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-io.h" 22 #include "block/block_int.h" 23 #include "block/qdict.h" 24 #include "crypto/secret.h" 25 #include "qemu/cutils.h" 26 #include "sysemu/replay.h" 27 #include "qapi/qmp/qstring.h" 28 #include "qapi/qmp/qdict.h" 29 #include "qapi/qmp/qjson.h" 30 #include "qapi/qmp/qlist.h" 31 #include "qapi/qobject-input-visitor.h" 32 #include "qapi/qapi-visit-block-core.h" 33 34 /* 35 * When specifying the image filename use: 36 * 37 * rbd:poolname/devicename[@snapshotname][:option1=value1[:option2=value2...]] 38 * 39 * poolname must be the name of an existing rados pool. 40 * 41 * devicename is the name of the rbd image. 42 * 43 * Each option given is used to configure rados, and may be any valid 44 * Ceph option, "id", or "conf". 45 * 46 * The "id" option indicates what user we should authenticate as to 47 * the Ceph cluster. If it is excluded we will use the Ceph default 48 * (normally 'admin'). 49 * 50 * The "conf" option specifies a Ceph configuration file to read. If 51 * it is not specified, we will read from the default Ceph locations 52 * (e.g., /etc/ceph/ceph.conf). To avoid reading _any_ configuration 53 * file, specify conf=/dev/null. 54 * 55 * Configuration values containing :, @, or = can be escaped with a 56 * leading "\". 57 */ 58 59 #define OBJ_MAX_SIZE (1UL << OBJ_DEFAULT_OBJ_ORDER) 60 61 #define RBD_MAX_SNAPS 100 62 63 #define RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN 8 64 65 static const char rbd_luks_header_verification[ 66 RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN] = { 67 'L', 'U', 'K', 'S', 0xBA, 0xBE, 0, 1 68 }; 69 70 static const char rbd_luks2_header_verification[ 71 RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN] = { 72 'L', 'U', 'K', 'S', 0xBA, 0xBE, 0, 2 73 }; 74 75 typedef enum { 76 RBD_AIO_READ, 77 RBD_AIO_WRITE, 78 RBD_AIO_DISCARD, 79 RBD_AIO_FLUSH, 80 RBD_AIO_WRITE_ZEROES 81 } RBDAIOCmd; 82 83 typedef struct BDRVRBDState { 84 rados_t cluster; 85 rados_ioctx_t io_ctx; 86 rbd_image_t image; 87 char *image_name; 88 char *snap; 89 char *namespace; 90 uint64_t image_size; 91 uint64_t object_size; 92 } BDRVRBDState; 93 94 typedef struct RBDTask { 95 BlockDriverState *bs; 96 Coroutine *co; 97 bool complete; 98 int64_t ret; 99 } RBDTask; 100 101 typedef struct RBDDiffIterateReq { 102 uint64_t offs; 103 uint64_t bytes; 104 bool exists; 105 } RBDDiffIterateReq; 106 107 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, 108 BlockdevOptionsRbd *opts, bool cache, 109 const char *keypairs, const char *secretid, 110 Error **errp); 111 112 static char *qemu_rbd_strchr(char *src, char delim) 113 { 114 char *p; 115 116 for (p = src; *p; ++p) { 117 if (*p == delim) { 118 return p; 119 } 120 if (*p == '\\' && p[1] != '\0') { 121 ++p; 122 } 123 } 124 125 return NULL; 126 } 127 128 129 static char *qemu_rbd_next_tok(char *src, char delim, char **p) 130 { 131 char *end; 132 133 *p = NULL; 134 135 end = qemu_rbd_strchr(src, delim); 136 if (end) { 137 *p = end + 1; 138 *end = '\0'; 139 } 140 return src; 141 } 142 143 static void qemu_rbd_unescape(char *src) 144 { 145 char *p; 146 147 for (p = src; *src; ++src, ++p) { 148 if (*src == '\\' && src[1] != '\0') { 149 src++; 150 } 151 *p = *src; 152 } 153 *p = '\0'; 154 } 155 156 static void qemu_rbd_parse_filename(const char *filename, QDict *options, 157 Error **errp) 158 { 159 const char *start; 160 char *p, *buf; 161 QList *keypairs = NULL; 162 char *found_str, *image_name; 163 164 if (!strstart(filename, "rbd:", &start)) { 165 error_setg(errp, "File name must start with 'rbd:'"); 166 return; 167 } 168 169 buf = g_strdup(start); 170 p = buf; 171 172 found_str = qemu_rbd_next_tok(p, '/', &p); 173 if (!p) { 174 error_setg(errp, "Pool name is required"); 175 goto done; 176 } 177 qemu_rbd_unescape(found_str); 178 qdict_put_str(options, "pool", found_str); 179 180 if (qemu_rbd_strchr(p, '@')) { 181 image_name = qemu_rbd_next_tok(p, '@', &p); 182 183 found_str = qemu_rbd_next_tok(p, ':', &p); 184 qemu_rbd_unescape(found_str); 185 qdict_put_str(options, "snapshot", found_str); 186 } else { 187 image_name = qemu_rbd_next_tok(p, ':', &p); 188 } 189 /* Check for namespace in the image_name */ 190 if (qemu_rbd_strchr(image_name, '/')) { 191 found_str = qemu_rbd_next_tok(image_name, '/', &image_name); 192 qemu_rbd_unescape(found_str); 193 qdict_put_str(options, "namespace", found_str); 194 } else { 195 qdict_put_str(options, "namespace", ""); 196 } 197 qemu_rbd_unescape(image_name); 198 qdict_put_str(options, "image", image_name); 199 if (!p) { 200 goto done; 201 } 202 203 /* The following are essentially all key/value pairs, and we treat 204 * 'id' and 'conf' a bit special. Key/value pairs may be in any order. */ 205 while (p) { 206 char *name, *value; 207 name = qemu_rbd_next_tok(p, '=', &p); 208 if (!p) { 209 error_setg(errp, "conf option %s has no value", name); 210 break; 211 } 212 213 qemu_rbd_unescape(name); 214 215 value = qemu_rbd_next_tok(p, ':', &p); 216 qemu_rbd_unescape(value); 217 218 if (!strcmp(name, "conf")) { 219 qdict_put_str(options, "conf", value); 220 } else if (!strcmp(name, "id")) { 221 qdict_put_str(options, "user", value); 222 } else { 223 /* 224 * We pass these internally to qemu_rbd_set_keypairs(), so 225 * we can get away with the simpler list of [ "key1", 226 * "value1", "key2", "value2" ] rather than a raw dict 227 * { "key1": "value1", "key2": "value2" } where we can't 228 * guarantee order, or even a more correct but complex 229 * [ { "key1": "value1" }, { "key2": "value2" } ] 230 */ 231 if (!keypairs) { 232 keypairs = qlist_new(); 233 } 234 qlist_append_str(keypairs, name); 235 qlist_append_str(keypairs, value); 236 } 237 } 238 239 if (keypairs) { 240 qdict_put(options, "=keyvalue-pairs", 241 