blockdev.c

   1 /*
   2  * QEMU host block devices
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * This work is licensed under the terms of the GNU GPL, version 2 or
   7  * later.  See the COPYING file in the top-level directory.
   8  *
   9  * This file incorporates work covered by the following copyright and
  10  * permission notice:
  11  *
  12  * Copyright (c) 2003-2008 Fabrice Bellard
  13  *
  14  * Permission is hereby granted, free of charge, to any person obtaining a copy
  15  * of this software and associated documentation files (the "Software"), to deal
  16  * in the Software without restriction, including without limitation the rights
  17  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  18  * copies of the Software, and to permit persons to whom the Software is
  19  * furnished to do so, subject to the following conditions:
  20  *
  21  * The above copyright notice and this permission notice shall be included in
  22  * all copies or substantial portions of the Software.
  23  *
  24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  25  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  26  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  27  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  28  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  29  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  30  * THE SOFTWARE.
  31  */
  32
  33 #include "sysemu/blockdev.h"
  34 #include "hw/block/block.h"
  35 #include "block/blockjob.h"
  36 #include "monitor/monitor.h"
  37 #include "qapi/qmp/qerror.h"
  38 #include "qemu/option.h"
  39 #include "qemu/config-file.h"
  40 #include "qapi/qmp/types.h"
  41 #include "qapi-visit.h"
  42 #include "qapi/qmp-output-visitor.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_put_ref(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 static void drive_uninit(DriveInfo *dinfo)
 218 {
 219     if (dinfo->opts) {
 220         qemu_opts_del(dinfo->opts);
 221     }
 222
 223     bdrv_unref(dinfo->bdrv);
 224     g_free(dinfo->id);
 225     QTAILQ_REMOVE(&drives, dinfo, next);
 226     g_free(dinfo->serial);
 227     g_free(dinfo);
 228 }
 229
 230 void drive_put_ref(DriveInfo *dinfo)
 231 {
 232     assert(dinfo->refcount);
 233     if (--dinfo->refcount == 0) {
 234         drive_uninit(dinfo);
 235     }
 236 }
 237
 238 void drive_get_ref(DriveInfo *dinfo)
 239 {
 240     dinfo->refcount++;
 241 }
 242
 243 typedef struct {
 244     QEMUBH *bh;
 245     BlockDriverState *bs;
 246 } BDRVPutRefBH;
 247
 248 static void bdrv_put_ref_bh(void *opaque)
 249 {
 250     BDRVPutRefBH *s = opaque;
 251
 252     bdrv_unref(s->bs);
 253     qemu_bh_delete(s->bh);
 254     g_free(s);
 255 }
 256
 257 /*
 258  * Release a BDS reference in a BH
 259  *
 260  * It is not safe to use bdrv_unref() from a callback function when the callers
 261  * still need the BlockDriverState.  In such cases we schedule a BH to release
 262  * the reference.
 263  */
 264 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
 265 {
 266     BDRVPutRefBH *s;
 267
 268     s = g_new(BDRVPutRefBH, 1);
 269     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
 270     s->bs = bs;
 271     qemu_bh_schedule(s->bh);
 272 }
 273
 274 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
 275 {
 276     if (!strcmp(buf, "ignore")) {
 277         return BLOCKDEV_ON_ERROR_IGNORE;
 278     } else if (!is_read && !strcmp(buf, "enospc")) {
 279         return BLOCKDEV_ON_ERROR_ENOSPC;
 280     } else if (!strcmp(buf, "stop")) {
 281         return BLOCKDEV_ON_ERROR_STOP;
 282     } else if (!strcmp(buf, "report")) {
 283         return BLOCKDEV_ON_ERROR_REPORT;
 284     } else {
 285         error_setg(errp, "'%s' invalid %s error action",
 286                    buf, is_read ? "read" : "write");
 287         return -1;
 288     }
 289 }
 290
 291 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
 292 {
 293     if (throttle_conflicting(cfg)) {
 294         error_setg(errp, "bps/iops/max total values and read/write values"
 295                          " cannot be used at the same time");
 296         return false;
 297     }
 298
 299     if (!throttle_is_valid(cfg)) {
 300         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
 301         return false;
 302     }
 303
 304     return true;
 305 }
 306
 307 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
 308
 309 /* Takes the ownership of bs_opts */
 310 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
 311                                 Error **errp)
 312 {
 313     const char *buf;
 314     const char *serial;
 315     int ro = 0;
 316     int bdrv_flags = 0;
 317     int on_read_error, on_write_error;
 318     DriveInfo *dinfo;
 319     ThrottleConfig cfg;
 320     int snapshot = 0;
 321     bool copy_on_read;
 322     int ret;
 323     Error *error = NULL;
 324     QemuOpts *opts;
 325     const char *id;
 326     bool has_driver_specific_opts;
 327     BlockDriver *drv = NULL;
 328
 329     /* Check common options by copying from bs_opts to opts, all other options
 330      * stay in bs_opts for processing by bdrv_open(). */
 331     id = qdict_get_try_str(bs_opts, "id");
 332     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
 333     if (error) {
 334         error_propagate(errp, error);
 335         return NULL;
 336     }
 337
 338     qemu_opts_absorb_qdict(opts, bs_opts, &error);
 339     if (error) {
 340         error_propagate(errp, error);
 341         goto early_err;
 342     }
 343
 344     if (id) {
 345         qdict_del(bs_opts, "id");
 346     }
 347
 348     has_driver_specific_opts = !!qdict_size(bs_opts);
 349
 350     /* extract parameters */
 351     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 352     ro = qemu_opt_get_bool(opts, "read-only", 0);
 353     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
 354
 355     serial = qemu_opt_get(opts, "serial");
 356
 357     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
 358         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
 359             error_setg(errp, "invalid discard option");
 360             goto early_err;
 361         }
 362     }
 363
 364     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
 365         bdrv_flags |= BDRV_O_CACHE_WB;
 366     }
 367     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
 368         bdrv_flags |= BDRV_O_NOCACHE;
 369     }
 370     if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
 371         bdrv_flags |= BDRV_O_NO_FLUSH;
 372     }
 373
 374 #ifdef CONFIG_LINUX_AIO
 375     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
 376         if (!strcmp(buf, "native")) {
 377             bdrv_flags |= BDRV_O_NATIVE_AIO;
 378         } else if (!strcmp(buf, "threads")) {
 379             /* this is the default */
 380         } else {
 381            error_setg(errp, "invalid aio option");
 382            goto early_err;
 383         }
 384     }
 385 #endif
 386
 387     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
 388         if (is_help_option(buf)) {
 389             error_printf("Supported formats:");
 390             bdrv_iterate_format(bdrv_format_print, NULL);
 391             error_printf("\n");
 392             goto early_err;
 393         }
 394
 395         drv = bdrv_find_format(buf);
 396         if (!drv) {
 397             error_setg(errp, "'%s' invalid format", buf);
 398             goto early_err;
 399         }
 400     }
 401
 402     /* disk I/O throttling */
 403     memset(&cfg, 0, sizeof(cfg));
 404     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
 405         qemu_opt_get_number(opts, "throttling.bps-total", 0);
 406     cfg.buckets[THROTTLE_BPS_READ].avg  =
 407         qemu_opt_get_number(opts, "throttling.bps-read", 0);
 408     cfg.buckets[THROTTLE_BPS_WRITE].avg =
 409         qemu_opt_get_number(opts, "throttling.bps-write", 0);
 410     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
 411         qemu_opt_get_number(opts, "throttling.iops-total", 0);
 412     cfg.buckets[THROTTLE_OPS_READ].avg =
 413         qemu_opt_get_number(opts, "throttling.iops-read", 0);
 414     cfg.buckets[THROTTLE_OPS_WRITE].avg =
 415         qemu_opt_get_number(opts, "throttling.iops-write", 0);
 416
 417     cfg.buckets[THROTTLE_BPS_TOTAL].max =
 418         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
 419     cfg.buckets[THROTTLE_BPS_READ].max  =
 420         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
 421     cfg.buckets[THROTTLE_BPS_WRITE].max =
 422         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
 423     cfg.buckets[THROTTLE_OPS_TOTAL].max =
 424         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
 425     cfg.buckets[THROTTLE_OPS_READ].max =
 426         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
 427     cfg.buckets[THROTTLE_OPS_WRITE].max =
 428         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
 429
 430     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
 431
 432     if (!check_throttle_config(&cfg, &error)) {
 433         error_propagate(errp, error);
 434         goto early_err;
 435     }
 436
 437     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
 438     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
 439         on_write_error = parse_block_error_action(buf, 0, &error);
 440         if (error) {
 441             error_propagate(errp, error);
 442             goto early_err;
 443         }
 444     }
 445
 446     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
 447     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
 448         on_read_error = parse_block_error_action(buf, 1, &error);
