block/rbd.c

   1 /*
   2  * QEMU Block driver for RADOS (Ceph)
   3  *
   4  * Copyright (C) 2010-2011 Christian Brunner <chb@muc.de>,
   5  *                         Josh Durgin <josh.durgin@dreamhost.com>
   6  *
   7  * This work is licensed under the terms of the GNU GPL, version 2.  See
   8  * the COPYING file in the top-level directory.
   9  *
  10  * Contributions after 2012-01-13 are licensed under the terms of the
  11  * GNU GPL, version 2 or (at your option) any later version.
  12  */
  13
  14 #include "qemu/osdep.h"
  15
  16 #include "qemu-common.h"
  17 #include "qemu/error-report.h"
  18 #include "block/block_int.h"
  19
  20 #include <rbd/librbd.h>
  21
  22 /*
  23  * When specifying the image filename use:
  24  *
  25  * rbd:poolname/devicename[@snapshotname][:option1=value1[:option2=value2...]]
  26  *
  27  * poolname must be the name of an existing rados pool.
  28  *
  29  * devicename is the name of the rbd image.
  30  *
  31  * Each option given is used to configure rados, and may be any valid
  32  * Ceph option, "id", or "conf".
  33  *
  34  * The "id" option indicates what user we should authenticate as to
  35  * the Ceph cluster.  If it is excluded we will use the Ceph default
  36  * (normally 'admin').
  37  *
  38  * The "conf" option specifies a Ceph configuration file to read.  If
  39  * it is not specified, we will read from the default Ceph locations
  40  * (e.g., /etc/ceph/ceph.conf).  To avoid reading _any_ configuration
  41  * file, specify conf=/dev/null.
  42  *
  43  * Configuration values containing :, @, or = can be escaped with a
  44  * leading "\".
  45  */
  46
  47 /* rbd_aio_discard added in 0.1.2 */
  48 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 2)
  49 #define LIBRBD_SUPPORTS_DISCARD
  50 #else
  51 #undef LIBRBD_SUPPORTS_DISCARD
  52 #endif
  53
  54 #define OBJ_MAX_SIZE (1UL << OBJ_DEFAULT_OBJ_ORDER)
  55
  56 #define RBD_MAX_CONF_NAME_SIZE 128
  57 #define RBD_MAX_CONF_VAL_SIZE 512
  58 #define RBD_MAX_CONF_SIZE 1024
  59 #define RBD_MAX_POOL_NAME_SIZE 128
  60 #define RBD_MAX_SNAP_NAME_SIZE 128
  61 #define RBD_MAX_SNAPS 100
  62
  63 typedef enum {
  64     RBD_AIO_READ,
  65     RBD_AIO_WRITE,
  66     RBD_AIO_DISCARD,
  67     RBD_AIO_FLUSH
  68 } RBDAIOCmd;
  69
  70 typedef struct RBDAIOCB {
  71     BlockAIOCB common;
  72     QEMUBH *bh;
  73     int64_t ret;
  74     QEMUIOVector *qiov;
  75     char *bounce;
  76     RBDAIOCmd cmd;
  77     int error;
  78     struct BDRVRBDState *s;
  79 } RBDAIOCB;
  80
  81 typedef struct RADOSCB {
  82     RBDAIOCB *acb;
  83     struct BDRVRBDState *s;
  84     int64_t size;
  85     char *buf;
  86     int64_t ret;
  87 } RADOSCB;
  88
  89 typedef struct BDRVRBDState {
  90     rados_t cluster;
  91     rados_ioctx_t io_ctx;
  92     rbd_image_t image;
  93     char name[RBD_MAX_IMAGE_NAME_SIZE];
  94     char *snap;
  95 } BDRVRBDState;
  96
  97 static int qemu_rbd_next_tok(char *dst, int dst_len,
  98                              char *src, char delim,
  99                              const char *name,
 100                              char **p, Error **errp)
 101 {
 102     int l;
 103     char *end;
 104
 105     *p = NULL;
 106
 107     if (delim != '\0') {
 108         for (end = src; *end; ++end) {
 109             if (*end == delim) {
 110                 break;
 111             }
 112             if (*end == '\\' && end[1] != '\0') {
 113                 end++;
 114             }
 115         }
 116         if (*end == delim) {
 117             *p = end + 1;
 118             *end = '\0';
 119         }
 120     }
 121     l = strlen(src);
 122     if (l >= dst_len) {
 123         error_setg(errp, "%s too long", name);
 124         return -EINVAL;
 125     } else if (l == 0) {
 126         error_setg(errp, "%s too short", name);
 127         return -EINVAL;
 128     }
 129
 130     pstrcpy(dst, dst_len, src);
 131
