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 <rbd/librbd.h> 17 #include "qapi/error.h" 18 #include "qemu/error-report.h" 19 #include "qemu/module.h" 20 #include "qemu/option.h" 21 #include "block/block_int.h" 22 #include "block/qdict.h" 23 #include "crypto/secret.h" 24 #include "qemu/cutils.h" 25 #include "sysemu/replay.h" 26 #include "qapi/qmp/qstring.h" 27 #include "qapi/qmp/qdict.h" 28 #include "qapi/qmp/qjson.h" 29 #include "qapi/qmp/qlist.h" 30 #include "qapi/qobject-input-visitor.h" 31 #include "qapi/qapi-visit-block-core.h" 32 33 /* 34 * When specifying the image filename use: 35 * 36 * rbd:poolname/devicename[@snapshotname][:option1=value1[:option2=value2...]] 37 * 38 * poolname must be the name of an existing rados pool. 39 * 40 * devicename is the name of the rbd image. 41 * 42 * Each option given is used to configure rados, and may be any valid 43 * Ceph option, "id", or "conf". 44 * 45 * The "id" option indicates what user we should authenticate as to 46 * the Ceph cluster. If it is excluded we will use the Ceph default 47 * (normally 'admin'). 48 * 49 * The "conf" option specifies a Ceph configuration file to read. If 50 * it is not specified, we will read from the default Ceph locations 51 * (e.g., /etc/ceph/ceph.conf). To avoid reading _any_ configuration 52 * file, specify conf=/dev/null. 53 * 54 * Configuration values containing :, @, or = can be escaped with a 55 * leading "\". 56 */ 57 58 /* rbd_aio_discard added in 0.1.2 */ 59 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 2) 60 #define LIBRBD_SUPPORTS_DISCARD 61 #else 62 #undef LIBRBD_SUPPORTS_DISCARD 63 #endif 64 65 #define OBJ_MAX_SIZE (1UL << OBJ_DEFAULT_OBJ_ORDER) 66 67 #define RBD_MAX_SNAPS 100 68 69 /* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */ 70 #ifdef LIBRBD_SUPPORTS_IOVEC 71 #define LIBRBD_USE_IOVEC 1 72 #else 73 #define LIBRBD_USE_IOVEC 0 74 #endif 75 76 typedef enum { 77 RBD_AIO_READ, 78 RBD_AIO_WRITE, 79 RBD_AIO_DISCARD, 80 RBD_AIO_FLUSH 81 } RBDAIOCmd; 82 83 typedef struct RBDAIOCB { 84 BlockAIOCB common; 85 int64_t ret; 86 QEMUIOVector *qiov; 87 char *bounce; 88 RBDAIOCmd cmd; 89 int error; 90 struct BDRVRBDState *s; 91 } RBDAIOCB; 92 93 typedef struct RADOSCB { 94 RBDAIOCB *acb; 95 struct BDRVRBDState *s; 96 int64_t size; 97 char *buf; 98 int64_t ret; 99 } RADOSCB; 100 101 typedef struct BDRVRBDState { 102 rados_t cluster; 103 rados_ioctx_t io_ctx; 104 rbd_image_t image; 105 char *image_name; 106 char *snap; 107 char *namespace; 108 uint64_t image_size; 109 } BDRVRBDState; 110 111 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, 112 BlockdevOptionsRbd *opts, bool cache, 113 const char *keypairs, const char *secretid, 114 Error **errp); 115 116 static char *qemu_rbd_strchr(char *src, char delim) 117 { 118 char *p; 119 120 for (p = src; *p; ++p) { 121 if (*p == delim) { 122 return p; 123 } 124 if (*p == '\\' && p[1] != '\0') { 125 ++p; 126 } 127 } 128 129 return NULL; 130 } 131 132 133 static char *qemu_rbd_next_tok(char *src, char delim, char **p) 134 { 135 char *end; 136 137 *p = NULL; 138 139 end = qemu_rbd_strchr(src, delim); 140 if (end) { 141 *p = end + 1; 142 *end = '\0'; 143 } 144 return src; 145 } 146 147 static void qemu_rbd_unescape(char *src) 148 { 149 char *p; 150 151 for (p = src; *src; ++src, ++p) { 152 if (*src == '\\' && src[1] != '\0') { 153 src++; 154 } 155 *p = *src; 156 } 157 *p = '\0'; 158 } 159 160 static void qemu_rbd_parse_filename(const char *filename, QDict *options, 161 Error **errp) 162 { 163 const char *start; 164 char *p, *buf; 165 QList *keypairs = NULL; 166 char *found_str, *image_name; 167 168 if (!strstart(filename, "rbd:", &start)) { 169 error_setg(errp, "File name must start with 'rbd:'"); 170 return; 171 } 172 173 buf = g_strdup(start); 174 p = buf; 175 176 found_str = qemu_rbd_next_tok(p, '/', &p); 177 if (!p) { 178 error_setg(errp, "Pool name is required"); 179 goto done; 180 } 181 qemu_rbd_unescape(found_str); 182 qdict_put_str(options, "pool", found_str); 183 184 if (qemu_rbd_strchr(p, '@')) { 185 image_name = qemu_rbd_next_tok(p, '@', &p); 186 187 found_str = qemu_rbd_next_tok(p, ':', &p); 188 qemu_rbd_unescape(found_str); 189 qdict_put_str(options, "snapshot", found_str); 190 } else { 191 image_name = qemu_rbd_next_tok(p, ':', &p); 