 449         if (error) {
 450             error_propagate(errp, error);
 451             goto early_err;
 452         }
 453     }
 454
 455     if (bdrv_find_node(qemu_opts_id(opts))) {
 456         error_setg(errp, "device id=%s is conflicting with a node-name",
 457                    qemu_opts_id(opts));
 458         goto early_err;
 459     }
 460
 461     /* init */
 462     dinfo = g_malloc0(sizeof(*dinfo));
 463     dinfo->id = g_strdup(qemu_opts_id(opts));
 464     dinfo->bdrv = bdrv_new(dinfo->id);
 465     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
 466     dinfo->bdrv->read_only = ro;
 467     dinfo->refcount = 1;
 468     if (serial != NULL) {
 469         dinfo->serial = g_strdup(serial);
 470     }
 471     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
 472
 473     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
 474
 475     /* disk I/O throttling */
 476     if (throttle_enabled(&cfg)) {
 477         bdrv_io_limits_enable(dinfo->bdrv);
 478         bdrv_set_io_limits(dinfo->bdrv, &cfg);
 479     }
 480
 481     if (!file || !*file) {
 482         if (has_driver_specific_opts) {
 483             file = NULL;
 484         } else {
 485             QDECREF(bs_opts);
 486             qemu_opts_del(opts);
 487             return dinfo;
 488         }
 489     }
 490     if (snapshot) {
 491         /* always use cache=unsafe with snapshot */
 492         bdrv_flags &= ~BDRV_O_CACHE_MASK;
 493         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
 494     }
 495
 496     if (copy_on_read) {
 497         bdrv_flags |= BDRV_O_COPY_ON_READ;
 498     }
 499
 500     if (runstate_check(RUN_STATE_INMIGRATE)) {
 501         bdrv_flags |= BDRV_O_INCOMING;
 502     }
 503
 504     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
 505
 506     QINCREF(bs_opts);
 507     ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
 508
 509     if (ret < 0) {
 510         error_setg(errp, "could not open disk image %s: %s",
 511                    file ?: dinfo->id, error_get_pretty(error));
 512         error_free(error);
 513         goto err;
 514     }
 515
 516     if (bdrv_key_required(dinfo->bdrv))
 517         autostart = 0;
 518
 519     QDECREF(bs_opts);
 520     qemu_opts_del(opts);
 521
 522     return dinfo;
 523
 524 err:
 525     bdrv_unref(dinfo->bdrv);
 526     g_free(dinfo->id);
 527     QTAILQ_REMOVE(&drives, dinfo, next);
 528     g_free(dinfo);
 529 early_err:
 530     QDECREF(bs_opts);
 531     qemu_opts_del(opts);
 532     return NULL;
 533 }
 534
 535 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
 536 {
 537     const char *value;
 538
 539     value = qemu_opt_get(opts, from);
 540     if (value) {
 541         qemu_opt_set(opts, to, value);
 542         qemu_opt_unset(opts, from);
 543     }
 544 }
 545
 546 QemuOptsList qemu_legacy_drive_opts = {
 547     .name = "drive",
 548     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
 549     .desc = {
 550         {
 551             .name = "bus",
 552             .type = QEMU_OPT_NUMBER,
 553             .help = "bus number",
 554         },{
 555             .name = "unit",
 556             .type = QEMU_OPT_NUMBER,
 557             .help = "unit number (i.e. lun for scsi)",
 558         },{
 559             .name = "index",
 560             .type = QEMU_OPT_NUMBER,
 561             .help = "index number",
 562         },{
 563             .name = "media",
 564             .type = QEMU_OPT_STRING,
 565             .help = "media type (disk, cdrom)",
 566         },{
 567             .name = "if",
 568             .type = QEMU_OPT_STRING,
 569             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
 570         },{
 571             .name = "cyls",
 572             .type = QEMU_OPT_NUMBER,
 573             .help = "number of cylinders (ide disk geometry)",
 574         },{
 575             .name = "heads",
 576             .type = QEMU_OPT_NUMBER,
 577             .help = "number of heads (ide disk geometry)",
 578         },{
 579             .name = "secs",
 580             .type = QEMU_OPT_NUMBER,
 581             .help = "number of sectors (ide disk geometry)",
 582         },{
 583             .name = "trans",
 584             .type = QEMU_OPT_STRING,
 585             .help = "chs translation (auto, lba, none)",
 586         },{
 587             .name = "boot",
 588             .type = QEMU_OPT_BOOL,
 589             .help = "(deprecated, ignored)",
 590         },{
 591             .name = "addr",
 592             .type = QEMU_OPT_STRING,
 593             .help = "pci address (virtio only)",
 594         },{
 595             .name = "file",
 596             .type = QEMU_OPT_STRING,
 597             .help = "file name",
 598         },
 599
 600         /* Options that are passed on, but have special semantics with -drive */
 601         {
 602             .name = "read-only",
 603             .type = QEMU_OPT_BOOL,
 604             .help = "open drive file as read-only",
 605         },{
 606             .name = "rerror",
 607             .type = QEMU_OPT_STRING,
 608             .help = "read error action",
 609         },{
 610             .name = "werror",
 611             .type = QEMU_OPT_STRING,
 612             .help = "write error action",
 613         },{
 614             .name = "copy-on-read",
 615             .type = QEMU_OPT_BOOL,
 616             .help = "copy read data from backing file into image file",
 617         },
 618
 619         { /* end of list */ }
 620     },
 621 };
 622
 623 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
 624 {
 625     const char *value;
 626     DriveInfo *dinfo = NULL;
 627     QDict *bs_opts;
 628     QemuOpts *legacy_opts;
 629     DriveMediaType media = MEDIA_DISK;
 630     BlockInterfaceType type;
 631     int cyls, heads, secs, translation;
 632     int max_devs, bus_id, unit_id, index;
 633     const char *devaddr;
 634     const char *werror, *rerror;
 635     bool read_only = false;
 636     bool copy_on_read;
 637     const char *filename;
 638     Error *local_err = NULL;
 639
 640     /* Change legacy command line options into QMP ones */
 641     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
 642     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
 643     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
 644
 645     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
 646     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
 647     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
 648
 649     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
 650     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
 651     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
 652
 653     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
 654     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
 655     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
 656
 657     qemu_opt_rename(all_opts,
 658                     "iops_size", "throttling.iops-size");
 659
 660     qemu_opt_rename(all_opts, "readonly", "read-only");
 661
 662     value = qemu_opt_get(all_opts, "cache");
 663     if (value) {
 664         int flags = 0;
 665
 666         if (bdrv_parse_cache_flags(value, &flags) != 0) {
 667             error_report("invalid cache option");
 668             return NULL;
 669         }
 670
 671         /* Specific options take precedence */
 672         if (!qemu_opt_get(all_opts, "cache.writeback")) {
 673             qemu_opt_set_bool(all_opts, "cache.writeback",
 674                               !!(flags & BDRV_O_CACHE_WB));
 675         }
 676         if (!qemu_opt_get(all_opts, "cache.direct")) {
 677             qemu_opt_set_bool(all_opts, "cache.direct",
 678                               !!(flags & BDRV_O_NOCACHE));
 679         }
 680         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
 681             qemu_opt_set_bool(all_opts, "cache.no-flush",
 682                               !!(flags & BDRV_O_NO_FLUSH));
 683         }
 684         qemu_opt_unset(all_opts, "cache");
 685     }
 686
 687     /* Get a QDict for processing the options */
 688     bs_opts = qdict_new();
 689     qemu_opts_to_qdict(all_opts, bs_opts);
 690
 691     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
 692                                    &error_abort);
 693     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
 694     if (local_err) {
 695         qerror_report_err(local_err);
 696         error_free(local_err);
 697         goto fail;
 698     }
 699
 700     /* Deprecated option boot=[on|off] */
 701     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
 702         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
 703                 "ignored. Future versions will reject this parameter. Please "
 704                 "update your scripts.\n");
 705     }
 706
 707     /* Media type */
 708     value = qemu_opt_get(legacy_opts, "media");
 709     if (value) {
 710         if (!strcmp(value, "disk")) {
 711             media = MEDIA_DISK;
 712         } else if (!strcmp(value, "cdrom")) {
 713             media = MEDIA_CDROM;
 714             read_only = true;
 715         } else {
 716             error_report("'%s' invalid media", value);
 717             goto fail;
 718         }
 719     }
 720
 721     /* copy-on-read is disabled with a warning for read-only devices */
 722     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
 723     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
 724
 725     if (read_only && copy_on_read) {
 726         error_report("warning: disabling copy-on-read on read-only drive");
 727         copy_on_read = false;
 728     }
 729
 730     qdict_put(bs_opts, "read-only",
