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 "qemu/option.h"
  38 #include "qemu/config-file.h"
  39 #include "qapi/qmp/types.h"
  40 #include "qapi-visit.h"
  41 #include "qapi/qmp-output-visitor.h"
  42 #include "qapi/util.h"
  43 #include "sysemu/sysemu.h"
  44 #include "block/block_int.h"
  45 #include "qmp-commands.h"
  46 #include "trace.h"
  47 #include "sysemu/arch_init.h"
  48
  49 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
  50
  51 static const char *const if_name[IF_COUNT] = {
  52     [IF_NONE] = "none",
  53     [IF_IDE] = "ide",
  54     [IF_SCSI] = "scsi",
  55     [IF_FLOPPY] = "floppy",
  56     [IF_PFLASH] = "pflash",
  57     [IF_MTD] = "mtd",
  58     [IF_SD] = "sd",
  59     [IF_VIRTIO] = "virtio",
  60     [IF_XEN] = "xen",
  61 };
  62
  63 static const int if_max_devs[IF_COUNT] = {
  64     /*
  65      * Do not change these numbers!  They govern how drive option
  66      * index maps to unit and bus.  That mapping is ABI.
  67      *
  68      * All controllers used to imlement if=T drives need to support
  69      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
  70      * Otherwise, some index values map to "impossible" bus, unit
  71      * values.
  72      *
  73      * For instance, if you change [IF_SCSI] to 255, -drive
  74      * if=scsi,index=12 no longer means bus=1,unit=5, but
  75      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
  76      * the drive can't be set up.  Regression.
  77      */
  78     [IF_IDE] = 2,
  79     [IF_SCSI] = 7,
  80 };
  81
  82 /*
  83  * We automatically delete the drive when a device using it gets
  84  * unplugged.  Questionable feature, but we can't just drop it.
  85  * Device models call blockdev_mark_auto_del() to schedule the
  86  * automatic deletion, and generic qdev code calls blockdev_auto_del()
  87  * when deletion is actually safe.
  88  */
  89 void blockdev_mark_auto_del(BlockDriverState *bs)
  90 {
  91     DriveInfo *dinfo = drive_get_by_blockdev(bs);
  92
  93     if (dinfo && !dinfo->enable_auto_del) {
  94         return;
  95     }
  96
  97     if (bs->job) {
  98         block_job_cancel(bs->job);
  99     }
 100     if (dinfo) {
 101         dinfo->auto_del = 1;
 102     }
 103 }
 104
 105 void blockdev_auto_del(BlockDriverState *bs)
 106 {
 107     DriveInfo *dinfo = drive_get_by_blockdev(bs);
 108
 109     if (dinfo && dinfo->auto_del) {
 110         drive_del(dinfo);
 111     }
 112 }
 113
 114 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
 115 {
 116     int max_devs = if_max_devs[type];
 117     return max_devs ? index / max_devs : 0;
 118 }
 119
 120 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
 121 {
 122     int max_devs = if_max_devs[type];
 123     return max_devs ? index % max_devs : index;
 124 }
 125
 126 QemuOpts *drive_def(const char *optstr)
 127 {
 128     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
 129 }
 130
 131 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
 132                     const char *optstr)
 133 {
 134     QemuOpts *opts;
 135     char buf[32];
 136
 137     opts = drive_def(optstr);
 138     if (!opts) {
 139         return NULL;
 140     }
 141     if (type != IF_DEFAULT) {
 142         qemu_opt_set(opts, "if", if_name[type]);
 143     }
 144     if (index >= 0) {
 145         snprintf(buf, sizeof(buf), "%d", index);
 146         qemu_opt_set(opts, "index", buf);
 147     }
 148     if (file)
 149         qemu_opt_set(opts, "file", file);
 150     return opts;
 151 }
 152
 153 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
 154 {
 155     DriveInfo *dinfo;
 156
 157     /* seek interface, bus and unit */
 158
 159     QTAILQ_FOREACH(dinfo, &drives, next) {
 160         if (dinfo->type == type &&
 161             dinfo->bus == bus &&
 162             dinfo->unit == unit)
 163             return dinfo;
 164     }
 165
 166     return NULL;
 167 }
 168
 169 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
 170 {
 171     return drive_get(type,
 172                      drive_index_to_bus_id(type, index),
 173                      drive_index_to_unit_id(type, index));
 174 }
 175
 176 int drive_get_max_bus(BlockInterfaceType type)
 177 {
 178     int max_bus;
 179     DriveInfo *dinfo;
 180
 181     max_bus = -1;
 182     QTAILQ_FOREACH(dinfo, &drives, next) {
 183         if(dinfo->type == type &&
 184            dinfo->bus > max_bus)
 185             max_bus = dinfo->bus;
 186     }
 187     return max_bus;
 188 }
 189
 190 /* Get a block device.  This should only be used for single-drive devices
 191    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
 192    appropriate bus.  */
 193 DriveInfo *drive_get_next(BlockInterfaceType type)
 194 {
 195     static int next_block_unit[IF_COUNT];
 196
 197     return drive_get(type, 0, next_block_unit[type]++);
 198 }
 199
 200 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
 201 {
 202     DriveInfo *dinfo;
 203
 204     QTAILQ_FOREACH(dinfo, &drives, next) {
 205         if (dinfo->bdrv == bs) {
 206             return dinfo;
 207         }
 208     }
 209     return NULL;
 210 }
 211
 212 static void bdrv_format_print(void *opaque, const char *name)
 213 {
 214     error_printf(" %s", name);
 215 }
 216
 217 void drive_del(DriveInfo *dinfo)
 218 {
 219     bdrv_unref(dinfo->bdrv);
 220 }
 221
 222 void drive_info_del(DriveInfo *dinfo)
 223 {
 224     if (!dinfo) {
 225         return;
 226     }
 227     if (dinfo->opts) {
 228         qemu_opts_del(dinfo->opts);
 229     }
 230     g_free(dinfo->id);
 231     QTAILQ_REMOVE(&drives, dinfo, next);
 232     g_free(dinfo->serial);
 233     g_free(dinfo);
 234 }
 235
 236 typedef struct {
 237     QEMUBH *bh;
 238     BlockDriverState *bs;
 239 } BDRVPutRefBH;
 240
 241 static void bdrv_put_ref_bh(void *opaque)
 242 {
 243     BDRVPutRefBH *s = opaque;
 244
 245     bdrv_unref(s->bs);
 246     qemu_bh_delete(s->bh);
 247     g_free(s);
 248 }
 249
 250 /*
 251  * Release a BDS reference in a BH
 252  *
 253  * It is not safe to use bdrv_unref() from a callback function when the callers
 254  * still need the BlockDriverState.  In such cases we schedule a BH to release
 255  * the reference.
 256  */
 257 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
 258 {
 259     BDRVPutRefBH *s;
 260
 261     s = g_new(BDRVPutRefBH, 1);
 262     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
 263     s->bs = bs;
 264     qemu_bh_schedule(s->bh);
 265 }
 266
 267 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
 268 {
 269     if (!strcmp(buf, "ignore")) {
 270         return BLOCKDEV_ON_ERROR_IGNORE;
 271     } else if (!is_read && !strcmp(buf, "enospc")) {
 272         return BLOCKDEV_ON_ERROR_ENOSPC;
 273     } else if (!strcmp(buf, "stop")) {
 274         return BLOCKDEV_ON_ERROR_STOP;
 275     } else if (!strcmp(buf, "report")) {
 276         return BLOCKDEV_ON_ERROR_REPORT;
 277     } else {
 278         error_setg(errp, "'%s' invalid %s error action",
 279                    buf, is_read ? "read" : "write");
 280         return -1;
 281     }
 282 }
 283
 284 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
 285 {
 286     if (throttle_conflicting(cfg)) {
 287         error_setg(errp, "bps/iops/max total values and read/write values"
 288                          " cannot be used at the same time");
 289         return false;
 290     }
 291
 292     if (!throttle_is_valid(cfg)) {
 293         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
 294         return false;
 295     }
 296
 297     return true;
 298 }
 299
 300 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
 301
 302 /* Takes the ownership of bs_opts */
 303 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
 304                                 Error **errp)
 305 {
 306     const char *buf;
 307     int ro = 0;
 308     int bdrv_flags = 0;
 309     int on_read_error, on_write_error;
 310     BlockDriverState *bs;
 311     DriveInfo *dinfo;
 312     ThrottleConfig cfg;
 313     int snapshot = 0;
 314     bool copy_on_read;
 315     int ret;
 316     Error *error = NULL;
 317     QemuOpts *opts;
 318     const char *id;
 319     bool has_driver_specific_opts;
 320     BlockdevDetectZeroesOptions detect_zeroes;
 321     BlockDriver *drv = NULL;
 322
 323     /* Check common options by copying from bs_opts to opts, all other options
 324      * stay in bs_opts for processing by bdrv_open(). */
 325     id = qdict_get_try_str(bs_opts, "id");
 326     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
 327     if (error) {
 328         error_propagate(errp, error);
 329         goto err_no_opts;
 330     }
 331
 332     qemu_opts_absorb_qdict(opts, bs_opts, &error);
 333     if (error) {
 334         error_propagate(errp, error);
 335         goto early_err;
 336     }
 337
 338     if (id) {
 339         qdict_del(bs_opts, "id");
 340     }
 341
 342     has_driver_specific_opts = !!qdict_size(bs_opts);
 343
 344     /* extract parameters */
 345     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 346     ro = qemu_opt_get_bool(opts, "read-only", 0);
 347     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
 348
 349     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
 350         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
 351             error_setg(errp, "invalid discard option");
 352             goto early_err;
 353         }
 354     }
 355
 356     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
 357         bdrv_flags |= BDRV_O_CACHE_WB;
 358     }
 359     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
 360         bdrv_flags |= BDRV_O_NOCACHE;
 361     }
 362     if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
 363         bdrv_flags |= BDRV_O_NO_FLUSH;
 364     }
 365
 366 #ifdef CONFIG_LINUX_AIO
 367     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 368         if (!strcmp(buf, "native")) {
 369             bdrv_flags |= BDRV_O_NATIVE_AIO;
 370         } else if (!strcmp(buf, "threads")) {
 371             /* this is the default */
 372         } else {
 373            error_setg(errp, "invalid aio option");
 374            goto early_err;
 375         }
 376     }
 377 #endif
 378
 379     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 380         if (is_help_option(buf)) {
 381             error_printf("Supported formats:");
 382             bdrv_iterate_format(bdrv_format_print, NULL);
 383             error_printf("\n");
 384             goto early_err;
 385         }
 386
 387         drv = bdrv_find_format(buf);
 388         if (!drv) {
 389             error_setg(errp, "'%s' invalid format", buf);
 390             goto early_err;
 391         }
 392     }
 393
 394     /* disk I/O throttling */
 395     memset(&cfg, 0, sizeof(cfg));
 396     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
 397         qemu_opt_get_number(opts, "throttling.bps-total", 0);