 132     return 0;
 133 }
 134
 135 static void qemu_rbd_unescape(char *src)
 136 {
 137     char *p;
 138
 139     for (p = src; *src; ++src, ++p) {
 140         if (*src == '\\' && src[1] != '\0') {
 141             src++;
 142         }
 143         *p = *src;
 144     }
 145     *p = '\0';
 146 }
 147
 148 static int qemu_rbd_parsename(const char *filename,
 149                               char *pool, int pool_len,
 150                               char *snap, int snap_len,
 151                               char *name, int name_len,
 152                               char *conf, int conf_len,
 153                               Error **errp)
 154 {
 155     const char *start;
 156     char *p, *buf;
 157     int ret;
 158
 159     if (!strstart(filename, "rbd:", &start)) {
 160         error_setg(errp, "File name must start with 'rbd:'");
 161         return -EINVAL;
 162     }
 163
 164     buf = g_strdup(start);
 165     p = buf;
 166     *snap = '\0';
 167     *conf = '\0';
 168
 169     ret = qemu_rbd_next_tok(pool, pool_len, p,
 170                             '/', "pool name", &p, errp);
 171     if (ret < 0 || !p) {
 172         ret = -EINVAL;
 173         goto done;
 174     }
 175     qemu_rbd_unescape(pool);
 176
 177     if (strchr(p, '@')) {
 178         ret = qemu_rbd_next_tok(name, name_len, p,
 179                                 '@', "object name", &p, errp);
 180         if (ret < 0) {
 181             goto done;
 182         }
 183         ret = qemu_rbd_next_tok(snap, snap_len, p,
 184                                 ':', "snap name", &p, errp);
 185         qemu_rbd_unescape(snap);
 186     } else {
 187         ret = qemu_rbd_next_tok(name, name_len, p,
 188                                 ':', "object name", &p, errp);
 189     }
 190     qemu_rbd_unescape(name);
 191     if (ret < 0 || !p) {
 192         goto done;
 193     }
 194
 195     ret = qemu_rbd_next_tok(conf, conf_len, p,
 196                             '\0', "configuration", &p, errp);
 197
 198 done:
 199     g_free(buf);
 200     return ret;
 201 }
 202
 203 static char *qemu_rbd_parse_clientname(const char *conf, char *clientname)
 204 {
 205     const char *p = conf;
 206
 207     while (*p) {
 208         int len;
 209         const char *end = strchr(p, ':');
 210
 211         if (end) {
 212             len = end - p;
 213         } else {
 214             len = strlen(p);
 215         }
 216
 217         if (strncmp(p, "id=", 3) == 0) {
 218             len -= 3;
 219             strncpy(clientname, p + 3, len);
 220             clientname[len] = '\0';
 221             return clientname;
 222         }
 223         if (end == NULL) {
 224             break;
 225         }
 226         p = end + 1;
 227     }
 228     return NULL;
 229 }
 230
 231 static int qemu_rbd_set_conf(rados_t cluster, const char *conf,
 232                              bool only_read_conf_file,
 233                              Error **errp)
 234 {
 235     char *p, *buf;
 236     char name[RBD_MAX_CONF_NAME_SIZE];
 237     char value[RBD_MAX_CONF_VAL_SIZE];
 238     int ret = 0;
 239
 240     buf = g_strdup(conf);
 241     p = buf;
 242
 243     while (p) {
 244         ret = qemu_rbd_next_tok(name, sizeof(name), p,
 245                                 '=', "conf option name", &p, errp);
 246         if (ret < 0) {
 247             break;
 248         }
 249         qemu_rbd_unescape(name);
 250
 251         if (!p) {
 252             error_setg(errp, "conf option %s has no value", name);
 253             ret = -EINVAL;
 254             break;
 255         }
 256
 257         ret = qemu_rbd_next_tok(value, sizeof(value), p,
 258                                 ':', "conf option value", &p, errp);
 259         if (ret < 0) {
 260             break;
 261         }
 262         qemu_rbd_unescape(value);
 263
 264         if (strcmp(name, "conf") == 0) {
 265             /* read the conf file alone, so it doesn't override more