 731               qstring_from_str(read_only ? "on" : "off"));
 732     qdict_put(bs_opts, "copy-on-read",
 733               qstring_from_str(copy_on_read ? "on" :"off"));
 734
 735     /* Controller type */
 736     value = qemu_opt_get(legacy_opts, "if");
 737     if (value) {
 738         for (type = 0;
 739              type < IF_COUNT && strcmp(value, if_name[type]);
 740              type++) {
 741         }
 742         if (type == IF_COUNT) {
 743             error_report("unsupported bus type '%s'", value);
 744             goto fail;
 745         }
 746     } else {
 747         type = block_default_type;
 748     }
 749
 750     /* Geometry */
 751     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
 752     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
 753     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
 754
 755     if (cyls || heads || secs) {
 756         if (cyls < 1) {
 757             error_report("invalid physical cyls number");
 758             goto fail;
 759         }
 760         if (heads < 1) {
 761             error_report("invalid physical heads number");
 762             goto fail;
 763         }
 764         if (secs < 1) {
 765             error_report("invalid physical secs number");
 766             goto fail;
 767         }
 768     }
 769
 770     translation = BIOS_ATA_TRANSLATION_AUTO;
 771     value = qemu_opt_get(legacy_opts, "trans");
 772     if (value != NULL) {
 773         if (!cyls) {
 774             error_report("'%s' trans must be used with cyls, heads and secs",
 775                          value);
 776             goto fail;
 777         }
 778         if (!strcmp(value, "none")) {
 779             translation = BIOS_ATA_TRANSLATION_NONE;
 780         } else if (!strcmp(value, "lba")) {
 781             translation = BIOS_ATA_TRANSLATION_LBA;
 782         } else if (!strcmp(value, "large")) {
 783             translation = BIOS_ATA_TRANSLATION_LARGE;
 784         } else if (!strcmp(value, "rechs")) {
 785             translation = BIOS_ATA_TRANSLATION_RECHS;
 786         } else if (!strcmp(value, "auto")) {
 787             translation = BIOS_ATA_TRANSLATION_AUTO;
 788         } else {
 789             error_report("'%s' invalid translation type", value);
 790             goto fail;
 791         }
 792     }
 793
 794     if (media == MEDIA_CDROM) {
 795         if (cyls || secs || heads) {
 796             error_report("CHS can't be set with media=cdrom");
 797             goto fail;
 798         }
 799     }
 800
 801     /* Device address specified by bus/unit or index.
 802      * If none was specified, try to find the first free one. */
 803     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
 804     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
 805     index   = qemu_opt_get_number(legacy_opts, "index", -1);
 806
 807     max_devs = if_max_devs[type];
 808
 809     if (index != -1) {
 810         if (bus_id != 0 || unit_id != -1) {
 811             error_report("index cannot be used with bus and unit");
 812             goto fail;
 813         }
 814         bus_id = drive_index_to_bus_id(type, index);
 815         unit_id = drive_index_to_unit_id(type, index);
 816     }
 817
 818     if (unit_id == -1) {
 819        unit_id = 0;
 820        while (drive_get(type, bus_id, unit_id) != NULL) {
 821            unit_id++;
 822            if (max_devs && unit_id >= max_devs) {
 823                unit_id -= max_devs;
 824                bus_id++;
 825            }
 826        }
 827     }
 828
 829     if (max_devs && unit_id >= max_devs) {
 830         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
 831         goto fail;
 832     }
 833
 834     if (drive_get(type, bus_id, unit_id) != NULL) {
 835         error_report("drive with bus=%d, unit=%d (index=%d) exists",
 836                      bus_id, unit_id, index);
 837         goto fail;
 838     }
 839
 840     /* no id supplied -> create one */
 841     if (qemu_opts_id(all_opts) == NULL) {
 842         char *new_id;
 843         const char *mediastr = "";
 844         if (type == IF_IDE || type == IF_SCSI) {
 845             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
 846         }
 847         if (max_devs) {
 848             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
 849                                      mediastr, unit_id);
 850         } else {
 851             new_id = g_strdup_printf("%s%s%i", if_name[type],
 852                                      mediastr, unit_id);
 853         }
 854         qdict_put(bs_opts, "id", qstring_from_str(new_id));
 855         g_free(new_id);
 856     }
 857
 858     /* Add virtio block device */
 859     devaddr = qemu_opt_get(legacy_opts, "addr");
 860     if (devaddr && type != IF_VIRTIO) {
 861         error_report("addr is not supported by this bus type");
 862         goto fail;
 863     }
 864
 865     if (type == IF_VIRTIO) {
 866         QemuOpts *devopts;
 867         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
 868                                    &error_abort);
 869         if (arch_type == QEMU_ARCH_S390X) {
 870             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
 871         } else {
 872             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
 873         }
 874         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
 875         if (devaddr) {
 876             qemu_opt_set(devopts, "addr", devaddr);
 877         }
 878     }
 879
 880     filename = qemu_opt_get(legacy_opts, "file");
 881
 882     /* Check werror/rerror compatibility with if=... */
 883     werror = qemu_opt_get(legacy_opts, "werror");
 884     if (werror != NULL) {
 885         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
 886             type != IF_NONE) {
 887             error_report("werror is not supported by this bus type");
 888             goto fail;
 889         }
 890         qdict_put(bs_opts, "werror", qstring_from_str(werror));
 891     }
 892
 893     rerror = qemu_opt_get(legacy_opts, "rerror");
 894     if (rerror != NULL) {
 895         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
 896             type != IF_NONE) {
 897             error_report("rerror is not supported by this bus type");
 898             goto fail;
 899         }
 900         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
 901     }
 902
 903     /* Actual block device init: Functionality shared with blockdev-add */
 904     dinfo = blockdev_init(filename, bs_opts, &local_err);
 905     if (dinfo == NULL) {
 906         if (local_err) {
 907             qerror_report_err(local_err);
 908             error_free(local_err);
 909         }
 910         goto fail;
 911     } else {
 912         assert(!local_err);
 913     }
 914
 915     /* Set legacy DriveInfo fields */
 916     dinfo->enable_auto_del = true;
 917     dinfo->opts = all_opts;
 918
 919     dinfo->cyls = cyls;
 920     dinfo->heads = heads;
 921     dinfo->secs = secs;
 922     dinfo->trans = translation;
 923
 924     dinfo->type = type;
 925     dinfo->bus = bus_id;
 926     dinfo->unit = unit_id;
 927     dinfo->devaddr = devaddr;
 928
 929     switch(type) {
 930     case IF_IDE:
 931     case IF_SCSI:
 932     case IF_XEN:
 933     case IF_NONE:
 934         dinfo->media_cd = media == MEDIA_CDROM;
 935         break;
 936     default:
 937         break;
 938     }
 939
 940 fail:
 941     qemu_opts_del(legacy_opts);
 942     return dinfo;
 943 }
 944
 945 void do_commit(Monitor *mon, const QDict *qdict)
 946 {
 947     const char *device = qdict_get_str(qdict, "device");
 948     BlockDriverState *bs;
 949     int ret;
 950
 951     if (!strcmp(device, "all")) {
 952         ret = bdrv_commit_all();
 953     } else {
 954         bs = bdrv_find(device);
 955         if (!bs) {
 956             monitor_printf(mon, "Device '%s' not found\n", device);
 957             return;
 958         }
 959         ret = bdrv_commit(bs);
 960     }
 961     if (ret < 0) {
 962         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
 963                        strerror(-ret));
 964     }
 965 }
 966
 967 static void blockdev_do_action(int kind, void *data, Error **errp)
 968 {
 969     TransactionAction action;
 970     TransactionActionList list;
 971
 972     action.kind = kind;
 973     action.data = data;
 974     list.value = &action;
 975     list.next = NULL;
 976     qmp_transaction(&list, errp);
 977 }
 978
 979 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
 980                                 bool has_node_name, const char *node_name,
 981                                 const char *snapshot_file,
 982                                 bool has_snapshot_node_name,
 983                                 const char *snapshot_node_name,
 984                                 bool has_format, const char *format,
 985                                 bool has_mode, NewImageMode mode, Error **errp)
 986 {
 987     BlockdevSnapshot snapshot = {
 988         .has_device = has_device,
 989         .device = (char *) device,
 990         .has_node_name = has_node_name,
 991         .node_name = (char *) node_name,
 992         .snapshot_file = (char *) snapshot_file,
 993         .has_snapshot_node_name = has_snapshot_node_name,
 994         .snapshot_node_name = (char *) snapshot_node_name,
 995         .has_format = has_format,
 996         .format = (char *) format,
 997         .has_mode = has_mode,
 998         .mode = mode,
 999     };
1000     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1001                        &snapshot, errp);
1002 }
1003
1004 void qmp_blockdev_snapshot_internal_sync(const char *device,