 398     cfg.buckets[THROTTLE_BPS_READ].avg  =
 399         qemu_opt_get_number(opts, "throttling.bps-read", 0);
 400     cfg.buckets[THROTTLE_BPS_WRITE].avg =
 401         qemu_opt_get_number(opts, "throttling.bps-write", 0);
 402     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
 403         qemu_opt_get_number(opts, "throttling.iops-total", 0);
 404     cfg.buckets[THROTTLE_OPS_READ].avg =
 405         qemu_opt_get_number(opts, "throttling.iops-read", 0);
 406     cfg.buckets[THROTTLE_OPS_WRITE].avg =
 407         qemu_opt_get_number(opts, "throttling.iops-write", 0);
 408
 409     cfg.buckets[THROTTLE_BPS_TOTAL].max =
 410         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
 411     cfg.buckets[THROTTLE_BPS_READ].max  =
 412         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
 413     cfg.buckets[THROTTLE_BPS_WRITE].max =
 414         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
 415     cfg.buckets[THROTTLE_OPS_TOTAL].max =
 416         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
 417     cfg.buckets[THROTTLE_OPS_READ].max =
 418         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
 419     cfg.buckets[THROTTLE_OPS_WRITE].max =
 420         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
 421
 422     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
 423
 424     if (!check_throttle_config(&cfg, &error)) {
 425         error_propagate(errp, error);
 426         goto early_err;
 427     }
 428
 429     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
 430     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 431         on_write_error = parse_block_error_action(buf, 0, &error);
 432         if (error) {
 433             error_propagate(errp, error);
 434             goto early_err;
 435         }
 436     }
 437
 438     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
 439     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 440         on_read_error = parse_block_error_action(buf, 1, &error);
 441         if (error) {
 442             error_propagate(errp, error);
 443             goto early_err;
 444         }
 445     }
 446
 447     detect_zeroes =
 448         qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
 449                         qemu_opt_get(opts, "detect-zeroes"),
 450                         BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
 451                         BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
 452                         &error);
 453     if (error) {
 454         error_propagate(errp, error);
 455         goto early_err;
 456     }
 457
 458     if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
 459         !(bdrv_flags & BDRV_O_UNMAP)) {
 460         error_setg(errp, "setting detect-zeroes to unmap is not allowed "
 461                          "without setting discard operation to unmap");
 462         goto early_err;
 463     }
 464
 465     /* init */
 466     bs = bdrv_new(qemu_opts_id(opts), errp);
 467     if (!bs) {
 468         goto early_err;
 469     }
 470     bs->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
 471     bs->read_only = ro;
 472     bs->detect_zeroes = detect_zeroes;
 473
 474     bdrv_set_on_error(bs, on_read_error, on_write_error);
 475
 476     /* disk I/O throttling */
 477     if (throttle_enabled(&cfg)) {
 478         bdrv_io_limits_enable(bs);
 479         bdrv_set_io_limits(bs, &cfg);
 480     }
 481
 482     dinfo = g_malloc0(sizeof(*dinfo));
 483     dinfo->id = g_strdup(qemu_opts_id(opts));
 484     dinfo->bdrv = bs;
 485     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 486
 487     if (!file || !*file) {
 488         if (has_driver_specific_opts) {
 489             file = NULL;
 490         } else {
 491             QDECREF(bs_opts);
 492             qemu_opts_del(opts);
 493             return dinfo;
 494         }
 495     }
 496     if (snapshot) {
 497         /* always use cache=unsafe with snapshot */
 498         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 499         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 500     }
 501
 502     if (copy_on_read) {
 503         bdrv_flags |= BDRV_O_COPY_ON_READ;
 504     }
 505
 506     if (runstate_check(RUN_STATE_INMIGRATE)) {
 507         bdrv_flags |= BDRV_O_INCOMING;
 508     }
 509
 510     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 511
 512     QINCREF(bs_opts);
 513     ret = bdrv_open(&bs, file, NULL, bs_opts, bdrv_flags, drv, &error);
 514     assert(bs == dinfo->bdrv);
 515
 516     if (ret < 0) {
 517         error_setg(errp, "could not open disk image %s: %s",
 518                    file ?: dinfo->id, error_get_pretty(error));
 519         error_free(error);
 520         goto err;
 521     }
 522
 523     if (bdrv_key_required(bs)) {
 524         autostart = 0;
 525     }
 526
 527     QDECREF(bs_opts);
 528     qemu_opts_del(opts);
 529
 530     return dinfo;
 531
 532 err:
 533     bdrv_unref(bs);
 534 early_err:
 535     qemu_opts_del(opts);
 536 err_no_opts:
 537     QDECREF(bs_opts);
 538     return NULL;
 539 }
 540
 541 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
 542 {
 543     const char *value;
 544
 545     value = qemu_opt_get(opts, from);
 546     if (value) {
 547         qemu_opt_set(opts, to, value);
 548         qemu_opt_unset(opts, from);
 549     }
 550 }
 551
 552 QemuOptsList qemu_legacy_drive_opts = {
 553     .name = "drive",
 554     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
 555     .desc = {
 556         {
 557             .name = "bus",
 558             .type = QEMU_OPT_NUMBER,
 559             .help = "bus number",
 560         },{
 561             .name = "unit",
 562             .type = QEMU_OPT_NUMBER,
 563             .help = "unit number (i.e. lun for scsi)",
 564         },{
 565             .name = "index",
 566             .type = QEMU_OPT_NUMBER,
 567             .help = "index number",
 568         },{
 569             .name = "media",
 570             .type = QEMU_OPT_STRING,
 571             .help = "media type (disk, cdrom)",
 572         },{
 573             .name = "if",
 574             .type = QEMU_OPT_STRING,
 575             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
 576         },{
 577             .name = "cyls",
 578             .type = QEMU_OPT_NUMBER,
 579             .help = "number of cylinders (ide disk geometry)",
 580         },{
 581             .name = "heads",
 582             .type = QEMU_OPT_NUMBER,
 583             .help = "number of heads (ide disk geometry)",
 584         },{
 585             .name = "secs",
 586             .type = QEMU_OPT_NUMBER,
 587             .help = "number of sectors (ide disk geometry)",
 588         },{
 589             .name = "trans",
 590             .type = QEMU_OPT_STRING,
 591             .help = "chs translation (auto, lba, none)",
 592         },{
 593             .name = "boot",
 594             .type = QEMU_OPT_BOOL,
 595             .help = "(deprecated, ignored)",
 596         },{
 597             .name = "addr",
 598             .type = QEMU_OPT_STRING,
 599             .help = "pci address (virtio only)",
 600         },{
 601             .name = "serial",
 602             .type = QEMU_OPT_STRING,
 603             .help = "disk serial number",
 604         },{
 605             .name = "file",
 606             .type = QEMU_OPT_STRING,
 607             .help = "file name",
 608         },
 609
 610         /* Options that are passed on, but have special semantics with -drive */
 611         {
 612             .name = "read-only",
 613             .type = QEMU_OPT_BOOL,
 614             .help = "open drive file as read-only",
 615         },{
 616             .name = "rerror",
 617             .type = QEMU_OPT_STRING,
 618             .help = "read error action",
 619         },{
 620             .name = "werror",
 621             .type = QEMU_OPT_STRING,
 622             .help = "write error action",
 623         },{
 624             .name = "copy-on-read",
 625             .type = QEMU_OPT_BOOL,
 626             .help = "copy read data from backing file into image file",
 627         },
 628
 629         { /* end of list */ }
 630     },
 631 };
 632
 633 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
 634 {
 635     const char *value;
 636     DriveInfo *dinfo = NULL;
 637     QDict *bs_opts;
 638     QemuOpts *legacy_opts;
 639     DriveMediaType media = MEDIA_DISK;
 640     BlockInterfaceType type;
 641     int cyls, heads, secs, translation;
 642     int max_devs, bus_id, unit_id, index;
 643     const char *devaddr;
 644     const char *werror, *rerror;
 645     bool read_only = false;
 646     bool copy_on_read;
 647     const char *serial;
 648     const char *filename;
 649     Error *local_err = NULL;
 650
 651     /* Change legacy command line options into QMP ones */
 652     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
 653     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
 654     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
 655
 656     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
 657     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
 658     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
 659
 660     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
 661     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
 662     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
 663
 664     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
 665     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
 666     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
 667
 668     qemu_opt_rename(all_opts,
 669                     "iops_size", "throttling.iops-size");
 670
 671     qemu_opt_rename(all_opts, "readonly", "read-only");
 672
 673     value = qemu_opt_get(all_opts, "cache");
 674     if (value) {
 675         int flags = 0;
 676
 677         if (bdrv_parse_cache_flags(value, &flags) != 0) {
 678             error_report("invalid cache option");
 679             return NULL;
 680         }
 681
 682         /* Specific options take precedence */
 683         if (!qemu_opt_get(all_opts, "cache.writeback")) {
 684             qemu_opt_set_bool(all_opts, "cache.writeback",
 685                               !!(flags & BDRV_O_CACHE_WB));
 686         }
 687         if (!qemu_opt_get(all_opts, "cache.direct")) {
 688             qemu_opt_set_bool(all_opts, "cache.direct",
 689                               !!(flags & BDRV_O_NOCACHE));
 690         }
 691         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
 692             qemu_opt_set_bool(all_opts, "cache.no-flush",
 693                               !!(flags & BDRV_O_NO_FLUSH));
 694         }
 695         qemu_opt_unset(all_opts, "cache");
 696     }
 697
 698     /* Get a QDict for processing the options */
