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