1005                                          const char *name,
1006                                          Error **errp)
1007 {
1008     BlockdevSnapshotInternal snapshot = {
1009         .device = (char *) device,
1010         .name = (char *) name
1011     };
1012
1013     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1014                        &snapshot, errp);
1015 }
1016
1017 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1018                                                          bool has_id,
1019                                                          const char *id,
1020                                                          bool has_name,
1021                                                          const char *name,
1022                                                          Error **errp)
1023 {
1024     BlockDriverState *bs = bdrv_find(device);
1025     QEMUSnapshotInfo sn;
1026     Error *local_err = NULL;
1027     SnapshotInfo *info = NULL;
1028     int ret;
1029
1030     if (!bs) {
1031         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1032         return NULL;
1033     }
1034
1035     if (!has_id) {
1036         id = NULL;
1037     }
1038
1039     if (!has_name) {
1040         name = NULL;
1041     }
1042
1043     if (!id && !name) {
1044         error_setg(errp, "Name or id must be provided");
1045         return NULL;
1046     }
1047
1048     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1049     if (local_err) {
1050         error_propagate(errp, local_err);
1051         return NULL;
1052     }
1053     if (!ret) {
1054         error_setg(errp,
1055                    "Snapshot with id '%s' and name '%s' does not exist on "
1056                    "device '%s'",
1057                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1058         return NULL;
1059     }
1060
1061     bdrv_snapshot_delete(bs, id, name, &local_err);
1062     if (local_err) {
1063         error_propagate(errp, local_err);
1064         return NULL;
1065     }
1066
1067     info = g_malloc0(sizeof(SnapshotInfo));
1068     info->id = g_strdup(sn.id_str);
1069     info->name = g_strdup(sn.name);
1070     info->date_nsec = sn.date_nsec;
1071     info->date_sec = sn.date_sec;
1072     info->vm_state_size = sn.vm_state_size;
1073     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1074     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1075
1076     return info;
1077 }
1078
1079 /* New and old BlockDriverState structs for group snapshots */
1080
1081 typedef struct BlkTransactionState BlkTransactionState;
1082
1083 /* Only prepare() may fail. In a single transaction, only one of commit() or
1084    abort() will be called, clean() will always be called if it present. */
1085 typedef struct BdrvActionOps {
1086     /* Size of state struct, in bytes. */
1087     size_t instance_size;
1088     /* Prepare the work, must NOT be NULL. */
1089     void (*prepare)(BlkTransactionState *common, Error **errp);
1090     /* Commit the changes, can be NULL. */
1091     void (*commit)(BlkTransactionState *common);
1092     /* Abort the changes on fail, can be NULL. */
1093     void (*abort)(BlkTransactionState *common);
1094     /* Clean up resource in the end, can be NULL. */
1095     void (*clean)(BlkTransactionState *common);
1096 } BdrvActionOps;
1097
1098 /*
1099  * This structure must be arranged as first member in child type, assuming
1100  * that compiler will also arrange it to the same address with parent instance.
1101  * Later it will be used in free().
1102  */
1103 struct BlkTransactionState {
1104     TransactionAction *action;
1105     const BdrvActionOps *ops;
1106     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1107 };
1108
1109 /* internal snapshot private data */
1110 typedef struct InternalSnapshotState {
1111     BlkTransactionState common;
1112     BlockDriverState *bs;
1113     QEMUSnapshotInfo sn;
1114 } InternalSnapshotState;
1115
1116 static void internal_snapshot_prepare(BlkTransactionState *common,
1117                                       Error **errp)
1118 {
1119     const char *device;
1120     const char *name;
1121     BlockDriverState *bs;
1122     QEMUSnapshotInfo old_sn, *sn;
1123     bool ret;
1124     qemu_timeval tv;
1125     BlockdevSnapshotInternal *internal;
1126     InternalSnapshotState *state;
1127     int ret1;
1128
1129     g_assert(common->action->kind ==
1130              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1131     internal = common->action->blockdev_snapshot_internal_sync;
1132     state = DO_UPCAST(InternalSnapshotState, common, common);
1133
1134     /* 1. parse input */
1135     device = internal->device;
1136     name = internal->name;
1137
1138     /* 2. check for validation */
1139     bs = bdrv_find(device);
1140     if (!bs) {
1141         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1142         return;
1143     }
1144
1145     if (!bdrv_is_inserted(bs)) {
1146         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1147         return;
1148     }
1149
1150     if (bdrv_is_read_only(bs)) {
1151         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1152         return;
1153     }
1154
1155     if (!bdrv_can_snapshot(bs)) {
1156         error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1157                   bs->drv->format_name, device, "internal snapshot");
1158         return;
1159     }
1160
1161     if (!strlen(name)) {
1162         error_setg(errp, "Name is empty");
1163         return;
1164     }
1165
1166     /* check whether a snapshot with name exist */
1167     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn, errp);
1168     if (error_is_set(errp)) {
1169         return;
1170     } else if (ret) {
1171         error_setg(errp,
1172                    "Snapshot with name '%s' already exists on device '%s'",
1173                    name, device);
1174         return;
1175     }
1176
1177     /* 3. take the snapshot */
1178     sn = &state->sn;
1179     pstrcpy(sn->name, sizeof(sn->name), name);
1180     qemu_gettimeofday(&tv);
1181     sn->date_sec = tv.tv_sec;
1182     sn->date_nsec = tv.tv_usec * 1000;
1183     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1184
1185     ret1 = bdrv_snapshot_create(bs, sn);
1186     if (ret1 < 0) {
1187         error_setg_errno(errp, -ret1,
1188                          "Failed to create snapshot '%s' on device '%s'",
1189                          name, device);
1190         return;
1191     }
1192
1193     /* 4. succeed, mark a snapshot is created */
1194     state->bs = bs;
1195 }
1196
1197 static void internal_snapshot_abort(BlkTransactionState *common)
1198 {
1199     InternalSnapshotState *state =
1200                              DO_UPCAST(InternalSnapshotState, common, common);
1201     BlockDriverState *bs = state->bs;
1202     QEMUSnapshotInfo *sn = &state->sn;
1203     Error *local_error = NULL;
1204
1205     if (!bs) {
1206         return;
1207     }
1208
1209     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1210         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1211                      "device '%s' in abort: %s",
1212                      sn->id_str,
1213                      sn->name,
1214                      bdrv_get_device_name(bs),
1215                      error_get_pretty(local_error));
1216         error_free(local_error);
1217     }
1218 }
1219
1220 /* external snapshot private data */
1221 typedef struct ExternalSnapshotState {
1222     BlkTransactionState common;
1223     BlockDriverState *old_bs;
1224     BlockDriverState *new_bs;
1225 } ExternalSnapshotState;
1226
1227 static void external_snapshot_prepare(BlkTransactionState *common,
1228                                       Error **errp)
1229 {
1230     BlockDriver *drv;
1231     int flags, ret;
1232     QDict *options = NULL;
1233     Error *local_err = NULL;
1234     bool has_device = false;
1235     const char *device;
1236     bool has_node_name = false;
1237     const char *node_name;
1238     bool has_snapshot_node_name = false;
1239     const char *snapshot_node_name;
1240     const char *new_image_file;
1241     const char *format = "qcow2";
1242     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1243     ExternalSnapshotState *state =
1244                              DO_UPCAST(ExternalSnapshotState, common, common);
1245     TransactionAction *action = common->action;
1246
1247     /* get parameters */
1248     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1249
1250     has_device = action->blockdev_snapshot_sync->has_device;
1251     device = action->blockdev_snapshot_sync->device;
1252     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1253     node_name = action->blockdev_snapshot_sync->node_name;
1254     has_snapshot_node_name =
1255         action->blockdev_snapshot_sync->has_snapshot_node_name;
1256     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1257
1258     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1259     if (action->blockdev_snapshot_sync->has_format) {
1260         format = action->blockdev_snapshot_sync->format;
1261     }
1262     if (action->blockdev_snapshot_sync->has_mode) {
1263         mode = action->blockdev_snapshot_sync->mode;
1264     }
1265
1266     /* start processing */
1267     drv = bdrv_find_format(format);
1268     if (!drv) {
1269         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1270         return;
1271     }
1272
1273     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1274                                    has_node_name ? node_name : NULL,
1275                                    &local_err);
1276     if (local_err) {
1277         error_propagate(errp, local_err);