qstring_from_gstring(qobject_to_json(QOBJECT(keypairs)))); 242 } 243 244 done: 245 g_free(buf); 246 qobject_unref(keypairs); 247 return; 248 } 249 250 static int qemu_rbd_set_auth(rados_t cluster, BlockdevOptionsRbd *opts, 251 Error **errp) 252 { 253 char *key, *acr; 254 int r; 255 GString *accu; 256 RbdAuthModeList *auth; 257 258 if (opts->key_secret) { 259 key = qcrypto_secret_lookup_as_base64(opts->key_secret, errp); 260 if (!key) { 261 return -EIO; 262 } 263 r = rados_conf_set(cluster, "key", key); 264 g_free(key); 265 if (r < 0) { 266 error_setg_errno(errp, -r, "Could not set 'key'"); 267 return r; 268 } 269 } 270 271 if (opts->has_auth_client_required) { 272 accu = g_string_new(""); 273 for (auth = opts->auth_client_required; auth; auth = auth->next) { 274 if (accu->str[0]) { 275 g_string_append_c(accu, ';'); 276 } 277 g_string_append(accu, RbdAuthMode_str(auth->value)); 278 } 279 acr = g_string_free(accu, FALSE); 280 r = rados_conf_set(cluster, "auth_client_required", acr); 281 g_free(acr); 282 if (r < 0) { 283 error_setg_errno(errp, -r, 284 "Could not set 'auth_client_required'"); 285 return r; 286 } 287 } 288 289 return 0; 290 } 291 292 static int qemu_rbd_set_keypairs(rados_t cluster, const char *keypairs_json, 293 Error **errp) 294 { 295 QList *keypairs; 296 QString *name; 297 QString *value; 298 const char *key; 299 size_t remaining; 300 int ret = 0; 301 302 if (!keypairs_json) { 303 return ret; 304 } 305 keypairs = qobject_to(QList, 306 qobject_from_json(keypairs_json, &error_abort)); 307 remaining = qlist_size(keypairs) / 2; 308 assert(remaining); 309 310 while (remaining--) { 311 name = qobject_to(QString, qlist_pop(keypairs)); 312 value = qobject_to(QString, qlist_pop(keypairs)); 313 assert(name && value); 314 key = qstring_get_str(name); 315 316 ret = rados_conf_set(cluster, key, qstring_get_str(value)); 317 qobject_unref(value); 318 if (ret < 0) { 319 error_setg_errno(errp, -ret, "invalid conf option %s", key); 320 qobject_unref(name); 321 ret = -EINVAL; 322 break; 323 } 324 qobject_unref(name); 325 } 326 327 qobject_unref(keypairs); 328 return ret; 329 } 330 331 #ifdef LIBRBD_SUPPORTS_ENCRYPTION 332 static int qemu_rbd_convert_luks_options( 333 RbdEncryptionOptionsLUKSBase *luks_opts, 334 char **passphrase, 335 size_t *passphrase_len, 336 Error **errp) 337 { 338 return qcrypto_secret_lookup(luks_opts->key_secret, (uint8_t **)passphrase, 339 passphrase_len, errp); 340 } 341 342 static int qemu_rbd_convert_luks_create_options( 343 RbdEncryptionCreateOptionsLUKSBase *luks_opts, 344 rbd_encryption_algorithm_t *alg, 345 char **passphrase, 346 size_t *passphrase_len, 347 Error **errp) 348 { 349 int r = 0; 350 351 r = qemu_rbd_convert_luks_options( 352 qapi_RbdEncryptionCreateOptionsLUKSBase_base(luks_opts), 353 passphrase, passphrase_len, errp); 354 if (r < 0) { 355 return r; 356 } 357 358 if (luks_opts->has_cipher_alg) { 359 switch (luks_opts->cipher_alg) { 360 case QCRYPTO_CIPHER_ALG_AES_128: { 361 *alg = RBD_ENCRYPTION_ALGORITHM_AES128; 362 break; 363 } 364 case QCRYPTO_CIPHER_ALG_AES_256: { 365 *alg = RBD_ENCRYPTION_ALGORITHM_AES256; 366 break; 367 } 368 default: { 369 r = -ENOTSUP; 370 error_setg_errno(errp, -r, "unknown encryption algorithm: %u", 371 luks_opts->cipher_alg); 372 return r; 373 } 374 } 375 } else { 376 /* default alg */ 377 *alg = RBD_ENCRYPTION_ALGORITHM_AES256; 378 } 379 380 return 0; 381 } 382 383 static int qemu_rbd_encryption_format(rbd_image_t image, 384 RbdEncryptionCreateOptions *encrypt, 385 Error **errp) 386 { 387 int r = 0; 388 g_autofree char *passphrase = NULL; 389 size_t passphrase_len; 390 rbd_encryption_format_t format; 391 rbd_encryption_options_t opts; 392 rbd_encryption_luks1_format_options_t luks_opts; 393 rbd_encryption_luks2_format_options_t luks2_opts; 394 size_t opts_size; 395 uint64_t raw_size, effective_size; 396 397 r = rbd_get_size(image, &raw_size); 398 if (r < 0) { 399 error_setg_errno(errp, -r, "cannot get raw image size"); 400 return r; 401 } 402 403 switch (encrypt->format) { 404 case RBD_IMAGE_ENCRYPTION_FORMAT_LUKS: { 405 memset(&luks_opts, 0, sizeof(luks_opts)); 406 format = RBD_ENCRYPTION_FORMAT_LUKS1; 407 opts = &luks_opts; 408 opts_size = sizeof(luks_opts); 409 r = qemu_rbd_convert_luks_create_options( 410 qapi_RbdEncryptionCreateOptionsLUKS_base(&encrypt->u.luks), 411 &luks_opts.alg, &passphrase, &passphrase_len, errp); 412 if (r < 0) { 413 return r; 414 } 415 luks_opts.passphrase = passphrase; 416 luks_opts.passphrase_size = passphrase_len; 417 break; 418 } 419 case RBD_IMAGE_ENCRYPTION_FORMAT_LUKS2: { 420 memset(&luks2_opts, 0, sizeof(luks2_opts)); 421 format = RBD_ENCRYPTION_FORMAT_LUKS2; 422 opts = &luks2_opts; 423 opts_size = sizeof(luks2_opts); 424 r = qemu_rbd_convert_luks_create_options( 425 qapi_RbdEncryptionCreateOptionsLUKS2_base( 426 &encrypt->u.luks2), 427 &luks2_opts.alg, &passphrase, &passphrase_len, errp); 428 if (r < 0) { 429 return r; 430 } 431 luks2_opts.passphrase = passphrase; 432 luks2_opts.passphrase_size = passphrase_len; 433 break; 434 } 435 default: { 436 r = -ENOTSUP; 437 error_setg_errno( 438 errp, -r, "unknown image encryption format: %u", 439 encrypt->format); 440 return r; 441 } 442 } 443 444 r = rbd_encryption_format(image, format, opts, opts_size); 445 if (r < 0) { 446 error_setg_errno(errp, -r, "encryption format fail"); 447 return r; 448 } 449 450 r = rbd_get_size(image, &effective_size); 451 if (r < 0) { 452 error_setg_errno(errp, -r, "cannot get effective image size"); 453 return r; 454 } 455 456 r = rbd_resize(image, raw_size + (raw_size - effective_size)); 457 if (r < 0) { 458 error_setg_errno(errp, -r, "cannot resize image after format"); 459 return r; 460 } 461 462 return 0; 463 } 464 465 static int qemu_rbd_encryption_load(rbd_image_t image, 466 