1 /* 2 * QEMU Block driver for NBD 3 * 4 * Copyright (c) 2019 Virtuozzo International GmbH. 5 * Copyright (C) 2016 Red Hat, Inc. 6 * Copyright (C) 2008 Bull S.A.S. 7 * Author: Laurent Vivier <Laurent.Vivier@bull.net> 8 * 9 * Some parts: 10 * Copyright (C) 2007 Anthony Liguori <anthony@codemonkey.ws> 11 * 12 * Permission is hereby granted, free of charge, to any person obtaining a copy 13 * of this software and associated documentation files (the "Software"), to deal 14 * in the Software without restriction, including without limitation the rights 15 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 16 * copies of the Software, and to permit persons to whom the Software is 17 * furnished to do so, subject to the following conditions: 18 * 19 * The above copyright notice and this permission notice shall be included in 20 * all copies or substantial portions of the Software. 21 * 22 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 23 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 24 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 25 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 26 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 27 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 28 * THE SOFTWARE. 29 */ 30 31 #include "qemu/osdep.h" 32 33 #include "trace.h" 34 #include "qemu/uri.h" 35 #include "qemu/option.h" 36 #include "qemu/cutils.h" 37 #include "qemu/main-loop.h" 38 #include "qemu/atomic.h" 39 40 #include "qapi/qapi-visit-sockets.h" 41 #include "qapi/qmp/qstring.h" 42 #include "qapi/clone-visitor.h" 43 44 #include "block/qdict.h" 45 #include "block/nbd.h" 46 #include "block/block_int.h" 47 #include "block/coroutines.h" 48 49 #include "qemu/yank.h" 50 51 #define EN_OPTSTR ":exportname=" 52 #define MAX_NBD_REQUESTS 16 53 54 #define HANDLE_TO_INDEX(bs, handle) ((handle) ^ (uint64_t)(intptr_t)(bs)) 55 #define INDEX_TO_HANDLE(bs, index) ((index) ^ (uint64_t)(intptr_t)(bs)) 56 57 typedef struct { 58 Coroutine *coroutine; 59 uint64_t offset; /* original offset of the request */ 60 bool receiving; /* waiting for connection_co? */ 61 } NBDClientRequest; 62 63 typedef enum NBDClientState { 64 NBD_CLIENT_CONNECTING_WAIT, 65 NBD_CLIENT_CONNECTING_NOWAIT, 66 NBD_CLIENT_CONNECTED, 67 NBD_CLIENT_QUIT 68 } NBDClientState; 69 70 typedef struct BDRVNBDState { 71 QIOChannel *ioc; /* The current I/O channel */ 72 NBDExportInfo info; 73 74 CoMutex send_mutex; 75 CoQueue free_sema; 76 Coroutine *connection_co; 77 Coroutine *teardown_co; 78 QemuCoSleep reconnect_sleep; 79 bool drained; 80 bool wait_drained_end; 81 int in_flight; 82 NBDClientState state; 83 bool wait_in_flight; 84 85 QEMUTimer *reconnect_delay_timer; 86 87 NBDClientRequest requests[MAX_NBD_REQUESTS]; 88 NBDReply reply; 89 BlockDriverState *bs; 90 91 /* Connection parameters */ 92 uint32_t reconnect_delay; 93 SocketAddress *saddr; 94 char *export, *tlscredsid; 95 QCryptoTLSCreds *tlscreds; 96 const char *hostname; 97 char *x_dirty_bitmap; 98 bool alloc_depth; 99 100 NBDClientConnection *conn; 101 } BDRVNBDState; 102 103 static void nbd_yank(void *opaque); 104 105 static void nbd_clear_bdrvstate(BlockDriverState *bs) 106 { 107 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 108 109 nbd_client_connection_release(s->conn); 110 s->conn = NULL; 111 112 yank_unregister_instance(BLOCKDEV_YANK_INSTANCE(bs->node_name)); 113 114 object_unref(OBJECT(s->tlscreds)); 115 qapi_free_SocketAddress(s->saddr); 116 s->saddr = NULL; 117 g_free(s->export); 118 s->export = NULL; 119 g_free(s->tlscredsid); 120 s->tlscredsid = NULL; 121 g_free(s->x_dirty_bitmap); 122 s->x_dirty_bitmap = NULL; 123 } 124 125 static bool nbd_client_connected(BDRVNBDState *s) 126 { 127 return qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTED; 128 } 129 130 static void nbd_channel_error(BDRVNBDState *s, int ret) 131 { 132 if (ret == -EIO) { 133 if (nbd_client_connected(s)) { 134 s->state = s->reconnect_delay ? NBD_CLIENT_CONNECTING_WAIT : 135 NBD_CLIENT_CONNECTING_NOWAIT; 136 } 137 } else { 138 if (nbd_client_connected(s)) { 139 qio_channel_shutdown(s->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL); 140 } 141 s->state = NBD_CLIENT_QUIT; 142 } 143 } 144 145 static void nbd_recv_coroutines_wake_all(BDRVNBDState *s) 146 { 147 int i; 148 149 for (i = 0; i < MAX_NBD_REQUESTS; i++) { 150 NBDClientRequest *req = &s->requests[i]; 151 152 if (req->coroutine && req->receiving) { 153 req->receiving = false; 154 aio_co_wake(req->coroutine); 155 } 156 } 157 } 158 159 static void reconnect_delay_timer_del(BDRVNBDState *s) 160 { 161 if (s->reconnect_delay_timer) { 162 timer_free(s->reconnect_delay_timer); 163 s->reconnect_delay_timer = NULL; 164 } 165 } 166 167 static void reconnect_delay_timer_cb(void *opaque) 168 { 169 BDRVNBDState *s = opaque; 170 171 if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) { 172 s->state = NBD_CLIENT_CONNECTING_NOWAIT; 173 while (qemu_co_enter_next(&s->free_sema, NULL)) { 174 /* Resume all queued requests */ 175 } 176 } 177 178 reconnect_delay_timer_del(s); 179 } 180 181 static void reconnect_delay_timer_init(BDRVNBDState *s, uint64_t expire_time_ns) 182 { 183 if (qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTING_WAIT) { 184 return; 185 } 186 187 assert(!s->reconnect_delay_timer); 188 s->reconnect_delay_timer = aio_timer_new(bdrv_get_aio_context(s->bs), 189 QEMU_CLOCK_REALTIME, 190 SCALE_NS, 191 reconnect_delay_timer_cb, s); 192 timer_mod(s->reconnect_delay_timer, expire_time_ns); 193 } 194 195 static void nbd_client_detach_aio_context(BlockDriverState *bs) 196 { 197 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 198 199 /* Timer is deleted in nbd_client_co_drain_begin() */ 200 assert(!s->reconnect_delay_timer); 201 /* 202 * If reconnect is in progress we may have no ->ioc. It will be 203 * re-instantiated in the proper aio context once the connection is 204 * reestablished. 205 */ 206 if (s->ioc) { 207 qio_channel_detach_aio_context(QIO_CHANNEL(s->ioc)); 208 } 209 } 210 211 static void nbd_client_attach_aio_context_bh(void *opaque) 212 { 213 BlockDriverState *bs = opaque; 214 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 215 216 if (s->connection_co) { 217 /* 218 * The node is still drained, so we know the coroutine has yielded in 219 * nbd_read_eof(), the only place where bs->in_flight can reach 0, or 220 * it is entered for the first time. Both places are safe for entering 221 * the coroutine. 222 */ 223 qemu_aio_coroutine_enter(bs->aio_context, s->connection_co); 224 } 225 bdrv_dec_in_flight(bs); 226 } 227 228 static void nbd_client_attach_aio_context(BlockDriverState *bs, 229 AioContext *new_context) 230 { 231 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 232 233 /* 234 * s->connection_co is either yielded from nbd_receive_reply or from 235 * nbd_co_reconnect_loop() 236 */ 237 if (nbd_client_connected(s)) { 238 qio_channel_attach_aio_context(QIO_CHANNEL(s->ioc), new_context); 239 } 240 241 bdrv_inc_in_flight(bs); 242 243 /* 244 * Need to wait here for the BH to run because the BH must run while the 245 * node is still drained. 246 */ 247 aio_wait_bh_oneshot(new_context, nbd_client_attach_aio_context_bh, bs); 248 } 249 250 static void coroutine_fn nbd_client_co_drain_begin(BlockDriverState *bs) 251 { 252 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 253 254 s->drained = true; 255 qemu_co_sleep_wake(&s->reconnect_sleep); 256 257 nbd_co_establish_connection_cancel(s->conn); 258 259 reconnect_delay_timer_del(s); 260 261 if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) { 262 s->state = NBD_CLIENT_CONNECTING_NOWAIT; 263 qemu_co_queue_restart_all(&s->free_sema); 264 } 265 } 266 267 static void coroutine_fn nbd_client_co_drain_end(BlockDriverState *bs) 268 { 269 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 270 271 s->drained = false; 272 if (s->wait_drained_end) { 273 s->wait_drained_end = false; 274 aio_co_wake(s->connection_co); 275 } 276 } 277 278 279 static void nbd_teardown_connection(BlockDriverState *bs) 280 { 281 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 282 283 if (s->ioc) { 284 /* finish any pending coroutines */ 285 qio_channel_shutdown(s->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL); 286 } 287 288 s->state = NBD_CLIENT_QUIT; 289 if (s->connection_co) { 290 qemu_co_sleep_wake(&s->reconnect_sleep); 291 nbd_co_establish_connection_cancel(s->conn); 292 } 293 if (qemu_in_coroutine()) { 294 s->teardown_co = qemu_coroutine_self(); 295 /* connection_co resumes us when it terminates */ 296 qemu_coroutine_yield(); 297 s->teardown_co = NULL; 298 } else { 299 BDRV_POLL_WHILE(bs, s->connection_co); 300 } 301 assert(!s->connection_co); 302 } 303 304 static bool nbd_client_connecting(BDRVNBDState *s) 305 { 306 NBDClientState state = qatomic_load_acquire(&s->state); 307 return state == NBD_CLIENT_CONNECTING_WAIT || 308 state == NBD_CLIENT_CONNECTING_NOWAIT; 309 } 310 311 static bool nbd_client_connecting_wait(BDRVNBDState *s) 312 { 313 return qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT; 314 } 315 316 /* 317 * Update @bs with information learned during a completed negotiation process. 