1278         return;
1279     }
1280
1281     if (has_node_name && !has_snapshot_node_name) {
1282         error_setg(errp, "New snapshot node name missing");
1283         return;
1284     }
1285
1286     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1287         error_setg(errp, "New snapshot node name already existing");
1288         return;
1289     }
1290
1291     if (!bdrv_is_inserted(state->old_bs)) {
1292         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1293         return;
1294     }
1295
1296     if (bdrv_in_use(state->old_bs)) {
1297         error_set(errp, QERR_DEVICE_IN_USE, device);
1298         return;
1299     }
1300
1301     if (!bdrv_is_read_only(state->old_bs)) {
1302         if (bdrv_flush(state->old_bs)) {
1303             error_set(errp, QERR_IO_ERROR);
1304             return;
1305         }
1306     }
1307
1308     if (!bdrv_is_first_non_filter(state->old_bs)) {
1309         error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1310         return;
1311     }
1312
1313     flags = state->old_bs->open_flags;
1314
1315     /* create new image w/backing file */
1316     if (mode != NEW_IMAGE_MODE_EXISTING) {
1317         bdrv_img_create(new_image_file, format,
1318                         state->old_bs->filename,
1319                         state->old_bs->drv->format_name,
1320                         NULL, -1, flags, &local_err, false);
1321         if (local_err) {
1322             error_propagate(errp, local_err);
1323             return;
1324         }
1325     }
1326
1327     if (has_snapshot_node_name) {
1328         options = qdict_new();
1329         qdict_put(options, "node-name",
1330                   qstring_from_str(snapshot_node_name));
1331     }
1332
1333     /* TODO Inherit bs->options or only take explicit options with an
1334      * extended QMP command? */
1335     assert(state->new_bs == NULL);
1336     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1337                     flags | BDRV_O_NO_BACKING, drv, &local_err);
1338     /* We will manually add the backing_hd field to the bs later */
1339     if (ret != 0) {
1340         error_propagate(errp, local_err);
1341     }
1342 }
1343
1344 static void external_snapshot_commit(BlkTransactionState *common)
1345 {
1346     ExternalSnapshotState *state =
1347                              DO_UPCAST(ExternalSnapshotState, common, common);
1348
1349     /* This removes our old bs and adds the new bs */
1350     bdrv_append(state->new_bs, state->old_bs);
1351     /* We don't need (or want) to use the transactional
1352      * bdrv_reopen_multiple() across all the entries at once, because we
1353      * don't want to abort all of them if one of them fails the reopen */
1354     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1355                 NULL);
1356 }
1357
1358 static void external_snapshot_abort(BlkTransactionState *common)
1359 {
1360     ExternalSnapshotState *state =
1361                              DO_UPCAST(ExternalSnapshotState, common, common);
1362     if (state->new_bs) {
1363         bdrv_unref(state->new_bs);
1364     }
1365 }
1366
1367 typedef struct DriveBackupState {
1368     BlkTransactionState common;
1369     BlockDriverState *bs;
1370     BlockJob *job;
1371 } DriveBackupState;
1372
1373 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1374 {
1375     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1376     DriveBackup *backup;
1377     Error *local_err = NULL;
1378
1379     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1380     backup = common->action->drive_backup;
1381
1382     qmp_drive_backup(backup->device, backup->target,
1383                      backup->has_format, backup->format,
1384                      backup->sync,
1385                      backup->has_mode, backup->mode,
1386                      backup->has_speed, backup->speed,
1387                      backup->has_on_source_error, backup->on_source_error,
1388                      backup->has_on_target_error, backup->on_target_error,
1389                      &local_err);
1390     if (local_err) {
1391         error_propagate(errp, local_err);
1392         state->bs = NULL;
1393         state->job = NULL;
1394         return;
1395     }
1396
1397     state->bs = bdrv_find(backup->device);
1398     state->job = state->bs->job;
1399 }
1400
1401 static void drive_backup_abort(BlkTransactionState *common)
1402 {
1403     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1404     BlockDriverState *bs = state->bs;
1405
1406     /* Only cancel if it's the job we started */
1407     if (bs && bs->job && bs->job == state->job) {
1408         block_job_cancel_sync(bs->job);
1409     }
1410 }
1411
1412 static void abort_prepare(BlkTransactionState *common, Error **errp)
1413 {
1414     error_setg(errp, "Transaction aborted using Abort action");
1415 }
1416
1417 static void abort_commit(BlkTransactionState *common)
1418 {
1419     g_assert_not_reached(); /* this action never succeeds */
1420 }
1421
1422 static const BdrvActionOps actions[] = {
1423     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1424         .instance_size = sizeof(ExternalSnapshotState),
1425         .prepare  = external_snapshot_prepare,
1426         .commit   = external_snapshot_commit,
1427         .abort = external_snapshot_abort,
1428     },
1429     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1430         .instance_size = sizeof(DriveBackupState),
1431         .prepare = drive_backup_prepare,
1432         .abort = drive_backup_abort,
1433     },
1434     [TRANSACTION_ACTION_KIND_ABORT] = {
1435         .instance_size = sizeof(BlkTransactionState),
1436         .prepare = abort_prepare,
1437         .commit = abort_commit,
1438     },
1439     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1440         .instance_size = sizeof(InternalSnapshotState),
1441         .prepare  = internal_snapshot_prepare,
1442         .abort = internal_snapshot_abort,
1443     },
1444 };
1445
1446 /*
1447  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1448  *  then we do not pivot any of the devices in the group, and abandon the
1449  *  snapshots
1450  */
1451 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1452 {
1453     TransactionActionList *dev_entry = dev_list;
1454     BlkTransactionState *state, *next;
1455     Error *local_err = NULL;
1456
1457     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1458     QSIMPLEQ_INIT(&snap_bdrv_states);
1459
1460     /* drain all i/o before any snapshots */
1461     bdrv_drain_all();
1462
1463     /* We don't do anything in this loop that commits us to the snapshot */
1464     while (NULL != dev_entry) {
1465         TransactionAction *dev_info = NULL;
1466         const BdrvActionOps *ops;
1467
1468         dev_info = dev_entry->value;
1469         dev_entry = dev_entry->next;
1470
1471         assert(dev_info->kind < ARRAY_SIZE(actions));
1472
1473         ops = &actions[dev_info->kind];
1474         assert(ops->instance_size > 0);
1475
1476         state = g_malloc0(ops->instance_size);
1477         state->ops = ops;
1478         state->action = dev_info;
1479         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1480
1481         state->ops->prepare(state, &local_err);
1482         if (local_err) {
1483             error_propagate(errp, local_err);
1484             goto delete_and_fail;
1485         }
1486     }
1487
1488     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1489         if (state->ops->commit) {
1490             state->ops->commit(state);
1491         }
1492     }
1493
1494     /* success */
1495     goto exit;
1496
1497 delete_and_fail:
1498     /*
1499     * failure, and it is all-or-none; abandon each new bs, and keep using
1500     * the original bs for all images
1501     */
1502     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1503         if (state->ops->abort) {
1504             state->ops->abort(state);
1505         }
1506     }
1507 exit:
1508     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1509         if (state->ops->clean) {
1510             state->ops->clean(state);
1511         }
1512         g_free(state);
1513     }
1514 }
1515
1516
1517 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1518 {
1519     if (bdrv_in_use(bs)) {
1520         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1521         return;
1522     }
1523     if (!bdrv_dev_has_removable_media(bs)) {
1524         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1525         return;
1526     }
1527
1528     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1529         bdrv_dev_eject_request(bs, force);
1530         if (!force) {
1531             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1532             return;
1533         }
1534     }
1535
1536     bdrv_close(bs);
1537 }
1538
1539 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1540 {
1541     BlockDriverState *bs;
1542
1543     bs = bdrv_find(device);
1544     if (!bs) {
1545         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1546         return;
1547     }
1548
1549     eject_device(bs, force, errp);
1550 }
1551
1552 void qmp_block_passwd(bool has_device, const char *device,
1553                       bool has_node_name, const char *node_name,
1554                       const char *password, Error **errp)
1555 {
1556     Error *local_err = NULL;
1557     BlockDriverState *bs;
1558     int err;
1559
1560     bs = bdrv_lookup_bs(has_device ? device : NULL,
1561                         has_node_name ? node_name : NULL,
1562                         &local_err);
1563     if (local_err) {
1564         error_propagate(errp, local_err);
1565         return;
1566     }
1567