 266                specific settings for a particular device */
 267             if (only_read_conf_file) {
 268                 ret = rados_conf_read_file(cluster, value);
 269                 if (ret < 0) {
 270                     error_setg(errp, "error reading conf file %s", value);
 271                     break;
 272                 }
 273             }
 274         } else if (strcmp(name, "id") == 0) {
 275             /* ignore, this is parsed by qemu_rbd_parse_clientname() */
 276         } else if (!only_read_conf_file) {
 277             ret = rados_conf_set(cluster, name, value);
 278             if (ret < 0) {
 279                 error_setg(errp, "invalid conf option %s", name);
 280                 ret = -EINVAL;
 281                 break;
 282             }
 283         }
 284     }
 285
 286     g_free(buf);
 287     return ret;
 288 }
 289
 290 static int qemu_rbd_create(const char *filename, QemuOpts *opts, Error **errp)
 291 {
 292     Error *local_err = NULL;
 293     int64_t bytes = 0;
 294     int64_t objsize;
 295     int obj_order = 0;
 296     char pool[RBD_MAX_POOL_NAME_SIZE];
 297     char name[RBD_MAX_IMAGE_NAME_SIZE];
 298     char snap_buf[RBD_MAX_SNAP_NAME_SIZE];
 299     char conf[RBD_MAX_CONF_SIZE];
 300     char clientname_buf[RBD_MAX_CONF_SIZE];
 301     char *clientname;
 302     rados_t cluster;
 303     rados_ioctx_t io_ctx;
 304     int ret;
 305
 306     if (qemu_rbd_parsename(filename, pool, sizeof(pool),
 307                            snap_buf, sizeof(snap_buf),
 308                            name, sizeof(name),
 309                            conf, sizeof(conf), &local_err) < 0) {
 310         error_propagate(errp, local_err);
 311         return -EINVAL;
 312     }
 313
 314     /* Read out options */
 315     bytes = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
 316                      BDRV_SECTOR_SIZE);
 317     objsize = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, 0);
 318     if (objsize) {
 319         if ((objsize - 1) & objsize) {    /* not a power of 2? */
 320             error_setg(errp, "obj size needs to be power of 2");
 321             return -EINVAL;
 322         }
 323         if (objsize < 4096) {
 324             error_setg(errp, "obj size too small");
 325             return -EINVAL;
 326         }
 327         obj_order = ctz32(objsize);
 328     }
 329
 330     clientname = qemu_rbd_parse_clientname(conf, clientname_buf);
 331     if (rados_create(&cluster, clientname) < 0) {
 332         error_setg(errp, "error initializing");
 333         return -EIO;
 334     }
 335
 336     if (strstr(conf, "conf=") == NULL) {
 337         /* try default location, but ignore failure */
 338         rados_conf_read_file(cluster, NULL);
 339     } else if (conf[0] != '\0' &&
 340                qemu_rbd_set_conf(cluster, conf, true, &local_err) < 0) {
 341         rados_shutdown(cluster);
 342         error_propagate(errp, local_err);
 343         return -EIO;
 344     }
 345
 346     if (conf[0] != '\0' &&
 347         qemu_rbd_set_conf(cluster, conf, false, &local_err) < 0) {
 348         rados_shutdown(cluster);
 349         error_propagate(errp, local_err);
 350         return -EIO;
 351     }
 352
 353     if (rados_connect(cluster) < 0) {
 354         error_setg(errp, "error connecting");
 355         rados_shutdown(cluster);
 356         return -EIO;
 357     }
 358
 359     if (rados_ioctx_create(cluster, pool, &io_ctx) < 0) {
 360         error_setg(errp, "error opening pool %s", pool);
 361         rados_shutdown(cluster);
 362         return -EIO;
 363     }
 364
 365     ret = rbd_create(io_ctx, name, bytes, &obj_order);
 366     rados_ioctx_destroy(io_ctx);
 367     rados_shutdown(cluster);
 368
 369     return ret;
 370 }
 371
 372 /*
 373  * This aio completion is being called from rbd_finish_bh() and runs in qemu
 374  * BH context.