318 * Return failure if the server's advertised options are incompatible with the 319 * client's needs. 320 */ 321 static int nbd_handle_updated_info(BlockDriverState *bs, Error **errp) 322 { 323 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 324 int ret; 325 326 if (s->x_dirty_bitmap) { 327 if (!s->info.base_allocation) { 328 error_setg(errp, "requested x-dirty-bitmap %s not found", 329 s->x_dirty_bitmap); 330 return -EINVAL; 331 } 332 if (strcmp(s->x_dirty_bitmap, "qemu:allocation-depth") == 0) { 333 s->alloc_depth = true; 334 } 335 } 336 337 if (s->info.flags & NBD_FLAG_READ_ONLY) { 338 ret = bdrv_apply_auto_read_only(bs, "NBD export is read-only", errp); 339 if (ret < 0) { 340 return ret; 341 } 342 } 343 344 if (s->info.flags & NBD_FLAG_SEND_FUA) { 345 bs->supported_write_flags = BDRV_REQ_FUA; 346 bs->supported_zero_flags |= BDRV_REQ_FUA; 347 } 348 349 if (s->info.flags & NBD_FLAG_SEND_WRITE_ZEROES) { 350 bs->supported_zero_flags |= BDRV_REQ_MAY_UNMAP; 351 if (s->info.flags & NBD_FLAG_SEND_FAST_ZERO) { 352 bs->supported_zero_flags |= BDRV_REQ_NO_FALLBACK; 353 } 354 } 355 356 trace_nbd_client_handshake_success(s->export); 357 358 return 0; 359 } 360 361 int coroutine_fn nbd_co_do_establish_connection(BlockDriverState *bs, 362 Error **errp) 363 { 364 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 365 int ret; 366 367 assert(!s->ioc); 368 369 s->ioc = nbd_co_establish_connection(s->conn, &s->info, true, errp); 370 if (!s->ioc) { 371 return -ECONNREFUSED; 372 } 373 374 yank_register_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), nbd_yank, 375 bs); 376 377 ret = nbd_handle_updated_info(s->bs, NULL); 378 if (ret < 0) { 379 /* 380 * We have connected, but must fail for other reasons. 381 * Send NBD_CMD_DISC as a courtesy to the server. 382 */ 383 NBDRequest request = { .type = NBD_CMD_DISC }; 384 385 nbd_send_request(s->ioc, &request); 386 387 yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), 388 nbd_yank, bs); 389 object_unref(OBJECT(s->ioc)); 390 s->ioc = NULL; 391 392 return ret; 393 } 394 395 qio_channel_set_blocking(s->ioc, false, NULL); 396 qio_channel_attach_aio_context(s->ioc, bdrv_get_aio_context(bs)); 397 398 /* successfully connected */ 399 s->state = NBD_CLIENT_CONNECTED; 400 qemu_co_queue_restart_all(&s->free_sema); 401 402 return 0; 403 } 404 405 static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState *s) 406 { 407 if (!nbd_client_connecting(s)) { 408 return; 409 } 410 411 /* Wait for completion of all in-flight requests */ 412 413 qemu_co_mutex_lock(&s->send_mutex); 414 415 while (s->in_flight > 0) { 416 qemu_co_mutex_unlock(&s->send_mutex); 417 nbd_recv_coroutines_wake_all(s); 418 s->wait_in_flight = true; 419 qemu_coroutine_yield(); 420 s->wait_in_flight = false; 421 qemu_co_mutex_lock(&s->send_mutex); 422 } 423 424 qemu_co_mutex_unlock(&s->send_mutex); 425 426 if (!nbd_client_connecting(s)) { 427 return; 428 } 429 430 /* 431 * Now we are sure that nobody is accessing the channel, and no one will 432 * try until we set the state to CONNECTED. 433 */ 434 435 /* Finalize previous connection if any */ 436 if (s->ioc) { 437 qio_channel_detach_aio_context(QIO_CHANNEL(s->ioc)); 438 yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), 439 nbd_yank, s->bs); 440 object_unref(OBJECT(s->ioc)); 441 s->ioc = NULL; 442 } 443 444 nbd_co_do_establish_connection(s->bs, NULL); 445 } 446 447 static coroutine_fn void nbd_co_reconnect_loop(BDRVNBDState *s) 448 { 449 uint64_t timeout = 1 * NANOSECONDS_PER_SECOND; 450 uint64_t max_timeout = 16 * NANOSECONDS_PER_SECOND; 451 452 if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) { 453 reconnect_delay_timer_init(s, qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + 454 s->reconnect_delay * NANOSECONDS_PER_SECOND); 455 } 456 457 nbd_reconnect_attempt(s); 458 459 while (nbd_client_connecting(s)) { 460 if (s->drained) { 461 bdrv_dec_in_flight(s->bs); 462 s->wait_drained_end = true; 463 while (s->drained) { 464 /* 465 * We may be entered once from nbd_client_attach_aio_context_bh 466 * and then from nbd_client_co_drain_end. So here is a loop. 467 */ 468 qemu_coroutine_yield(); 469 } 470 bdrv_inc_in_flight(s->bs); 471 } else { 472 qemu_co_sleep_ns_wakeable(&s->reconnect_sleep, 473 QEMU_CLOCK_REALTIME, timeout); 474 if (s->drained) { 475 continue; 476 } 477 if (timeout < max_timeout) { 478 timeout *= 2; 479 } 480 } 481 482 nbd_reconnect_attempt(s); 483 } 484 485 reconnect_delay_timer_del(s); 486 } 487 488 static coroutine_fn void nbd_connection_entry(void *opaque) 489 { 490 BDRVNBDState *s = opaque; 491 uint64_t i; 492 int ret = 0; 493 Error *local_err = NULL; 494 495 while (qatomic_load_acquire(&s->state) != NBD_CLIENT_QUIT) { 496 /* 497 * The NBD client can only really be considered idle when it has 498 * yielded from qio_channel_readv_all_eof(), waiting for data. This is 499 * the point where the additional scheduled coroutine entry happens 500 * after nbd_client_attach_aio_context(). 501 * 502 * Therefore we keep an additional in_flight reference all the time and 503 * only drop it temporarily here. 504 */ 505 506 if (nbd_client_connecting(s)) { 507 nbd_co_reconnect_loop(s); 508 } 509 510 if (!nbd_client_connected(s)) { 511 continue; 512 } 513 514 assert(s->reply.handle == 0); 515 ret = nbd_receive_reply(s->bs, s->ioc, &s->reply, &local_err); 516 517 if (local_err) { 518 trace_nbd_read_reply_entry_fail(ret, error_get_pretty(local_err)); 519 error_free(local_err); 520 local_err = NULL; 521 } 522 if (ret <= 0) { 523 nbd_channel_error(s, ret ? ret : -EIO); 524 continue; 525 } 526 527 /* 528 * There's no need for a mutex on the receive side, because the 529 * handler acts as a synchronization point and ensures that only 530 * one coroutine is called until the reply finishes. 531 */ 532 i = HANDLE_TO_INDEX(s, s->reply.handle); 533 if (i >= MAX_NBD_REQUESTS || 534 !s->requests[i].coroutine || 535 !s->requests[i].receiving || 536 (nbd_reply_is_structured(&s->reply) && !s->info.structured_reply)) 537 { 538 nbd_channel_error(s, -EINVAL); 539 continue; 540 } 541 542 /* 543 * We're woken up again by the request itself. Note that there 544 * is no race between yielding and reentering connection_co. This 545 * is because: 546 * 547 * - if the request runs on the same AioContext, it is only 548 * entered after we yield 549 * 550 * - if the request runs on a different AioContext, reentering 551 * connection_co happens through a bottom half, which can only 552 * run after we yield. 