1568     err = bdrv_set_key(bs, password);
1569     if (err == -EINVAL) {
1570         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1571         return;
1572     } else if (err < 0) {
1573         error_set(errp, QERR_INVALID_PASSWORD);
1574         return;
1575     }
1576 }
1577
1578 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1579                                     int bdrv_flags, BlockDriver *drv,
1580                                     const char *password, Error **errp)
1581 {
1582     Error *local_err = NULL;
1583     int ret;
1584
1585     ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1586     if (ret < 0) {
1587         error_propagate(errp, local_err);
1588         return;
1589     }
1590
1591     if (bdrv_key_required(bs)) {
1592         if (password) {
1593             if (bdrv_set_key(bs, password) < 0) {
1594                 error_set(errp, QERR_INVALID_PASSWORD);
1595             }
1596         } else {
1597             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1598                       bdrv_get_encrypted_filename(bs));
1599         }
1600     } else if (password) {
1601         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1602     }
1603 }
1604
1605 void qmp_change_blockdev(const char *device, const char *filename,
1606                          const char *format, Error **errp)
1607 {
1608     BlockDriverState *bs;
1609     BlockDriver *drv = NULL;
1610     int bdrv_flags;
1611     Error *err = NULL;
1612
1613     bs = bdrv_find(device);
1614     if (!bs) {
1615         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1616         return;
1617     }
1618
1619     if (format) {
1620         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1621         if (!drv) {
1622             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1623             return;
1624         }
1625     }
1626
1627     eject_device(bs, 0, &err);
1628     if (err) {
1629         error_propagate(errp, err);
1630         return;
1631     }
1632
1633     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1634     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1635
1636     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1637 }
1638
1639 /* throttling disk I/O limits */
1640 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1641                                int64_t bps_wr,
1642                                int64_t iops,
1643                                int64_t iops_rd,
1644                                int64_t iops_wr,
1645                                bool has_bps_max,
1646                                int64_t bps_max,
1647                                bool has_bps_rd_max,
1648                                int64_t bps_rd_max,
1649                                bool has_bps_wr_max,
1650                                int64_t bps_wr_max,
1651                                bool has_iops_max,
1652                                int64_t iops_max,
1653                                bool has_iops_rd_max,
1654                                int64_t iops_rd_max,
1655                                bool has_iops_wr_max,
1656                                int64_t iops_wr_max,
1657                                bool has_iops_size,
1658                                int64_t iops_size, Error **errp)
1659 {
1660     ThrottleConfig cfg;
1661     BlockDriverState *bs;
1662
1663     bs = bdrv_find(device);
1664     if (!bs) {
1665         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1666         return;
1667     }
1668
1669     memset(&cfg, 0, sizeof(cfg));
1670     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1671     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1672     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1673
1674     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1675     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1676     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1677
1678     if (has_bps_max) {
1679         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1680     }
1681     if (has_bps_rd_max) {
1682         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1683     }
1684     if (has_bps_wr_max) {
1685         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1686     }
1687     if (has_iops_max) {
1688         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1689     }
1690     if (has_iops_rd_max) {
1691         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1692     }
1693     if (has_iops_wr_max) {
1694         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1695     }
1696
1697     if (has_iops_size) {
1698         cfg.op_size = iops_size;
1699     }
1700
1701     if (!check_throttle_config(&cfg, errp)) {
1702         return;
1703     }
1704
1705     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1706         bdrv_io_limits_enable(bs);
1707     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1708         bdrv_io_limits_disable(bs);
1709     }
1710
1711     if (bs->io_limits_enabled) {
1712         bdrv_set_io_limits(bs, &cfg);
1713     }
1714 }
1715
1716 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1717 {
1718     const char *id = qdict_get_str(qdict, "id");
1719     BlockDriverState *bs;
1720
1721     bs = bdrv_find(id);
1722     if (!bs) {
1723         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1724         return -1;
1725     }
1726     if (bdrv_in_use(bs)) {
1727         qerror_report(QERR_DEVICE_IN_USE, id);
1728         return -1;
1729     }
1730
1731     /* quiesce block driver; prevent further io */
1732     bdrv_drain_all();
1733     bdrv_flush(bs);
1734     bdrv_close(bs);
1735
1736     /* if we have a device attached to this BlockDriverState
1737      * then we need to make the drive anonymous until the device
1738      * can be removed.  If this is a drive with no device backing
1739      * then we can just get rid of the block driver state right here.
1740      */
1741     if (bdrv_get_attached_dev(bs)) {
1742         bdrv_make_anon(bs);
1743
1744         /* Further I/O must not pause the guest */
1745         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1746                           BLOCKDEV_ON_ERROR_REPORT);
1747     } else {
1748         drive_uninit(drive_get_by_blockdev(bs));
1749     }
1750
1751     return 0;
1752 }
1753
1754 void qmp_block_resize(bool has_device, const char *device,
1755                       bool has_node_name, const char *node_name,
1756                       int64_t size, Error **errp)
1757 {
1758     Error *local_err = NULL;
1759     BlockDriverState *bs;
1760     int ret;
1761
1762     bs = bdrv_lookup_bs(has_device ? device : NULL,
1763                         has_node_name ? node_name : NULL,
1764                         &local_err);
1765     if (local_err) {
1766         error_propagate(errp, local_err);
1767         return;
1768     }
1769
1770     if (!bdrv_is_first_non_filter(bs)) {
1771         error_set(errp, QERR_FEATURE_DISABLED, "resize");
1772         return;
1773     }
1774
1775     if (size < 0) {
1776         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1777         return;
1778     }
1779
1780     /* complete all in-flight operations before resizing the device */
1781     bdrv_drain_all();
1782
1783     ret = bdrv_truncate(bs, size);
1784     switch (ret) {
1785     case 0:
1786         break;
1787     case -ENOMEDIUM:
1788         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1789         break;
1790     case -ENOTSUP:
1791         error_set(errp, QERR_UNSUPPORTED);
1792         break;
1793     case -EACCES:
1794         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1795         break;
1796     case -EBUSY:
1797         error_set(errp, QERR_DEVICE_IN_USE, device);
1798         break;
1799     default:
1800         error_setg_errno(errp, -ret, "Could not resize");
1801         break;
1802     }
1803 }
1804
1805 static void block_job_cb(void *opaque, int ret)
1806 {
1807     BlockDriverState *bs = opaque;
1808     QObject *obj;
1809
1810     trace_block_job_cb(bs, bs->job, ret);
1811
1812     assert(bs->job);
1813     obj = qobject_from_block_job(bs->job);
1814     if (ret < 0) {
1815         QDict *dict = qobject_to_qdict(obj);
1816         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1817     }
1818
1819     if (block_job_is_cancelled(bs->job)) {
1820         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1821     } else {
1822         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1823     }
1824     qobject_decref(obj);
1825
1826     bdrv_put_ref_bh_schedule(bs);
1827 }
1828
1829 void qmp_block_stream(const char *device, bool has_base,
1830                       const char *base, bool has_speed, int64_t speed,
1831                       bool has_on_error, BlockdevOnError on_error,
1832                       Error **errp)
1833 {
1834     BlockDriverState *bs;
1835     BlockDriverState *base_bs = NULL;
1836     Error *local_err = NULL;
1837
1838     if (!has_on_error) {
1839         on_error = BLOCKDEV_ON_ERROR_REPORT;
1840     }
1841
1842     bs = bdrv_find(device);
1843     if (!bs) {
1844         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1845         return;
1846     }
1847
1848     if (base) {
1849         base_bs = bdrv_find_backing_image(bs, base);
1850         if (base_bs == NULL) {
1851             error_set(errp, QERR_BASE_NOT_FOUND, base);
1852             return;
1853         }
1854     }
1855
1856     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1857                  on_error, block_job_cb, bs, &local_err);
1858     if (local_err) {
1859         error_propagate(errp, local_err);
1860         return;
1861     }
1862