 375  */
 376 static void qemu_rbd_complete_aio(RADOSCB *rcb)
 377 {
 378     RBDAIOCB *acb = rcb->acb;
 379     int64_t r;
 380
 381     r = rcb->ret;
 382
 383     if (acb->cmd != RBD_AIO_READ) {
 384         if (r < 0) {
 385             acb->ret = r;
 386             acb->error = 1;
 387         } else if (!acb->error) {
 388             acb->ret = rcb->size;
 389         }
 390     } else {
 391         if (r < 0) {
 392             memset(rcb->buf, 0, rcb->size);
 393             acb->ret = r;
 394             acb->error = 1;
 395         } else if (r < rcb->size) {
 396             memset(rcb->buf + r, 0, rcb->size - r);
 397             if (!acb->error) {
 398                 acb->ret = rcb->size;
 399             }
 400         } else if (!acb->error) {
 401             acb->ret = r;
 402         }
 403     }
 404
 405     g_free(rcb);
 406
 407     if (acb->cmd == RBD_AIO_READ) {
 408         qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
 409     }
 410     qemu_vfree(acb->bounce);
 411     acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
 412
 413     qemu_aio_unref(acb);
 414 }
 415
 416 /* TODO Convert to fine grained options */
 417 static QemuOptsList runtime_opts = {
 418     .name = "rbd",
 419     .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
 420     .desc = {
 421         {
 422             .name = "filename",
 423             .type = QEMU_OPT_STRING,
 424             .help = "Specification of the rbd image",
 425         },
 426         { /* end of list */ }
 427     },
 428 };
 429
 430 static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
 431                          Error **errp)
 432 {
 433     BDRVRBDState *s = bs->opaque;
 434     char pool[RBD_MAX_POOL_NAME_SIZE];
 435     char snap_buf[RBD_MAX_SNAP_NAME_SIZE];
 436     char conf[RBD_MAX_CONF_SIZE];
 437     char clientname_buf[RBD_MAX_CONF_SIZE];
 438     char *clientname;
 439     QemuOpts *opts;
 440     Error *local_err = NULL;
 441     const char *filename;
 442     int r;
 443
 444     opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
 445     qemu_opts_absorb_qdict(opts, options, &local_err);
 446     if (local_err) {
 447         error_propagate(errp, local_err);
 448         qemu_opts_del(opts);
 449         return -EINVAL;
 450     }
 451
 452     filename = qemu_opt_get(opts, "filename");
 453
 454     if (qemu_rbd_parsename(filename, pool, sizeof(pool),
 455                            snap_buf, sizeof(snap_buf),
 456                            s->name, sizeof(s->name),
 457                            conf, sizeof(conf), errp) < 0) {
 458         r = -EINVAL;
 459         goto failed_opts;
 460     }
 461
 462     clientname = qemu_rbd_parse_clientname(conf, clientname_buf);
 463     r = rados_create(&s->cluster, clientname);
 464     if (r < 0) {
 465         error_setg(errp, "error initializing");
 466         goto failed_opts;
 467     }
 468
 469     s->snap = NULL;
 470     if (snap_buf[0] != '\0') {
 471         s->snap = g_strdup(snap_buf);
 472     }
 473
 474     if (strstr(conf, "conf=") == NULL) {
 475         /* try default location, but ignore failure */
 476         rados_conf_read_file(s->cluster, NULL);
 477     } else if (conf[0] != '\0') {
 478         r = qemu_rbd_set_conf(s->cluster, conf, true, errp);
 479         if (r < 0) {
 480             goto failed_shutdown;
 481         }
 482     }
 483
 484     if (conf[0] != '\0') {
 485         r = qemu_rbd_set_conf(s->cluster, conf, false, errp);
 486         if (r < 0) {
 487             goto failed_shutdown;
 488         }
 489     }
 490
 491     /*
 492      * Fallback to more conservative semantics if setting cache
 493      * options fails. Ignore errors from setting rbd_cache because the
 494      * only possible error is that the option does not exist, and
 495      * librbd defaults to no caching. If write through caching cannot
 496      * be set up, fall back to no caching.