RbdEncryptionOptions *encrypt, 467 Error **errp) 468 { 469 int r = 0; 470 g_autofree char *passphrase = NULL; 471 size_t passphrase_len; 472 rbd_encryption_luks1_format_options_t luks_opts; 473 rbd_encryption_luks2_format_options_t luks2_opts; 474 rbd_encryption_format_t format; 475 rbd_encryption_options_t opts; 476 size_t opts_size; 477 478 switch (encrypt->format) { 479 case RBD_IMAGE_ENCRYPTION_FORMAT_LUKS: { 480 memset(&luks_opts, 0, sizeof(luks_opts)); 481 format = RBD_ENCRYPTION_FORMAT_LUKS1; 482 opts = &luks_opts; 483 opts_size = sizeof(luks_opts); 484 r = qemu_rbd_convert_luks_options( 485 qapi_RbdEncryptionOptionsLUKS_base(&encrypt->u.luks), 486 &passphrase, &passphrase_len, errp); 487 if (r < 0) { 488 return r; 489 } 490 luks_opts.passphrase = passphrase; 491 luks_opts.passphrase_size = passphrase_len; 492 break; 493 } 494 case RBD_IMAGE_ENCRYPTION_FORMAT_LUKS2: { 495 memset(&luks2_opts, 0, sizeof(luks2_opts)); 496 format = RBD_ENCRYPTION_FORMAT_LUKS2; 497 opts = &luks2_opts; 498 opts_size = sizeof(luks2_opts); 499 r = qemu_rbd_convert_luks_options( 500 qapi_RbdEncryptionOptionsLUKS2_base(&encrypt->u.luks2), 501 &passphrase, &passphrase_len, errp); 502 if (r < 0) { 503 return r; 504 } 505 luks2_opts.passphrase = passphrase; 506 luks2_opts.passphrase_size = passphrase_len; 507 break; 508 } 509 default: { 510 r = -ENOTSUP; 511 error_setg_errno( 512 errp, -r, "unknown image encryption format: %u", 513 encrypt->format); 514 return r; 515 } 516 } 517 518 r = rbd_encryption_load(image, format, opts, opts_size); 519 if (r < 0) { 520 error_setg_errno(errp, -r, "encryption load fail"); 521 return r; 522 } 523 524 return 0; 525 } 526 #endif 527 528 /* FIXME Deprecate and remove keypairs or make it available in QMP. */ 529 static int qemu_rbd_do_create(BlockdevCreateOptions *options, 530 const char *keypairs, const char *password_secret, 531 Error **errp) 532 { 533 BlockdevCreateOptionsRbd *opts = &options->u.rbd; 534 rados_t cluster; 535 rados_ioctx_t io_ctx; 536 int obj_order = 0; 537 int ret; 538 539 assert(options->driver == BLOCKDEV_DRIVER_RBD); 540 if (opts->location->snapshot) { 541 error_setg(errp, "Can't use snapshot name for image creation"); 542 return -EINVAL; 543 } 544 545 #ifndef LIBRBD_SUPPORTS_ENCRYPTION 546 if (opts->encrypt) { 547 error_setg(errp, "RBD library does not support image encryption"); 548 return -ENOTSUP; 549 } 550 #endif 551 552 if (opts->has_cluster_size) { 553 int64_t objsize = opts->cluster_size; 554 if ((objsize - 1) & objsize) { /* not a power of 2? */ 555 error_setg(errp, "obj size needs to be power of 2"); 556 return -EINVAL; 557 } 558 if (objsize < 4096) { 559 error_setg(errp, "obj size too small"); 560 return -EINVAL; 561 } 562 obj_order = ctz32(objsize); 563 } 564 565 ret = qemu_rbd_connect(&cluster, &io_ctx, opts->location, false, keypairs, 566 password_secret, errp); 567 if (ret < 0) { 568 return ret; 569 } 570 571 ret = rbd_create(io_ctx, opts->location->image, opts->size, &obj_order); 572 if (ret < 0) { 573 error_setg_errno(errp, -ret, "error rbd create"); 574 goto out; 575 } 576 577 #ifdef LIBRBD_SUPPORTS_ENCRYPTION 578 if (opts->encrypt) { 579 rbd_image_t image; 580 581 ret = rbd_open(io_ctx, opts->location->image, &image, NULL); 582 if (ret < 0) { 583 error_setg_errno(errp, -ret, 584 "error opening image '%s' for encryption format", 585 opts->location->image); 586 goto out; 587 } 588 589 ret = qemu_rbd_encryption_format(image, opts->encrypt, errp); 590 rbd_close(image); 591 if (ret < 0) { 592 /* encryption format fail, try removing the image */ 593 rbd_remove(io_ctx, opts->location->image); 594 goto out; 595 } 596 } 597 #endif 598 599 ret = 0; 600 out: 601 rados_ioctx_destroy(io_ctx); 602 rados_shutdown(cluster); 603 return ret; 604 } 605 606 static int qemu_rbd_co_create(BlockdevCreateOptions *options, Error **errp) 607 { 608 return qemu_rbd_do_create(options, NULL, NULL, errp); 609 } 610 611 static int qemu_rbd_extract_encryption_create_options( 612 QemuOpts *opts, 613 RbdEncryptionCreateOptions **spec, 614 Error **errp) 615 { 616 QDict *opts_qdict; 617 QDict *encrypt_qdict; 618 Visitor *v; 619 int ret = 0; 620 621 opts_qdict = qemu_opts_to_qdict(opts, NULL); 622 qdict_extract_subqdict(opts_qdict, &encrypt_qdict, "encrypt."); 623 qobject_unref(opts_qdict); 624 if (!qdict_size(encrypt_qdict)) { 625 *spec = NULL; 626 goto exit; 627 } 628 629 /* Convert options into a QAPI object */ 630 v = qobject_input_visitor_new_flat_confused(encrypt_qdict, errp); 631 if (!v) { 632 ret = -EINVAL; 633 goto exit; 634 } 635 636 visit_type_RbdEncryptionCreateOptions(v, NULL, spec, errp); 637 visit_free(v); 638 if (!*spec) { 639 ret = -EINVAL; 640 goto exit; 641 } 642 643 exit: 644 qobject_unref(encrypt_qdict); 645 return ret; 646 } 647 648 static int coroutine_fn qemu_rbd_co_create_opts(BlockDriver *drv, 649 const char *filename, 650 QemuOpts *opts, 651 Error **errp) 652 { 653 BlockdevCreateOptions *create_options; 654 BlockdevCreateOptionsRbd *rbd_opts; 655 BlockdevOptionsRbd *loc; 656 RbdEncryptionCreateOptions *encrypt = NULL; 657 Error *local_err = NULL; 658 const char *keypairs, *password_secret; 659 QDict *options = NULL; 660 int ret = 0; 661 662 create_options = g_new0(BlockdevCreateOptions, 1); 663 create_options->driver = BLOCKDEV_DRIVER_RBD; 664 rbd_opts = &create_options->u.rbd; 665 666 rbd_opts->location = g_new0(BlockdevOptionsRbd, 1); 667 668 password_secret = qemu_opt_get(opts, "password-secret"); 669 670 /* Read out options */ 671 rbd_opts->size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), 672 BDRV_SECTOR_SIZE); 673 rbd_opts->cluster_size = qemu_opt_get_size_del(opts, 674 BLOCK_OPT_CLUSTER_SIZE, 0); 675 rbd_opts->has_cluster_size = (rbd_opts->cluster_size != 0); 676 677 options = qdict_new(); 678 qemu_rbd_parse_filename(filename, options, &local_err); 679 if (local_err) { 680 ret = -EINVAL; 681 error_propagate(errp, local_err); 682 goto exit; 683 } 684 685 ret = qemu_rbd_extract_encryption_create_options(opts, &encrypt, errp); 686 if (ret < 0) { 687 goto exit; 688 } 689 rbd_opts->encrypt = encrypt; 690 691 /* 692 * Caution: while qdict_get_try_str() is fine, getting non-string 693 * types would require more care. When @options come from -blockdev 694 * or blockdev_add, its members are typed according to the QAPI 695 * schema, but when they come from -drive, they're all QString. 