 699     bs_opts = qdict_new();
 700     qemu_opts_to_qdict(all_opts, bs_opts);
 701
 702     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
 703                                    &error_abort);
 704     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
 705     if (local_err) {
 706         error_report("%s", error_get_pretty(local_err));
 707         error_free(local_err);
 708         goto fail;
 709     }
 710
 711     /* Deprecated option boot=[on|off] */
 712     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
 713         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 714                 "ignored. Future versions will reject this parameter. Please "
 715                 "update your scripts.\n");
 716     }
 717
 718     /* Media type */
 719     value = qemu_opt_get(legacy_opts, "media");
 720     if (value) {
 721         if (!strcmp(value, "disk")) {
 722             media = MEDIA_DISK;
 723         } else if (!strcmp(value, "cdrom")) {
 724             media = MEDIA_CDROM;
 725             read_only = true;
 726         } else {
 727             error_report("'%s' invalid media", value);
 728             goto fail;
 729         }
 730     }
 731
 732     /* copy-on-read is disabled with a warning for read-only devices */
 733     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
 734     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
 735
 736     if (read_only && copy_on_read) {
 737         error_report("warning: disabling copy-on-read on read-only drive");
 738         copy_on_read = false;
 739     }
 740
 741     qdict_put(bs_opts, "read-only",
 742               qstring_from_str(read_only ? "on" : "off"));
 743     qdict_put(bs_opts, "copy-on-read",
 744               qstring_from_str(copy_on_read ? "on" :"off"));
 745
 746     /* Controller type */
 747     value = qemu_opt_get(legacy_opts, "if");
 748     if (value) {
 749         for (type = 0;
 750              type < IF_COUNT && strcmp(value, if_name[type]);
 751              type++) {
 752         }
 753         if (type == IF_COUNT) {
 754             error_report("unsupported bus type '%s'", value);
 755             goto fail;
 756         }
 757     } else {
 758         type = block_default_type;
 759     }
 760
 761     /* Geometry */
 762     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
 763     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
 764     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
 765
 766     if (cyls || heads || secs) {
 767         if (cyls < 1) {
 768             error_report("invalid physical cyls number");
 769             goto fail;
 770         }
 771         if (heads < 1) {
 772             error_report("invalid physical heads number");
 773             goto fail;
 774         }
 775         if (secs < 1) {
 776             error_report("invalid physical secs number");
 777             goto fail;
 778         }
 779     }
 780
 781     translation = BIOS_ATA_TRANSLATION_AUTO;
 782     value = qemu_opt_get(legacy_opts, "trans");
 783     if (value != NULL) {
 784         if (!cyls) {
 785             error_report("'%s' trans must be used with cyls, heads and secs",
 786                          value);
 787             goto fail;
 788         }
 789         if (!strcmp(value, "none")) {
 790             translation = BIOS_ATA_TRANSLATION_NONE;
 791         } else if (!strcmp(value, "lba")) {
 792             translation = BIOS_ATA_TRANSLATION_LBA;
 793         } else if (!strcmp(value, "large")) {
 794             translation = BIOS_ATA_TRANSLATION_LARGE;
 795         } else if (!strcmp(value, "rechs")) {
 796             translation = BIOS_ATA_TRANSLATION_RECHS;
 797         } else if (!strcmp(value, "auto")) {
 798             translation = BIOS_ATA_TRANSLATION_AUTO;
 799         } else {
 800             error_report("'%s' invalid translation type", value);
 801             goto fail;
 802         }
 803     }
 804
 805     if (media == MEDIA_CDROM) {
 806         if (cyls || secs || heads) {
 807             error_report("CHS can't be set with media=cdrom");
 808             goto fail;
 809         }
 810     }
 811
 812     /* Device address specified by bus/unit or index.
 813      * If none was specified, try to find the first free one. */
 814     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
 815     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
 816     index   = qemu_opt_get_number(legacy_opts, "index", -1);
 817
 818     max_devs = if_max_devs[type];
 819
 820     if (index != -1) {
 821         if (bus_id != 0 || unit_id != -1) {
 822             error_report("index cannot be used with bus and unit");
 823             goto fail;
 824         }
 825         bus_id = drive_index_to_bus_id(type, index);
 826         unit_id = drive_index_to_unit_id(type, index);
 827     }
 828
 829     if (unit_id == -1) {
 830        unit_id = 0;
 831        while (drive_get(type, bus_id, unit_id) != NULL) {
 832            unit_id++;
 833            if (max_devs && unit_id >= max_devs) {
 834                unit_id -= max_devs;
 835                bus_id++;
 836            }
 837        }
 838     }
 839
 840     if (max_devs && unit_id >= max_devs) {
 841         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
 842         goto fail;
 843     }
 844
 845     if (drive_get(type, bus_id, unit_id) != NULL) {
 846         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 847                      bus_id, unit_id, index);
 848         goto fail;
 849     }
 850
 851     /* Serial number */
 852     serial = qemu_opt_get(legacy_opts, "serial");
 853
 854     /* no id supplied -> create one */
 855     if (qemu_opts_id(all_opts) == NULL) {
 856         char *new_id;
 857         const char *mediastr = "";
 858         if (type == IF_IDE || type == IF_SCSI) {
 859             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 860         }
 861         if (max_devs) {
 862             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
 863                                      mediastr, unit_id);
 864         } else {
 865             new_id = g_strdup_printf("%s%s%i", if_name[type],
 866                                      mediastr, unit_id);
 867         }
 868         qdict_put(bs_opts, "id", qstring_from_str(new_id));
 869         g_free(new_id);
 870     }
 871
 872     /* Add virtio block device */
 873     devaddr = qemu_opt_get(legacy_opts, "addr");
 874     if (devaddr && type != IF_VIRTIO) {
 875         error_report("addr is not supported by this bus type");
 876         goto fail;
 877     }
 878
 879     if (type == IF_VIRTIO) {
 880         QemuOpts *devopts;
 881         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
 882                                    &error_abort);
 883         if (arch_type == QEMU_ARCH_S390X) {
 884             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
 885         } else {
 886             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
 887         }
 888         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
 889         if (devaddr) {
 890             qemu_opt_set(devopts, "addr", devaddr);
 891         }
 892     }
 893
 894     filename = qemu_opt_get(legacy_opts, "file");
 895
 896     /* Check werror/rerror compatibility with if=... */
 897     werror = qemu_opt_get(legacy_opts, "werror");
 898     if (werror != NULL) {
 899         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
 900             type != IF_NONE) {
 901             error_report("werror is not supported by this bus type");
 902             goto fail;
 903         }
 904         qdict_put(bs_opts, "werror", qstring_from_str(werror));
 905     }
 906
 907     rerror = qemu_opt_get(legacy_opts, "rerror");
 908     if (rerror != NULL) {
 909         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
 910             type != IF_NONE) {
 911             error_report("rerror is not supported by this bus type");
 912             goto fail;
 913         }
 914         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
 915     }
 916
 917     /* Actual block device init: Functionality shared with blockdev-add */
 918     dinfo = blockdev_init(filename, bs_opts, &local_err);
 919     bs_opts = NULL;
 920     if (dinfo == NULL) {
 921         if (local_err) {
 922             error_report("%s", error_get_pretty(local_err));
 923             error_free(local_err);
 924         }
 925         goto fail;
 926     } else {
 927         assert(!local_err);
 928     }
 929
 930     /* Set legacy DriveInfo fields */
 931     dinfo->enable_auto_del = true;
 932     dinfo->opts = all_opts;
 933
 934     dinfo->cyls = cyls;
 935     dinfo->heads = heads;
 936     dinfo->secs = secs;
 937     dinfo->trans = translation;
 938
 939     dinfo->type = type;
 940     dinfo->bus = bus_id;
 941     dinfo->unit = unit_id;
 942     dinfo->devaddr = devaddr;
 943
 944     dinfo->serial = g_strdup(serial);
 945
 946     switch(type) {
 947     case IF_IDE:
 948     case IF_SCSI:
 949     case IF_XEN:
 950     case IF_NONE:
 951         dinfo->media_cd = media == MEDIA_CDROM;
 952         break;
 953     default:
 954         break;
 955     }
 956
 957 fail:
 958     qemu_opts_del(legacy_opts);
 959     QDECREF(bs_opts);
 960     return dinfo;
 961 }
 962
 963 void do_commit(Monitor *mon, const QDict *qdict)
 964 {
 965     const char *device = qdict_get_str(qdict, "device");
 966     BlockDriverState *bs;
 967     int ret;
 968
 969     if (!strcmp(device, "all")) {
 970         ret = bdrv_commit_all();
 971     } else {
 972         bs = bdrv_find(device);
 973         if (!bs) {
 974             monitor_printf(mon, "Device '%s' not found\n", device);
 975             return;
 976         }
 977         ret = bdrv_commit(bs);
 978     }
 979     if (ret < 0) {
 980         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
 981                        strerror(-ret));
 982     }
 983 }
 984
 985 static void blockdev_do_action(int kind, void *data, Error **errp)
 986 {
 987     TransactionAction action;
 988     TransactionActionList list;
 989
 990     action.kind = kind;
 991     action.data = data;
 992     list.value = &action;
 993     list.next = NULL;
 994     qmp_transaction(&list, errp);
 995 }
 996
 997 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
 998                                 bool has_node_name, const char *node_name,
 999                                 const char *snapshot_file,
1000                                 bool has_snapshot_node_name,
1001                                 const char *snapshot_node_name,
1002                                 bool has_format, const char *format,
1003                                 bool has_mode, NewImageMode mode, Error **errp)
1004 {
1005     BlockdevSnapshot snapshot = {
1006         .has_device = has_device,
1007         .device = (char *) device,