192 } 193 /* Check for namespace in the image_name */ 194 if (qemu_rbd_strchr(image_name, '/')) { 195 found_str = qemu_rbd_next_tok(image_name, '/', &image_name); 196 qemu_rbd_unescape(found_str); 197 qdict_put_str(options, "namespace", found_str); 198 } else { 199 qdict_put_str(options, "namespace", ""); 200 } 201 qemu_rbd_unescape(image_name); 202 qdict_put_str(options, "image", image_name); 203 if (!p) { 204 goto done; 205 } 206 207 /* The following are essentially all key/value pairs, and we treat 208 * 'id' and 'conf' a bit special. Key/value pairs may be in any order. */ 209 while (p) { 210 char *name, *value; 211 name = qemu_rbd_next_tok(p, '=', &p); 212 if (!p) { 213 error_setg(errp, "conf option %s has no value", name); 214 break; 215 } 216 217 qemu_rbd_unescape(name); 218 219 value = qemu_rbd_next_tok(p, ':', &p); 220 qemu_rbd_unescape(value); 221 222 if (!strcmp(name, "conf")) { 223 qdict_put_str(options, "conf", value); 224 } else if (!strcmp(name, "id")) { 225 qdict_put_str(options, "user", value); 226 } else { 227 /* 228 * We pass these internally to qemu_rbd_set_keypairs(), so 229 * we can get away with the simpler list of [ "key1", 230 * "value1", "key2", "value2" ] rather than a raw dict 231 * { "key1": "value1", "key2": "value2" } where we can't 232 * guarantee order, or even a more correct but complex 233 * [ { "key1": "value1" }, { "key2": "value2" } ] 234 */ 235 if (!keypairs) { 236 keypairs = qlist_new(); 237 } 238 qlist_append_str(keypairs, name); 239 qlist_append_str(keypairs, value); 240 } 241 } 242 243 if (keypairs) { 244 qdict_put(options, "=keyvalue-pairs", 245 qstring_from_gstring(qobject_to_json(QOBJECT(keypairs)))); 246 } 247 248 done: 249 g_free(buf); 250 qobject_unref(keypairs); 251 return; 252 } 253 254 255 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp) 256 { 257 /* XXX Does RBD support AIO on less than 512-byte alignment? */ 258 bs->bl.request_alignment = 512; 259 } 260 261 262 static int qemu_rbd_set_auth(rados_t cluster, BlockdevOptionsRbd *opts, 263 Error **errp) 264 { 265 char *key, *acr; 266 int r; 267 GString *accu; 268 RbdAuthModeList *auth; 269 270 if (opts->key_secret) { 271 key = qcrypto_secret_lookup_as_base64(opts->key_secret, errp); 272 if (!key) { 273 return -EIO; 274 } 275 r = rados_conf_set(cluster, "key", key); 276 g_free(key); 277 if (r < 0) { 278 error_setg_errno(errp, -r, "Could not set 'key'"); 279 return r; 280 } 281 } 282 283 if (opts->has_auth_client_required) { 284 accu = g_string_new(""); 285 for (auth = opts->auth_client_required; auth; auth = auth->next) { 286 if (accu->str[0]) { 287 g_string_append_c(accu, ';'); 288 } 289 g_string_append(accu, RbdAuthMode_str(auth->value)); 290 } 291 acr = g_string_free(accu, FALSE); 292 r = rados_conf_set(cluster, "auth_client_required", acr); 293 g_free(acr); 294 if (r < 0) { 295 error_setg_errno(errp, -r, 296 "Could not set 'auth_client_required'"); 297 return r; 298 } 299 } 300 301 return 0; 302 } 303 304 static int qemu_rbd_set_keypairs(rados_t cluster, const char *keypairs_json, 305 Error **errp) 306 { 307 QList *keypairs; 308 QString *name; 309 QString *value; 310 const char *key; 311 size_t remaining; 312 int ret = 0; 313 314 if (!keypairs_json) { 315 return ret; 316 } 317 keypairs = qobject_to(QList, 318 qobject_from_json(keypairs_json, &error_abort)); 319 remaining = qlist_size(keypairs) / 2; 320 assert(remaining); 321 322 while (remaining--) { 323 name = qobject_to(QString, qlist_pop(keypairs)); 324 value = qobject_to(QString, qlist_pop(keypairs)); 325 assert(name && value); 326 key = qstring_get_str(name); 327 328 ret = rados_conf_set(cluster, key, qstring_get_str(value)); 329 qobject_unref(value); 330 if (ret < 0) { 331 error_setg_errno(errp, -ret, "invalid conf option %s", key); 332 qobject_unref(name); 333 ret = -EINVAL; 334 break; 335 } 336 qobject_unref(name); 337 } 338 339 qobject_unref(keypairs); 340 return ret; 341 } 342 343 static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs) 344 { 345 if (LIBRBD_USE_IOVEC) { 346 RBDAIOCB *acb = rcb->acb; 347 iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0, 348 acb->qiov->size - offs); 349 } else { 350 memset(rcb->buf + offs, 0, rcb->size - offs); 351 } 352 } 353 354 /* FIXME Deprecate and remove