696 */ 697 loc = rbd_opts->location; 698 loc->pool = g_strdup(qdict_get_try_str(options, "pool")); 699 loc->conf = g_strdup(qdict_get_try_str(options, "conf")); 700 loc->user = g_strdup(qdict_get_try_str(options, "user")); 701 loc->q_namespace = g_strdup(qdict_get_try_str(options, "namespace")); 702 loc->image = g_strdup(qdict_get_try_str(options, "image")); 703 keypairs = qdict_get_try_str(options, "=keyvalue-pairs"); 704 705 ret = qemu_rbd_do_create(create_options, keypairs, password_secret, errp); 706 if (ret < 0) { 707 goto exit; 708 } 709 710 exit: 711 qobject_unref(options); 712 qapi_free_BlockdevCreateOptions(create_options); 713 return ret; 714 } 715 716 static char *qemu_rbd_mon_host(BlockdevOptionsRbd *opts, Error **errp) 717 { 718 const char **vals; 719 const char *host, *port; 720 char *rados_str; 721 InetSocketAddressBaseList *p; 722 int i, cnt; 723 724 if (!opts->has_server) { 725 return NULL; 726 } 727 728 for (cnt = 0, p = opts->server; p; p = p->next) { 729 cnt++; 730 } 731 732 vals = g_new(const char *, cnt + 1); 733 734 for (i = 0, p = opts->server; p; p = p->next, i++) { 735 host = p->value->host; 736 port = p->value->port; 737 738 if (strchr(host, ':')) { 739 vals[i] = g_strdup_printf("[%s]:%s", host, port); 740 } else { 741 vals[i] = g_strdup_printf("%s:%s", host, port); 742 } 743 } 744 vals[i] = NULL; 745 746 rados_str = i ? g_strjoinv(";", (char **)vals) : NULL; 747 g_strfreev((char **)vals); 748 return rados_str; 749 } 750 751 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, 752 BlockdevOptionsRbd *opts, bool cache, 753 const char *keypairs, const char *secretid, 754 Error **errp) 755 { 756 char *mon_host = NULL; 757 Error *local_err = NULL; 758 int r; 759 760 if (secretid) { 761 if (opts->key_secret) { 762 error_setg(errp, 763 "Legacy 'password-secret' clashes with 'key-secret'"); 764 return -EINVAL; 765 } 766 opts->key_secret = g_strdup(secretid); 767 } 768 769 mon_host = qemu_rbd_mon_host(opts, &local_err); 770 if (local_err) { 771 error_propagate(errp, local_err); 772 r = -EINVAL; 773 goto out; 774 } 775 776 r = rados_create(cluster, opts->user); 777 if (r < 0) { 778 error_setg_errno(errp, -r, "error initializing"); 779 goto out; 780 } 781 782 /* try default location when conf=NULL, but ignore failure */ 783 r = rados_conf_read_file(*cluster, opts->conf); 784 if (opts->conf && r < 0) { 785 error_setg_errno(errp, -r, "error reading conf file %s", opts->conf); 786 goto failed_shutdown; 787 } 788 789 r = qemu_rbd_set_keypairs(*cluster, keypairs, errp); 790 if (r < 0) { 791 goto failed_shutdown; 792 } 793 794 if (mon_host) { 795 r = rados_conf_set(*cluster, "mon_host", mon_host); 796 if (r < 0) { 797 goto failed_shutdown; 798 } 799 } 800 801 r = qemu_rbd_set_auth(*cluster, opts, errp); 802 if (r < 0) { 803 goto failed_shutdown; 804 } 805 806 /* 807 * Fallback to more conservative semantics if setting cache 808 * options fails. Ignore errors from setting rbd_cache because the 809 * only possible error is that the option does not exist, and 810 * librbd defaults to no caching. If write through caching cannot 811 * be set up, fall back to no caching. 812 */ 813 if (cache) { 814 rados_conf_set(*cluster, "rbd_cache", "true"); 815 } else { 816 rados_conf_set(*cluster, "rbd_cache", "false"); 817 } 818 819 r = rados_connect(*cluster); 820 if (r < 0) { 821 error_setg_errno(errp, -r, "error connecting"); 822 goto failed_shutdown; 823 } 824 825 r = rados_ioctx_create(*cluster, opts->pool, io_ctx); 826 if (r < 0) { 827 error_setg_errno(errp, -r, "error opening pool %s", opts->pool); 828 goto failed_shutdown; 829 } 830 831 #ifdef HAVE_RBD_NAMESPACE_EXISTS 832 if (opts->q_namespace && strlen(opts->q_namespace) > 0) { 833 bool exists; 834 835 r = rbd_namespace_exists(*io_ctx, opts->q_namespace, &exists); 836 if (r < 0) { 837 error_setg_errno(errp, -r, "error checking namespace"); 838 goto failed_ioctx_destroy; 839 } 840 841 if (!exists) { 842 error_setg(errp, "namespace '%s' does not exist", 843 opts->q_namespace); 844 r = -ENOENT; 845 goto failed_ioctx_destroy; 846 } 847 } 848 #endif 849 850 /* 851 * Set the namespace after opening the io context on the pool, 852 * if nspace == NULL or if nspace == "", it is just as we did nothing 853 */ 854 rados_ioctx_set_namespace(*io_ctx, opts->q_namespace); 855 856 r = 0; 857 goto out; 858 859 #ifdef HAVE_RBD_NAMESPACE_EXISTS 860 failed_ioctx_destroy: 861 rados_ioctx_destroy(*io_ctx); 862 #endif 863 failed_shutdown: 864 rados_shutdown(*cluster); 865 out: 866 g_free(mon_host); 867 return r; 868 } 869 870 static int qemu_rbd_convert_options(QDict *options, BlockdevOptionsRbd **opts, 871 Error **errp) 872 { 873 Visitor *v; 874 875 /* Convert the remaining options into a QAPI object */ 876 v = qobject_input_visitor_new_flat_confused(options, errp); 877 if (!v) { 878 return -EINVAL; 879 } 880 881 visit_type_BlockdevOptionsRbd(v, NULL, opts, errp); 882 visit_free(v); 883 if (!opts) { 884 return -EINVAL; 885 } 886 887 return 0; 888 } 889 890 static int qemu_rbd_attempt_legacy_options(QDict *options, 891 BlockdevOptionsRbd **opts, 892 char **keypairs) 893 { 894 char *filename; 895 int r; 896 897 filename = g_strdup(qdict_get_try_str(options, "filename")); 