1008         .has_node_name = has_node_name,
1009         .node_name = (char *) node_name,
1010         .snapshot_file = (char *) snapshot_file,
1011         .has_snapshot_node_name = has_snapshot_node_name,
1012         .snapshot_node_name = (char *) snapshot_node_name,
1013         .has_format = has_format,
1014         .format = (char *) format,
1015         .has_mode = has_mode,
1016         .mode = mode,
1017     };
1018     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1019                        &snapshot, errp);
1020 }
1021
1022 void qmp_blockdev_snapshot_internal_sync(const char *device,
1023                                          const char *name,
1024                                          Error **errp)
1025 {
1026     BlockdevSnapshotInternal snapshot = {
1027         .device = (char *) device,
1028         .name = (char *) name
1029     };
1030
1031     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1032                        &snapshot, errp);
1033 }
1034
1035 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1036                                                          bool has_id,
1037                                                          const char *id,
1038                                                          bool has_name,
1039                                                          const char *name,
1040                                                          Error **errp)
1041 {
1042     BlockDriverState *bs = bdrv_find(device);
1043     QEMUSnapshotInfo sn;
1044     Error *local_err = NULL;
1045     SnapshotInfo *info = NULL;
1046     int ret;
1047
1048     if (!bs) {
1049         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1050         return NULL;
1051     }
1052
1053     if (!has_id) {
1054         id = NULL;
1055     }
1056
1057     if (!has_name) {
1058         name = NULL;
1059     }
1060
1061     if (!id && !name) {
1062         error_setg(errp, "Name or id must be provided");
1063         return NULL;
1064     }
1065
1066     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1067     if (local_err) {
1068         error_propagate(errp, local_err);
1069         return NULL;
1070     }
1071     if (!ret) {
1072         error_setg(errp,
1073                    "Snapshot with id '%s' and name '%s' does not exist on "
1074                    "device '%s'",
1075                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1076         return NULL;
1077     }
1078
1079     bdrv_snapshot_delete(bs, id, name, &local_err);
1080     if (local_err) {
1081         error_propagate(errp, local_err);
1082         return NULL;
1083     }
1084
1085     info = g_new0(SnapshotInfo, 1);
1086     info->id = g_strdup(sn.id_str);
1087     info->name = g_strdup(sn.name);
1088     info->date_nsec = sn.date_nsec;
1089     info->date_sec = sn.date_sec;
1090     info->vm_state_size = sn.vm_state_size;
1091     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1092     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1093
1094     return info;
1095 }
1096
1097 /* New and old BlockDriverState structs for group snapshots */
1098
1099 typedef struct BlkTransactionState BlkTransactionState;
1100
1101 /* Only prepare() may fail. In a single transaction, only one of commit() or
1102    abort() will be called, clean() will always be called if it present. */
1103 typedef struct BdrvActionOps {
1104     /* Size of state struct, in bytes. */
1105     size_t instance_size;
1106     /* Prepare the work, must NOT be NULL. */
1107     void (*prepare)(BlkTransactionState *common, Error **errp);
1108     /* Commit the changes, can be NULL. */
1109     void (*commit)(BlkTransactionState *common);
1110     /* Abort the changes on fail, can be NULL. */
1111     void (*abort)(BlkTransactionState *common);
1112     /* Clean up resource in the end, can be NULL. */
1113     void (*clean)(BlkTransactionState *common);
1114 } BdrvActionOps;
1115
1116 /*
1117  * This structure must be arranged as first member in child type, assuming
1118  * that compiler will also arrange it to the same address with parent instance.
1119  * Later it will be used in free().
1120  */
1121 struct BlkTransactionState {
1122     TransactionAction *action;
1123     const BdrvActionOps *ops;
1124     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1125 };
1126
1127 /* internal snapshot private data */
1128 typedef struct InternalSnapshotState {
1129     BlkTransactionState common;
1130     BlockDriverState *bs;
1131     QEMUSnapshotInfo sn;
1132 } InternalSnapshotState;
1133
1134 static void internal_snapshot_prepare(BlkTransactionState *common,
1135                                       Error **errp)
1136 {
1137     Error *local_err = NULL;
1138     const char *device;
1139     const char *name;
1140     BlockDriverState *bs;
1141     QEMUSnapshotInfo old_sn, *sn;
1142     bool ret;
1143     qemu_timeval tv;
1144     BlockdevSnapshotInternal *internal;
1145     InternalSnapshotState *state;
1146     int ret1;
1147
1148     g_assert(common->action->kind ==
1149              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1150     internal = common->action->blockdev_snapshot_internal_sync;
1151     state = DO_UPCAST(InternalSnapshotState, common, common);
1152
1153     /* 1. parse input */
1154     device = internal->device;
1155     name = internal->name;
1156
1157     /* 2. check for validation */
1158     bs = bdrv_find(device);
1159     if (!bs) {
1160         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1161         return;
1162     }
1163
1164     if (!bdrv_is_inserted(bs)) {
1165         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1166         return;
1167     }
1168
1169     if (bdrv_is_read_only(bs)) {
1170         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1171         return;
1172     }
1173
1174     if (!bdrv_can_snapshot(bs)) {
1175         error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1176                   bs->drv->format_name, device, "internal snapshot");
1177         return;
1178     }
1179
1180     if (!strlen(name)) {
1181         error_setg(errp, "Name is empty");
1182         return;
1183     }
1184
1185     /* check whether a snapshot with name exist */
1186     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1187                                             &local_err);
1188     if (local_err) {
1189         error_propagate(errp, local_err);
1190         return;
1191     } else if (ret) {
1192         error_setg(errp,
1193                    "Snapshot with name '%s' already exists on device '%s'",
1194                    name, device);
1195         return;
1196     }
1197
1198     /* 3. take the snapshot */
1199     sn = &state->sn;
1200     pstrcpy(sn->name, sizeof(sn->name), name);
1201     qemu_gettimeofday(&tv);
1202     sn->date_sec = tv.tv_sec;
1203     sn->date_nsec = tv.tv_usec * 1000;
1204     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1205
1206     ret1 = bdrv_snapshot_create(bs, sn);
1207     if (ret1 < 0) {
1208         error_setg_errno(errp, -ret1,
1209                          "Failed to create snapshot '%s' on device '%s'",
1210                          name, device);
1211         return;
1212     }
1213
1214     /* 4. succeed, mark a snapshot is created */
1215     state->bs = bs;
1216 }
1217
1218 static void internal_snapshot_abort(BlkTransactionState *common)
1219 {
1220     InternalSnapshotState *state =
1221                              DO_UPCAST(InternalSnapshotState, common, common);
1222     BlockDriverState *bs = state->bs;
1223     QEMUSnapshotInfo *sn = &state->sn;
1224     Error *local_error = NULL;
1225
1226     if (!bs) {
1227         return;
1228     }
1229
1230     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1231         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1232                      "device '%s' in abort: %s",
1233                      sn->id_str,
1234                      sn->name,
1235                      bdrv_get_device_name(bs),
1236                      error_get_pretty(local_error));
1237         error_free(local_error);
1238     }
1239 }
1240
1241 /* external snapshot private data */
1242 typedef struct ExternalSnapshotState {
1243     BlkTransactionState common;
1244     BlockDriverState *old_bs;
1245     BlockDriverState *new_bs;
1246 } ExternalSnapshotState;
1247
1248 static void external_snapshot_prepare(BlkTransactionState *common,
1249                                       Error **errp)
1250 {
1251     BlockDriver *drv;
1252     int flags, ret;
1253     QDict *options = NULL;
1254     Error *local_err = NULL;
1255     bool has_device = false;
1256     const char *device;
1257     bool has_node_name = false;
1258     const char *node_name;
1259     bool has_snapshot_node_name = false;
1260     const char *snapshot_node_name;
1261     const char *new_image_file;
1262     const char *format = "qcow2";
1263     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1264     ExternalSnapshotState *state =
1265                              DO_UPCAST(ExternalSnapshotState, common, common);
1266     TransactionAction *action = common->action;
1267
1268     /* get parameters */
1269     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1270
1271     has_device = action->blockdev_snapshot_sync->has_device;
1272     device = action->blockdev_snapshot_sync->device;
1273     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1274     node_name = action->blockdev_snapshot_sync->node_name;
1275     has_snapshot_node_name =
1276         action->blockdev_snapshot_sync->has_snapshot_node_name;
1277     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1278
1279     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1280     if (action->blockdev_snapshot_sync->has_format) {
1281         format = action->blockdev_snapshot_sync->format;
1282     }
1283     if (action->blockdev_snapshot_sync->has_mode) {
1284         mode = action->blockdev_snapshot_sync->mode;
1285     }
1286
1287     /* start processing */
1288     drv = bdrv_find_format(format);
1289     if (!drv) {
1290         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1291         return;
1292     }
1293
1294     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1295                                    has_node_name ? node_name : NULL,
1296                                    &local_err);
1297     if (local_err) {
1298         error_propagate(errp, local_err);
1299         return;
1300     }
1301
1302     if (has_node_name && !has_snapshot_node_name) {
1303         error_setg(errp, "New snapshot node name missing");
1304         return;
1305     }
1306
1307     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1308         error_setg(errp, "New snapshot node name already existing");
1309         return;
1310     }
1311
1312     if (!bdrv_is_inserted(state->old_bs)) {