553 */ 554 s->requests[i].receiving = false; 555 aio_co_wake(s->requests[i].coroutine); 556 qemu_coroutine_yield(); 557 } 558 559 qemu_co_queue_restart_all(&s->free_sema); 560 nbd_recv_coroutines_wake_all(s); 561 bdrv_dec_in_flight(s->bs); 562 563 s->connection_co = NULL; 564 if (s->ioc) { 565 qio_channel_detach_aio_context(QIO_CHANNEL(s->ioc)); 566 yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), 567 nbd_yank, s->bs); 568 object_unref(OBJECT(s->ioc)); 569 s->ioc = NULL; 570 } 571 572 if (s->teardown_co) { 573 aio_co_wake(s->teardown_co); 574 } 575 aio_wait_kick(); 576 } 577 578 static int nbd_co_send_request(BlockDriverState *bs, 579 NBDRequest *request, 580 QEMUIOVector *qiov) 581 { 582 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 583 int rc, i = -1; 584 585 qemu_co_mutex_lock(&s->send_mutex); 586 while (s->in_flight == MAX_NBD_REQUESTS || nbd_client_connecting_wait(s)) { 587 qemu_co_queue_wait(&s->free_sema, &s->send_mutex); 588 } 589 590 if (!nbd_client_connected(s)) { 591 rc = -EIO; 592 goto err; 593 } 594 595 s->in_flight++; 596 597 for (i = 0; i < MAX_NBD_REQUESTS; i++) { 598 if (s->requests[i].coroutine == NULL) { 599 break; 600 } 601 } 602 603 g_assert(qemu_in_coroutine()); 604 assert(i < MAX_NBD_REQUESTS); 605 606 s->requests[i].coroutine = qemu_coroutine_self(); 607 s->requests[i].offset = request->from; 608 s->requests[i].receiving = false; 609 610 request->handle = INDEX_TO_HANDLE(s, i); 611 612 assert(s->ioc); 613 614 if (qiov) { 615 qio_channel_set_cork(s->ioc, true); 616 rc = nbd_send_request(s->ioc, request); 617 if (nbd_client_connected(s) && rc >= 0) { 618 if (qio_channel_writev_all(s->ioc, qiov->iov, qiov->niov, 619 NULL) < 0) { 620 rc = -EIO; 621 } 622 } else if (rc >= 0) { 623 rc = -EIO; 624 } 625 qio_channel_set_cork(s->ioc, false); 626 } else { 627 rc = nbd_send_request(s->ioc, request); 628 } 629 630 err: 631 if (rc < 0) { 632 nbd_channel_error(s, rc); 633 if (i != -1) { 634 s->requests[i].coroutine = NULL; 635 s->in_flight--; 636 } 637 if (s->in_flight == 0 && s->wait_in_flight) { 638 aio_co_wake(s->connection_co); 639 } else { 640 qemu_co_queue_next(&s->free_sema); 641 } 642 } 643 qemu_co_mutex_unlock(&s->send_mutex); 644 return rc; 645 } 646 647 static inline uint16_t payload_advance16(uint8_t **payload) 648 { 649 *payload += 2; 650 return lduw_be_p(*payload - 2); 651 } 652 653 static inline uint32_t payload_advance32(uint8_t **payload) 654 { 655 *payload += 4; 656 return ldl_be_p(*payload - 4); 657 } 658 659 static inline uint64_t payload_advance64(uint8_t **payload) 660 { 661 *payload += 8; 662 return ldq_be_p(*payload - 8); 663 } 664 665 static int nbd_parse_offset_hole_payload(BDRVNBDState *s, 666 NBDStructuredReplyChunk *chunk, 667 uint8_t *payload, uint64_t orig_offset, 668 QEMUIOVector *qiov, Error **errp) 669 { 670 uint64_t offset; 671 uint32_t hole_size; 672 673 if (chunk->length != sizeof(offset) + sizeof(hole_size)) { 674 error_setg(errp, "Protocol error: invalid payload for " 675 "NBD_REPLY_TYPE_OFFSET_HOLE"); 676 return -EINVAL; 677 } 678 679 offset = payload_advance64(&payload); 680 hole_size = payload_advance32(&payload); 681 682 if (!hole_size || offset < orig_offset || hole_size > qiov->size || 683 offset > orig_offset + qiov->size - hole_size) { 684 error_setg(errp, "Protocol error: server sent chunk exceeding requested" 685 " region"); 686 return -EINVAL; 687 } 688 if (s->info.min_block && 689 !QEMU_IS_ALIGNED(hole_size, s->info.min_block)) { 690 trace_nbd_structured_read_compliance("hole"); 691 } 692 693 qemu_iovec_memset(qiov, offset - orig_offset, 0, hole_size); 694 695 return 0; 696 } 697 698 /* 699 * nbd_parse_blockstatus_payload 700 * Based on our request, we expect only one extent in reply, for the 701 * base:allocation context. 702 */ 703 static int nbd_parse_blockstatus_payload(BDRVNBDState *s, 704 NBDStructuredReplyChunk *chunk, 705 uint8_t *payload, uint64_t orig_length, 706 NBDExtent *extent, Error **errp) 707 { 708 uint32_t context_id; 709 710 /* The server succeeded, so it must have sent [at least] one extent */ 711 if (chunk->length < sizeof(context_id) + sizeof(*extent)) { 712 error_setg(errp, "Protocol error: invalid payload for " 713 "NBD_REPLY_TYPE_BLOCK_STATUS"); 714 return -EINVAL; 715 } 716 717 context_id = payload_advance32(&payload); 718 if (s->info.context_id != context_id) { 719 error_setg(errp, "Protocol error: unexpected context id %d for " 720 "NBD_REPLY_TYPE_BLOCK_STATUS, when negotiated context " 721 "id is %d", context_id, 722 s->info.context_id); 723 return -EINVAL; 724 } 725 726 extent->length = payload_advance32(&payload); 727 extent->flags = payload_advance32(&payload); 728 729 if (extent->length == 0) { 730 error_setg(errp, "Protocol error: server sent status chunk with " 731 "zero length"); 732 return -EINVAL; 733 } 734 735 /* 736 * A server sending unaligned block status is in violation of the 737 * protocol, but as qemu-nbd 3.1 is such a server (at least for 738 * POSIX files that are not a multiple of 512 bytes, since qemu 739 * rounds files up to 512-byte multiples but lseek(SEEK_HOLE) 740 * still sees an implicit hole beyond the real EOF), it's nicer to 741 * work around the misbehaving server. If the request included 742 * more than the final unaligned block, truncate it back to an 743 * aligned result; if the request was only the final block, round 744 * up to the full block and change the status to fully-allocated 745 * (always a safe status, even if it loses information). 746 */ 747 if (s->info.min_block && !QEMU_IS_ALIGNED(extent->length, 748 s->info.min_block)) { 749 trace_nbd_parse_blockstatus_compliance("extent length is unaligned"); 750 if (extent->length > s->info.min_block) { 751 extent->length = QEMU_ALIGN_DOWN(extent->length, 752 s->info.min_block); 753 } else { 754 extent->length = s->info.min_block; 755 extent->flags = 0; 756 } 757 } 758 759 /* 760 * We used NBD_CMD_FLAG_REQ_ONE, so the server should not have 761 * sent us any more than one extent, nor should it have included 762 * status beyond our request in that extent. However, it's easy 763 * enough to ignore the server's noncompliance without killing the 764 * connection; just ignore trailing extents, and clamp things to 765 * the length of our request. 766 */ 767 if (chunk->length > sizeof(context_id) + sizeof(*extent)) { 768 trace_nbd_parse_blockstatus_compliance("more than one extent"); 769 } 770 if (extent->length > orig_length) { 771 extent->length = orig_length; 772 trace_nbd_parse_blockstatus_compliance("extent length too large"); 773 } 774 775 /* 776 * HACK: if we are using x-dirty-bitmaps to access 777 * qemu:allocation-depth, treat all depths > 2 the same as 2, 778 * since nbd_client_co_block_status is only expecting the low two 779 * bits to be set. 780 */ 781 if (s->alloc_depth && extent->flags > 2) { 782 extent->flags = 2; 783 } 784 785 return 0; 786 } 787 788 /* 789 * nbd_parse_error_payload 790 * on success @errp contains message describing nbd error reply 791 */ 792 static int nbd_parse_error_payload(NBDStructuredReplyChunk *chunk, 793 uint8_t *payload, int *request_ret, 794 Error **errp) 795 { 796 uint32_t error; 797 uint16_t message_size; 798 799 assert(chunk->type & (1 << 15)); 800 801 if (chunk->length < sizeof(error) + sizeof(message_size)) { 802 error_setg(errp, 803 "Protocol error: invalid payload for structured error"); 804 return -EINVAL; 805 } 806 807 error = nbd_errno_to_system_errno(payload_advance32(&payload)); 808 if (error == 0) { 809 error_setg(errp, "Protocol error: server sent structured error chunk " 810 "with error = 0"); 811 return -EINVAL; 812 } 813 814 *request_ret = -error; 815 message_size = payload_advance16(&payload); 816 817 if (message_size > chunk->length - sizeof(error) - sizeof(message_size)) { 818 error_setg(errp, "Protocol error: server sent structured error chunk " 819 "with incorrect message size"); 820 return -EINVAL; 821 } 822 823 /* TODO: Add a trace point to mention the server complaint */ 824 825 /* TODO handle ERROR_OFFSET */ 826 827 return 0; 828 } 829 830 static int nbd_co_receive_offset_data_payload(BDRVNBDState *s, 831 uint64_t orig_offset, 832 QEMUIOVector *qiov, Error **errp) 833 { 834 QEMUIOVector sub_qiov; 835 uint64_t offset; 836 size_t data_size; 837 int ret; 838 NBDStructuredReplyChunk *chunk = &s->reply.structured; 839 840 assert(nbd_reply_is_structured(&s->reply)); 841 842 /* The NBD spec requires at least one byte of payload */ 843 if (chunk->length <= sizeof(offset)) { 844 error_setg(errp, "Protocol error: invalid payload for " 845 "NBD_REPLY_TYPE_OFFSET_DATA"); 846 return -EINVAL; 847 } 848 849 if (nbd_read64(s->ioc, &offset, "OFFSET_DATA offset", errp) < 0) { 850 return -EIO; 851 } 852 853 data_size = chunk->length - sizeof(offset); 854 assert(data_size); 855 if (offset < orig_offset || data_size > qiov->size || 856 offset > orig_offset + qiov->size - data_size) { 857 error_setg(errp, "Protocol error: server sent chunk exceeding requested" 858 " region"); 859 return -EINVAL; 860 } 861 if (s->info.min_block && !QEMU_IS_ALIGNED(data_size, s->info.min_block)) { 862 trace_nbd_structured_read_compliance("data"); 863 } 864 865 qemu_iovec_init(&sub_qiov, qiov->niov); 866 qemu_iovec_concat(&sub_qiov, qiov, offset - orig_offset, data_size); 867 ret = qio_channel_readv_all(s->ioc, sub_qiov.iov, sub_qiov.niov, errp); 868 qemu_iovec_destroy(&sub_qiov); 869 870 return ret < 0 ? -EIO : 0; 871 } 872 873 #define NBD_MAX_MALLOC_PAYLOAD 1000 874 static coroutine_fn int nbd_co_receive_structured_payload( 875 BDRVNBDState *s, void **payload, Error **errp) 876 { 877 int ret; 878 uint32_t len; 879 880 assert(nbd_reply_is_structured(&s->reply)); 881 882 len = s->reply.structured.length; 883 884 if (len == 0) { 885 return 0; 886 } 887 888 if (payload == NULL) { 889 error_setg(errp, "Unexpected structured payload"); 890 return -EINVAL; 891 } 892 893 if (len > NBD_MAX_MALLOC_PAYLOAD) { 894 error_setg(errp, "Payload too large"); 895 return -EINVAL; 896 } 897 898 *payload = g_new(char, len); 899 ret = nbd_read(s->ioc, *payload, len, "structured payload", errp); 900 if (ret < 0) { 901 g_free(*payload); 902 *payload = NULL; 903 return ret; 904 } 905 906 return 0; 907 } 908 909 /* 910 * nbd_co_do_receive_one_chunk 911 * for simple reply: 912 * set request_ret to received reply error 913 * if qiov is not NULL: read payload to @qiov 914 * for structured reply chunk: 915 * if error chunk: read payload, set @request_ret, do not set @payload 916 * else if offset_data chunk: read payload data to @qiov, do not set @payload 917 * else: read payload to @payload 918 * 919 * If function fails, @errp contains corresponding error message, and the 920 * connection with the server is suspect. If it returns 0, then the 921 * transaction succeeded (although @request_ret may be a negative errno 922 * corresponding to the server's error reply), and errp is unchanged. 923 */ 924 static coroutine_fn int nbd_co_do_receive_one_chunk( 925 BDRVNBDState *s, uint64_t handle, bool only_structured, 926 int *request_ret, QEMUIOVector *qiov, void **payload, Error **errp) 927 { 928 int ret; 929 int i = HANDLE_TO_INDEX(s, handle); 930 void *local_payload = NULL; 931 NBDStructuredReplyChunk *chunk; 932 933 if (payload) { 934 *payload = NULL; 935 } 936 *request_ret = 0; 937 938 /* Wait until we're woken up by nbd_connection_entry. */ 939 s->requests[i].receiving = true; 940 qemu_coroutine_yield(); 941 assert(!s->requests[i].receiving); 942 if (!nbd_client_connected(s)) { 943 error_setg(errp, "Connection closed"); 944 return -EIO; 945 } 946 assert(s->ioc); 947 948 assert(s->reply.handle == handle); 949 950 if (nbd_reply_is_simple(&s->reply)) { 951 if (only_structured) { 952 error_setg(errp, "Protocol error: simple reply when structured " 953 "reply chunk was expected"); 954 return -EINVAL; 955 } 956 957 *request_ret = -nbd_errno_to_system_errno(s->reply.simple.error); 958 if (*request_ret < 0 || !qiov) { 959 return 0; 960 } 961 962 return qio_channel_readv_all(s->ioc, qiov->iov, qiov->niov, 963 errp) < 0 ? -EIO : 0; 964 } 965 966 /* handle structured reply chunk */ 967 assert(s->info.structured_reply); 968 chunk = &s->reply.structured; 969 970 if (chunk->type == NBD_REPLY_TYPE_NONE) { 971 if (!(chunk->flags & NBD_REPLY_FLAG_DONE)) { 972 error_setg(errp, "Protocol error: NBD_REPLY_TYPE_NONE chunk without" 973 " NBD_REPLY_FLAG_DONE flag set"); 974 return -EINVAL; 975 } 976 if (chunk->length) { 977 error_setg(errp, "Protocol error: NBD_REPLY_TYPE_NONE chunk with" 978 " nonzero length"); 979 return -EINVAL; 980 } 981 return 0; 982 } 983 984 if (chunk->type == NBD_REPLY_TYPE_OFFSET_DATA) { 985 if (!qiov) { 986 error_setg(errp, "Unexpected NBD_REPLY_TYPE_OFFSET_DATA chunk"); 987 return -EINVAL; 988 } 989 990 return nbd_co_receive_offset_data_payload(s, s->requests[i].offset, 991 qiov, errp); 992 } 993 994 if (nbd_reply_type_is_error(chunk->type)) { 995 payload = &local_payload; 996 } 997 998 ret = nbd_co_receive_structured_payload(s, payload, errp); 999 if (ret < 0) { 1000 return ret; 1001 } 1002 1003 if (nbd_reply_type_is_error(chunk->type)) { 1004 ret = nbd_parse_error_payload(chunk, local_payload, request_ret, errp); 1005 g_free(local_payload); 1006 return ret; 1007 } 1008 1009 return 0; 1010 } 1011 1012 /* 1013 * nbd_co_receive_one_chunk 1014 * Read reply, wake up connection_co and set s->quit if needed. 1015 * Return value is a fatal error code or normal nbd reply error code 1016 */ 1017 static coroutine_fn int nbd_co_receive_one_chunk( 1018 BDRVNBDState *s, uint64_t handle, bool only_structured, 1019 int *request_ret, QEMUIOVector *qiov, NBDReply *reply, void **payload, 1020 Error **errp) 1021 { 1022 int ret = nbd_co_do_receive_one_chunk(s, handle, only_structured, 1023 request_ret, qiov, payload, errp); 1024 1025 if (ret < 0) { 1026 memset(reply, 0, sizeof(*reply)); 1027 nbd_channel_error(s, ret); 1028 } else { 1029 /* For assert at loop start in nbd_connection_entry */ 1030 *reply = s->reply; 1031 } 1032 s->reply.handle = 0; 1033 1034 if (s->connection_co && !s->wait_in_flight) { 1035 /* 1036 * We must check s->wait_in_flight, because we may entered by 1037 * nbd_recv_coroutines_wake_all(), in this case we should not 1038 * wake connection_co here, it will woken by last request. 1039 */ 1040 aio_co_wake(s->connection_co); 1041 } 1042 1043 return ret; 1044 } 1045 1046 typedef struct NBDReplyChunkIter { 1047 int ret; 1048 int request_ret; 1049 Error *err; 1050 bool done, only_structured; 1051 } NBDReplyChunkIter; 1052 1053 static void nbd_iter_channel_error(NBDReplyChunkIter *iter, 1054 int ret, Error **local_err) 1055 { 1056 assert(local_err && *local_err); 1057 assert(ret < 0); 1058 1059 if (!iter->ret) { 1060 iter->ret = ret; 1061 error_propagate(&iter->err, *local_err); 1062 } else { 1063 error_free(*local_err); 1064 } 1065 1066 *local_err = NULL; 1067 } 1068 1069 static void nbd_iter_request_error(NBDReplyChunkIter *iter, int ret) 1070 { 1071 assert(ret < 0); 1072 1073 if (!iter->request_ret) { 1074 iter->request_ret = ret; 1075 } 1076 } 1077 1078 /* 1079 * NBD_FOREACH_REPLY_CHUNK 1080 * The pointer stored in @payload requires g_free() to free it. 1081 */ 1082 #define NBD_FOREACH_REPLY_CHUNK(s, iter, handle, structured, \ 1083 qiov, reply, payload) \ 1084 for (iter = (NBDReplyChunkIter) { .only_structured = structured }; \ 1085 nbd_reply_chunk_iter_receive(s, &iter, handle, qiov, reply, payload);) 1086 1087 /* 1088 * nbd_reply_chunk_iter_receive 1089 * The pointer stored in @payload requires g_free() to free it. 1090 */ 1091 static bool nbd_reply_chunk_iter_receive(BDRVNBDState *s, 1092 NBDReplyChunkIter *iter, 1093 uint64_t handle, 1094 QEMUIOVector *qiov, NBDReply *reply, 1095 void **payload) 1096 { 1097 int ret, request_ret; 1098 NBDReply local_reply; 1099 NBDStructuredReplyChunk *chunk; 1100 Error *local_err = NULL; 1101 if (!nbd_client_connected(s)) { 1102 error_setg(&local_err, "Connection closed"); 1103 nbd_iter_channel_error(iter, -EIO, &local_err); 1104 goto break_loop; 1105 } 1106 1107 if (iter->done) { 1108 /* Previous iteration was last. */ 1109 goto break_loop; 1110 } 1111 1112 if (reply == NULL) { 1113 reply = &local_reply; 1114 } 1115 1116 ret = nbd_co_receive_one_chunk(s, handle, iter->only_structured, 1117 &request_ret, qiov, reply, payload, 1118 &local_err); 1119 if (ret < 0) { 1120 nbd_iter_channel_error(iter, ret, &local_err); 1121 } else if (request_ret < 0) { 1122 nbd_iter_request_error(iter, request_ret); 1123 } 1124 1125 /* Do not execute the body of NBD_FOREACH_REPLY_CHUNK for simple reply. */ 1126 if (nbd_reply_is_simple(reply) || !nbd_client_connected(s)) { 1127 goto break_loop; 1128 } 1129 1130 chunk = &reply->structured; 1131 iter->only_structured = true; 1132 1133 if (chunk->type == NBD_REPLY_TYPE_NONE) { 1134 /* NBD_REPLY_FLAG_DONE is already checked in nbd_co_receive_one_chunk */ 1135 assert(chunk->flags & NBD_REPLY_FLAG_DONE); 1136 goto break_loop; 1137 } 1138 1139 if (chunk->flags & NBD_REPLY_FLAG_DONE) { 1140 /* This iteration is last. */ 1141 iter->done = true; 1142 } 1143 1144 /* Execute the loop body */ 1145 return true; 1146 1147 break_loop: 1148 s->requests[HANDLE_TO_INDEX(s, handle)].coroutine = NULL; 1149 1150 qemu_co_mutex_lock(&s->send_mutex); 1151 s->in_flight--; 1152 if (s->in_flight == 0 && s->wait_in_flight) { 1153 aio_co_wake(s->connection_co); 1154 } else { 1155 qemu_co_queue_next(&s->free_sema); 1156 } 1157 qemu_co_mutex_unlock(&s->send_mutex); 1158 1159 return false; 1160 } 1161 1162 static int nbd_co_receive_return_code(BDRVNBDState *s, uint64_t handle, 1163 int *request_ret, Error **errp) 1164 { 1165 NBDReplyChunkIter iter; 1166 1167 NBD_FOREACH_REPLY_CHUNK(s, iter, handle, false, NULL, NULL, NULL) { 1168 /* nbd_reply_chunk_iter_receive does all the work */ 1169 } 1170 1171 error_propagate(errp, iter.err); 1172 *request_ret = iter.request_ret; 1173 return iter.ret; 1174 } 1175 1176 static int nbd_co_receive_cmdread_reply(BDRVNBDState *s, uint64_t handle, 1177 uint64_t offset, QEMUIOVector *qiov, 1178 int *request_ret, Error **errp) 1179 { 1180 NBDReplyChunkIter iter; 1181 NBDReply reply; 1182 void *payload = NULL; 1183 Error *local_err = NULL; 1184 1185 NBD_FOREACH_REPLY_CHUNK(s, iter, handle, s->info.structured_reply, 1186 qiov, &reply, &payload) 1187 { 1188 int ret; 1189 NBDStructuredReplyChunk *chunk = &reply.structured; 1190 1191 assert(nbd_reply_is_structured(&reply)); 1192 1193 switch (chunk->type) { 1194 case NBD_REPLY_TYPE_OFFSET_DATA: 1195 /* 1196 * special cased in nbd_co_receive_one_chunk, data is already 1197 * in qiov 1198 */ 1199 break; 1200 case NBD_REPLY_TYPE_OFFSET_HOLE: 1201 ret = nbd_parse_offset_hole_payload(s, &reply.structured, payload, 1202 offset, qiov, &local_err); 1203 if (ret < 0) { 1204 nbd_channel_error(s, ret); 1205 nbd_iter_channel_error(&iter, ret, &local_err); 1206 } 1207 break; 1208 default: 1209 if (!nbd_reply_type_is_error(chunk->type)) { 1210 /* not allowed reply type */ 1211 nbd_channel_error(s, -EINVAL); 1212 error_setg(&local_err, 1213 "Unexpected reply type: %d (%s) for CMD_READ", 1214 chunk->type, nbd_reply_type_lookup(chunk->type)); 1215 nbd_iter_channel_error(&iter, -EINVAL, &local_err); 1216 } 1217 } 1218 1219 g_free(payload); 1220 payload = NULL; 1221 } 1222 1223 error_propagate(errp, iter.err); 1224 *request_ret = iter.request_ret; 1225 return iter.ret; 1226 } 1227 1228 static int nbd_co_receive_blockstatus_reply(BDRVNBDState *s, 1229 uint64_t handle, uint64_t length, 1230 NBDExtent *extent, 1231 int *request_ret, Error **errp) 1232 { 1233 NBDReplyChunkIter iter; 1234 NBDReply reply; 1235 void *payload = NULL; 1236 Error *local_err = NULL; 1237 bool received = false; 1238 1239 assert(!extent->length); 1240 NBD_FOREACH_REPLY_CHUNK(s, iter, handle, false, NULL, &reply, &payload) { 1241 int ret; 1242 NBDStructuredReplyChunk *chunk = &reply.structured; 1243 1244 assert(nbd_reply_is_structured(&reply)); 1245 1246 switch (chunk->type) { 1247 case NBD_REPLY_TYPE_BLOCK_STATUS: 1248 if (received) { 1249 nbd_channel_error(s, -EINVAL); 1250 error_setg(&local_err, "Several BLOCK_STATUS chunks in reply"); 1251 nbd_iter_channel_error(&iter, -EINVAL, &local_err); 1252 } 1253 received = true; 1254 1255 ret = nbd_parse_blockstatus_payload(s, &reply.structured, 1256 payload, length, extent, 1257 &local_err); 1258 if (ret < 0) { 1259 nbd_channel_error(s, ret); 1260 nbd_iter_channel_error(&iter, ret, &local_err); 1261 } 1262 break; 1263 default: 1264 if (!nbd_reply_type_is_error(chunk->type)) { 1265 nbd_channel_error(s, -EINVAL); 1266 error_setg(&local_err, 1267 "Unexpected reply type: %d (%s) " 1268 "for CMD_BLOCK_STATUS", 1269 chunk->type, nbd_reply_type_lookup(chunk->type)); 1270 nbd_iter_channel_error(&iter, -EINVAL, &local_err); 1271 } 1272 } 1273 1274 g_free(payload); 1275 payload = NULL; 1276 } 1277 1278 if (!extent->length && !iter.request_ret) { 1279 error_setg(&local_err, "Server did not reply with any status extents"); 1280 nbd_iter_channel_error(&iter, -EIO, &local_err); 1281 } 1282 1283 error_propagate(errp, iter.err); 1284 *request_ret = iter.request_ret; 1285 return iter.ret; 1286 } 1287 1288 static int nbd_co_request(BlockDriverState *bs, NBDRequest *request, 1289 QEMUIOVector *write_qiov) 1290 { 1291 int ret, request_ret; 1292 Error *local_err = NULL; 1293 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1294 1295 assert(request->type != NBD_CMD_READ); 1296 if (write_qiov) { 1297 assert(request->type == NBD_CMD_WRITE); 1298 assert(request->len == iov_size(write_qiov->iov, write_qiov->niov)); 1299 } else { 1300 assert(request->type != NBD_CMD_WRITE); 1301 } 1302 1303 do { 1304 ret = nbd_co_send_request(bs, request, write_qiov); 1305 if (ret < 0) { 1306 continue; 1307 } 1308 1309 ret = nbd_co_receive_return_code(s, request->handle, 1310 &request_ret, &local_err); 1311 if (local_err) { 1312 trace_nbd_co_request_fail(request->from, request->len, 1313 request->handle, request->flags, 1314 request->type, 1315 nbd_cmd_lookup(request->type), 1316 ret, error_get_pretty(local_err)); 1317 error_free(local_err); 1318 local_err = NULL; 1319 } 1320 } while (ret < 0 && nbd_client_connecting_wait(s)); 1321 1322 return ret ? ret : request_ret; 1323 } 1324 1325 static int nbd_client_co_preadv(BlockDriverState *bs, uint64_t offset, 1326 uint64_t bytes, QEMUIOVector *qiov, int flags) 1327 { 1328 int ret, request_ret; 1329 Error *local_err = NULL; 1330 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1331 NBDRequest request = { 1332 .type = NBD_CMD_READ, 1333 .from = offset, 1334 .len = bytes, 1335 }; 1336 1337 assert(bytes <= NBD_MAX_BUFFER_SIZE); 1338 assert(!flags); 1339 1340 if (!bytes) { 1341 return 0; 1342 } 1343 /* 1344 * Work around the fact that the block layer doesn't do 1345 * byte-accurate sizing yet - if the read exceeds the server's 1346 * advertised size because the block layer rounded size up, then 1347 * truncate the request to the server and tail-pad with zero. 1348 */ 1349 if (offset >= s->info.size) { 1350 assert(bytes < BDRV_SECTOR_SIZE); 1351 qemu_iovec_memset(qiov, 0, 0, bytes); 1352 return 0; 1353 } 1354 if (offset + bytes > s->info.size) { 1355 uint64_t slop = offset + bytes - s->info.size; 1356 1357 assert(slop < BDRV_SECTOR_SIZE); 1358 qemu_iovec_memset(qiov, bytes - slop, 0, slop); 1359 request.len -= slop; 1360 } 1361 1362 do { 1363 ret = nbd_co_send_request(bs, &request, NULL); 1364 if (ret < 0) { 1365 continue; 1366 } 1367 1368 ret = nbd_co_receive_cmdread_reply(s, request.handle, offset, qiov, 1369 &request_ret, &local_err); 1370 if (local_err) { 1371 trace_nbd_co_request_fail(request.from, request.len, request.handle, 1372 request.flags, request.type, 1373 nbd_cmd_lookup(request.type), 1374 ret, error_get_pretty(local_err)); 1375 error_free(local_err); 1376 local_err = NULL; 1377 } 1378 } while (ret < 0 && nbd_client_connecting_wait(s)); 1379 1380 return ret ? ret : request_ret; 1381 } 1382 1383 static int nbd_client_co_pwritev(BlockDriverState *bs, uint64_t offset, 1384 uint64_t bytes, QEMUIOVector *qiov, int flags) 1385 { 1386 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1387 NBDRequest request = { 1388 .type = NBD_CMD_WRITE, 1389 .from = offset, 1390 .len = bytes, 1391 }; 1392 1393 assert(!(s->info.flags & NBD_FLAG_READ_ONLY)); 1394 if (flags & BDRV_REQ_FUA) { 1395 assert(s->info.flags & NBD_FLAG_SEND_FUA); 1396 request.flags |= NBD_CMD_FLAG_FUA; 1397 } 1398 1399 assert(bytes <= NBD_MAX_BUFFER_SIZE); 1400 1401 if (!bytes) { 1402 return 0; 1403 } 1404 return nbd_co_request(bs, &request, qiov); 1405 } 1406 1407 static int nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, 1408 int bytes, BdrvRequestFlags flags) 1409 { 1410 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1411 NBDRequest request = { 1412 .type = NBD_CMD_WRITE_ZEROES, 1413 .from = offset, 1414 .len = bytes, 1415 }; 1416 1417 assert(!(s->info.flags & NBD_FLAG_READ_ONLY)); 1418 if (!(s->info.flags & NBD_FLAG_SEND_WRITE_ZEROES)) { 1419 return -ENOTSUP; 1420 } 1421 1422 if (flags & BDRV_REQ_FUA) { 1423 assert(s->info.flags & NBD_FLAG_SEND_FUA); 1424 request.flags |= NBD_CMD_FLAG_FUA; 1425 } 1426 if (!