keypairs or make it available in QMP. */ 355 static int qemu_rbd_do_create(BlockdevCreateOptions *options, 356 const char *keypairs, const char *password_secret, 357 Error **errp) 358 { 359 BlockdevCreateOptionsRbd *opts = &options->u.rbd; 360 rados_t cluster; 361 rados_ioctx_t io_ctx; 362 int obj_order = 0; 363 int ret; 364 365 assert(options->driver == BLOCKDEV_DRIVER_RBD); 366 if (opts->location->has_snapshot) { 367 error_setg(errp, "Can't use snapshot name for image creation"); 368 return -EINVAL; 369 } 370 371 if (opts->has_cluster_size) { 372 int64_t objsize = opts->cluster_size; 373 if ((objsize - 1) & objsize) { /* not a power of 2? */ 374 error_setg(errp, "obj size needs to be power of 2"); 375 return -EINVAL; 376 } 377 if (objsize < 4096) { 378 error_setg(errp, "obj size too small"); 379 return -EINVAL; 380 } 381 obj_order = ctz32(objsize); 382 } 383 384 ret = qemu_rbd_connect(&cluster, &io_ctx, opts->location, false, keypairs, 385 password_secret, errp); 386 if (ret < 0) { 387 return ret; 388 } 389 390 ret = rbd_create(io_ctx, opts->location->image, opts->size, &obj_order); 391 if (ret < 0) { 392 error_setg_errno(errp, -ret, "error rbd create"); 393 goto out; 394 } 395 396 ret = 0; 397 out: 398 rados_ioctx_destroy(io_ctx); 399 rados_shutdown(cluster); 400 return ret; 401 } 402 403 static int qemu_rbd_co_create(BlockdevCreateOptions *options, Error **errp) 404 { 405 return qemu_rbd_do_create(options, NULL, NULL, errp); 406 } 407 408 static int coroutine_fn qemu_rbd_co_create_opts(BlockDriver *drv, 409 const char *filename, 410 QemuOpts *opts, 411 Error **errp) 412 { 413 BlockdevCreateOptions *create_options; 414 BlockdevCreateOptionsRbd *rbd_opts; 415 BlockdevOptionsRbd *loc; 416 Error *local_err = NULL; 417 const char *keypairs, *password_secret; 418 QDict *options = NULL; 419 int ret = 0; 420 421 create_options = g_new0(BlockdevCreateOptions, 1); 422 create_options->driver = BLOCKDEV_DRIVER_RBD; 423 rbd_opts = &create_options->u.rbd; 424 425 rbd_opts->location = g_new0(BlockdevOptionsRbd, 1); 426 427 password_secret = qemu_opt_get(opts, "password-secret"); 428 429 /* Read out options */ 430 rbd_opts->size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), 431 BDRV_SECTOR_SIZE); 432 rbd_opts->cluster_size = qemu_opt_get_size_del(opts, 433 BLOCK_OPT_CLUSTER_SIZE, 0); 434 rbd_opts->has_cluster_size = (rbd_opts->cluster_size != 0); 435 436 options = qdict_new(); 437 qemu_rbd_parse_filename(filename, options, &local_err); 438 if (local_err) { 439 ret = -EINVAL; 440 error_propagate(errp, local_err); 441 goto exit; 442 } 443 444 /* 445 * Caution: while qdict_get_try_str() is fine, getting non-string 446 * types would require more care. When @options come from -blockdev 447 * or blockdev_add, its members are typed according to the QAPI 448 * schema, but when they come from -drive, they're all QString. 449 */ 450 loc = rbd_opts->location; 451 loc->pool = g_strdup(qdict_get_try_str(options, "pool")); 452 loc->conf = g_strdup(qdict_get_try_str(options, "conf")); 453 loc->has_conf = !!loc->conf; 454 loc->user = g_strdup(qdict_get_try_str(options, "user")); 455 loc->has_user = !!loc->user; 456 loc->q_namespace = g_strdup(qdict_get_try_str(options, "namespace")); 457 loc->has_q_namespace = !!loc->q_namespace; 458 loc->image = g_strdup(qdict_get_try_str(options, "image")); 459 keypairs = qdict_get_try_str(options, "=keyvalue-pairs"); 460 461 ret = qemu_rbd_do_create(create_options, keypairs, password_secret, errp); 462 if (ret < 0) { 463 goto exit; 464 } 465 466 exit: 467 qobject_unref(options); 468 qapi_free_BlockdevCreateOptions(create_options); 469 return ret; 470 } 471 472 /* 473 * This aio completion is being called from rbd_finish_bh() and runs in qemu 474 * BH context. 