1313         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1314         return;
1315     }
1316
1317     if (bdrv_op_is_blocked(state->old_bs,
1318                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1319         return;
1320     }
1321
1322     if (!bdrv_is_read_only(state->old_bs)) {
1323         if (bdrv_flush(state->old_bs)) {
1324             error_set(errp, QERR_IO_ERROR);
1325             return;
1326         }
1327     }
1328
1329     if (!bdrv_is_first_non_filter(state->old_bs)) {
1330         error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1331         return;
1332     }
1333
1334     flags = state->old_bs->open_flags;
1335
1336     /* create new image w/backing file */
1337     if (mode != NEW_IMAGE_MODE_EXISTING) {
1338         bdrv_img_create(new_image_file, format,
1339                         state->old_bs->filename,
1340                         state->old_bs->drv->format_name,
1341                         NULL, -1, flags, &local_err, false);
1342         if (local_err) {
1343             error_propagate(errp, local_err);
1344             return;
1345         }
1346     }
1347
1348     if (has_snapshot_node_name) {
1349         options = qdict_new();
1350         qdict_put(options, "node-name",
1351                   qstring_from_str(snapshot_node_name));
1352     }
1353
1354     /* TODO Inherit bs->options or only take explicit options with an
1355      * extended QMP command? */
1356     assert(state->new_bs == NULL);
1357     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1358                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1359     /* We will manually add the backing_hd field to the bs later */
1360     if (ret != 0) {
1361         error_propagate(errp, local_err);
1362     }
1363 }
1364
1365 static void external_snapshot_commit(BlkTransactionState *common)
1366 {
1367     ExternalSnapshotState *state =
1368                              DO_UPCAST(ExternalSnapshotState, common, common);
1369
1370     /* This removes our old bs and adds the new bs */
1371     bdrv_append(state->new_bs, state->old_bs);
1372     /* We don't need (or want) to use the transactional
1373      * bdrv_reopen_multiple() across all the entries at once, because we
1374      * don't want to abort all of them if one of them fails the reopen */
1375     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1376                 NULL);
1377 }
1378
1379 static void external_snapshot_abort(BlkTransactionState *common)
1380 {
1381     ExternalSnapshotState *state =
1382                              DO_UPCAST(ExternalSnapshotState, common, common);
1383     if (state->new_bs) {
1384         bdrv_unref(state->new_bs);
1385     }
1386 }
1387
1388 typedef struct DriveBackupState {
1389     BlkTransactionState common;
1390     BlockDriverState *bs;
1391     BlockJob *job;
1392 } DriveBackupState;
1393
1394 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1395 {
1396     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1397     DriveBackup *backup;
1398     Error *local_err = NULL;
1399
1400     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1401     backup = common->action->drive_backup;
1402
1403     qmp_drive_backup(backup->device, backup->target,
1404                      backup->has_format, backup->format,
1405                      backup->sync,
1406                      backup->has_mode, backup->mode,
1407                      backup->has_speed, backup->speed,
1408                      backup->has_on_source_error, backup->on_source_error,
1409                      backup->has_on_target_error, backup->on_target_error,
1410                      &local_err);
1411     if (local_err) {
1412         error_propagate(errp, local_err);
1413         state->bs = NULL;
1414         state->job = NULL;
1415         return;
1416     }
1417
1418     state->bs = bdrv_find(backup->device);
1419     state->job = state->bs->job;
1420 }
1421
1422 static void drive_backup_abort(BlkTransactionState *common)
1423 {
1424     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1425     BlockDriverState *bs = state->bs;
1426
1427     /* Only cancel if it's the job we started */
1428     if (bs && bs->job && bs->job == state->job) {
1429         block_job_cancel_sync(bs->job);
1430     }
1431 }
1432
1433 static void abort_prepare(BlkTransactionState *common, Error **errp)
1434 {
1435     error_setg(errp, "Transaction aborted using Abort action");
1436 }
1437
1438 static void abort_commit(BlkTransactionState *common)
1439 {
1440     g_assert_not_reached(); /* this action never succeeds */
1441 }
1442
1443 static const BdrvActionOps actions[] = {
1444     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1445         .instance_size = sizeof(ExternalSnapshotState),
1446         .prepare  = external_snapshot_prepare,
1447         .commit   = external_snapshot_commit,
1448         .abort = external_snapshot_abort,
1449     },
1450     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1451         .instance_size = sizeof(DriveBackupState),
1452         .prepare = drive_backup_prepare,
1453         .abort = drive_backup_abort,
1454     },
1455     [TRANSACTION_ACTION_KIND_ABORT] = {
1456         .instance_size = sizeof(BlkTransactionState),
1457         .prepare = abort_prepare,
1458         .commit = abort_commit,
1459     },
1460     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1461         .instance_size = sizeof(InternalSnapshotState),
1462         .prepare  = internal_snapshot_prepare,
1463         .abort = internal_snapshot_abort,
1464     },
1465 };
1466
1467 /*
1468  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1469  *  then we do not pivot any of the devices in the group, and abandon the
1470  *  snapshots
1471  */
1472 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1473 {
1474     TransactionActionList *dev_entry = dev_list;
1475     BlkTransactionState *state, *next;
1476     Error *local_err = NULL;
1477
1478     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1479     QSIMPLEQ_INIT(&snap_bdrv_states);
1480
1481     /* drain all i/o before any snapshots */
1482     bdrv_drain_all();
1483
1484     /* We don't do anything in this loop that commits us to the snapshot */
1485     while (NULL != dev_entry) {
1486         TransactionAction *dev_info = NULL;
1487         const BdrvActionOps *ops;
1488
1489         dev_info = dev_entry->value;
1490         dev_entry = dev_entry->next;
1491
1492         assert(dev_info->kind < ARRAY_SIZE(actions));
1493
1494         ops = &actions[dev_info->kind];
1495         assert(ops->instance_size > 0);
1496
1497         state = g_malloc0(ops->instance_size);
1498         state->ops = ops;
1499         state->action = dev_info;
1500         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1501
1502         state->ops->prepare(state, &local_err);
1503         if (local_err) {
1504             error_propagate(errp, local_err);
1505             goto delete_and_fail;
1506         }
1507     }
1508
1509     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1510         if (state->ops->commit) {
1511             state->ops->commit(state);
1512         }
1513     }
1514
1515     /* success */
1516     goto exit;
1517
1518 delete_and_fail:
1519     /*
1520     * failure, and it is all-or-none; abandon each new bs, and keep using
1521     * the original bs for all images
1522     */
1523     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1524         if (state->ops->abort) {
1525             state->ops->abort(state);
1526         }
1527     }
1528 exit:
1529     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1530         if (state->ops->clean) {
1531             state->ops->clean(state);
1532         }
1533         g_free(state);
1534     }
1535 }
1536
1537
1538 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1539 {
1540     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1541         return;
1542     }
1543     if (!bdrv_dev_has_removable_media(bs)) {
1544         error_setg(errp, "Device '%s' is not removable",
1545                    bdrv_get_device_name(bs));
1546         return;
1547     }
1548
1549     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1550         bdrv_dev_eject_request(bs, force);
1551         if (!force) {
1552             error_setg(errp, "Device '%s' is locked",
1553                        bdrv_get_device_name(bs));
1554             return;
1555         }
1556     }
1557
1558     bdrv_close(bs);
1559 }
1560
1561 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1562 {
1563     BlockDriverState *bs;
1564
1565     bs = bdrv_find(device);
1566     if (!bs) {
1567         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1568         return;
1569     }
1570
1571     eject_device(bs, force, errp);
1572 }
1573
1574 void qmp_block_passwd(bool has_device, const char *device,
1575                       bool has_node_name, const char *node_name,
1576                       const char *password, Error **errp)
1577 {
1578     Error *local_err = NULL;
1579     BlockDriverState *bs;
1580     int err;
1581
1582     bs = bdrv_lookup_bs(has_device ? device : NULL,
1583                         has_node_name ? node_name : NULL,
1584                         &local_err);
1585     if (local_err) {
1586         error_propagate(errp, local_err);
1587         return;
1588     }
1589
1590     err = bdrv_set_key(bs, password);
1591     if (err == -EINVAL) {
1592         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1593         return;
1594     } else if (err < 0) {
1595         error_set(errp, QERR_INVALID_PASSWORD);
1596         return;
1597     }
1598 }
1599
1600 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1601                                     int bdrv_flags, BlockDriver *drv,
1602                                     const char *password, Error **errp)
1603 {
1604     Error *local_err = NULL;
1605     int ret;
1606
1607     ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1608     if (ret < 0) {
1609         error_propagate(errp, local_err);
1610         return;
1611     }
1612
1613     if (bdrv_key_required(bs)) {
1614         if (password) {
1615             if (bdrv_set_key(bs, password) < 0) {
1616                 error_set(errp, QERR_INVALID_PASSWORD);
1617             }
1618         } else {
1619             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1620                       bdrv_get_encrypted_filename(bs));
1621         }
1622     } else if (password) {
1623         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1624     }
1625 }
1626
1627 void qmp_change_blockdev(const char *device, const char *filename,
1628                          const char *format, Error **errp)
1629 {
1630     BlockDriverState *bs;
1631     BlockDriver *drv = NULL;
1632     int bdrv_flags;
1633     Error *err = NULL;
1634
1635     bs = bdrv_find(device);
1636     if (!bs) {