(flags & BDRV_REQ_MAY_UNMAP)) { 1427 request.flags |= NBD_CMD_FLAG_NO_HOLE; 1428 } 1429 if (flags & BDRV_REQ_NO_FALLBACK) { 1430 assert(s->info.flags & NBD_FLAG_SEND_FAST_ZERO); 1431 request.flags |= NBD_CMD_FLAG_FAST_ZERO; 1432 } 1433 1434 if (!bytes) { 1435 return 0; 1436 } 1437 return nbd_co_request(bs, &request, NULL); 1438 } 1439 1440 static int nbd_client_co_flush(BlockDriverState *bs) 1441 { 1442 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1443 NBDRequest request = { .type = NBD_CMD_FLUSH }; 1444 1445 if (!(s->info.flags & NBD_FLAG_SEND_FLUSH)) { 1446 return 0; 1447 } 1448 1449 request.from = 0; 1450 request.len = 0; 1451 1452 return nbd_co_request(bs, &request, NULL); 1453 } 1454 1455 static int nbd_client_co_pdiscard(BlockDriverState *bs, int64_t offset, 1456 int bytes) 1457 { 1458 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1459 NBDRequest request = { 1460 .type = NBD_CMD_TRIM, 1461 .from = offset, 1462 .len = bytes, 1463 }; 1464 1465 assert(!(s->info.flags & NBD_FLAG_READ_ONLY)); 1466 if (!(s->info.flags & NBD_FLAG_SEND_TRIM) || !bytes) { 1467 return 0; 1468 } 1469 1470 return nbd_co_request(bs, &request, NULL); 1471 } 1472 1473 static int coroutine_fn nbd_client_co_block_status( 1474 BlockDriverState *bs, bool want_zero, int64_t offset, int64_t bytes, 1475 int64_t *pnum, int64_t *map, BlockDriverState **file) 1476 { 1477 int ret, request_ret; 1478 NBDExtent extent = { 0 }; 1479 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1480 Error *local_err = NULL; 1481 1482 NBDRequest request = { 1483 .type = NBD_CMD_BLOCK_STATUS, 1484 .from = offset, 1485 .len = MIN(QEMU_ALIGN_DOWN(INT_MAX, bs->bl.request_alignment), 1486 MIN(bytes, s->info.size - offset)), 1487 .flags = NBD_CMD_FLAG_REQ_ONE, 1488 }; 1489 1490 if (!s->info.base_allocation) { 1491 *pnum = bytes; 1492 *map = offset; 1493 *file = bs; 1494 return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; 1495 } 1496 1497 /* 1498 * Work around the fact that the block layer doesn't do 1499 * byte-accurate sizing yet - if the status request exceeds the 1500 * server's advertised size because the block layer rounded size 1501 * up, we truncated the request to the server (above), or are 1502 * called on just the hole. 1503 */ 1504 if (offset >= s->info.size) { 1505 *pnum = bytes; 1506 assert(bytes < BDRV_SECTOR_SIZE); 1507 /* Intentionally don't report offset_valid for the hole */ 1508 return BDRV_BLOCK_ZERO; 1509 } 1510 1511 if (s->info.min_block) { 1512 assert(QEMU_IS_ALIGNED(request.len, s->info.min_block)); 1513 } 1514 do { 1515 ret = nbd_co_send_request(bs, &request, NULL); 1516 if (ret < 0) { 1517 continue; 1518 } 1519 1520 ret = nbd_co_receive_blockstatus_reply(s, request.handle, bytes, 1521 &extent, &request_ret, 1522 &local_err); 1523 if (local_err) { 1524 trace_nbd_co_request_fail(request.from, request.len, request.handle, 1525 request.flags, request.type, 1526 nbd_cmd_lookup(request.type), 1527 ret, error_get_pretty(local_err)); 1528 error_free(local_err); 1529 local_err = NULL; 1530 } 1531 } while (ret < 0 && nbd_client_connecting_wait(s)); 1532 1533 if (ret < 0 || request_ret < 0) { 1534 return ret ? ret : request_ret; 1535 } 1536 1537 assert(extent.length); 1538 *pnum = extent.length; 1539 *map = offset; 1540 *file = bs; 1541 return (extent.flags & NBD_STATE_HOLE ? 0 : BDRV_BLOCK_DATA) | 1542 (extent.flags & NBD_STATE_ZERO ? BDRV_BLOCK_ZERO : 0) | 1543 BDRV_BLOCK_OFFSET_VALID; 1544 } 1545 1546 static int nbd_client_reopen_prepare(BDRVReopenState *state, 1547 BlockReopenQueue *queue, Error **errp) 1548 { 1549 BDRVNBDState *s = (BDRVNBDState *)state->bs->opaque; 1550 1551 if ((state->flags & BDRV_O_RDWR) && (s->info.flags & NBD_FLAG_READ_ONLY)) { 1552 error_setg(errp, "Can't reopen read-only NBD mount as read/write"); 1553 return -EACCES; 1554 } 1555 return 0; 1556 } 1557 1558 static void nbd_yank(void *opaque) 1559 { 1560 BlockDriverState *bs = opaque; 1561 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1562 1563 qatomic_store_release(&s->state, NBD_CLIENT_QUIT); 1564 qio_channel_shutdown(QIO_CHANNEL(s->ioc), QIO_CHANNEL_SHUTDOWN_BOTH, NULL); 1565 } 1566 1567 static void nbd_client_close(BlockDriverState *bs) 1568 { 1569 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1570 NBDRequest request = { .type = NBD_CMD_DISC }; 1571 1572 if (s->ioc) { 1573 nbd_send_request(s->ioc, &request); 1574 } 1575 1576 nbd_teardown_connection(bs); 1577 } 1578 1579 1580 /* 1581 * Parse nbd_open options 1582 */ 1583 1584 static int nbd_parse_uri(const char *filename, QDict *options) 1585 { 1586 URI *uri; 1587 const char *p; 1588 QueryParams *qp = NULL; 1589 int ret = 0; 1590 bool is_unix; 1591 1592 uri = uri_parse(filename); 1593 if (!uri) { 1594 return -EINVAL; 1595 } 1596 1597 /* transport */ 1598 if (!g_strcmp0(uri->scheme, "nbd")) { 1599 is_unix = false; 1600 } else if (!g_strcmp0(uri->scheme, "nbd+tcp")) { 1601 is_unix = false; 1602 } else if (!g_strcmp0(uri->scheme, "nbd+unix")) { 1603 is_unix = true; 1604 } else { 1605 ret = -EINVAL; 1606 goto out; 1607 } 1608 1609 p = uri->path ? uri->path : ""; 1610 if (p[0] == '/') { 1611 p++; 1612 } 1613 if (p[0]) { 1614 qdict_put_str(options, "export", p); 1615 } 1616 1617 qp = query_params_parse(uri->query); 1618 if (qp->n > 1 || (is_unix && !qp->n) || (!is_unix && qp->n)) { 1619 ret = -EINVAL; 1620 goto out; 1621 } 1622 1623 if (is_unix) { 1624 /* nbd+unix:///export?socket=path */ 1625 if (uri->server || uri->port || strcmp(qp->p[0].name, "socket")) { 1626 ret = -EINVAL; 1627 goto out; 1628 } 1629 qdict_put_str(options, "server.type", "unix"); 1630 qdict_put_str(options, "server.path", qp->p[0].value); 1631 } else { 1632 QString *host; 1633 char *port_str; 1634 1635 /* nbd[+tcp]://host[:port]/export */ 1636 if (!uri->server) { 1637 ret = -EINVAL; 1638 goto out; 1639 } 1640 1641 /* strip braces from literal IPv6 address */ 1642 if (uri->server[0] == '[') { 1643 host = qstring_from_substr(uri->server, 1, 1644 strlen(uri->server) - 1); 1645 } else { 1646 host = qstring_from_str(uri->server); 1647 } 1648 1649 qdict_put_str(options, "server.type", "inet"); 1650 qdict_put(options, "server.host", host); 1651 1652 port_str = g_strdup_printf("%d", uri->port ?: NBD_DEFAULT_PORT); 1653 qdict_put_str(options, "server.port", port_str); 1654 g_free(port_str); 1655 } 1656 1657 out: 1658 if (qp) { 1659 query_params_free(qp); 1660 } 1661 uri_free(uri); 1662 return ret; 1663 } 1664 1665 static bool nbd_has_filename_options_conflict(QDict *options, Error **errp) 1666 { 1667 const QDictEntry *e; 1668 1669 for (e = qdict_first(options); e; e = qdict_next(options, e)) { 1670 if (!strcmp(e->key, "host") || 1671 !strcmp(e->key, "port") || 1672 !strcmp(e->key, "path") || 1673 !strcmp(e->key, "export") || 1674 strstart(e->key, "server.", NULL)) 1675 { 1676 error_setg(errp, "Option '%s' cannot be used with a file name", 1677 e->key); 1678 return true; 1679 } 1680 } 1681 1682 return false; 1683 } 1684 1685 static void nbd_parse_filename(const char *filename, QDict *options, 1686 Error **errp) 1687 { 1688 g_autofree char *file = NULL; 1689 char *export_name; 1690 const char *host_spec; 1691 const char *unixpath; 1692 1693 if (nbd_has_filename_options_conflict(options, errp)) { 1694 return; 1695 } 1696 1697 if (strstr(filename, "://")) { 1698 int ret = nbd_parse_uri(filename, options); 1699 if (ret < 0) { 1700 error_setg(errp, "No valid URL specified"); 1701 } 1702 return; 1703 } 1704 1705 file = g_strdup(filename); 1706 1707 export_name = strstr(file, EN_OPTSTR); 1708 if (export_name) { 1709 if (export_name[strlen(EN_OPTSTR)] == 0) { 1710 return; 1711 } 1712 export_name[0] = 0; /* truncate 'file' */ 1713 export_name += strlen(EN_OPTSTR); 1714 1715 qdict_put_str(options, "export", export_name); 1716 } 1717 1718 /* extract the host_spec - fail if it's not nbd:... */ 1719 if (!strstart(file, "nbd:", &host_spec)) { 1720 error_setg(errp, "File name string for NBD must start with 'nbd:'"); 1721 return; 1722 } 1723 1724 if (!