1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * Copyright 2023 Red Hat 4 */ 5 6 #include "data-vio.h" 7 8 #include <linux/atomic.h> 9 #include <linux/bio.h> 10 #include <linux/blkdev.h> 11 #include <linux/delay.h> 12 #include <linux/device-mapper.h> 13 #include <linux/jiffies.h> 14 #include <linux/kernel.h> 15 #include <linux/list.h> 16 #include <linux/lz4.h> 17 #include <linux/minmax.h> 18 #include <linux/sched.h> 19 #include <linux/spinlock.h> 20 #include <linux/wait.h> 21 22 #include "logger.h" 23 #include "memory-alloc.h" 24 #include "murmurhash3.h" 25 #include "permassert.h" 26 27 #include "block-map.h" 28 #include "dump.h" 29 #include "encodings.h" 30 #include "int-map.h" 31 #include "io-submitter.h" 32 #include "logical-zone.h" 33 #include "packer.h" 34 #include "recovery-journal.h" 35 #include "slab-depot.h" 36 #include "status-codes.h" 37 #include "types.h" 38 #include "vdo.h" 39 #include "vio.h" 40 #include "wait-queue.h" 41 42 /** 43 * DOC: Bio flags. 44 * 45 * For certain flags set on user bios, if the user bio has not yet been acknowledged, setting those 46 * flags on our own bio(s) for that request may help underlying layers better fulfill the user 47 * bio's needs. This constant contains the aggregate of those flags; VDO strips all the other 48 * flags, as they convey incorrect information. 49 * 50 * These flags are always irrelevant if we have already finished the user bio as they are only 51 * hints on IO importance. If VDO has finished the user bio, any remaining IO done doesn't care how 52 * important finishing the finished bio was. 53 * 54 * Note that bio.c contains the complete list of flags we believe may be set; the following list 55 * explains the action taken with each of those flags VDO could receive: 56 * 57 * * REQ_SYNC: Passed down if the user bio is not yet completed, since it indicates the user bio 58 * completion is required for further work to be done by the issuer. 59 * * REQ_META: Passed down if the user bio is not yet completed, since it may mean the lower layer 60 * treats it as more urgent, similar to REQ_SYNC. 61 * * REQ_PRIO: Passed down if the user bio is not yet completed, since it indicates the user bio is 62 * important. 63 * * REQ_NOMERGE: Set only if the incoming bio was split; irrelevant to VDO IO. 64 * * REQ_IDLE: Set if the incoming bio had more IO quickly following; VDO's IO pattern doesn't 65 * match incoming IO, so this flag is incorrect for it. 66 * * REQ_FUA: Handled separately, and irrelevant to VDO IO otherwise. 67 * * REQ_RAHEAD: Passed down, as, for reads, it indicates trivial importance. 68 * * REQ_BACKGROUND: Not passed down, as VIOs are a limited resource and VDO needs them recycled 69 * ASAP to service heavy load, which is the only place where REQ_BACKGROUND might aid in load 70 * prioritization. 71 */ 72 static blk_opf_t PASSTHROUGH_FLAGS = (REQ_PRIO | REQ_META | REQ_SYNC | REQ_RAHEAD); 73 74 /** 75 * DOC: 76 * 77 * The data_vio_pool maintains the pool of data_vios which a vdo uses to service incoming bios. For 78 * correctness, and in order to avoid potentially expensive or blocking memory allocations during 79 * normal operation, the number of concurrently active data_vios is capped. Furthermore, in order 80 * to avoid starvation of reads and writes, at most 75% of the data_vios may be used for 81 * discards. The data_vio_pool is responsible for enforcing these limits. Threads submitting bios 82 * for which a data_vio or discard permit are not available will block until the necessary 83 * resources are available. The pool is also responsible for distributing resources to blocked 84 * threads and waking them. Finally, the pool attempts to batch the work of recycling data_vios by 85 * performing the work of actually assigning resources to blocked threads or placing data_vios back 86 * into the pool on a single cpu at a time. 87 * 88 * The pool contains two "limiters", one for tracking data_vios and one for tracking discard 89 * permits. The limiters also provide safe cross-thread access to pool statistics without the need 90 * to take the pool's lock. When a thread submits a bio to a vdo device, it will first attempt to 91 * get a discard permit if it is a discard, and then to get a data_vio. If the necessary resources 92 * are available, the incoming bio will be assigned to the acquired data_vio, and it will be 93 * launched. However, if either of these are unavailable, the arrival time of the bio is recorded 94 * in the bio's bi_private field, the bio and its submitter are both queued on the appropriate 95 * limiter and the submitting thread will then put itself to sleep. (note that this mechanism will 96 * break if jiffies are only 32 bits.) 97 * 98 * Whenever a data_vio has completed processing for the bio it was servicing, release_data_vio() 99 * will be called on it. This function will add the data_vio to a funnel queue, and then check the 100 * state of the pool. If the pool is not currently processing released data_vios, the pool's 101 * completion will be enqueued on a cpu queue. This obviates the need for the releasing threads to 102 * hold the pool's lock, and also batches release work while avoiding starvation of the cpu 103 * threads. 