1637         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1638         return;
1639     }
1640
1641     if (format) {
1642         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1643         if (!drv) {
1644             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1645             return;
1646         }
1647     }
1648
1649     eject_device(bs, 0, &err);
1650     if (err) {
1651         error_propagate(errp, err);
1652         return;
1653     }
1654
1655     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1656     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1657
1658     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1659 }
1660
1661 /* throttling disk I/O limits */
1662 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1663                                int64_t bps_wr,
1664                                int64_t iops,
1665                                int64_t iops_rd,
1666                                int64_t iops_wr,
1667                                bool has_bps_max,
1668                                int64_t bps_max,
1669                                bool has_bps_rd_max,
1670                                int64_t bps_rd_max,
1671                                bool has_bps_wr_max,
1672                                int64_t bps_wr_max,
1673                                bool has_iops_max,
1674                                int64_t iops_max,
1675                                bool has_iops_rd_max,
1676                                int64_t iops_rd_max,
1677                                bool has_iops_wr_max,
1678                                int64_t iops_wr_max,
1679                                bool has_iops_size,
1680                                int64_t iops_size, Error **errp)
1681 {
1682     ThrottleConfig cfg;
1683     BlockDriverState *bs;
1684     AioContext *aio_context;
1685
1686     bs = bdrv_find(device);
1687     if (!bs) {
1688         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1689         return;
1690     }
1691
1692     memset(&cfg, 0, sizeof(cfg));
1693     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1694     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1695     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1696
1697     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1698     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1699     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1700
1701     if (has_bps_max) {
1702         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1703     }
1704     if (has_bps_rd_max) {
1705         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1706     }
1707     if (has_bps_wr_max) {
1708         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1709     }
1710     if (has_iops_max) {
1711         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1712     }
1713     if (has_iops_rd_max) {
1714         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1715     }
1716     if (has_iops_wr_max) {
1717         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1718     }
1719
1720     if (has_iops_size) {
1721         cfg.op_size = iops_size;
1722     }
1723
1724     if (!check_throttle_config(&cfg, errp)) {
1725         return;
1726     }
1727
1728     aio_context = bdrv_get_aio_context(bs);
1729     aio_context_acquire(aio_context);
1730
1731     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1732         bdrv_io_limits_enable(bs);
1733     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1734         bdrv_io_limits_disable(bs);
1735     }
1736
1737     if (bs->io_limits_enabled) {
1738         bdrv_set_io_limits(bs, &cfg);
1739     }
1740
1741     aio_context_release(aio_context);
1742 }
1743
1744 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1745 {
1746     const char *id = qdict_get_str(qdict, "id");
1747     BlockDriverState *bs;
1748     DriveInfo *dinfo;
1749     AioContext *aio_context;
1750     Error *local_err = NULL;
1751
1752     bs = bdrv_find(id);
1753     if (!bs) {
1754         error_report("Device '%s' not found", id);
1755         return -1;
1756     }
1757
1758     dinfo = drive_get_by_blockdev(bs);
1759     if (dinfo && !dinfo->enable_auto_del) {
1760         error_report("Deleting device added with blockdev-add"
1761                      " is not supported");
1762         return -1;
1763     }
1764
1765     aio_context = bdrv_get_aio_context(bs);
1766     aio_context_acquire(aio_context);
1767
1768     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1769         error_report("%s", error_get_pretty(local_err));
1770         error_free(local_err);
1771         aio_context_release(aio_context);
1772         return -1;
1773     }
1774
1775     /* quiesce block driver; prevent further io */
1776     bdrv_drain_all();
1777     bdrv_flush(bs);
1778     bdrv_close(bs);
1779
1780     /* if we have a device attached to this BlockDriverState
1781      * then we need to make the drive anonymous until the device
1782      * can be removed.  If this is a drive with no device backing
1783      * then we can just get rid of the block driver state right here.
1784      */
1785     if (bdrv_get_attached_dev(bs)) {
1786         bdrv_make_anon(bs);
1787
1788         /* Further I/O must not pause the guest */
1789         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1790                           BLOCKDEV_ON_ERROR_REPORT);
1791     } else {
1792         drive_del(dinfo);
1793     }
1794
1795     aio_context_release(aio_context);
1796     return 0;
1797 }
1798
1799 void qmp_block_resize(bool has_device, const char *device,
1800                       bool has_node_name, const char *node_name,
1801                       int64_t size, Error **errp)
1802 {
1803     Error *local_err = NULL;
1804     BlockDriverState *bs;
1805     AioContext *aio_context;
1806     int ret;
1807
1808     bs = bdrv_lookup_bs(has_device ? device : NULL,
1809                         has_node_name ? node_name : NULL,
1810                         &local_err);
1811     if (local_err) {
1812         error_propagate(errp, local_err);
1813         return;
1814     }
1815
1816     aio_context = bdrv_get_aio_context(bs);
1817     aio_context_acquire(aio_context);
1818
1819     if (!bdrv_is_first_non_filter(bs)) {
1820         error_set(errp, QERR_FEATURE_DISABLED, "resize");
1821         goto out;
1822     }
1823
1824     if (size < 0) {
1825         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1826         goto out;
1827     }
1828
1829     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
1830         error_set(errp, QERR_DEVICE_IN_USE, device);
1831         goto out;
1832     }
1833
1834     /* complete all in-flight operations before resizing the device */
1835     bdrv_drain_all();
1836
1837     ret = bdrv_truncate(bs, size);
1838     switch (ret) {
1839     case 0:
1840         break;
1841     case -ENOMEDIUM:
1842         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1843         break;
1844     case -ENOTSUP:
1845         error_set(errp, QERR_UNSUPPORTED);
1846         break;
1847     case -EACCES:
1848         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1849         break;
1850     case -EBUSY:
1851         error_set(errp, QERR_DEVICE_IN_USE, device);
1852         break;
1853     default:
1854         error_setg_errno(errp, -ret, "Could not resize");
1855         break;
1856     }
1857
1858 out:
1859     aio_context_release(aio_context);
1860 }
1861
1862 static void block_job_cb(void *opaque, int ret)
1863 {
1864     BlockDriverState *bs = opaque;
1865     const char *msg = NULL;
1866
1867     trace_block_job_cb(bs, bs->job, ret);
1868
1869     assert(bs->job);
1870
1871     if (ret < 0) {
1872         msg = strerror(-ret);
1873     }
1874
1875     if (block_job_is_cancelled(bs->job)) {
1876         block_job_event_cancelled(bs->job);
1877     } else {
1878         block_job_event_completed(bs->job, msg);
1879     }
1880
1881     bdrv_put_ref_bh_schedule(bs);
1882 }
1883
1884 void qmp_block_stream(const char *device,
1885                       bool has_base, const char *base,
1886                       bool has_backing_file, const char *backing_file,
1887                       bool has_speed, int64_t speed,
1888                       bool has_on_error, BlockdevOnError on_error,
1889                       Error **errp)
1890 {
1891     BlockDriverState *bs;
1892     BlockDriverState *base_bs = NULL;
1893     Error *local_err = NULL;
1894     const char *base_name = NULL;
1895
1896     if (!has_on_error) {
1897         on_error = BLOCKDEV_ON_ERROR_REPORT;
1898     }
1899
1900     bs = bdrv_find(device);
1901     if (!bs) {
1902         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1903         return;
1904     }
1905
1906     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
1907         return;
1908     }
1909
1910     if (has_base) {
1911         base_bs = bdrv_find_backing_image(bs, base);
1912         if (base_bs == NULL) {
1913             error_set(errp, QERR_BASE_NOT_FOUND, base);
1914             return;
1915         }
1916         base_name = base;
1917     }
1918
1919     /* if we are streaming the entire chain, the result will have no backing
1920      * file, and specifying one is therefore an error */
1921     if (base_bs == NULL && has_backing_file) {
1922         error_setg(errp, "backing file specified, but streaming the "
1923                          "entire chain");
1924         return;
1925     }
1926
1927     /* backing_file string overrides base bs filename */
1928     base_name = has_backing_file ? backing_file : base_name;
1929
1930     stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
1931                  on_error, block_job_cb, bs, &local_err);
1932     if (local_err) {
1933         error_propagate(errp, local_err);
1934         return;
1935     }
1936
1937     trace_qmp_block_stream(bs, bs->job);
1938 }
1939
1940 void qmp_block_commit(const char *device,
1941                       bool has_base, const char *base,
1942                       bool has_top, const char *top,
1943                       bool has_backing_file, const char *backing_file,
1944                       bool has_speed, int64_t speed,
1945                       Error **errp)
1946 {
1947     BlockDriverState *bs;
1948     BlockDriverState *base_bs, *top_bs;
1949     Error *local_err = NULL;
1950     /* This will be part of the QMP command, if/when the
1951      * BlockdevOnError change for blkmirror makes it in
1952      */
1953     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1954
1955     if (!has_speed) {
1956         speed = 0;
1957     }
1958
1959     /* drain all i/o before commits */
1960     bdrv_drain_all();
1961
1962     /* Important Note:
1963      *  libvirt relies on the DeviceNotFound error class in order to probe for
1964      *  live commit feature versions; for this to work, we must make sure to
