, const char *format, Error **errp)
 935 {
 936     BlockDriverState *bs;
 937     BlockDriver *drv = NULL;
 938     int bdrv_flags;
 939     Error *err = NULL;
 940
 941     bs = bdrv_find(device);
 942     if (!bs) {
 943         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 944         return;
 945     }
 946
 947     if (format) {
 948         drv = bdrv_find_whitelisted_format(format);
 949         if (!drv) {
 950             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
 951             return;
 952         }
 953     }
 954
 955     eject_device(bs, 0, &err);
 956     if (error_is_set(&err)) {
 957         error_propagate(errp, err);
 958         return;
 959     }
 960
 961     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
 962     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
 963
 964     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
 965 }
 966
 967 /* throttling disk I/O limits */
 968 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
 969                                int64_t bps_wr, int64_t iops, int64_t iops_rd,
 970                                int64_t iops_wr, Error **errp)
 971 {
 972     BlockIOLimit io_limits;
 973     BlockDriverState *bs;
 974
 975     bs = bdrv_find(device);
 976     if (!bs) {
 977         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
 978         return;
 979     }
 980
 981     io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
 982     io_limits.bps[BLOCK_IO_LIMIT_READ]  = bps_rd;
 983     io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
 984     io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
 985     io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
 986     io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
 987
 988     if (!do_check_io_limits(&io_limits)) {
 989         error_set(errp, QERR_INVALID_PARAMETER_COMBINATION);
 990         return;
 991     }
 992
 993     bs->io_limits = io_limits;
 994     bs->slice_time = BLOCK_IO_SLICE_TIME;
 995
 996     if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
 997         bdrv_io_limits_enable(bs);
 998     } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
 999         bdrv_io_limits_disable(bs);
1000     } else {
1001         if (bs->block_timer) {
1002             qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
1003         }
1004     }
1005 }
1006
1007 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1008 {
1009     const char *id = qdict_get_str(qdict, "id");
1010     BlockDriverState *bs;
1011
1012     bs = bdrv_find(id);
1013     if (!bs) {
1014         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1015         return -1;
1016     }
1017     if (bdrv_in_use(bs)) {
1018         qerror_report(QERR_DEVICE_IN_USE, id);
1019         return -1;
1020     }
1021
1022     /* quiesce block driver; prevent further io */
1023     bdrv_drain_all();
1024     bdrv_flush(bs);
1025     bdrv_close(bs);
1026
1027     /* if we have a device attached to this BlockDriverState
1028      * then we need to make the drive anonymous until the device
1029      * can be removed.  If this is a drive with no device backing
1030      * then we can just get rid of the block driver state right here.
1031      */
1032     if (bdrv_get_attached_dev(bs)) {
1033         bdrv_make_anon(bs);
1034     } else {
1035         drive_uninit(drive_get_by_blockdev(bs));
1036     }
1037
1038     return 0;
1039 }
1040
1041 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1042 {
1043     BlockDriverState *bs;
1044
1045     bs = bdrv_find(device);
1046     if (!bs) {
1047         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1048         return;
1049     }
1050
1051     if (size < 0) {
1052         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1053         return;
1054     }
1055
1056     switch (bdrv_truncate(bs, size)) {
1057     case 0:
1058         break;
1059     case -ENOMEDIUM:
1060         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1061         break;
1062     case -ENOTSUP:
1063         error_set(errp, QERR_UNSUPPORTED);
1064         break;
1065     case -EACCES:
1066         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1067         break;
1068     case -EBUSY:
1069         error_set(errp, QERR_DEVICE_IN_USE, device);
1070         break;
1071     default:
1072         error_set(errp, QERR_UNDEFINED_ERROR);
1073         break;
1074     }
1075 }
1076
1077 static QObject *qobject_from_block_job(BlockJob *job)
1078 {