898 if (!filename) { 899 return -EINVAL; 900 } 901 qdict_del(options, "filename"); 902 903 qemu_rbd_parse_filename(filename, options, NULL); 904 905 /* keypairs freed by caller */ 906 *keypairs = g_strdup(qdict_get_try_str(options, "=keyvalue-pairs")); 907 if (*keypairs) { 908 qdict_del(options, "=keyvalue-pairs"); 909 } 910 911 r = qemu_rbd_convert_options(options, opts, NULL); 912 913 g_free(filename); 914 return r; 915 } 916 917 static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, 918 Error **errp) 919 { 920 BDRVRBDState *s = bs->opaque; 921 BlockdevOptionsRbd *opts = NULL; 922 const QDictEntry *e; 923 Error *local_err = NULL; 924 char *keypairs, *secretid; 925 rbd_image_info_t info; 926 int r; 927 928 keypairs = g_strdup(qdict_get_try_str(options, "=keyvalue-pairs")); 929 if (keypairs) { 930 qdict_del(options, "=keyvalue-pairs"); 931 } 932 933 secretid = g_strdup(qdict_get_try_str(options, "password-secret")); 934 if (secretid) { 935 qdict_del(options, "password-secret"); 936 } 937 938 r = qemu_rbd_convert_options(options, &opts, &local_err); 939 if (local_err) { 940 /* If keypairs are present, that means some options are present in 941 * the modern option format. Don't attempt to parse legacy option 942 * formats, as we won't support mixed usage. */ 943 if (keypairs) { 944 error_propagate(errp, local_err); 945 goto out; 946 } 947 948 /* If the initial attempt to convert and process the options failed, 949 * we may be attempting to open an image file that has the rbd options 950 * specified in the older format consisting of all key/value pairs 951 * encoded in the filename. Go ahead and attempt to parse the 952 * filename, and see if we can pull out the required options. */ 953 r = qemu_rbd_attempt_legacy_options(options, &opts, &keypairs); 954 if (r < 0) { 955 /* Propagate the original error, not the legacy parsing fallback 956 * error, as the latter was just a best-effort attempt. */ 957 error_propagate(errp, local_err); 958 goto out; 959 } 960 /* Take care whenever deciding to actually deprecate; once this ability 961 * is removed, we will not be able to open any images with legacy-styled 962 * backing image strings. */ 963 warn_report("RBD options encoded in the filename as keyvalue pairs " 964 "is deprecated"); 965 } 966 967 /* Remove the processed options from the QDict (the visitor processes 968 * _all_ options in the QDict) */ 969 while ((e = qdict_first(options))) { 970 qdict_del(options, e->key); 971 } 972 973 r = qemu_rbd_connect(&s->cluster, &s->io_ctx, opts, 974 !(flags & BDRV_O_NOCACHE), keypairs, secretid, errp); 975 if (r < 0) { 976 goto out; 977 } 978 979 s->snap = g_strdup(opts->snapshot); 980 s->image_name = g_strdup(opts->image); 981 982 /* rbd_open is always r/w */ 983 r = rbd_open(s->io_ctx, s->image_name, &s->image, s->snap); 984 if (r < 0) { 985 error_setg_errno(errp, -r, "error reading header from %s", 986 s->image_name); 987 goto failed_open; 988 } 989 990 if (opts->encrypt) { 991 #ifdef LIBRBD_SUPPORTS_ENCRYPTION 992 r = qemu_rbd_encryption_load(s->image, opts->encrypt, errp); 993 if (r < 0) { 994 goto failed_post_open; 995 } 996 #else 997 r = -ENOTSUP; 998 error_setg(errp, "RBD library does not support image encryption"); 999 goto failed_post_open; 1000 #endif 1001 } 1002 1003 r = rbd_stat(s->image, &info, sizeof(info)); 1004 if (r < 0) { 1005 error_setg_errno(errp, -r, "error getting image info from %s", 1006 s->image_name); 1007 goto failed_post_open; 1008 } 1009 s->image_size = info.size; 1010 s->object_size = info.obj_size; 1011 1012 /* If we are using an rbd snapshot, we must be r/o, otherwise 1013 * leave as-is */ 1014 if (s->snap != NULL) { 1015 r = bdrv_apply_auto_read_only(bs, "rbd snapshots are read-only", errp); 1016 if (r < 0) { 1017 goto failed_post_open; 1018 } 1019 } 1020 1021 #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES 1022 bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK; 1023 #endif 1024 1025 /* When extending regular files, we get zeros from the OS */ 1026 bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE; 1027 1028 r = 0; 1029 goto out; 1030 1031 failed_post_open: 1032 rbd_close(s->image); 1033 failed_open: 1034 rados_ioctx_destroy(s->io_ctx); 1035 g_free(s->snap); 1036 g_free(s->image_name); 1037 rados_shutdown(s->cluster); 1038 out: 1039 qapi_free_BlockdevOptionsRbd(opts); 1040 g_free(keypairs); 1041 g_free(secretid); 1042 return r; 1043 } 1044 1045 1046 /* Since RBD is currently always opened R/W via the API, 1047 * we just need to check if we are using a snapshot or not, in 1048 * order to determine if we will allow it to be R/W */ 1049 static int qemu_rbd_reopen_prepare(BDRVReopenState *state, 1050 BlockReopenQueue *queue, Error **errp) 1051 { 1052 BDRVRBDState *s = state->bs->opaque; 1053 int ret = 0; 1054 1055 if (s->snap && state->flags & BDRV_O_RDWR) { 1056 error_setg(errp, 1057 "Cannot change node '%s' to r/w when using RBD snapshot", 1058 bdrv_get_device_or_node_name(state->bs)); 1059 ret = -EINVAL; 1060 } 1061 1062 return ret; 1063 } 1064 1065 static void qemu_rbd_close(BlockDriverState *bs) 1066 { 1067 BDRVRBDState *s = bs->opaque; 1068 1069 rbd_close(s->image); 1070 rados_ioctx_destroy(s->io_ctx); 1071 g_free(s->snap); 1072 g_free(s->image_name); 1073 rados_shutdown(s->cluster); 1074 } 1075 1076 /* Resize the RBD image and update the 'image_size' with the current size */ 1077 static int qemu_rbd_resize(BlockDriverState *bs, uint64_t size) 1078 { 1079 BDRVRBDState *s = bs->opaque; 1080 int r; 1081 1082 r = rbd_resize(s->image, size); 1083 if (r < 0) { 1084 return r; 1085 } 1086 1087 s->image_size = size; 1088 1089 return 0; 1090 } 1091 1092 static void qemu_rbd_finish_bh(void *opaque) 1093 { 1094 RBDTask *task = opaque; 1095 task->complete = true; 1096 aio_co_wake(task->co); 1097 } 1098 1099 /* 1100 * This is the completion callback function for all rbd aio calls 1101 * started from qemu_rbd_start_co(). 