 497      */
 498     if (flags & BDRV_O_NOCACHE) {
 499         rados_conf_set(s->cluster, "rbd_cache", "false");
 500     } else {
 501         rados_conf_set(s->cluster, "rbd_cache", "true");
 502     }
 503
 504     r = rados_connect(s->cluster);
 505     if (r < 0) {
 506         error_setg(errp, "error connecting");
 507         goto failed_shutdown;
 508     }
 509
 510     r = rados_ioctx_create(s->cluster, pool, &s->io_ctx);
 511     if (r < 0) {
 512         error_setg(errp, "error opening pool %s", pool);
 513         goto failed_shutdown;
 514     }
 515
 516     r = rbd_open(s->io_ctx, s->name, &s->image, s->snap);
 517     if (r < 0) {
 518         error_setg(errp, "error reading header from %s", s->name);
 519         goto failed_open;
 520     }
 521
 522     bs->read_only = (s->snap != NULL);
 523
 524     qemu_opts_del(opts);
 525     return 0;
 526
 527 failed_open:
 528     rados_ioctx_destroy(s->io_ctx);
 529 failed_shutdown:
 530     rados_shutdown(s->cluster);
 531     g_free(s->snap);
 532 failed_opts:
 533     qemu_opts_del(opts);
 534     return r;
 535 }
 536
 537 static void qemu_rbd_close(BlockDriverState *bs)
 538 {
 539     BDRVRBDState *s = bs->opaque;
 540
 541     rbd_close(s->image);
 542     rados_ioctx_destroy(s->io_ctx);
 543     g_free(s->snap);
 544     rados_shutdown(s->cluster);
 545 }
 546
 547 static const AIOCBInfo rbd_aiocb_info = {
 548     .aiocb_size = sizeof(RBDAIOCB),
 549 };
 550
 551 static void rbd_finish_bh(void *opaque)
 552 {
 553     RADOSCB *rcb = opaque;
 554     qemu_bh_delete(rcb->acb->bh);
 555     qemu_rbd_complete_aio(rcb);
 556 }
 557
 558 /*
 559  * This is the callback function for rbd_aio_read and _write
 560  *
 561  * Note: this function is being called from a non qemu thread so
 562  * we need to be careful about what we do here. Generally we only
 563  * schedule a BH, and do the rest of the io completion handling
 564  * from rbd_finish_bh() which runs in a qemu context.
 565  */
 566 static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
 567 {
 568     RBDAIOCB *acb = rcb->acb;
 569
 570     rcb->ret = rbd_aio_get_return_value(c);
 571     rbd_aio_release(c);
 572
 573     acb->bh = aio_bh_new(bdrv_get_aio_context(acb->common.bs),
 574                          rbd_finish_bh, rcb);
 575     qemu_bh_schedule(acb->bh);
 576 }
 577
 578 static int rbd_aio_discard_wrapper(rbd_image_t image,
 579                                    uint64_t off,
 580                                    uint64_t len,
 581                                    rbd_completion_t comp)
 582 {
 583 #ifdef LIBRBD_SUPPORTS_DISCARD
 584     return rbd_aio_discard(image, off, len, comp);
 585 #else
 586     return -ENOTSUP;
 587 #endif
 588 }
 589
 590 static int rbd_aio_flush_wrapper(rbd_image_t image,
 591                                  rbd_completion_t comp)
 592 {
 593 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
 594     return rbd_aio_flush(image, comp);
 595 #else
 596     return -ENOTSUP;
 597 #endif
 598 }
 599
 600 static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
 601                                  int64_t sector_num,
 602                                  QEMUIOVector *qiov,
 603                                  int nb_sectors,
 604                                  BlockCompletionFunc *cb,
 605                                  void *opaque,
 606                                  RBDAIOCmd cmd)
 607 {
 608     RBDAIOCB *acb;
 609     RADOSCB *rcb = NULL;
 610     rbd_completion_t c;
 611     int64_t off, size;
 612     char *buf;
 613     int r;
 614
 615     BDRVRBDState *s = bs->opaque;
 616