1965      *  perform the device lookup before any generic errors that may occur in a
1966      *  scenario in which all optional arguments are omitted. */
1967     bs = bdrv_find(device);
1968     if (!bs) {
1969         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1970         return;
1971     }
1972
1973     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, errp)) {
1974         return;
1975     }
1976
1977     /* default top_bs is the active layer */
1978     top_bs = bs;
1979
1980     if (has_top && top) {
1981         if (strcmp(bs->filename, top) != 0) {
1982             top_bs = bdrv_find_backing_image(bs, top);
1983         }
1984     }
1985
1986     if (top_bs == NULL) {
1987         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1988         return;
1989     }
1990
1991     if (has_base && base) {
1992         base_bs = bdrv_find_backing_image(top_bs, base);
1993     } else {
1994         base_bs = bdrv_find_base(top_bs);
1995     }
1996
1997     if (base_bs == NULL) {
1998         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1999         return;
2000     }
2001
2002     /* Do not allow attempts to commit an image into itself */
2003     if (top_bs == base_bs) {
2004         error_setg(errp, "cannot commit an image into itself");
2005         return;
2006     }
2007
2008     if (top_bs == bs) {
2009         if (has_backing_file) {
2010             error_setg(errp, "'backing-file' specified,"
2011                              " but 'top' is the active layer");
2012             return;
2013         }
2014         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2015                             bs, &local_err);
2016     } else {
2017         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2018                      has_backing_file ? backing_file : NULL, &local_err);
2019     }
2020     if (local_err != NULL) {
2021         error_propagate(errp, local_err);
2022         return;
2023     }
2024 }
2025
2026 void qmp_drive_backup(const char *device, const char *target,
2027                       bool has_format, const char *format,
2028                       enum MirrorSyncMode sync,
2029                       bool has_mode, enum NewImageMode mode,
2030                       bool has_speed, int64_t speed,
2031                       bool has_on_source_error, BlockdevOnError on_source_error,
2032                       bool has_on_target_error, BlockdevOnError on_target_error,
2033                       Error **errp)
2034 {
2035     BlockDriverState *bs;
2036     BlockDriverState *target_bs;
2037     BlockDriverState *source = NULL;
2038     BlockDriver *drv = NULL;
2039     Error *local_err = NULL;
2040     int flags;
2041     int64_t size;
2042     int ret;
2043
2044     if (!has_speed) {
2045         speed = 0;
2046     }
2047     if (!has_on_source_error) {
2048         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2049     }
2050     if (!has_on_target_error) {
2051         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2052     }
2053     if (!has_mode) {
2054         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2055     }
2056
2057     bs = bdrv_find(device);
2058     if (!bs) {
2059         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2060         return;
2061     }
2062
2063     if (!bdrv_is_inserted(bs)) {
2064         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2065         return;
2066     }
2067
2068     if (!has_format) {
2069         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2070     }
2071     if (format) {
2072         drv = bdrv_find_format(format);
2073         if (!drv) {
2074             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2075             return;
2076         }
2077     }
2078
2079     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2080         return;
2081     }
2082
2083     flags = bs->open_flags | BDRV_O_RDWR;
2084
2085     /* See if we have a backing HD we can use to create our new image
2086      * on top of. */
2087     if (sync == MIRROR_SYNC_MODE_TOP) {
2088         source = bs->backing_hd;
2089         if (!source) {
2090             sync = MIRROR_SYNC_MODE_FULL;
2091         }
2092     }
2093     if (sync == MIRROR_SYNC_MODE_NONE) {
2094         source = bs;
2095     }
2096
2097     size = bdrv_getlength(bs);
2098     if (size < 0) {
2099         error_setg_errno(errp, -size, "bdrv_getlength failed");
2100         return;
2101     }
2102
2103     if (mode != NEW_IMAGE_MODE_EXISTING) {
2104         assert(format && drv);
2105         if (source) {
2106             bdrv_img_create(target, format, source->filename,
2107                             source->drv->format_name, NULL,
2108                             size, flags, &local_err, false);
2109         } else {
2110             bdrv_img_create(target, format, NULL, NULL, NULL,
2111                             size, flags, &local_err, false);
2112         }
2113     }
2114
2115     if (local_err) {
2116         error_propagate(errp, local_err);
2117         return;
2118     }
2119
2120     target_bs = NULL;
2121     ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2122     if (ret < 0) {
2123         error_propagate(errp, local_err);
2124         return;
2125     }
2126
2127     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2128                  block_job_cb, bs, &local_err);
2129     if (local_err != NULL) {
2130         bdrv_unref(target_bs);
2131         error_propagate(errp, local_err);
2132         return;
2133     }
2134 }
2135
2136 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2137 {
2138     return bdrv_named_nodes_list();
2139 }
2140
2141 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
2142
2143 void qmp_drive_mirror(const char *device, const char *target,
2144                       bool has_format, const char *format,
2145                       bool has_node_name, const char *node_name,
2146                       bool has_replaces, const char *replaces,
2147                       enum MirrorSyncMode sync,
2148                       bool has_mode, enum NewImageMode mode,
2149                       bool has_speed, int64_t speed,
2150                       bool has_granularity, uint32_t granularity,
2151                       bool has_buf_size, int64_t buf_size,
2152                       bool has_on_source_error, BlockdevOnError on_source_error,
2153                       bool has_on_target_error, BlockdevOnError on_target_error,
2154                       Error **errp)
2155 {
2156     BlockDriverState *bs;
2157     BlockDriverState *source, *target_bs;
2158     BlockDriver *drv = NULL;
2159     Error *local_err = NULL;
2160     QDict *options = NULL;
2161     int flags;
2162     int64_t size;
2163     int ret;
2164
2165     if (!has_speed) {
2166         speed = 0;
2167     }
2168     if (!has_on_source_error) {
2169         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2170     }
2171     if (!has_on_target_error) {
2172         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2173     }
2174     if (!has_mode) {
2175         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2176     }
2177     if (!has_granularity) {
2178         granularity = 0;
2179     }
2180     if (!has_buf_size) {
2181         buf_size = DEFAULT_MIRROR_BUF_SIZE;
2182     }
2183
2184     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2185         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2186                   "a value in range [512B, 64MB]");
2187         return;
2188     }
2189     if (granularity & (granularity - 1)) {
2190         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity", "power of 2");
2191         return;
2192     }
2193
2194     bs = bdrv_find(device);
2195     if (!bs) {
2196         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2197         return;
2198     }
2199
2200     if (!bdrv_is_inserted(bs)) {
2201         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2202         return;
2203     }
2204
2205     if (!has_format) {
2206         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2207     }
2208     if (format) {
2209         drv = bdrv_find_format(format);
2210         if (!drv) {
2211             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2212             return;
2213         }
2214     }
2215
2216     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2217         return;
2218     }
2219
2220     flags = bs->open_flags | BDRV_O_RDWR;
2221     source = bs->backing_hd;
2222     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2223         sync = MIRROR_SYNC_MODE_FULL;
2224     }
2225     if (sync == MIRROR_SYNC_MODE_NONE) {
2226         source = bs;
2227     }
2228
2229     size = bdrv_getlength(bs);
2230     if (size < 0) {
2231         error_setg_errno(errp, -size, "bdrv_getlength failed");
2232         return;
2233     }
2234
2235     if (has_replaces) {
2236         BlockDriverState *to_replace_bs;
2237
2238         if (!has_node_name) {
2239             error_setg(errp, "a node-name must be provided when replacing a"
2240                              " named node of the graph");
2241             return;
2242         }
2243
2244         to_replace_bs = check_to_replace_node(replaces, &local_err);
2245
2246         if (!to_replace_bs) {
2247             error_propagate(errp, local_err);
2248             return;
2249         }
2250
2251         if (size != bdrv_getlength(to_replace_bs)) {
2252             error_setg(errp, "cannot replace image with a mirror image of "
2253                              "different size");
2254             return;
2255         }
2256     }
2257
2258     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2259         && mode != NEW_IMAGE_MODE_EXISTING)
2260     {
2261         /* create new image w/o backing file */
2262         assert(format && drv);
2263         bdrv_img_create(target, format,
2264                         NULL, NULL, NULL, size, flags, &local_err, false);
2265     } else {
2266         switch (mode) {
2267         case NEW_IMAGE_MODE_EXISTING:
2268             break;
2269         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2270             /* create new image with backing file */
2271             bdrv_img_create(target, format,
2272                             source->filename,
2273                             source->drv->format_name,
2274                             NULL, size, flags, &local_err, false);
2275             break;
2276         default:
2277             abort();
2278         }
2279     }
2280
2281     if (local_err) {
2282         error_propagate(errp, local_err);
2283         return;
2284     }
2285
2286     if (has_node_name) {
2287         options = qdict_new();
2288         qdict_put(options, "node-name", qstring_from_str(node_name));
2289     }
2290
2291     /* Mirroring takes care of copy-on-write using the source's backing
2292      * file.