1102 * 1103 * Note: this function is being called from a non qemu thread so 1104 * we need to be careful about what we do here. Generally we only 1105 * schedule a BH, and do the rest of the io completion handling 1106 * from qemu_rbd_finish_bh() which runs in a qemu context. 1107 */ 1108 static void qemu_rbd_completion_cb(rbd_completion_t c, RBDTask *task) 1109 { 1110 task->ret = rbd_aio_get_return_value(c); 1111 rbd_aio_release(c); 1112 aio_bh_schedule_oneshot(bdrv_get_aio_context(task->bs), 1113 qemu_rbd_finish_bh, task); 1114 } 1115 1116 static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs, 1117 uint64_t offset, 1118 uint64_t bytes, 1119 QEMUIOVector *qiov, 1120 int flags, 1121 RBDAIOCmd cmd) 1122 { 1123 BDRVRBDState *s = bs->opaque; 1124 RBDTask task = { .bs = bs, .co = qemu_coroutine_self() }; 1125 rbd_completion_t c; 1126 int r; 1127 1128 assert(!qiov || qiov->size == bytes); 1129 1130 if (cmd == RBD_AIO_WRITE || cmd == RBD_AIO_WRITE_ZEROES) { 1131 /* 1132 * RBD APIs don't allow us to write more than actual size, so in order 1133 * to support growing images, we resize the image before write 1134 * operations that exceed the current size. 1135 */ 1136 if (offset + bytes > s->image_size) { 1137 int r = qemu_rbd_resize(bs, offset + bytes); 1138 if (r < 0) { 1139 return r; 1140 } 1141 } 1142 } 1143 1144 r = rbd_aio_create_completion(&task, 1145 (rbd_callback_t) qemu_rbd_completion_cb, &c); 1146 if (r < 0) { 1147 return r; 1148 } 1149 1150 switch (cmd) { 1151 case RBD_AIO_READ: 1152 r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, offset, c); 1153 break; 1154 case RBD_AIO_WRITE: 1155 r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, offset, c); 1156 break; 1157 case RBD_AIO_DISCARD: 1158 r = rbd_aio_discard(s->image, offset, bytes, c); 1159 break; 1160 case RBD_AIO_FLUSH: 1161 r = rbd_aio_flush(s->image, c); 1162 break; 1163 #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES 1164 case RBD_AIO_WRITE_ZEROES: { 1165 int zero_flags = 0; 1166 #ifdef RBD_WRITE_ZEROES_FLAG_THICK_PROVISION 1167 if (!(flags & BDRV_REQ_MAY_UNMAP)) { 1168 zero_flags = RBD_WRITE_ZEROES_FLAG_THICK_PROVISION; 1169 } 1170 #endif 1171 r = rbd_aio_write_zeroes(s->image, offset, bytes, c, zero_flags, 0); 1172 break; 1173 } 1174 #endif 1175 default: 1176 r = -EINVAL; 1177 } 1178 1179 if (r < 0) { 1180 error_report("rbd request failed early: cmd %d offset %" PRIu64 1181 " bytes %" PRIu64 " flags %d r %d (%s)", cmd, offset, 1182 bytes, flags, r, strerror(-r)); 1183 rbd_aio_release(c); 1184 return r; 1185 } 1186 1187 while (!task.complete) { 1188 qemu_coroutine_yield(); 1189 } 1190 1191 if (task.ret < 0) { 1192 error_report("rbd request failed: cmd %d offset %" PRIu64 " bytes %" 1193 PRIu64 " flags %d task.ret %" PRIi64 " (%s)", cmd, offset, 1194 bytes, flags, task.ret, strerror(-task.ret)); 1195 return task.ret; 1196 } 1197 1198 /* zero pad short reads */ 1199 if (cmd == RBD_AIO_READ && task.ret < qiov->size) { 1200 qemu_iovec_memset(qiov, task.ret, 0, qiov->size - task.ret); 1201 } 1202 1203 return 0; 1204 } 1205 1206 static int 1207 coroutine_fn qemu_rbd_co_preadv(BlockDriverState *bs, int64_t offset, 1208 int64_t bytes, QEMUIOVector *qiov, 1209 BdrvRequestFlags flags) 1210 { 1211 return qemu_rbd_start_co(bs, offset, bytes, qiov, flags, RBD_AIO_READ); 1212 } 1213 1214 static int 1215 coroutine_fn qemu_rbd_co_pwritev(BlockDriverState *bs, int64_t offset, 1216 int64_t bytes, QEMUIOVector *qiov, 1217 BdrvRequestFlags flags) 1218 { 1219 return qemu_rbd_start_co(bs, offset, bytes, qiov, flags, RBD_AIO_WRITE); 1220 } 1221 1222 static int coroutine_fn qemu_rbd_co_flush(BlockDriverState *bs) 1223 { 1224 return qemu_rbd_start_co(bs, 0, 0, NULL, 0, RBD_AIO_FLUSH); 1225 } 1226 1227 static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs, 1228 int64_t offset, int64_t bytes) 1229 { 1230 return qemu_rbd_start_co(bs, offset, bytes, NULL, 0, RBD_AIO_DISCARD); 1231 } 1232 1233 #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES 1234 static int 1235 coroutine_fn qemu_rbd_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, 1236 int64_t bytes, BdrvRequestFlags flags) 1237 { 1238 return qemu_rbd_start_co(bs, offset, bytes, NULL, flags, 1239 RBD_AIO_WRITE_ZEROES); 1240 } 1241 #endif 1242 1243 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi) 1244 { 1245 BDRVRBDState *s = bs->opaque; 1246 bdi->cluster_size = s->object_size; 1247 return 0; 1248 } 1249 1250 static ImageInfoSpecific *qemu_rbd_get_specific_info(BlockDriverState *bs, 1251 Error **errp) 1252 { 1253 BDRVRBDState *s = bs->opaque; 1254 ImageInfoSpecific *spec_info; 1255 char buf[RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN] = {0}; 1256 int r; 1257 1258 if (s->image_size >= RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN) { 1259 r = rbd_read(s->image, 0, 1260 RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN, buf); 1261 if (r < 0) { 1262 error_setg_errno(errp, -r, "cannot read image start for probe"); 1263 return NULL; 1264 } 1265 } 1266 1267 spec_info = g_new(ImageInfoSpecific, 1); 1268 *spec_info = (ImageInfoSpecific){ 1269 .type = IMAGE_INFO_SPECIFIC_KIND_RBD, 1270 .u.rbd.data = g_new0(ImageInfoSpecificRbd, 1), 1271 }; 1272 1273 if (memcmp(buf, rbd_luks_header_verification, 1274 RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN) == 0) { 1275 spec_info->u.rbd.data->encryption_format = 1276 RBD_IMAGE_ENCRYPTION_FORMAT_LUKS; 1277 spec_info->u.rbd.data->has_encryption_format = true; 1278 } else if (memcmp(buf, rbd_luks2_header_verification, 1279 RBD_ENCRYPTION_LUKS_HEADER_VERIFICATION_LEN) == 0) { 1280 spec_info->u.rbd.data->encryption_format = 1281 RBD_IMAGE_ENCRYPTION_FORMAT_LUKS2; 1282 spec_info->u.rbd.data->has_encryption_format = true; 1283 } else { 1284 spec_info->u.rbd.data->has_encryption_format = false; 1285 } 1286 1287 return spec_info; 1288 } 1289 1290 /* 1291 * rbd_diff_iterate2 allows to interrupt the exection by returning a negative 1292 * value in the callback routine. Choose a value that does not conflict with 1293 * an existing exitcode and return it if we want to prematurely stop the 1294 * execution because we detected a change in the allocation status. 