 617     acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque);
 618     acb->cmd = cmd;
 619     acb->qiov = qiov;
 620     if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
 621         acb->bounce = NULL;
 622     } else {
 623         acb->bounce = qemu_try_blockalign(bs, qiov->size);
 624         if (acb->bounce == NULL) {
 625             goto failed;
 626         }
 627     }
 628     acb->ret = 0;
 629     acb->error = 0;
 630     acb->s = s;
 631     acb->bh = NULL;
 632
 633     if (cmd == RBD_AIO_WRITE) {
 634         qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
 635     }
 636
 637     buf = acb->bounce;
 638
 639     off = sector_num * BDRV_SECTOR_SIZE;
 640     size = nb_sectors * BDRV_SECTOR_SIZE;
 641
 642     rcb = g_new(RADOSCB, 1);
 643     rcb->acb = acb;
 644     rcb->buf = buf;
 645     rcb->s = acb->s;
 646     rcb->size = size;
 647     r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
 648     if (r < 0) {
 649         goto failed;
 650     }
 651
 652     switch (cmd) {
 653     case RBD_AIO_WRITE:
 654         r = rbd_aio_write(s->image, off, size, buf, c);
 655         break;
 656     case RBD_AIO_READ:
 657         r = rbd_aio_read(s->image, off, size, buf, c);
 658         break;
 659     case RBD_AIO_DISCARD:
 660         r = rbd_aio_discard_wrapper(s->image, off, size, c);
 661         break;
 662     case RBD_AIO_FLUSH:
 663         r = rbd_aio_flush_wrapper(s->image, c);
 664         break;
 665     default:
 666         r = -EINVAL;
 667     }
 668
 669     if (r < 0) {
 670         goto failed_completion;
 671     }
 672
 673     return &acb->common;
 674
 675 failed_completion:
 676     rbd_aio_release(c);
 677 failed:
 678     g_free(rcb);
 679     qemu_vfree(acb->bounce);
 680     qemu_aio_unref(acb);
 681     return NULL;
 682 }
 683
 684 static BlockAIOCB *qemu_rbd_aio_readv(BlockDriverState *bs,
 685                                       int64_t sector_num,
 686                                       QEMUIOVector *qiov,
 687                                       int nb_sectors,
 688                                       BlockCompletionFunc *cb,
 689                                       void *opaque)
 690 {
 691     return rbd_start_aio(bs, sector_num, qiov, nb_sectors, cb, opaque,
 692                          RBD_AIO_READ);
 693 }
 694
 695 static BlockAIOCB *qemu_rbd_aio_writev(BlockDriverState *bs,
 696                                        int64_t sector_num,
 697                                        QEMUIOVector *qiov,
 698                                        int nb_sectors,
 699                                        BlockCompletionFunc *cb,
 700                                        void *opaque)
 701 {
 702     return rbd_start_aio(bs, sector_num, qiov, nb_sectors, cb, opaque,
 703                          RBD_AIO_WRITE);
 704 }
 705
 706 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
 707 static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs,
 708                                       BlockCompletionFunc *cb,
 709                                       void *opaque)
 710 {
 711     return rbd_start_aio(bs, 0, NULL, 0, cb, opaque, RBD_AIO_FLUSH);
 712 }
 713
 714 #else
 715
 716 static int qemu_rbd_co_flush(BlockDriverState *bs)
 717 {
 718 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 1)
 719     /* rbd_flush added in 0.1.1 */
 720     BDRVRBDState *s = bs->opaque;
 721     return rbd_flush(s->image);
 722 #else
 723     return 0;
 724 #endif
 725 }
 726 #endif
 727
 728 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 729 {
 730     BDRVRBDState *s = bs->opaque;
 731     rbd_image_info_t info;
 732     int r;
 733