1863     trace_qmp_block_stream(bs, bs->job);
1864 }
1865
1866 void qmp_block_commit(const char *device,
1867                       bool has_base, const char *base, const char *top,
1868                       bool has_speed, int64_t speed,
1869                       Error **errp)
1870 {
1871     BlockDriverState *bs;
1872     BlockDriverState *base_bs, *top_bs;
1873     Error *local_err = NULL;
1874     /* This will be part of the QMP command, if/when the
1875      * BlockdevOnError change for blkmirror makes it in
1876      */
1877     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1878
1879     /* drain all i/o before commits */
1880     bdrv_drain_all();
1881
1882     bs = bdrv_find(device);
1883     if (!bs) {
1884         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1885         return;
1886     }
1887
1888     /* default top_bs is the active layer */
1889     top_bs = bs;
1890
1891     if (top) {
1892         if (strcmp(bs->filename, top) != 0) {
1893             top_bs = bdrv_find_backing_image(bs, top);
1894         }
1895     }
1896
1897     if (top_bs == NULL) {
1898         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1899         return;
1900     }
1901
1902     if (has_base && base) {
1903         base_bs = bdrv_find_backing_image(top_bs, base);
1904     } else {
1905         base_bs = bdrv_find_base(top_bs);
1906     }
1907
1908     if (base_bs == NULL) {
1909         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1910         return;
1911     }
1912
1913     if (top_bs == bs) {
1914         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1915                             bs, &local_err);
1916     } else {
1917         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1918                     &local_err);
1919     }
1920     if (local_err != NULL) {
1921         error_propagate(errp, local_err);
1922         return;
1923     }
1924 }
1925
1926 void qmp_drive_backup(const char *device, const char *target,
1927                       bool has_format, const char *format,
1928                       enum MirrorSyncMode sync,
1929                       bool has_mode, enum NewImageMode mode,
1930                       bool has_speed, int64_t speed,
1931                       bool has_on_source_error, BlockdevOnError on_source_error,
1932                       bool has_on_target_error, BlockdevOnError on_target_error,
1933                       Error **errp)
1934 {
1935     BlockDriverState *bs;
1936     BlockDriverState *target_bs;
1937     BlockDriverState *source = NULL;
1938     BlockDriver *drv = NULL;
1939     Error *local_err = NULL;
1940     int flags;
1941     int64_t size;
1942     int ret;
1943
1944     if (!has_speed) {
1945         speed = 0;
1946     }
1947     if (!has_on_source_error) {
1948         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1949     }
1950     if (!has_on_target_error) {
1951         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1952     }
1953     if (!has_mode) {
1954         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1955     }
1956
1957     bs = bdrv_find(device);
1958     if (!bs) {
1959         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1960         return;
1961     }
1962
1963     if (!bdrv_is_inserted(bs)) {
1964         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1965         return;
1966     }
1967
1968     if (!has_format) {
1969         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1970     }
1971     if (format) {
1972         drv = bdrv_find_format(format);
1973         if (!drv) {
1974             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1975             return;
1976         }
1977     }
1978
1979     if (bdrv_in_use(bs)) {
1980         error_set(errp, QERR_DEVICE_IN_USE, device);
1981         return;
1982     }
1983
1984     flags = bs->open_flags | BDRV_O_RDWR;
1985
1986     /* See if we have a backing HD we can use to create our new image
1987      * on top of. */
1988     if (sync == MIRROR_SYNC_MODE_TOP) {
1989         source = bs->backing_hd;
1990         if (!source) {
1991             sync = MIRROR_SYNC_MODE_FULL;
1992         }
1993     }
1994     if (sync == MIRROR_SYNC_MODE_NONE) {
1995         source = bs;
1996     }
1997
1998     size = bdrv_getlength(bs);
1999     if (size < 0) {
2000         error_setg_errno(errp, -size, "bdrv_getlength failed");
2001         return;
2002     }
2003
2004     if (mode != NEW_IMAGE_MODE_EXISTING) {
2005         assert(format && drv);
2006         if (source) {
2007             bdrv_img_create(target, format, source->filename,
2008                             source->drv->format_name, NULL,
2009                             size, flags, &local_err, false);
2010         } else {
2011             bdrv_img_create(target, format, NULL, NULL, NULL,
2012                             size, flags, &local_err, false);
2013         }
2014     }
2015
2016     if (local_err) {
2017         error_propagate(errp, local_err);
2018         return;
2019     }
2020
2021     target_bs = NULL;
2022     ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2023     if (ret < 0) {
2024         error_propagate(errp, local_err);
2025         return;
2026     }
2027
2028     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2029                  block_job_cb, bs, &local_err);
2030     if (local_err != NULL) {
2031         bdrv_unref(target_bs);
2032         error_propagate(errp, local_err);
2033         return;
2034     }
2035 }
2036
2037 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2038 {
2039     return bdrv_named_nodes_list();
2040 }
2041
2042 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
2043
2044 void qmp_drive_mirror(const char *device, const char *target,
2045                       bool has_format, const char *format,
2046                       enum MirrorSyncMode sync,
2047                       bool has_mode, enum NewImageMode mode,
2048                       bool has_speed, int64_t speed,
2049                       bool has_granularity, uint32_t granularity,
2050                       bool has_buf_size, int64_t buf_size,
2051                       bool has_on_source_error, BlockdevOnError on_source_error,
2052                       bool has_on_target_error, BlockdevOnError on_target_error,
2053                       Error **errp)
2054 {
2055     BlockDriverState *bs;
2056     BlockDriverState *source, *target_bs;
2057     BlockDriver *drv = NULL;
2058     Error *local_err = NULL;
2059     int flags;
2060     int64_t size;
2061     int ret;
2062
2063     if (!has_speed) {
2064         speed = 0;
2065     }
2066     if (!has_on_source_error) {
2067         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2068     }
2069     if (!has_on_target_error) {
2070         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2071     }
2072     if (!has_mode) {
2073         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2074     }
2075     if (!has_granularity) {
2076         granularity = 0;
2077     }
2078     if (!has_buf_size) {
2079         buf_size = DEFAULT_MIRROR_BUF_SIZE;
2080     }
2081
2082     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2083         error_set(errp, QERR_INVALID_PARAMETER, device);
2084         return;
2085     }
2086     if (granularity & (granularity - 1)) {
2087         error_set(errp, QERR_INVALID_PARAMETER, device);
2088         return;
2089     }
2090
2091     bs = bdrv_find(device);
2092     if (!bs) {
2093         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2094         return;
2095     }
2096
2097     if (!bdrv_is_inserted(bs)) {
2098         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2099         return;
2100     }
2101
2102     if (!has_format) {
2103         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2104     }
2105     if (format) {
2106         drv = bdrv_find_format(format);
2107         if (!drv) {
2108             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2109             return;
2110         }
2111     }
2112
2113     if (bdrv_in_use(bs)) {
2114         error_set(errp, QERR_DEVICE_IN_USE, device);
2115         return;
2116     }
2117
2118     flags = bs->open_flags | BDRV_O_RDWR;
2119     source = bs->backing_hd;
2120     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2121         sync = MIRROR_SYNC_MODE_FULL;
2122     }
2123     if (sync == MIRROR_SYNC_MODE_NONE) {
2124         source = bs;
2125     }
2126
2127     size = bdrv_getlength(bs);
2128     if (size < 0) {
2129         error_setg_errno(errp, -size, "bdrv_getlength failed");
2130         return;
2131     }
2132
2133     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2134         && mode != NEW_IMAGE_MODE_EXISTING)
2135     {
2136         /* create new image w/o backing file */
2137         assert(format && drv);
2138         bdrv_img_create(target, format,
2139                         NULL, NULL, NULL, size, flags, &local_err, false);
2140     } else {
2141         switch (mode) {
2142         case NEW_IMAGE_MODE_EXISTING:
2143             break;
2144         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2145             /* create new image with backing file */
2146             bdrv_img_create(target, format,
2147                             source->filename,
2148                             source->drv->format_name,
2149                             NULL, size, flags, &local_err, false);
2150             break;
2151         default:
2152             abort();
2153         }
2154     }
2155
2156     if (local_err) {
2157         error_propagate(errp, local_err);
2158         return;
2159     }
2160
2161     /* Mirroring takes care of copy-on-write using the source's backing
2162      * file.