1079     return qobject_from_jsonf("{ 'type': %s,"
1080                               "'device': %s,"
1081                               "'len': %" PRId64 ","
1082                               "'offset': %" PRId64 ","
1083                               "'speed': %" PRId64 " }",
1084                               job->job_type->job_type,
1085                               bdrv_get_device_name(job->bs),
1086                               job->len,
1087                               job->offset,
1088                               job->speed);
1089 }
1090
1091 static void block_stream_cb(void *opaque, int ret)
1092 {
1093     BlockDriverState *bs = opaque;
1094     QObject *obj;
1095
1096     trace_block_stream_cb(bs, bs->job, ret);
1097
1098     assert(bs->job);
1099     obj = qobject_from_block_job(bs->job);
1100     if (ret < 0) {
1101         QDict *dict = qobject_to_qdict(obj);
1102         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1103     }
1104
1105     if (block_job_is_cancelled(bs->job)) {
1106         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1107     } else {
1108         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1109     }
1110     qobject_decref(obj);
1111
1112     drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1113 }
1114
1115 void qmp_block_stream(const char *device, bool has_base,
1116                       const char *base, Error **errp)
1117 {
1118     BlockDriverState *bs;
1119     BlockDriverState *base_bs = NULL;
1120     int ret;
1121
1122     bs = bdrv_find(device);
1123     if (!bs) {
1124         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1125         return;
1126     }
1127
1128     if (base) {
1129         base_bs = bdrv_find_backing_image(bs, base);
1130         if (base_bs == NULL) {
1131             error_set(errp, QERR_BASE_NOT_FOUND, base);
1132             return;
1133         }
1134     }
1135
1136     ret = stream_start(bs, base_bs, base, block_stream_cb, bs);
1137     if (ret < 0) {
1138         switch (ret) {
1139         case -EBUSY:
1140             error_set(errp, QERR_DEVICE_IN_USE, device);
1141             return;
1142         default:
1143             error_set(errp, QERR_NOT_SUPPORTED);
1144             return;
1145         }
1146     }
1147
1148     /* Grab a reference so hotplug does not delete the BlockDriverState from
1149      * underneath us.
1150      */
1151     drive_get_ref(drive_get_by_blockdev(bs));
1152
1153     trace_qmp_block_stream(bs, bs->job);
1154 }
1155
1156 static BlockJob *find_block_job(const char *device)
1157 {
1158     BlockDriverState *bs;
1159
1160     bs = bdrv_find(device);
1161     if (!bs || !bs->job) {
1162         return NULL;
1163     }
1164     return bs->job;
1165 }
1166
1167 void qmp_block_job_set_speed(const char *device, int64_t value, Error **errp)
1168 {
1169     BlockJob *job = find_block_job(device);
1170
1171     if (!job) {
1172         error_set(errp, QERR_DEVICE_NOT_ACTIVE, device);
1173         return;
1174     }
1175
1176     if (block_job_set_speed(job, value) < 0) {
1177         error_set(errp, QERR_NOT_SUPPORTED);
1178     }
1179 }
1180
1181 void qmp_block_job_cancel(const char *device, Error **errp)
1182 {
1183     BlockJob *job = find_block_job(device);
1184
1185     if (!job) {
1186         error_set(errp, QERR_DEVICE_NOT_ACTIVE, device);
1187         return;
1188     }
1189
1190     trace_qmp_block_job_cancel(job);
1191     block_job_cancel(job);
1192 }
1193
1194 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1195 {
1196     BlockJobInfoList **prev = opaque;
1197     BlockJob *job = bs->job;
1198
1199     if (job) {
1200         BlockJobInfoList *elem;
1201         BlockJobInfo *info = g_new(BlockJobInfo, 1);
1202         *info = (BlockJobInfo){
1203             .type   = g_strdup(job->job_type->job_type),
1204             .device = g_strdup(bdrv_get_device_name(bs)),
1205             .len    = job->len,
1206             .offset = job->offset,
1207             .speed  = job->speed,
1208         };
1209
1210         elem = g_new0(BlockJobInfoList, 1);
1211         elem->value = info;
1212
1213         (*prev)->next = elem;
1214         *prev = elem;
1215     }
1216 }
1217
1218 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1219 {
1220     /* Dummy is a fake list element for holding the head pointer */
1221     BlockJobInfoList dummy = {};
1222     BlockJobInfoList *prev = &dummy;
1223     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1224     return dummy.next;
1225 }