 734     r = rbd_stat(s->image, &info, sizeof(info));
 735     if (r < 0) {
 736         return r;
 737     }
 738
 739     bdi->cluster_size = info.obj_size;
 740     return 0;
 741 }
 742
 743 static int64_t qemu_rbd_getlength(BlockDriverState *bs)
 744 {
 745     BDRVRBDState *s = bs->opaque;
 746     rbd_image_info_t info;
 747     int r;
 748
 749     r = rbd_stat(s->image, &info, sizeof(info));
 750     if (r < 0) {
 751         return r;
 752     }
 753
 754     return info.size;
 755 }
 756
 757 static int qemu_rbd_truncate(BlockDriverState *bs, int64_t offset)
 758 {
 759     BDRVRBDState *s = bs->opaque;
 760     int r;
 761
 762     r = rbd_resize(s->image, offset);
 763     if (r < 0) {
 764         return r;
 765     }
 766
 767     return 0;
 768 }
 769
 770 static int qemu_rbd_snap_create(BlockDriverState *bs,
 771                                 QEMUSnapshotInfo *sn_info)
 772 {
 773     BDRVRBDState *s = bs->opaque;
 774     int r;
 775
 776     if (sn_info->name[0] == '\0') {
 777         return -EINVAL; /* we need a name for rbd snapshots */
 778     }
 779
 780     /*
 781      * rbd snapshots are using the name as the user controlled unique identifier
 782      * we can't use the rbd snapid for that purpose, as it can't be set
 783      */
 784     if (sn_info->id_str[0] != '\0' &&
 785         strcmp(sn_info->id_str, sn_info->name) != 0) {
 786         return -EINVAL;
 787     }
 788
 789     if (strlen(sn_info->name) >= sizeof(sn_info->id_str)) {
 790         return -ERANGE;
 791     }
 792
 793     r = rbd_snap_create(s->image, sn_info->name);
 794     if (r < 0) {
 795         error_report("failed to create snap: %s", strerror(-r));
 796         return r;
 797     }
 798
 799     return 0;
 800 }
 801
 802 static int qemu_rbd_snap_remove(BlockDriverState *bs,
 803                                 const char *snapshot_id,
 804                                 const char *snapshot_name,
 805                                 Error **errp)
 806 {
 807     BDRVRBDState *s = bs->opaque;
 808     int r;
 809
 810     if (!snapshot_name) {
 811         error_setg(errp, "rbd need a valid snapshot name");
 812         return -EINVAL;
 813     }
 814
 815     /* If snapshot_id is specified, it must be equal to name, see
 816        qemu_rbd_snap_list() */
 817     if (snapshot_id && strcmp(snapshot_id, snapshot_name)) {
 818         error_setg(errp,
 819                    "rbd do not support snapshot id, it should be NULL or "
 820                    "equal to snapshot name");
 821         return -EINVAL;
 822     }
 823
 824     r = rbd_snap_remove(s->image, snapshot_name);
 825     if (r < 0) {
 826         error_setg_errno(errp, -r, "Failed to remove the snapshot");
 827     }
 828     return r;
 829 }
 830
 831 static int qemu_rbd_snap_rollback(BlockDriverState *bs,
 832                                   const char *snapshot_name)
 833 {
 834     BDRVRBDState *s = bs->opaque;
 835     int r;
 836
 837     r = rbd_snap_rollback(s->image, snapshot_name);
 838     return r;
 839 }
 840
 841 static int qemu_rbd_snap_list(BlockDriverState *bs,
 842                               QEMUSnapshotInfo **psn_tab)
 843 {
 844     BDRVRBDState *s = bs->opaque;
 845     QEMUSnapshotInfo *sn_info, *sn_tab = NULL;
 846     int i, snap_count;
 847     rbd_snap_info_t *snaps;
 848     int max_snaps = RBD_MAX_SNAPS;
 849
 850     do {
 851         snaps = g_new(rbd_snap_info_t, max_snaps);
 852         snap_count = rbd_snap_list(s->image, snaps, &max_snaps);
 853         if (snap_count <= 0) {
 854             g_free(snaps);
 855         }
 856     } while (snap_count == -ERANGE);
 857
 858     if (snap_count <= 0) {