2163      */
2164     target_bs = NULL;
2165     ret = bdrv_open(&target_bs, target, NULL, NULL, flags | BDRV_O_NO_BACKING,
2166                     drv, &local_err);
2167     if (ret < 0) {
2168         error_propagate(errp, local_err);
2169         return;
2170     }
2171
2172     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2173                  on_source_error, on_target_error,
2174                  block_job_cb, bs, &local_err);
2175     if (local_err != NULL) {
2176         bdrv_unref(target_bs);
2177         error_propagate(errp, local_err);
2178         return;
2179     }
2180 }
2181
2182 static BlockJob *find_block_job(const char *device)
2183 {
2184     BlockDriverState *bs;
2185
2186     bs = bdrv_find(device);
2187     if (!bs || !bs->job) {
2188         return NULL;
2189     }
2190     return bs->job;
2191 }
2192
2193 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2194 {
2195     BlockJob *job = find_block_job(device);
2196
2197     if (!job) {
2198         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2199         return;
2200     }
2201
2202     block_job_set_speed(job, speed, errp);
2203 }
2204
2205 void qmp_block_job_cancel(const char *device,
2206                           bool has_force, bool force, Error **errp)
2207 {
2208     BlockJob *job = find_block_job(device);
2209
2210     if (!has_force) {
2211         force = false;
2212     }
2213
2214     if (!job) {
2215         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2216         return;
2217     }
2218     if (job->paused && !force) {
2219         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
2220         return;
2221     }
2222
2223     trace_qmp_block_job_cancel(job);
2224     block_job_cancel(job);
2225 }
2226
2227 void qmp_block_job_pause(const char *device, Error **errp)
2228 {
2229     BlockJob *job = find_block_job(device);
2230
2231     if (!job) {
2232         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2233         return;
2234     }
2235
2236     trace_qmp_block_job_pause(job);
2237     block_job_pause(job);
2238 }
2239
2240 void qmp_block_job_resume(const char *device, Error **errp)
2241 {
2242     BlockJob *job = find_block_job(device);
2243
2244     if (!job) {
2245         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2246         return;
2247     }
2248
2249     trace_qmp_block_job_resume(job);
2250     block_job_resume(job);
2251 }
2252
2253 void qmp_block_job_complete(const char *device, Error **errp)
2254 {
2255     BlockJob *job = find_block_job(device);
2256
2257     if (!job) {
2258         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2259         return;
2260     }
2261
2262     trace_qmp_block_job_complete(job);
2263     block_job_complete(job, errp);
2264 }
2265
2266 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2267 {
2268     QmpOutputVisitor *ov = qmp_output_visitor_new();
2269     DriveInfo *dinfo;
2270     QObject *obj;
2271     QDict *qdict;
2272     Error *local_err = NULL;
2273
2274     /* Require an ID in the top level */
2275     if (!options->has_id) {
2276         error_setg(errp, "Block device needs an ID");
2277         goto fail;
2278     }
2279
2280     /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2281      * cache.direct=false instead of silently switching to aio=threads, except
2282      * if called from drive_init.
2283      *
2284      * For now, simply forbidding the combination for all drivers will do. */
2285     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2286         bool direct = options->has_cache &&
2287                       options->cache->has_direct &&
2288                       options->cache->direct;
2289         if (!direct) {
2290             error_setg(errp, "aio=native requires cache.direct=true");
2291             goto fail;
2292         }
2293     }
2294
2295     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2296                                &options, NULL, &local_err);
2297     if (local_err) {
2298         error_propagate(errp, local_err);
2299         goto fail;
2300     }
2301
2302     obj = qmp_output_get_qobject(ov);
2303     qdict = qobject_to_qdict(obj);
2304
2305     qdict_flatten(qdict);
2306
2307     dinfo = blockdev_init(NULL, qdict, &local_err);
2308     if (local_err) {
2309         error_propagate(errp, local_err);
2310         goto fail;
2311     }
2312
2313     if (bdrv_key_required(dinfo->bdrv)) {
2314         drive_uninit(dinfo);
2315         error_setg(errp, "blockdev-add doesn't support encrypted devices");
2316         goto fail;
2317     }
2318
2319 fail:
2320     qmp_output_visitor_cleanup(ov);
2321 }
2322
2323 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2324 {
2325     BlockJobInfoList **prev = opaque;
2326     BlockJob *job = bs->job;
2327
2328     if (job) {
2329         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2330         elem->value = block_job_query(bs->job);
2331         (*prev)->next = elem;
2332         *prev = elem;
2333     }
2334 }
2335
2336 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2337 {
2338     /* Dummy is a fake list element for holding the head pointer */
2339     BlockJobInfoList dummy = {};
2340     BlockJobInfoList *prev = &dummy;
2341     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2342     return dummy.next;
2343 }
2344
2345 QemuOptsList qemu_common_drive_opts = {
2346     .name = "drive",
2347     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2348     .desc = {
2349         {
2350             .name = "snapshot",
2351             .type = QEMU_OPT_BOOL,
2352             .help = "enable/disable snapshot mode",
2353         },{
2354             .name = "discard",
2355             .type = QEMU_OPT_STRING,
2356             .help = "discard operation (ignore/off, unmap/on)",
2357         },{
2358             .name = "cache.writeback",
2359             .type = QEMU_OPT_BOOL,
2360             .help = "enables writeback mode for any caches",
2361         },{
2362             .name = "cache.direct",
2363             .type = QEMU_OPT_BOOL,
2364             .help = "enables use of O_DIRECT (bypass the host page cache)",
2365         },{
2366             .name = "cache.no-flush",
2367             .type = QEMU_OPT_BOOL,
2368             .help = "ignore any flush requests for the device",
2369         },{
2370             .name = "aio",
2371             .type = QEMU_OPT_STRING,
2372             .help = "host AIO implementation (threads, native)",
2373         },{
2374             .name = "format",
2375             .type = QEMU_OPT_STRING,
2376             .help = "disk format (raw, qcow2, ...)",
2377         },{
2378             .name = "serial",
2379             .type = QEMU_OPT_STRING,
2380             .help = "disk serial number",
2381         },{
2382             .name = "rerror",
2383             .type = QEMU_OPT_STRING,
2384             .help = "read error action",
2385         },{
2386             .name = "werror",
2387             .type = QEMU_OPT_STRING,
2388             .help = "write error action",
2389         },{
2390             .name = "read-only",
2391             .type = QEMU_OPT_BOOL,
2392             .help = "open drive file as read-only",
2393         },{
2394             .name = "throttling.iops-total",
2395             .type = QEMU_OPT_NUMBER,
2396             .help = "limit total I/O operations per second",
2397         },{
2398             .name = "throttling.iops-read",
2399             .type = QEMU_OPT_NUMBER,
2400             .help = "limit read operations per second",
2401         },{
2402             .name = "throttling.iops-write",
2403             .type = QEMU_OPT_NUMBER,
2404             .help = "limit write operations per second",
2405         },{
2406             .name = "throttling.bps-total",
2407             .type = QEMU_OPT_NUMBER,
2408             .help = "limit total bytes per second",
2409         },{
2410             .name = "throttling.bps-read",
2411             .type = QEMU_OPT_NUMBER,
2412             .help = "limit read bytes per second",
2413         },{
2414             .name = "throttling.bps-write",
2415             .type = QEMU_OPT_NUMBER,
2416             .help = "limit write bytes per second",
2417         },{
2418             .name = "throttling.iops-total-max",
2419             .type = QEMU_OPT_NUMBER,
2420             .help = "I/O operations burst",
2421         },{
2422             .name = "throttling.iops-read-max",
2423             .type = QEMU_OPT_NUMBER,
2424             .help = "I/O operations read burst",
2425         },{
2426             .name = "throttling.iops-write-max",
2427             .type = QEMU_OPT_NUMBER,
2428             .help = "I/O operations write burst",
2429         },{
2430             .name = "throttling.bps-total-max",
2431             .type = QEMU_OPT_NUMBER,
2432             .help = "total bytes burst",
2433         },{
2434             .name = "throttling.bps-read-max",
2435             .type = QEMU_OPT_NUMBER,
2436             .help = "total bytes read burst",
2437         },{
2438             .name = "throttling.bps-write-max",
2439             .type = QEMU_OPT_NUMBER,
2440             .help = "total bytes write burst",
2441         },{
2442             .name = "throttling.iops-size",
2443             .type = QEMU_OPT_NUMBER,
2444             .help = "when limiting by iops max size of an I/O in bytes",
2445         },{
2446             .name = "copy-on-read",
2447             .type = QEMU_OPT_BOOL,
2448             .help = "copy read data from backing file into image file",
2449         },
2450         { /* end of list */ }
2451     },
2452 };
2453
2454 QemuOptsList qemu_drive_opts = {
2455     .name = "drive",
2456     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2457     .desc = {
2458         /*
2459          * no elements => accept any params
2460          * validation will happen later
2461          */
2462         { /* end of list */ }
2463     },
2464 };