*host_spec) { 1725 return; 1726 } 1727 1728 /* are we a UNIX or TCP socket? */ 1729 if (strstart(host_spec, "unix:", &unixpath)) { 1730 qdict_put_str(options, "server.type", "unix"); 1731 qdict_put_str(options, "server.path", unixpath); 1732 } else { 1733 InetSocketAddress *addr = g_new(InetSocketAddress, 1); 1734 1735 if (inet_parse(addr, host_spec, errp)) { 1736 goto out_inet; 1737 } 1738 1739 qdict_put_str(options, "server.type", "inet"); 1740 qdict_put_str(options, "server.host", addr->host); 1741 qdict_put_str(options, "server.port", addr->port); 1742 out_inet: 1743 qapi_free_InetSocketAddress(addr); 1744 } 1745 } 1746 1747 static bool nbd_process_legacy_socket_options(QDict *output_options, 1748 QemuOpts *legacy_opts, 1749 Error **errp) 1750 { 1751 const char *path = qemu_opt_get(legacy_opts, "path"); 1752 const char *host = qemu_opt_get(legacy_opts, "host"); 1753 const char *port = qemu_opt_get(legacy_opts, "port"); 1754 const QDictEntry *e; 1755 1756 if (!path && !host && !port) { 1757 return true; 1758 } 1759 1760 for (e = qdict_first(output_options); e; e = qdict_next(output_options, e)) 1761 { 1762 if (strstart(e->key, "server.", NULL)) { 1763 error_setg(errp, "Cannot use 'server' and path/host/port at the " 1764 "same time"); 1765 return false; 1766 } 1767 } 1768 1769 if (path && host) { 1770 error_setg(errp, "path and host may not be used at the same time"); 1771 return false; 1772 } else if (path) { 1773 if (port) { 1774 error_setg(errp, "port may not be used without host"); 1775 return false; 1776 } 1777 1778 qdict_put_str(output_options, "server.type", "unix"); 1779 qdict_put_str(output_options, "server.path", path); 1780 } else if (host) { 1781 qdict_put_str(output_options, "server.type", "inet"); 1782 qdict_put_str(output_options, "server.host", host); 1783 qdict_put_str(output_options, "server.port", 1784 port ?: stringify(NBD_DEFAULT_PORT)); 1785 } 1786 1787 return true; 1788 } 1789 1790 static SocketAddress *nbd_config(BDRVNBDState *s, QDict *options, 1791 Error **errp) 1792 { 1793 SocketAddress *saddr = NULL; 1794 QDict *addr = NULL; 1795 Visitor *iv = NULL; 1796 1797 qdict_extract_subqdict(options, &addr, "server."); 1798 if (!qdict_size(addr)) { 1799 error_setg(errp, "NBD server address missing"); 1800 goto done; 1801 } 1802 1803 iv = qobject_input_visitor_new_flat_confused(addr, errp); 1804 if (!iv) { 1805 goto done; 1806 } 1807 1808 if (!visit_type_SocketAddress(iv, NULL, &saddr, errp)) { 1809 goto done; 1810 } 1811 1812 if (socket_address_parse_named_fd(saddr, errp) < 0) { 1813 qapi_free_SocketAddress(saddr); 1814 saddr = NULL; 1815 goto done; 1816 } 1817 1818 done: 1819 qobject_unref(addr); 1820 visit_free(iv); 1821 return saddr; 1822 } 1823 1824 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp) 1825 { 1826 Object *obj; 1827 QCryptoTLSCreds *creds; 1828 1829 obj = object_resolve_path_component( 1830 object_get_objects_root(), id); 1831 if (!obj) { 1832 error_setg(errp, "No TLS credentials with id '%s'", 1833 id); 1834 return NULL; 1835 } 1836 creds = (QCryptoTLSCreds *) 1837 object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS); 1838 if (!creds) { 1839 error_setg(errp, "Object with id '%s' is not TLS credentials", 1840 id); 1841 return NULL; 1842 } 1843 1844 if (!qcrypto_tls_creds_check_endpoint(creds, 1845 QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT, 1846 errp)) { 1847 return NULL; 1848 } 1849 object_ref(obj); 1850 return creds; 1851 } 1852 1853 1854 static QemuOptsList nbd_runtime_opts = { 1855 .name = "nbd", 1856 .head = QTAILQ_HEAD_INITIALIZER(nbd_runtime_opts.head), 1857 .desc = { 1858 { 1859 .name = "host", 1860 .type = QEMU_OPT_STRING, 1861 .help = "TCP host to connect to", 1862 }, 1863 { 1864 .name = "port", 1865 .type = QEMU_OPT_STRING, 1866 .help = "TCP port to connect to", 1867 }, 1868 { 1869 .name = "path", 1870 .type = QEMU_OPT_STRING, 1871 .help = "Unix socket path to connect to", 1872 }, 1873 { 1874 .name = "export", 1875 .type = QEMU_OPT_STRING, 1876 .help = "Name of the NBD export to open", 1877 }, 1878 { 1879 .name = "tls-creds", 1880 .type = QEMU_OPT_STRING, 1881 .help = "ID of the TLS credentials to use", 1882 }, 1883 { 1884 .name = "x-dirty-bitmap", 1885 .type = QEMU_OPT_STRING, 1886 .help = "experimental: expose named dirty bitmap in place of " 1887 "block status", 1888 }, 1889 { 1890 .name = "reconnect-delay", 1891 .type = QEMU_OPT_NUMBER, 1892 .help = "On an unexpected disconnect, the nbd client tries to " 1893 "connect again until succeeding or encountering a serious " 1894 "error. During the first @reconnect-delay seconds, all " 1895 "requests are paused and will be rerun on a successful " 1896 "reconnect. After that time, any delayed requests and all " 1897 "future requests before a successful reconnect will " 1898 "immediately fail. Default 0", 1899 }, 1900 { /* end of list */ } 1901 }, 1902 }; 1903 1904 static int nbd_process_options(BlockDriverState *bs, QDict *options, 1905 Error **errp) 1906 { 1907 BDRVNBDState *s = bs->opaque; 1908 QemuOpts *opts; 1909 int ret = -EINVAL; 1910 1911 opts = qemu_opts_create(&nbd_runtime_opts, NULL, 0, &error_abort); 1912 if (!qemu_opts_absorb_qdict(opts, options, errp)) { 1913 goto error; 1914 } 1915 1916 /* Translate @host, @port, and @path to a SocketAddress */ 1917 if (!nbd_process_legacy_socket_options(options, opts, errp)) { 1918 goto error; 1919 } 1920 1921 /* Pop the config into our state object. Exit if invalid. */ 1922 s->saddr = nbd_config(s, options, errp); 1923 if (!s->saddr) { 1924 goto error; 1925 } 1926 1927 s->export = g_strdup(qemu_opt_get(opts, "export")); 1928 if (s->export && strlen(s->export) > NBD_MAX_STRING_SIZE) { 1929 error_setg(errp, "export name too long to send to server"); 1930 goto error; 1931 } 1932 1933 s->tlscredsid = g_strdup(qemu_opt_get(opts, "tls-creds")); 1934 if (s->tlscredsid) { 1935 s->tlscreds = nbd_get_tls_creds(s->tlscredsid, errp); 1936 if (!s->tlscreds) { 1937 goto error; 1938 } 1939 1940 /* TODO SOCKET_ADDRESS_KIND_FD where fd has AF_INET or AF_INET6 */ 1941 if (s->saddr->type != SOCKET_ADDRESS_TYPE_INET) { 1942 error_setg(errp, "TLS only supported over IP sockets"); 1943 goto error; 1944 } 1945 s->hostname = s->saddr->u.inet.host; 1946 } 1947 1948 s->x_dirty_bitmap = g_strdup(qemu_opt_get(opts, "x-dirty-bitmap")); 1949 if (s->x_dirty_bitmap && strlen(s->x_dirty_bitmap) > NBD_MAX_STRING_SIZE) { 1950 error_setg(errp, "x-dirty-bitmap query too long to send to server"); 1951 goto error; 1952 } 1953 1954 s->reconnect_delay = qemu_opt_get_number(opts, "reconnect-delay", 0); 1955 1956 ret = 0; 1957 1958 error: 1959 qemu_opts_del(opts); 1960 return ret; 1961 } 1962 1963 static int nbd_open(BlockDriverState *bs, QDict *options, int flags, 1964 Error **errp) 1965 { 1966 int ret; 1967 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 1968 1969 s->bs = bs; 1970 qemu_co_mutex_init(&s->send_mutex); 1971 qemu_co_queue_init(&s->free_sema); 1972 1973 if (!yank_register_instance(BLOCKDEV_YANK_INSTANCE(bs->node_name), errp)) { 1974 return -EEXIST; 1975 } 1976 1977 ret = nbd_process_options(bs, options, errp); 1978 if (ret < 0) { 1979 goto fail; 1980 } 1981 1982 s->conn = nbd_client_connection_new(s->saddr, true, s->export, 1983 s->x_dirty_bitmap, s->tlscreds); 1984 1985 /* TODO: Configurable retry-until-timeout behaviour. */ 1986 ret = nbd_do_establish_connection(bs, errp); 1987 if (ret < 0) { 1988 goto fail; 1989 } 1990 1991 s->connection_co = qemu_coroutine_create(nbd_connection_entry, s); 1992 bdrv_inc_in_flight(bs); 1993 aio_co_schedule(bdrv_get_aio_context(bs), s->connection_co); 1994 1995 return 0; 1996 1997 fail: 1998 nbd_clear_bdrvstate(bs); 1999 return ret; 2000 } 2001 2002 static int nbd_co_flush(BlockDriverState *bs) 2003 { 2004 return nbd_client_co_flush(bs); 2005 } 2006 2007 static void nbd_refresh_limits(BlockDriverState *bs, Error **errp) 2008 { 2009 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 2010 uint32_t min = s->info.min_block; 2011 uint32_t max = MIN_NON_ZERO(NBD_MAX_BUFFER_SIZE, s->info.max_block); 2012 2013 /* 2014 * If the server did not advertise an alignment: 2015 * - a size that is not sector-aligned implies that an alignment 2016 * of 1 can be used to access those tail bytes 2017 * - advertisement of block status requires an alignment of 1, so 2018 * that we don't violate block layer constraints that block 2019 * status is always aligned (as we can't control whether the 2020 * server will report sub-sector extents, such as a hole at EOF 2021 * on an unaligned POSIX file) 2022 * - otherwise, assume the server is so old that we are safer avoiding 2023 * sub-sector requests 2024 */ 2025 if (!min) { 2026 min = (!QEMU_IS_ALIGNED(s->info.size, BDRV_SECTOR_SIZE) || 2027 s->info.base_allocation) ? 