2293      */
2294     target_bs = NULL;
2295     ret = bdrv_open(&target_bs, target, NULL, options,
2296                     flags | BDRV_O_NO_BACKING, drv, &local_err);
2297     if (ret < 0) {
2298         error_propagate(errp, local_err);
2299         return;
2300     }
2301
2302     /* pass the node name to replace to mirror start since it's loose coupling
2303      * and will allow to check whether the node still exist at mirror completion
2304      */
2305     mirror_start(bs, target_bs,
2306                  has_replaces ? replaces : NULL,
2307                  speed, granularity, buf_size, sync,
2308                  on_source_error, on_target_error,
2309                  block_job_cb, bs, &local_err);
2310     if (local_err != NULL) {
2311         bdrv_unref(target_bs);
2312         error_propagate(errp, local_err);
2313         return;
2314     }
2315 }
2316
2317 static BlockJob *find_block_job(const char *device)
2318 {
2319     BlockDriverState *bs;
2320
2321     bs = bdrv_find(device);
2322     if (!bs || !bs->job) {
2323         return NULL;
2324     }
2325     return bs->job;
2326 }
2327
2328 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2329 {
2330     BlockJob *job = find_block_job(device);
2331
2332     if (!job) {
2333         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2334         return;
2335     }
2336
2337     block_job_set_speed(job, speed, errp);
2338 }
2339
2340 void qmp_block_job_cancel(const char *device,
2341                           bool has_force, bool force, Error **errp)
2342 {
2343     BlockJob *job = find_block_job(device);
2344
2345     if (!has_force) {
2346         force = false;
2347     }
2348
2349     if (!job) {
2350         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2351         return;
2352     }
2353     if (job->paused && !force) {
2354         error_setg(errp, "The block job for device '%s' is currently paused",
2355                    device);
2356         return;
2357     }
2358
2359     trace_qmp_block_job_cancel(job);
2360     block_job_cancel(job);
2361 }
2362
2363 void qmp_block_job_pause(const char *device, Error **errp)
2364 {
2365     BlockJob *job = find_block_job(device);
2366
2367     if (!job) {
2368         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2369         return;
2370     }
2371
2372     trace_qmp_block_job_pause(job);
2373     block_job_pause(job);
2374 }
2375
2376 void qmp_block_job_resume(const char *device, Error **errp)
2377 {
2378     BlockJob *job = find_block_job(device);
2379
2380     if (!job) {
2381         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2382         return;
2383     }
2384
2385     trace_qmp_block_job_resume(job);
2386     block_job_resume(job);
2387 }
2388
2389 void qmp_block_job_complete(const char *device, Error **errp)
2390 {
2391     BlockJob *job = find_block_job(device);
2392
2393     if (!job) {
2394         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2395         return;
2396     }
2397
2398     trace_qmp_block_job_complete(job);
2399     block_job_complete(job, errp);
2400 }
2401
2402 void qmp_change_backing_file(const char *device,
2403                              const char *image_node_name,
2404                              const char *backing_file,
2405                              Error **errp)
2406 {
2407     BlockDriverState *bs = NULL;
2408     BlockDriverState *image_bs = NULL;
2409     Error *local_err = NULL;
2410     bool ro;
2411     int open_flags;
2412     int ret;
2413
2414     /* find the top layer BDS of the chain */
2415     bs = bdrv_find(device);
2416     if (!bs) {
2417         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2418         return;
2419     }
2420
2421     image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
2422     if (local_err) {
2423         error_propagate(errp, local_err);
2424         return;
2425     }
2426
2427     if (!image_bs) {
2428         error_setg(errp, "image file not found");
2429         return;
2430     }
2431
2432     if (bdrv_find_base(image_bs) == image_bs) {
2433         error_setg(errp, "not allowing backing file change on an image "
2434                          "without a backing file");
2435         return;
2436     }
2437
2438     /* even though we are not necessarily operating on bs, we need it to
2439      * determine if block ops are currently prohibited on the chain */
2440     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
2441         return;
2442     }
2443
2444     /* final sanity check */
2445     if (!bdrv_chain_contains(bs, image_bs)) {
2446         error_setg(errp, "'%s' and image file are not in the same chain",
2447                    device);
2448         return;
2449     }
2450
2451     /* if not r/w, reopen to make r/w */
2452     open_flags = image_bs->open_flags;
2453     ro = bdrv_is_read_only(image_bs);
2454
2455     if (ro) {
2456         bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
2457         if (local_err) {
2458             error_propagate(errp, local_err);
2459             return;
2460         }
2461     }
2462
2463     ret = bdrv_change_backing_file(image_bs, backing_file,
2464                                image_bs->drv ? image_bs->drv->format_name : "");
2465
2466     if (ret < 0) {
2467         error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
2468                          backing_file);
2469         /* don't exit here, so we can try to restore open flags if
2470          * appropriate */
2471     }
2472
2473     if (ro) {
2474         bdrv_reopen(image_bs, open_flags, &local_err);
2475         if (local_err) {
2476             error_propagate(errp, local_err); /* will preserve prior errp */
2477         }
2478     }
2479 }
2480
2481 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2482 {
2483     QmpOutputVisitor *ov = qmp_output_visitor_new();
2484     DriveInfo *dinfo;
2485     QObject *obj;
2486     QDict *qdict;
2487     Error *local_err = NULL;
2488
2489     /* Require an ID in the top level */
2490     if (!options->has_id) {
2491         error_setg(errp, "Block device needs an ID");
2492         goto fail;
2493     }
2494
2495     /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
2496      * cache.direct=false instead of silently switching to aio=threads, except
2497      * when called from drive_new().
2498      *
2499      * For now, simply forbidding the combination for all drivers will do. */
2500     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2501         bool direct = options->has_cache &&
2502                       options->cache->has_direct &&
2503                       options->cache->direct;
2504         if (!direct) {
2505             error_setg(errp, "aio=native requires cache.direct=true");
2506             goto fail;
2507         }
2508     }
2509
2510     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2511                                &options, NULL, &local_err);
2512     if (local_err) {
2513         error_propagate(errp, local_err);
2514         goto fail;
2515     }
2516
2517     obj = qmp_output_get_qobject(ov);
2518     qdict = qobject_to_qdict(obj);
2519
2520     qdict_flatten(qdict);
2521
2522     dinfo = blockdev_init(NULL, qdict, &local_err);
2523     if (local_err) {
2524         error_propagate(errp, local_err);
2525         goto fail;
2526     }
2527
2528     if (bdrv_key_required(dinfo->bdrv)) {
2529         drive_del(dinfo);
2530         error_setg(errp, "blockdev-add doesn't support encrypted devices");
2531         goto fail;
2532     }
2533
2534 fail:
2535     qmp_output_visitor_cleanup(ov);
2536 }
2537
2538 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2539 {
2540     BlockJobInfoList **prev = opaque;
2541     BlockJob *job = bs->job;
2542
2543     if (job) {
2544         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2545         elem->value = block_job_query(bs->job);
2546         (*prev)->next = elem;
2547         *prev = elem;
2548     }
2549 }
2550
2551 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2552 {
2553     /* Dummy is a fake list element for holding the head pointer */
2554     BlockJobInfoList dummy = {};
2555     BlockJobInfoList *prev = &dummy;
2556     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2557     return dummy.next;
2558 }
2559
2560 QemuOptsList qemu_common_drive_opts = {
2561     .name = "drive",
2562     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2563     .desc = {
2564         {
2565             .name = "snapshot",
2566             .type = QEMU_OPT_BOOL,
2567             .help = "enable/disable snapshot mode",
2568         },{
2569             .name = "discard",
2570             .type = QEMU_OPT_STRING,
2571             .help = "discard operation (ignore/off, unmap/on)",
2572         },{
2573             .name = "cache.writeback",
2574             .type = QEMU_OPT_BOOL,
2575             .help = "enables writeback mode for any caches",
2576         },{
2577             .name = "cache.direct",
2578             .type = QEMU_OPT_BOOL,
2579             .help = "enables use of O_DIRECT (bypass the host page cache)",
2580         },{
2581             .name = "cache.no-flush",
2582             .type = QEMU_OPT_BOOL,
2583             .help = "ignore any flush requests for the device",
2584         },{
2585             .name = "aio",
2586             .type = QEMU_OPT_STRING,
2587             .help = "host AIO implementation (threads, native)",
2588         },{
2589             .name = "format",
2590             .type = QEMU_OPT_STRING,
2591             .help = "disk format (raw, qcow2, ...)",
2592         },{
2593             .name = "rerror",
2594             .type = QEMU_OPT_STRING,
2595             .help = "read error action",
2596         },{
2597             .name = "werror",
2598             .type = QEMU_OPT_STRING,
2599             .help = "write error action",
2600         },{
2601             .name = "read-only",
2602             .type = QEMU_OPT_BOOL,
2603             .help = "open drive file as read-only",
2604         },{
2605             .name = "throttling.iops-total",
2606             .type = QEMU_OPT_NUMBER,
2607             .help = "limit total I/O operations per second",
2608         },{
2609             .name = "throttling.iops-read",
2610             .type = QEMU_OPT_NUMBER,
2611             .help = "limit read operations per second",
2612         },{
2613             .name = "throttling.iops-write",
2614             .type = QEMU_OPT_NUMBER,
2615             .help = "limit write operations per second",
2616         },{
2617             .name = "throttling.bps-total",
2618             .type = QEMU_OPT_NUMBER,
2619             .help = "limit total bytes per second",
2620         },{
2621             .name = "throttling.bps-read",
2622             .type = QEMU_OPT_NUMBER,
2623             .help = "limit read bytes per second",
2624         },{
2625             .name = "throttling.bps-write",
2626             .type = QEMU_OPT_NUMBER,
2627             .help = "limit write bytes per second",
2628         },{
2629             .name = "throttling.iops-total-max",
2630             .type = QEMU_OPT_NUMBER,
2631             .help = "I/O operations burst",
2632         },{
2633             .name = "throttling.iops-read-max",
2634             .type = QEMU_OPT_NUMBER,
2635             .help = "I/O operations read burst",
2636         },{
2637             .name = "throttling.iops-write-max",
2638             .type = QEMU_OPT_NUMBER,
2639             .help = "I/O operations write burst",
2640         },{
2641             .name = "throttling.bps-total-max",
2642             .type = QEMU_OPT_NUMBER,
2643             .help = "total bytes burst",
2644         },{
2645             .name = "throttling.bps-read-max",
2646             .type = QEMU_OPT_NUMBER,
2647             .help = "total bytes read burst",
2648         },{
2649             .name = "throttling.bps-write-max",
2650             .type = QEMU_OPT_NUMBER,
2651             .help = "total bytes write burst",
2652         },{
2653             .name = "throttling.iops-size",
2654             .type = QEMU_OPT_NUMBER,
2655             .help = "when limiting by iops max size of an I/O in bytes",
2656         },{
2657             .name = "copy-on-read",
2658             .type = QEMU_OPT_BOOL,
2659             .help = "copy read data from backing file into image file",
2660         },{
2661             .name = "detect-zeroes",
2662             .type = QEMU_OPT_STRING,
2663             .help = "try to optimize zero writes (off, on, unmap)",
2664         },
2665         { /* end of list */ }
2666     },
2667 };
2668
2669 QemuOptsList qemu_drive_opts = {
2670     .name = "drive",
2671     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2672     .desc = {
2673         /*
2674          * no elements => accept any params
2675          * validation will happen later
2676          */
2677         { /* end of list */ }
2678     },
2679 };