475 */ 476 static void qemu_rbd_complete_aio(RADOSCB *rcb) 477 { 478 RBDAIOCB *acb = rcb->acb; 479 int64_t r; 480 481 r = rcb->ret; 482 483 if (acb->cmd != RBD_AIO_READ) { 484 if (r < 0) { 485 acb->ret = r; 486 acb->error = 1; 487 } else if (!acb->error) { 488 acb->ret = rcb->size; 489 } 490 } else { 491 if (r < 0) { 492 qemu_rbd_memset(rcb, 0); 493 acb->ret = r; 494 acb->error = 1; 495 } else if (r < rcb->size) { 496 qemu_rbd_memset(rcb, r); 497 if (!acb->error) { 498 acb->ret = rcb->size; 499 } 500 } else if (!acb->error) { 501 acb->ret = r; 502 } 503 } 504 505 g_free(rcb); 506 507 if (!LIBRBD_USE_IOVEC) { 508 if (acb->cmd == RBD_AIO_READ) { 509 qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); 510 } 511 qemu_vfree(acb->bounce); 512 } 513 514 acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret)); 515 516 qemu_aio_unref(acb); 517 } 518 519 static char *qemu_rbd_mon_host(BlockdevOptionsRbd *opts, Error **errp) 520 { 521 const char **vals; 522 const char *host, *port; 523 char *rados_str; 524 InetSocketAddressBaseList *p; 525 int i, cnt; 526 527 if (!opts->has_server) { 528 return NULL; 529 } 530 531 for (cnt = 0, p = opts->server; p; p = p->next) { 532 cnt++; 533 } 534 535 vals = g_new(const char *, cnt + 1); 536 537 for (i = 0, p = opts->server; p; p = p->next, i++) { 538 host = p->value->host; 539 port = p->value->port; 540 541 if (strchr(host, ':')) { 542 vals[i] = g_strdup_printf("[%s]:%s", host, port); 543 } else { 544 vals[i] = g_strdup_printf("%s:%s", host, port); 545 } 546 } 547 vals[i] = NULL; 548 549 rados_str = i ? g_strjoinv(";", (char **)vals) : NULL; 550 g_strfreev((char **)vals); 551 return rados_str; 552 } 553 554 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, 555 BlockdevOptionsRbd *opts, bool cache, 556 const char *keypairs, const char *secretid, 557 Error **errp) 558 { 559 char *mon_host = NULL; 560 Error *local_err = NULL; 561 int r; 562 563 if (secretid) { 564 if (opts->key_secret) { 565 error_setg(errp, 566 "Legacy 'password-secret' clashes with 'key-secret'"); 567 return -EINVAL; 568 } 569 opts->key_secret = g_strdup(secretid); 570 opts->has_key_secret = true; 571 } 572 573 mon_host = qemu_rbd_mon_host(opts, &local_err); 574 if (local_err) { 575 error_propagate(errp, local_err); 576 r = -EINVAL; 577 goto out; 578 } 579 580 r = rados_create(cluster, opts->user); 581 if (r < 0) { 582 error_setg_errno(errp, -r, "error initializing"); 583 goto out; 584 } 585 586 /* try default location when conf=NULL, but ignore failure */ 587 r = rados_conf_read_file(*cluster, opts->conf); 588 if (opts->has_conf && r < 0) { 589 error_setg_errno(errp, -r, "error reading conf file %s", opts->conf); 590 goto failed_shutdown; 591 } 592 593 r = qemu_rbd_set_keypairs(*cluster, keypairs, errp); 594 if (r < 0) { 595 goto failed_shutdown; 596 } 597 598 if (mon_host) { 599 r = rados_conf_set(*cluster, "mon_host", mon_host); 600 if (r < 0) { 601 goto failed_shutdown; 602 } 603 } 604 605 r = qemu_rbd_set_auth(*cluster, opts, errp); 606 if (r < 0) { 607 goto failed_shutdown; 608 } 609 610 /* 611 * Fallback to more conservative semantics if setting cache 612 * options fails. Ignore errors from setting rbd_cache because the 613 * only possible error is that the option does not exist, and 614 * librbd defaults to no caching. If write through caching cannot 615 * be set up, fall back to no caching. 616 */ 617 if (cache) { 618 rados_conf_set(*cluster, "rbd_cache", "true"); 619 } else { 620 rados_conf_set(*cluster, "rbd_cache", "false"); 621 } 622 623 r = rados_connect(*cluster); 624 if (r < 0) { 625 error_setg_errno(errp, -r, "error connecting"); 626 goto failed_shutdown; 627 } 628 629 r = rados_ioctx_create(*cluster, opts->pool, io_ctx); 630 if (r < 0) { 631 error_setg_errno(errp, -r, "error opening pool %s", opts->pool); 632 goto failed_shutdown; 633 } 634 /* 635 * Set the namespace after opening the io context on the pool, 636 * if nspace == NULL or if nspace == "", it is just as we did nothing 637 */ 638 rados_ioctx_set_namespace(*io_ctx, opts->q_namespace); 639 640 r = 0; 641 goto out; 642 643 failed_shutdown: 644 rados_shutdown(*cluster); 645 out: 646 g_free(mon_host); 647 return r; 648 } 649 650 static int qemu_rbd_convert_options(QDict *options, BlockdevOptionsRbd **opts, 651 Error **errp) 652 { 653 Visitor *v; 654 655 /* Convert the remaining options into a QAPI object */ 656 v = qobject_input_visitor_new_flat_confused(options, errp); 657 if (!v) { 658 return -EINVAL; 659 } 660 661 visit_type_BlockdevOptionsRbd(v, NULL, opts, errp); 662 visit_free(v); 663 if (!opts) { 664 return -EINVAL; 665 } 666 667 return 0; 668 } 669 670 static int qemu_rbd_attempt_legacy_options(QDict *options, 671 BlockdevOptionsRbd **opts, 672 char **keypairs) 673 { 674 char *filename; 675 int r; 