 859         goto done;
 860     }
 861
 862     sn_tab = g_new0(QEMUSnapshotInfo, snap_count);
 863
 864     for (i = 0; i < snap_count; i++) {
 865         const char *snap_name = snaps[i].name;
 866
 867         sn_info = sn_tab + i;
 868         pstrcpy(sn_info->id_str, sizeof(sn_info->id_str), snap_name);
 869         pstrcpy(sn_info->name, sizeof(sn_info->name), snap_name);
 870
 871         sn_info->vm_state_size = snaps[i].size;
 872         sn_info->date_sec = 0;
 873         sn_info->date_nsec = 0;
 874         sn_info->vm_clock_nsec = 0;
 875     }
 876     rbd_snap_list_end(snaps);
 877     g_free(snaps);
 878
 879  done:
 880     *psn_tab = sn_tab;
 881     return snap_count;
 882 }
 883
 884 #ifdef LIBRBD_SUPPORTS_DISCARD
 885 static BlockAIOCB* qemu_rbd_aio_discard(BlockDriverState *bs,
 886                                         int64_t sector_num,
 887                                         int nb_sectors,
 888                                         BlockCompletionFunc *cb,
 889                                         void *opaque)
 890 {
 891     return rbd_start_aio(bs, sector_num, NULL, nb_sectors, cb, opaque,
 892                          RBD_AIO_DISCARD);
 893 }
 894 #endif
 895
 896 #ifdef LIBRBD_SUPPORTS_INVALIDATE
 897 static void qemu_rbd_invalidate_cache(BlockDriverState *bs,
 898                                       Error **errp)
 899 {
 900     BDRVRBDState *s = bs->opaque;
 901     int r = rbd_invalidate_cache(s->image);
 902     if (r < 0) {
 903         error_setg_errno(errp, -r, "Failed to invalidate the cache");
 904     }
 905 }
 906 #endif
 907
 908 static QemuOptsList qemu_rbd_create_opts = {
 909     .name = "rbd-create-opts",
 910     .head = QTAILQ_HEAD_INITIALIZER(qemu_rbd_create_opts.head),
 911     .desc = {
 912         {
 913             .name = BLOCK_OPT_SIZE,
 914             .type = QEMU_OPT_SIZE,
 915             .help = "Virtual disk size"
 916         },
 917         {
 918             .name = BLOCK_OPT_CLUSTER_SIZE,
 919             .type = QEMU_OPT_SIZE,
 920             .help = "RBD object size"
 921         },
 922         { /* end of list */ }
 923     }
 924 };
 925
 926 static BlockDriver bdrv_rbd = {
 927     .format_name        = "rbd",
 928     .instance_size      = sizeof(BDRVRBDState),
 929     .bdrv_needs_filename = true,
 930     .bdrv_file_open     = qemu_rbd_open,
 931     .bdrv_close         = qemu_rbd_close,
 932     .bdrv_create        = qemu_rbd_create,
 933     .bdrv_has_zero_init = bdrv_has_zero_init_1,
 934     .bdrv_get_info      = qemu_rbd_getinfo,
 935     .create_opts        = &qemu_rbd_create_opts,
 936     .bdrv_getlength     = qemu_rbd_getlength,
 937     .bdrv_truncate      = qemu_rbd_truncate,
 938     .protocol_name      = "rbd",
 939
 940     .bdrv_aio_readv         = qemu_rbd_aio_readv,
 941     .bdrv_aio_writev        = qemu_rbd_aio_writev,
 942
 943 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
 944     .bdrv_aio_flush         = qemu_rbd_aio_flush,
 945 #else
 946     .bdrv_co_flush_to_disk  = qemu_rbd_co_flush,
 947 #endif
 948
 949 #ifdef LIBRBD_SUPPORTS_DISCARD
 950     .bdrv_aio_discard       = qemu_rbd_aio_discard,
 951 #endif
 952
 953     .bdrv_snapshot_create   = qemu_rbd_snap_create,
 954     .bdrv_snapshot_delete   = qemu_rbd_snap_remove,
 955     .bdrv_snapshot_list     = qemu_rbd_snap_list,
 956     .bdrv_snapshot_goto     = qemu_rbd_snap_rollback,
 957 #ifdef LIBRBD_SUPPORTS_INVALIDATE
 958     .bdrv_invalidate_cache  = qemu_rbd_invalidate_cache,
 959 #endif
 960 };
 961
 962 static void bdrv_rbd_init(void)
 963 {
 964     bdrv_register(&bdrv_rbd);
 965 }
 966
 967 block_init(bdrv_rbd_init);