1295 */ 1296 #define QEMU_RBD_EXIT_DIFF_ITERATE2 -9000 1297 1298 static int qemu_rbd_diff_iterate_cb(uint64_t offs, size_t len, 1299 int exists, void *opaque) 1300 { 1301 RBDDiffIterateReq *req = opaque; 1302 1303 assert(req->offs + req->bytes <= offs); 1304 1305 /* treat a hole like an unallocated area and bail out */ 1306 if (!exists) { 1307 return 0; 1308 } 1309 1310 if (!req->exists && offs > req->offs) { 1311 /* 1312 * we started in an unallocated area and hit the first allocated 1313 * block. req->bytes must be set to the length of the unallocated area 1314 * before the allocated area. stop further processing. 1315 */ 1316 req->bytes = offs - req->offs; 1317 return QEMU_RBD_EXIT_DIFF_ITERATE2; 1318 } 1319 1320 if (req->exists && offs > req->offs + req->bytes) { 1321 /* 1322 * we started in an allocated area and jumped over an unallocated area, 1323 * req->bytes contains the length of the allocated area before the 1324 * unallocated area. stop further processing. 1325 */ 1326 return QEMU_RBD_EXIT_DIFF_ITERATE2; 1327 } 1328 1329 req->bytes += len; 1330 req->exists = true; 1331 1332 return 0; 1333 } 1334 1335 static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs, 1336 bool want_zero, int64_t offset, 1337 int64_t bytes, int64_t *pnum, 1338 int64_t *map, 1339 BlockDriverState **file) 1340 { 1341 BDRVRBDState *s = bs->opaque; 1342 int status, r; 1343 RBDDiffIterateReq req = { .offs = offset }; 1344 uint64_t features, flags; 1345 uint64_t head = 0; 1346 1347 assert(offset + bytes <= s->image_size); 1348 1349 /* default to all sectors allocated */ 1350 status = BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; 1351 *map = offset; 1352 *file = bs; 1353 *pnum = bytes; 1354 1355 /* check if RBD image supports fast-diff */ 1356 r = rbd_get_features(s->image, &features); 1357 if (r < 0) { 1358 return status; 1359 } 1360 if (!(features & RBD_FEATURE_FAST_DIFF)) { 1361 return status; 1362 } 1363 1364 /* check if RBD fast-diff result is valid */ 1365 r = rbd_get_flags(s->image, &flags); 1366 if (r < 0) { 1367 return status; 1368 } 1369 if (flags & RBD_FLAG_FAST_DIFF_INVALID) { 1370 return status; 1371 } 1372 1373 #if LIBRBD_VERSION_CODE < LIBRBD_VERSION(1, 17, 0) 1374 /* 1375 * librbd had a bug until early 2022 that affected all versions of ceph that 1376 * supported fast-diff. This bug results in reporting of incorrect offsets 1377 * if the offset parameter to rbd_diff_iterate2 is not object aligned. 1378 * Work around this bug by rounding down the offset to object boundaries. 1379 * This is OK because we call rbd_diff_iterate2 with whole_object = true. 1380 * However, this workaround only works for non cloned images with default 1381 * striping. 1382 * 1383 * See: https://tracker.ceph.com/issues/53784 1384 */ 1385 1386 /* check if RBD image has non-default striping enabled */ 1387 if (features & RBD_FEATURE_STRIPINGV2) { 1388 return status; 1389 } 1390 1391 #pragma GCC diagnostic push 1392 #pragma GCC diagnostic ignored "-Wdeprecated-declarations" 1393 /* 1394 * check if RBD image is a clone (= has a parent). 1395 * 1396 * rbd_get_parent_info is deprecated from Nautilus onwards, but the 1397 * replacement rbd_get_parent is not present in Luminous and Mimic. 1398 */ 1399 if (rbd_get_parent_info(s->image, NULL, 0, NULL, 0, NULL, 0) != -ENOENT) { 1400 return status; 1401 } 1402 #pragma GCC diagnostic pop 1403 1404 head = req.offs & (s->object_size - 1); 1405 req.offs -= head; 1406 bytes += head; 1407 #endif 1408 1409 r = rbd_diff_iterate2(s->image, NULL, req.offs, bytes, true, true, 1410 qemu_rbd_diff_iterate_cb, &req); 1411 if (r < 0 && r != QEMU_RBD_EXIT_DIFF_ITERATE2) { 1412 return status; 1413 } 1414 assert(req.bytes <= bytes); 1415 if (!req.exists) { 1416 if (r == 0) { 1417 /* 1418 * rbd_diff_iterate2 does not invoke callbacks for unallocated 1419 * areas. This here catches the case where no callback was 1420 * invoked at all (req.bytes == 0). 1421 */ 1422 assert(req.bytes == 0); 1423 req.bytes = bytes; 1424 } 1425 status = BDRV_BLOCK_ZERO | BDRV_BLOCK_OFFSET_VALID; 1426 } 1427 1428 assert(req.bytes > head); 1429 *pnum = req.bytes - head; 1430 return status; 1431 } 1432 1433 static int64_t qemu_rbd_getlength(BlockDriverState *bs) 1434 { 1435 BDRVRBDState *s = bs->opaque; 1436 int r; 1437 1438 r = rbd_get_size(s->image, &s->image_size); 1439 if (r < 0) { 1440 return r; 1441 } 1442 1443 return s->image_size; 1444 } 1445 1446 static int coroutine_fn qemu_rbd_co_truncate(BlockDriverState *bs, 1447 int64_t offset, 1448 bool exact, 1449 PreallocMode prealloc, 1450 BdrvRequestFlags flags, 1451 Error **errp) 1452 { 1453 int r; 1454 1455 if (prealloc != PREALLOC_MODE_OFF) { 1456 error_setg(errp, "Unsupported preallocation mode '%s'", 1457 PreallocMode_str(prealloc)); 1458 return -ENOTSUP; 1459 } 1460 1461 r = qemu_rbd_resize(bs, offset); 1462 if (r < 0) { 1463 error_setg_errno(errp, -r, "Failed to resize file"); 1464 return r; 1465 } 1466 1467 return 0; 1468 } 1469 1470 static int qemu_rbd_snap_create(BlockDriverState *bs, 1471 QEMUSnapshotInfo *sn_info) 1472 { 1473 BDRVRBDState *s = bs->opaque; 1474 int r; 1475 1476 if (sn_info->name[0] == '\0') { 1477 return -EINVAL; /* we need a name for rbd snapshots */ 1478 } 1479 1480 /* 1481 * rbd snapshots are using the name as the user controlled unique identifier 1482 * we can't use the rbd snapid for that purpose, as it can't be set 1483 */ 1484 if (sn_info->id_str[0] != '\0' && 1485 strcmp(sn_info->id_str, sn_info->name) != 0) { 1486 return -EINVAL; 1487 } 1488 1489 if (strlen(sn_info->name) >= sizeof(sn_info->id_str)) { 1490 return -ERANGE; 1491 } 1492 1493 r = rbd_snap_create(s->image, sn_info->name); 1494 if (r < 0) { 1495 error_report("failed to create snap: %s", strerror(-r)); 1496 return r; 1497 } 1498 1499 return 0; 1500 } 1501 1502 static int qemu_rbd_snap_remove(BlockDriverState *bs, 1503 const char *snapshot_id, 1504 const char *snapshot_name, 1505 Error **errp) 1506 { 1507 BDRVRBDState *s = bs->opaque; 1508 int r; 1509 1510 if (!snapshot_name) { 1511 error_setg(errp, "rbd need a valid snapshot name"); 1512 return -EINVAL; 1513 } 1514 1515 /* If snapshot_id is specified, it must be equal to name, see 1516 qemu_rbd_snap_list() */ 1517 if (snapshot_id && strcmp(snapshot_id, snapshot_name)) { 1518 error_setg(errp, 1519 "rbd do not support snapshot id, it should be NULL or " 1520 "equal to snapshot name"); 1521 return -EINVAL; 1522 } 1523 1524 r = rbd_snap_remove(s->image, snapshot_name); 1525 if (r < 0) { 1526 error_setg_errno(errp, -r, "Failed to remove the snapshot"); 1527 } 1528 return r; 1529 } 1530 1531 static int qemu_rbd_snap_rollback(BlockDriverState *bs, 1532 const char *snapshot_name) 1533 { 1534 BDRVRBDState *s = bs->opaque; 1535 1536 return rbd_snap_rollback(s->image, snapshot_name); 1537 } 1538 1539 static int qemu_rbd_snap_list(BlockDriverState *bs, 1540 QEMUSnapshotInfo **psn_tab) 1541 { 1542 BDRVRBDState *s = bs->opaque; 1543 QEMUSnapshotInfo *sn_info, *sn_tab = NULL; 1544 int i, snap_count; 1545 rbd_snap_info_t *snaps; 1546 int max_snaps = RBD_MAX_SNAPS; 1547 1548 do { 1549 snaps = g_new(rbd_snap_info_t, max_snaps); 1550 snap_count = rbd_snap_list(s->image, snaps, &max_snaps); 1551 if (snap_count <= 0) { 1552 g_free(snaps); 1553 } 1554 } while (snap_count == -ERANGE); 1555 1556 if (snap_count <= 0) { 1557 goto done; 1558 } 1559 1560 sn_tab = g_new0(QEMUSnapshotInfo, snap_count); 1561 1562 for (i = 0; i < snap_count; i++) { 1563 const char *snap_name = snaps[i].name; 1564 1565 sn_info = sn_tab + i; 1566 pstrcpy(sn_info->id_str, sizeof(sn_info->id_str), snap_name); 1567 pstrcpy(sn_info->name, sizeof(sn_info->name), snap_name); 1568 1569 sn_info->vm_state_size = snaps[i].size; 1570 sn_info->date_sec = 0; 1571 sn_info->date_nsec = 0; 1572 sn_info->vm_clock_nsec = 0; 1573 } 1574 rbd_snap_list_end(snaps); 1575 g_free(snaps); 1576 1577 done: 1578 *psn_tab = sn_tab; 1579 return snap_count; 1580 } 1581 1582 static void coroutine_fn qemu_rbd_co_invalidate_cache(BlockDriverState *bs, 1583 Error **errp) 1584 { 1585 BDRVRBDState *s = bs->opaque; 1586 int r = rbd_invalidate_cache(s->image); 1587 if (r < 0) { 1588 error_setg_errno(errp, -r, "Failed to invalidate the cache"); 1589 } 1590 } 1591 1592 static QemuOptsList qemu_rbd_create_opts = { 1593 .name = "rbd-create-opts", 1594 .head = QTAILQ_HEAD_INITIALIZER(qemu_rbd_create_opts.head), 1595 .desc = { 1596 { 1597 .name = BLOCK_OPT_SIZE, 1598 .type = QEMU_OPT_SIZE, 1599 .help = "Virtual disk size" 1600 }, 1601 { 1602 .name = BLOCK_OPT_CLUSTER_SIZE, 1603 .type = QEMU_OPT_SIZE, 1604 .help = "RBD object size" 1605 }, 1606 { 1607 .name = "password-secret", 1608 .type = QEMU_OPT_STRING, 1609 .help = "ID of secret providing the password", 1610 }, 1611 { 1612 .name = "encrypt.format", 1613 .type = QEMU_OPT_STRING, 1614 .help = "Encrypt the image, format choices: 'luks', 'luks2'", 1615 }, 1616 { 1617 .name = "encrypt.cipher-alg", 1618 .type = QEMU_OPT_STRING, 1619 .help = "Name of encryption cipher algorithm" 1620 " (allowed values: aes-128, aes-256)", 1621 }, 1622 { 1623 .name = "encrypt.key-secret", 1624 .type = QEMU_OPT_STRING, 1625 .help = "ID of secret providing LUKS passphrase", 1626 }, 1627 { /* end of list */ } 1628 } 1629 }; 1630 1631 static const char *const qemu_rbd_strong_runtime_opts[] = { 1632 "pool", 1633 "namespace", 1634 "image", 1635 "conf", 1636 "snapshot", 1637 "user", 1638 "server.", 1639 "password-secret", 1640 1641 NULL 1642 }; 1643 1644 static BlockDriver bdrv_rbd = { 1645 .format_name = "rbd", 1646 .instance_size = sizeof(BDRVRBDState), 1647 .bdrv_parse_filename = qemu_rbd_parse_filename, 1648 .bdrv_file_open = qemu_rbd_open, 1649 .bdrv_close = qemu_rbd_close, 1650 .bdrv_reopen_prepare = qemu_rbd_reopen_prepare, 1651 .bdrv_co_create = qemu_rbd_co_create, 1652 .bdrv_co_create_opts = qemu_rbd_co_create_opts, 1653 .bdrv_has_zero_init = bdrv_has_zero_init_1, 1654 .bdrv_get_info = qemu_rbd_getinfo, 1655 .bdrv_get_specific_info = qemu_rbd_get_specific_info, 1656 .create_opts = &qemu_rbd_create_opts, 1657 .bdrv_getlength = qemu_rbd_getlength, 1658 .bdrv_co_truncate = qemu_rbd_co_truncate, 1659 .protocol_name = "rbd", 1660 1661 .bdrv_co_preadv = qemu_rbd_co_preadv, 1662 .bdrv_co_pwritev = qemu_rbd_co_pwritev, 1663 .bdrv_co_flush_to_disk = qemu_rbd_co_flush, 1664 .bdrv_co_pdiscard = qemu_rbd_co_pdiscard, 1665 #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES 1666 .bdrv_co_pwrite_zeroes = qemu_rbd_co_pwrite_zeroes, 1667 #endif 1668 .bdrv_co_block_status = qemu_rbd_co_block_status, 1669 1670 .bdrv_snapshot_create = qemu_rbd_snap_create, 1671 .bdrv_snapshot_delete = qemu_rbd_snap_remove, 1672 .bdrv_snapshot_list = qemu_rbd_snap_list, 1673 .bdrv_snapshot_goto = qemu_rbd_snap_rollback, 1674 .bdrv_co_invalidate_cache = qemu_rbd_co_invalidate_cache, 1675 1676 .strong_runtime_opts = qemu_rbd_strong_runtime_opts, 1677 }; 1678 1679 static void bdrv_rbd_init(void) 1680 { 1681 bdrv_register(&bdrv_rbd); 1682 } 1683 1684 block_init(bdrv_rbd_init); 1685