676 677 filename = g_strdup(qdict_get_try_str(options, "filename")); 678 if (!filename) { 679 return -EINVAL; 680 } 681 qdict_del(options, "filename"); 682 683 qemu_rbd_parse_filename(filename, options, NULL); 684 685 /* keypairs freed by caller */ 686 *keypairs = g_strdup(qdict_get_try_str(options, "=keyvalue-pairs")); 687 if (*keypairs) { 688 qdict_del(options, "=keyvalue-pairs"); 689 } 690 691 r = qemu_rbd_convert_options(options, opts, NULL); 692 693 g_free(filename); 694 return r; 695 } 696 697 static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, 698 Error **errp) 699 { 700 BDRVRBDState *s = bs->opaque; 701 BlockdevOptionsRbd *opts = NULL; 702 const QDictEntry *e; 703 Error *local_err = NULL; 704 char *keypairs, *secretid; 705 int r; 706 707 keypairs = g_strdup(qdict_get_try_str(options, "=keyvalue-pairs")); 708 if (keypairs) { 709 qdict_del(options, "=keyvalue-pairs"); 710 } 711 712 secretid = g_strdup(qdict_get_try_str(options, "password-secret")); 713 if (secretid) { 714 qdict_del(options, "password-secret"); 715 } 716 717 r = qemu_rbd_convert_options(options, &opts, &local_err); 718 if (local_err) { 719 /* If keypairs are present, that means some options are present in 720 * the modern option format. Don't attempt to parse legacy option 721 * formats, as we won't support mixed usage. */ 722 if (keypairs) { 723 error_propagate(errp, local_err); 724 goto out; 725 } 726 727 /* If the initial attempt to convert and process the options failed, 728 * we may be attempting to open an image file that has the rbd options 729 * specified in the older format consisting of all key/value pairs 730 * encoded in the filename. Go ahead and attempt to parse the 731 * filename, and see if we can pull out the required options. */ 732 r = qemu_rbd_attempt_legacy_options(options, &opts, &keypairs); 733 if (r < 0) { 734 /* Propagate the original error, not the legacy parsing fallback 735 * error, as the latter was just a best-effort attempt. */ 736 error_propagate(errp, local_err); 737 goto out; 738 } 739 /* Take care whenever deciding to actually deprecate; once this ability 740 * is removed, we will not be able to open any images with legacy-styled 741 * backing image strings. */ 742 warn_report("RBD options encoded in the filename as keyvalue pairs " 743 "is deprecated"); 744 } 745 746 /* Remove the processed options from the QDict (the visitor processes 747 * _all_ options in the QDict) */ 748 while ((e = qdict_first(options))) { 749 qdict_del(options, e->key); 750 } 751 752 r = qemu_rbd_connect(&s->cluster, &s->io_ctx, opts, 753 !(flags & BDRV_O_NOCACHE), keypairs, secretid, errp); 754 if (r < 0) { 755 goto out; 756 } 757 758 s->snap = g_strdup(opts->snapshot); 759 s->image_name = g_strdup(opts->image); 760 761 /* rbd_open is always r/w */ 762 r = rbd_open(s->io_ctx, s->image_name, &s->image, s->snap); 763 if (r < 0) { 764 error_setg_errno(errp, -r, "error reading header from %s", 765 s->image_name); 766 goto failed_open; 767 } 768 769 r = rbd_get_size(s->image, &s->image_size); 770 if (r < 0) { 771 error_setg_errno(errp, -r, "error getting image size from %s", 772 s->image_name); 773 rbd_close(s->image); 774 goto failed_open; 775 } 776 777 /* If we are using an rbd snapshot, we must be r/o, otherwise 778 * leave as-is */ 779 if (s->snap != NULL) { 780 r = bdrv_apply_auto_read_only(bs, "rbd snapshots are read-only", errp); 781 if (r < 0) { 782 rbd_close(s->image); 783 goto failed_open; 784 } 785 } 786 787 /* When extending regular files, we get zeros from the OS */ 788 bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE; 789 790 r = 0; 791 goto out; 792 793 failed_open: 794 rados_ioctx_destroy(s->io_ctx); 795 g_free(s->snap); 796 g_free(s->image_name); 797 rados_shutdown(s->cluster); 798 out: 799 qapi_free_BlockdevOptionsRbd(opts); 800 g_free(keypairs); 801 g_free(secretid); 802 return r; 803 } 804 805 806 /* Since RBD is currently always opened R/W via the API, 807 * we just need to check if we are using a snapshot or not, in 808 * order to determine if we will allow it to be R/W */ 809 static int qemu_rbd_reopen_prepare(BDRVReopenState *state, 810 BlockReopenQueue *queue, Error **errp) 811 { 812 BDRVRBDState *s = state->bs->opaque; 813 int ret = 0; 814 815 if (s->snap && state->flags & BDRV_O_RDWR) { 816 error_setg(errp, 817 "Cannot change node '%s' to r/w when using RBD snapshot", 818 bdrv_get_device_or_node_name(state->bs)); 819 ret = -EINVAL; 820 } 821 822 return ret; 823 } 824 825 static void qemu_rbd_close(BlockDriverState *bs) 826 { 827 BDRVRBDState *s = bs->opaque; 828 829 rbd_close(s->image); 830 rados_ioctx_destroy(s->io_ctx); 831 g_free(s->snap); 832 g_free(s->image_name); 833 rados_shutdown(s->cluster); 834 } 835 836 /* Resize the RBD image and update the 'image_size' with the current size */ 837 static int qemu_rbd_resize(BlockDriverState *bs, uint64_t size) 838 { 839 BDRVRBDState *s = bs->opaque; 840 int r; 841 842 r = rbd_resize(s->image, size); 843 if (r < 0) { 844 return r; 845 } 846 847 s->image_size = size; 848 849 return 0; 850 } 851 852 static const AIOCBInfo rbd_aiocb_info = { 853 .aiocb_size = sizeof(RBDAIOCB), 854 }; 855 856 static void rbd_finish_bh(void *opaque) 857 { 858 RADOSCB *rcb = opaque; 859 qemu_rbd_complete_aio(rcb); 860 } 861 862 /* 863 * This is the callback function for rbd_aio_read and _write 864 * 865 * Note: this function is being called from a non qemu thread so 866 * we need to be careful about what we do here. Generally we only 867 * schedule a BH, and do the rest of the io completion handling 868 * from rbd_finish_bh() which runs in a qemu context. 869 */ 870 static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb) 871 { 872 RBDAIOCB *acb = rcb->acb; 873 874 rcb->ret = rbd_aio_get_return_value(c); 875 rbd_aio_release(c); 876 877 replay_bh_schedule_oneshot_event(bdrv_get_aio_context(acb->common.bs), 878 rbd_finish_bh, rcb); 879 } 880 881 static int rbd_aio_discard_wrapper(rbd_image_t image, 882 uint64_t off, 883 uint64_t len, 884 rbd_completion_t comp) 885 { 886 #ifdef LIBRBD_SUPPORTS_DISCARD 887 return rbd_aio_discard(image, off, len, comp); 888 #else 889 return -ENOTSUP; 890 #endif 891 } 892 893 static int rbd_aio_flush_wrapper(rbd_image_t image, 894 rbd_completion_t comp) 895 { 896 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH 897 return rbd_aio_flush(image, comp); 898 #else 899 return -ENOTSUP; 900 #endif 901 } 902 903 static BlockAIOCB *rbd_start_aio(BlockDriverState *bs, 904 int64_t off, 905 QEMUIOVector *qiov, 906 int64_t size, 907 BlockCompletionFunc *cb, 908 void *opaque, 909 RBDAIOCmd cmd) 910 { 911 RBDAIOCB *acb; 912 RADOSCB *rcb = NULL; 913 rbd_completion_t c; 914 int r; 915 916 BDRVRBDState *s = bs->opaque; 917 918 acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque); 919 acb->cmd = cmd; 920 acb->qiov = qiov; 921 assert(!qiov || qiov->size == size); 922 923 rcb = g_new(RADOSCB, 1); 924 925 if (!LIBRBD_USE_IOVEC) { 926 if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) { 927 acb->bounce = NULL; 928 } else { 929 acb->bounce = qemu_try_blockalign(bs, qiov->size); 930 if (acb->bounce == NULL) { 931 goto failed; 932 } 933 } 934 if (cmd == RBD_AIO_WRITE) { 935 qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); 936 } 937 rcb->buf = acb->bounce; 938 } 939 940 acb->ret = 0; 941 acb->error = 0; 942 acb->s = s; 943 944 rcb->acb = acb; 945 rcb->s = acb->s; 946 rcb->size = size; 947 r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c); 948 if (r < 0) { 949 goto failed; 950 } 951 952 switch (cmd) { 953 case RBD_AIO_WRITE: { 954 /* 955 * RBD APIs don't allow us to write more than actual size, so in order 956 * to support growing images, we resize the image before write 957 * operations that exceed the current size. 958 */ 959 if (off + size > s->image_size) { 960 r = qemu_rbd_resize(bs, off + size); 961 if (r < 0) { 962 goto failed_completion; 963 } 964 } 965 #ifdef LIBRBD_SUPPORTS_IOVEC 966 r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c); 967 #else 968 r = rbd_aio_write(s->image, off, size, rcb->buf, c); 969 #endif 970 break; 971 } 972 case RBD_AIO_READ: 973 #ifdef LIBRBD_SUPPORTS_IOVEC 974 r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c); 975 #else 976 r = rbd_aio_read(s->image, off, size, rcb->buf, c); 977 #endif 978 break; 979 case RBD_AIO_DISCARD: 980 r = rbd_aio_discard_wrapper(s->image, off, size, c); 981 break; 982 case RBD_AIO_FLUSH: 983 r = rbd_aio_flush_wrapper(s->image, c); 984 break; 985 default: 986 r = -EINVAL; 987 } 988 989 if (r < 0) { 990 goto