1 : BDRV_SECTOR_SIZE; 2028 } 2029 2030 bs->bl.request_alignment = min; 2031 bs->bl.max_pdiscard = QEMU_ALIGN_DOWN(INT_MAX, min); 2032 bs->bl.max_pwrite_zeroes = max; 2033 bs->bl.max_transfer = max; 2034 2035 if (s->info.opt_block && 2036 s->info.opt_block > bs->bl.opt_transfer) { 2037 bs->bl.opt_transfer = s->info.opt_block; 2038 } 2039 } 2040 2041 static void nbd_close(BlockDriverState *bs) 2042 { 2043 nbd_client_close(bs); 2044 nbd_clear_bdrvstate(bs); 2045 } 2046 2047 /* 2048 * NBD cannot truncate, but if the caller asks to truncate to the same size, or 2049 * to a smaller size with exact=false, there is no reason to fail the 2050 * operation. 2051 * 2052 * Preallocation mode is ignored since it does not seems useful to fail when 2053 * we never change anything. 2054 */ 2055 static int coroutine_fn nbd_co_truncate(BlockDriverState *bs, int64_t offset, 2056 bool exact, PreallocMode prealloc, 2057 BdrvRequestFlags flags, Error **errp) 2058 { 2059 BDRVNBDState *s = bs->opaque; 2060 2061 if (offset != s->info.size && exact) { 2062 error_setg(errp, "Cannot resize NBD nodes"); 2063 return -ENOTSUP; 2064 } 2065 2066 if (offset > s->info.size) { 2067 error_setg(errp, "Cannot grow NBD nodes"); 2068 return -EINVAL; 2069 } 2070 2071 return 0; 2072 } 2073 2074 static int64_t nbd_getlength(BlockDriverState *bs) 2075 { 2076 BDRVNBDState *s = bs->opaque; 2077 2078 return s->info.size; 2079 } 2080 2081 static void nbd_refresh_filename(BlockDriverState *bs) 2082 { 2083 BDRVNBDState *s = bs->opaque; 2084 const char *host = NULL, *port = NULL, *path = NULL; 2085 size_t len = 0; 2086 2087 if (s->saddr->type == SOCKET_ADDRESS_TYPE_INET) { 2088 const InetSocketAddress *inet = &s->saddr->u.inet; 2089 if (!inet->has_ipv4 && !inet->has_ipv6 && !inet->has_to) { 2090 host = inet->host; 2091 port = inet->port; 2092 } 2093 } else if (s->saddr->type == SOCKET_ADDRESS_TYPE_UNIX) { 2094 path = s->saddr->u.q_unix.path; 2095 } /* else can't represent as pseudo-filename */ 2096 2097 if (path && s->export) { 2098 len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), 2099 "nbd+unix:///%s?socket=%s", s->export, path); 2100 } else if (path && !s->export) { 2101 len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), 2102 "nbd+unix://?socket=%s", path); 2103 } else if (host && s->export) { 2104 len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), 2105 "nbd://%s:%s/%s", host, port, s->export); 2106 } else if (host && !s->export) { 2107 len = snprintf(bs->exact_filename, sizeof(bs->exact_filename), 2108 "nbd://%s:%s", host, port); 2109 } 2110 if (len >= sizeof(bs->exact_filename)) { 2111 /* Name is too long to represent exactly, so leave it empty. */ 2112 bs->exact_filename[0] = '\0'; 2113 } 2114 } 2115 2116 static char *nbd_dirname(BlockDriverState *bs, Error **errp) 2117 { 2118 /* The generic bdrv_dirname() implementation is able to work out some 2119 * directory name for NBD nodes, but that would be wrong. So far there is no 2120 * specification for how "export paths" would work, so NBD does not have 2121 * directory names. */ 2122 error_setg(errp, "Cannot generate a base directory for NBD nodes"); 2123 return NULL; 2124 } 2125 2126 static const char *const nbd_strong_runtime_opts[] = { 2127 "path", 2128 "host", 2129 "port", 2130 "export", 2131 "tls-creds", 2132 "server.", 2133 2134 NULL 2135 }; 2136 2137 static void nbd_cancel_in_flight(BlockDriverState *bs) 2138 { 2139 BDRVNBDState *s = (BDRVNBDState *)bs->opaque; 2140 2141 reconnect_delay_timer_del(s); 2142 2143 if (s->state == NBD_CLIENT_CONNECTING_WAIT) { 2144 s->state = NBD_CLIENT_CONNECTING_NOWAIT; 2145 qemu_co_queue_restart_all(&s->free_sema); 2146 } 2147 } 2148 2149 static BlockDriver bdrv_nbd = { 2150 .format_name = "nbd", 2151 .protocol_name = "nbd", 2152 .instance_size = sizeof(BDRVNBDState), 2153 .bdrv_parse_filename = nbd_parse_filename, 2154 .bdrv_co_create_opts = bdrv_co_create_opts_simple, 2155 .create_opts = &bdrv_create_opts_simple, 2156 .bdrv_file_open = nbd_open, 2157 .bdrv_reopen_prepare = nbd_client_reopen_prepare, 2158 .bdrv_co_preadv = nbd_client_co_preadv, 2159 .bdrv_co_pwritev = nbd_client_co_pwritev, 2160 .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, 2161 .bdrv_close = nbd_close, 2162 .bdrv_co_flush_to_os = nbd_co_flush, 2163 .bdrv_co_pdiscard = nbd_client_co_pdiscard, 2164 .bdrv_refresh_limits = nbd_refresh_limits, 2165 .bdrv_co_truncate = nbd_co_truncate, 2166 .bdrv_getlength = nbd_getlength, 2167 .bdrv_detach_aio_context = nbd_client_detach_aio_context, 2168 .bdrv_attach_aio_context = nbd_client_attach_aio_context, 2169 .bdrv_co_drain_begin = nbd_client_co_drain_begin, 2170 .bdrv_co_drain_end = nbd_client_co_drain_end, 2171 .bdrv_refresh_filename = nbd_refresh_filename, 2172 .bdrv_co_block_status = nbd_client_co_block_status, 2173 .bdrv_dirname = nbd_dirname, 2174 .strong_runtime_opts = nbd_strong_runtime_opts, 2175 .bdrv_cancel_in_flight = nbd_cancel_in_flight, 2176 }; 2177 2178 static BlockDriver bdrv_nbd_tcp = { 2179 .format_name = "nbd", 2180 .protocol_name = "nbd+tcp", 2181 .instance_size = sizeof(BDRVNBDState), 2182 .bdrv_parse_filename = nbd_parse_filename, 2183 .bdrv_co_create_opts = bdrv_co_create_opts_simple, 2184 .create_opts = &bdrv_create_opts_simple, 2185 .bdrv_file_open = nbd_open, 2186 .bdrv_reopen_prepare = nbd_client_reopen_prepare, 2187 .bdrv_co_preadv = nbd_client_co_preadv, 2188 .bdrv_co_pwritev = nbd_client_co_pwritev, 2189 .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, 2190 .bdrv_close = nbd_close, 2191 .bdrv_co_flush_to_os = nbd_co_flush, 2192 .bdrv_co_pdiscard = nbd_client_co_pdiscard, 2193 .bdrv_refresh_limits = nbd_refresh_limits, 2194 .bdrv_co_truncate = nbd_co_truncate, 2195 .bdrv_getlength = nbd_getlength, 2196 .bdrv_detach_aio_context = nbd_client_detach_aio_context, 2197 .bdrv_attach_aio_context = nbd_client_attach_aio_context, 2198 .bdrv_co_drain_begin = nbd_client_co_drain_begin, 2199 .bdrv_co_drain_end = nbd_client_co_drain_end, 2200 .bdrv_refresh_filename = nbd_refresh_filename, 2201 .bdrv_co_block_status = nbd_client_co_block_status, 2202 .bdrv_dirname = nbd_dirname, 2203 .strong_runtime_opts = nbd_strong_runtime_opts, 2204 .bdrv_cancel_in_flight = nbd_cancel_in_flight, 2205 }; 2206 2207 static BlockDriver bdrv_nbd_unix = { 2208 .format_name = "nbd", 2209 .protocol_name = "nbd+unix", 2210 .instance_size = sizeof(BDRVNBDState), 2211 .bdrv_parse_filename = nbd_parse_filename, 2212 .bdrv_co_create_opts = bdrv_co_create_opts_simple, 2213 .create_opts = &bdrv_create_opts_simple, 2214 .bdrv_file_open = nbd_open, 2215 .bdrv_reopen_prepare = nbd_client_reopen_prepare, 2216 .bdrv_co_preadv = nbd_client_co_preadv, 2217 .bdrv_co_pwritev = nbd_client_co_pwritev, 2218 .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, 2219 .bdrv_close = nbd_close, 2220 .bdrv_co_flush_to_os = nbd_co_flush, 2221 .bdrv_co_pdiscard = nbd_client_co_pdiscard, 2222 .bdrv_refresh_limits = nbd_refresh_limits, 2223 .bdrv_co_truncate = nbd_co_truncate, 2224 .bdrv_getlength = nbd_getlength, 2225 .bdrv_detach_aio_context = nbd_client_detach_aio_context, 2226 .bdrv_attach_aio_context = nbd_client_attach_aio_context, 2227 .bdrv_co_drain_begin = nbd_client_co_drain_begin, 2228 .bdrv_co_drain_end = nbd_client_co_drain_end, 2229 .bdrv_refresh_filename = nbd_refresh_filename, 2230 .bdrv_co_block_status = nbd_client_co_block_status, 2231 .bdrv_dirname = nbd_dirname, 2232 .strong_runtime_opts = nbd_strong_runtime_opts, 2233 .bdrv_cancel_in_flight = nbd_cancel_in_flight, 2234 }; 2235 2236 static void bdrv_nbd_init(void) 2237 { 2238 bdrv_register(&bdrv_nbd); 2239 bdrv_register(&bdrv_nbd_tcp); 2240 bdrv_register(&bdrv_nbd_unix); 2241 } 2242 2243 block_init(bdrv_nbd_init); 2244