failed_completion; 991 } 992 return &acb->common; 993 994 failed_completion: 995 rbd_aio_release(c); 996 failed: 997 g_free(rcb); 998 if (!LIBRBD_USE_IOVEC) { 999 qemu_vfree(acb->bounce); 1000 } 1001 1002 qemu_aio_unref(acb); 1003 return NULL; 1004 } 1005 1006 static BlockAIOCB *qemu_rbd_aio_preadv(BlockDriverState *bs, 1007 uint64_t offset, uint64_t bytes, 1008 QEMUIOVector *qiov, int flags, 1009 BlockCompletionFunc *cb, 1010 void *opaque) 1011 { 1012 return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque, 1013 RBD_AIO_READ); 1014 } 1015 1016 static BlockAIOCB *qemu_rbd_aio_pwritev(BlockDriverState *bs, 1017 uint64_t offset, uint64_t bytes, 1018 QEMUIOVector *qiov, int flags, 1019 BlockCompletionFunc *cb, 1020 void *opaque) 1021 { 1022 return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque, 1023 RBD_AIO_WRITE); 1024 } 1025 1026 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH 1027 static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs, 1028 BlockCompletionFunc *cb, 1029 void *opaque) 1030 { 1031 return rbd_start_aio(bs, 0, NULL, 0, cb, opaque, RBD_AIO_FLUSH); 1032 } 1033 1034 #else 1035 1036 static int qemu_rbd_co_flush(BlockDriverState *bs) 1037 { 1038 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 1) 1039 /* rbd_flush added in 0.1.1 */ 1040 BDRVRBDState *s = bs->opaque; 1041 return rbd_flush(s->image); 1042 #else 1043 return 0; 1044 #endif 1045 } 1046 #endif 1047 1048 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi) 1049 { 1050 BDRVRBDState *s = bs->opaque; 1051 rbd_image_info_t info; 1052 int r; 1053 1054 r = rbd_stat(s->image, &info, sizeof(info)); 1055 if (r < 0) { 1056 return r; 1057 } 1058 1059 bdi->cluster_size = info.obj_size; 1060 return 0; 1061 } 1062 1063 static int64_t qemu_rbd_getlength(BlockDriverState *bs) 1064 { 1065 BDRVRBDState *s = bs->opaque; 1066 rbd_image_info_t info; 1067 int r; 1068 1069 r = rbd_stat(s->image, &info, sizeof(info)); 1070 if (r < 0) { 1071 return r; 1072 } 1073 1074 return info.size; 1075 } 1076 1077 static int coroutine_fn qemu_rbd_co_truncate(BlockDriverState *bs, 1078 int64_t offset, 1079 bool exact, 1080 PreallocMode prealloc, 1081 BdrvRequestFlags flags, 1082 Error **errp) 1083 { 1084 int r; 1085 1086 if (prealloc != PREALLOC_MODE_OFF) { 1087 error_setg(errp, "Unsupported preallocation mode '%s'", 1088 PreallocMode_str(prealloc)); 1089 return -ENOTSUP; 1090 } 1091 1092 r = qemu_rbd_resize(bs, offset); 1093 if (r < 0) { 1094 error_setg_errno(errp, -r, "Failed to resize file"); 1095 return r; 1096 } 1097 1098 return 0; 1099 } 1100 1101 static int qemu_rbd_snap_create(BlockDriverState *bs, 1102 QEMUSnapshotInfo *sn_info) 1103 { 1104 BDRVRBDState *s = bs->opaque; 1105 int r; 1106 1107 if (sn_info->name[0] == '\0') { 1108 return -EINVAL; /* we need a name for rbd snapshots */ 1109 } 1110 1111 /* 1112 * rbd snapshots are using the name as the user controlled unique identifier 1113 * we can't use the rbd snapid for that purpose, as it can't be set 1114 */ 1115 if (sn_info->id_str[0] != '\0' && 1116 strcmp(sn_info->id_str, sn_info->name) != 0) { 1117 return -EINVAL; 1118 } 1119 1120 if (strlen(sn_info->name) >= sizeof(sn_info->id_str)) { 1121 return -ERANGE; 1122 } 1123 1124 r = rbd_snap_create(s->image, sn_info->name); 1125 if (r < 0) { 1126 error_report("failed to create snap: %s", strerror(-r)); 1127 return r; 1128 } 1129 1130 return 0; 1131 } 1132 1133 static int qemu_rbd_snap_remove(BlockDriverState *bs, 1134 const char *snapshot_id, 1135 const char *snapshot_name, 1136 Error **errp) 1137 { 1138 BDRVRBDState *s = bs->opaque; 1139 int r; 1140 1141 if (!snapshot_name) { 1142 error_setg(errp, "rbd need a valid snapshot name"); 1143 return -EINVAL; 1144 } 1145 1146 /* If snapshot_id is specified, it must be equal to name, see 1147 qemu_rbd_snap_list() */ 1148 if (snapshot_id && strcmp(snapshot_id, snapshot_name)) { 1149 error_setg(errp, 1150 "rbd do not support snapshot id, it should be NULL or " 1151 "equal to snapshot name"); 1152 return -EINVAL; 1153 } 1154 1155 r = rbd_snap_remove(s->image, snapshot_name); 1156 if (r < 0) { 1157 error_setg_errno(errp, -r, "Failed to remove the snapshot"); 1158 } 1159 return r; 1160 } 1161 1162 static int qemu_rbd_snap_rollback(BlockDriverState *bs, 1163 const char *snapshot_name) 1164 { 1165 BDRVRBDState *s = bs->opaque; 1166 1167 return rbd_snap_rollback(s->image, snapshot_name); 1168 } 1169 1170 static int qemu_rbd_snap_list(BlockDriverState *bs, 1171 QEMUSnapshotInfo **psn_tab) 1172 { 1173 BDRVRBDState *s = bs->opaque; 1174 QEMUSnapshotInfo *sn_info, *sn_tab = NULL; 1175 int i, snap_count; 1176 rbd_snap_info_t *snaps; 1177 int max_snaps = RBD_MAX_SNAPS; 1178 1179 do { 1180 snaps = g_new(rbd_snap_info_t, max_snaps); 1181 snap_count = rbd_snap_list(s->image, snaps, &max_snaps); 1182 if (snap_count <= 0) { 1183 g_free(snaps); 1184 } 1185 } while (snap_count == -ERANGE); 1186 1187 if (snap_count <= 0) { 1188 goto done; 1189 } 1190 1191 sn_tab = g_new0(QEMUSnapshotInfo, snap_count); 1192 1193 for (i = 0; i < snap_count; i++) { 1194 const char *snap_name = snaps[i].name; 1195 1196 sn_info = sn_tab + i; 1197 pstrcpy(sn_info->id_str, sizeof(sn_info->id_str), snap_name); 1198 pstrcpy(sn_info->name, sizeof(sn_info->name), snap_name); 1199 1200 sn_info->vm_state_size = snaps[i].size; 1201 sn_info->date_sec = 0; 1202 sn_info->date_nsec = 0; 1203 sn_info->vm_clock_nsec = 0; 1204 } 1205 rbd_snap_list_end(snaps); 1206 g_free(snaps); 1207 1208 done: 1209 *psn_tab = sn_tab; 1210 return snap_count; 1211 } 1212 1213 #ifdef LIBRBD_SUPPORTS_DISCARD 1214 static BlockAIOCB *qemu_rbd_aio_pdiscard(BlockDriverState *bs, 1215 int64_t offset, 1216 int bytes, 1217 BlockCompletionFunc *cb, 1218 void *opaque) 1219 { 1220 return rbd_start_aio(bs, offset, NULL, bytes, cb, opaque, 1221 RBD_AIO_DISCARD); 1222 } 1223 #endif 1224 1225 #ifdef LIBRBD_SUPPORTS_INVALIDATE 1226 static void coroutine_fn qemu_rbd_co_invalidate_cache(BlockDriverState *bs, 1227 Error **errp) 1228 { 1229 BDRVRBDState *s = bs->opaque; 1230 int r = rbd_invalidate_cache(s->image); 1231 if (r < 0) { 1232 error_setg_errno(errp, -r, "Failed to invalidate the cache"); 1233 } 1234 } 1235 #endif 1236 1237 static QemuOptsList qemu_rbd_create_opts = { 1238 .name = "rbd-create-opts", 1239 .head = QTAILQ_HEAD_INITIALIZER(qemu_rbd_create_opts.head), 1240 .desc = { 1241 { 1242 .name = BLOCK_OPT_SIZE, 1243 .type = QEMU_OPT_SIZE, 1244 .help = "Virtual disk size" 1245 }, 1246 { 1247 .name = BLOCK_OPT_CLUSTER_SIZE, 1248 .type = QEMU_OPT_SIZE, 1249 .help = "RBD object size" 1250 }, 1251 { 1252 .name = "password-secret", 1253 .type = QEMU_OPT_STRING, 1254 .help = "ID of secret providing the password", 1255 }, 1256 { /* end of list */ } 1257 } 1258 }; 1259 1260 static const char *const qemu_rbd_strong_runtime_opts[] = { 1261 "pool", 1262 "namespace", 1263 "image", 1264 "conf", 1265 "snapshot", 1266 "user", 1267 "server.", 1268 "password-secret", 1269 1270 NULL 1271 }; 1272 1273 static BlockDriver bdrv_rbd = { 1274 .format_name = "rbd", 1275 .instance_size = sizeof(BDRVRBDState), 1276 .bdrv_parse_filename = qemu_rbd_parse_filename, 1277 .bdrv_refresh_limits = qemu_rbd_refresh_limits, 1278 .bdrv_file_open = qemu_rbd_open, 1279 .bdrv_close = qemu_rbd_close, 1280 .bdrv_reopen_prepare = qemu_rbd_reopen_prepare, 1281 .bdrv_co_create = qemu_rbd_co_create, 1282 .bdrv_co_create_opts = qemu_rbd_co_create_opts, 1283 .bdrv_has_zero_init = bdrv_has_zero_init_1, 1284 .bdrv_get_info = qemu_rbd_getinfo, 1285 .create_opts = &qemu_rbd_create_opts, 1286 .bdrv_getlength = qemu_rbd_getlength, 1287 .bdrv_co_truncate = qemu_rbd_co_truncate, 1288 .protocol_name = "rbd", 1289 1290 .bdrv_aio_preadv = qemu_rbd_aio_preadv, 1291 .bdrv_aio_pwritev = qemu_rbd_aio_pwritev, 1292 1293 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH 1294 .bdrv_aio_flush = qemu_rbd_aio_flush, 1295 #else 1296 .bdrv_co_flush_to_disk = qemu_rbd_co_flush, 1297 #endif 1298 1299 #ifdef LIBRBD_SUPPORTS_DISCARD 1300 .bdrv_aio_pdiscard = qemu_rbd_aio_pdiscard, 1301 #endif 1302 1303 .bdrv_snapshot_create = qemu_rbd_snap_create, 1304 .bdrv_snapshot_delete = qemu_rbd_snap_remove, 1305 .bdrv_snapshot_list = qemu_rbd_snap_list, 1306 .bdrv_snapshot_goto = qemu_rbd_snap_rollback, 1307 #ifdef LIBRBD_SUPPORTS_INVALIDATE 1308 .bdrv_co_invalidate_cache = qemu_rbd_co_invalidate_cache, 1309 #endif 1310 1311 .strong_runtime_opts = qemu_rbd_strong_runtime_opts, 1312 }; 1313 1314 static void bdrv_rbd_init(void) 1315 { 1316 bdrv_register(&bdrv_rbd); 1317 } 1318 1319 block_init(bdrv_rbd_init); 1320