104 * 105 * Whenever the pool's completion is run on a cpu thread, it calls process_release_callback() which 106 * processes a batch of returned data_vios (currently at most 32) from the pool's funnel queue. For 107 * each data_vio, it first checks whether that data_vio was processing a discard. If so, and there 108 * is a blocked bio waiting for a discard permit, that permit is notionally transferred to the 109 * eldest discard waiter, and that waiter is moved to the end of the list of discard bios waiting 110 * for a data_vio. If there are no discard waiters, the discard permit is returned to the pool. 111 * Next, the data_vio is assigned to the oldest blocked bio which either has a discard permit, or 112 * doesn't need one and relaunched. If neither of these exist, the data_vio is returned to the 113 * pool. Finally, if any waiting bios were launched, the threads which blocked trying to submit 114 * them are awakened. 115 */ 116 117 #define DATA_VIO_RELEASE_BATCH_SIZE 128 118 119 static const unsigned int VDO_SECTORS_PER_BLOCK_MASK = VDO_SECTORS_PER_BLOCK - 1; 120 static const u32 COMPRESSION_STATUS_MASK = 0xff; 121 static const u32 MAY_NOT_COMPRESS_MASK = 0x80000000; 122 123 struct limiter; 124 typedef void (*assigner_fn)(struct limiter *limiter); 125 126 /* Bookkeeping structure for a single type of resource. */ 127 struct limiter { 128 /* The data_vio_pool to which this limiter belongs */ 129 struct data_vio_pool *pool; 130 /* The maximum number of data_vios available */ 131 data_vio_count_t limit; 132 /* The number of resources in use */ 133 data_vio_count_t busy; 134 /* The maximum number of resources ever simultaneously in use */ 135 data_vio_count_t max_busy; 136 /* The number of resources to release */ 137 data_vio_count_t release_count; 138 /* The number of waiters to wake */ 139 data_vio_count_t wake_count; 140 /* The list of waiting bios which are known to process_release_callback() */ 141 struct bio_list waiters; 142 /* The list of waiting bios which are not yet known to process_release_callback() */ 143 struct bio_list new_waiters; 144 /* The list of waiters which have their permits */ 145 struct bio_list *permitted_waiters; 146 /* The function for assigning a resource to a waiter */ 147 assigner_fn assigner; 148 /* The queue of blocked threads */ 149 wait_queue_head_t blocked_threads; 150 /* The arrival time of the eldest waiter */ 151 u64 arrival; 152 }; 153 154 /* 155 * A data_vio_pool is a collection of preallocated data_vios which may be acquired from any thread, 156 * and are released in batches. 157 */ 158 struct data_vio_pool { 159 /* Completion for scheduling releases */ 160 struct vdo_completion completion; 161 /* The administrative state of the pool */ 162 struct admin_state state; 163 /* Lock protecting the pool */ 164 spinlock_t lock; 165 /* The main limiter controlling the total data_vios in the pool. */ 166 struct limiter limiter; 167 /* The limiter controlling data_vios for discard */ 168 struct limiter discard_limiter; 169 /* The list of bios which have discard permits but still need a data_vio */ 170 struct bio_list permitted_discards; 171 /* The list of available data_vios */ 172 struct list_head available; 173 /* The queue of data_vios waiting to be returned to the pool */ 174 struct funnel_queue *queue; 175 /* Whether the pool is processing, or scheduled to process releases */ 176 atomic_t processing; 177 /* The data vios in the pool */ 178 struct data_vio data_vios[]; 179 }; 180 181 static const char * const ASYNC_OPERATION_NAMES[] = { 182 "launch", 183 "acknowledge_write", 184 "acquire_hash_lock", 185 "attempt_logical_block_lock", 186 "lock_duplicate_pbn", 187 "check_for_duplication", 188 "cleanup", 189 "compress_data_vio", 190 "find_block_map_slot", 191 "get_mapped_block_for_read", 192 "get_mapped_block_for_write", 193 "hash_data_vio", 194 "journal_remapping", 195 "vdo_attempt_packing", 196 "put_mapped_block", 197 "read_data_vio", 198 "update_dedupe_index", 199 "update_reference_counts", 200 "verify_duplication", 201 "write_data_vio", 202 }; 203 204 /* The steps taken cleaning up a VIO, in the order they are performed. */ 205 enum data_vio_cleanup_stage { 206 VIO_CLEANUP_START, 207 VIO_RELEASE_HASH_LOCK = VIO_CLEANUP_START, 208 VIO_RELEASE_ALLOCATED, 209 VIO_RELEASE_RECOVERY_LOCKS, 210 VIO_RELEASE_LOGICAL, 211 VIO_CLEANUP_DONE 212 }; 213 214 static inline struct data_vio_pool * __must_check 215 as_data_vio_pool(struct vdo_completion *completion) 216 { 217 vdo_assert_completion_type(completion, VDO_DATA_VIO_POOL_COMPLETION); 218 return container_of(completion, struct data_vio_pool, completion); 219 } 220 221 static inline u64 get_arrival_time(struct bio *bio) 222 { 223 return (u64) bio->bi_private; 224 } 225 226 /** 227 * check_for_drain_complete_locked() - Check whether a data_vio_pool has no outstanding data_vios 228 * or waiters while holding the pool's lock. 229 */ 230 static bool check_for_drain_complete_locked(struct data_vio_pool *pool) 231 { 232 if (pool->limiter.busy > 0) 233 return false; 234 235 VDO_ASSERT_LOG_ONLY((pool->discard_limiter.busy == 0), 236 "no outstanding discard permits"); 237 238 return (bio_list_empty(&pool->limiter.new_waiters) && 239 bio_list_empty(&pool->discard_limiter.new_waiters)); 240 } 241 242 static void initialize_lbn_lock(struct data_vio *data_vio, logical_block_number_t lbn) 243 { 244 struct vdo *vdo = vdo_from_data_vio(data_vio); 245 zone_count_t zone_number; 246 struct lbn_lock *lock = &data_vio->logical; 247 248 lock->lbn = lbn; 249 lock->locked = false; 250 vdo_waitq_init(&lock->waiters); 251 zone_number = vdo_compute_logical_zone(data_vio); 252 lock->zone = &vdo->logical_zones->zones[zone_number]; 253 } 254 255 static void launch_locked_request(struct data_vio *data_vio) 256 { 257 data_vio->logical.locked = true; 258 if (data_vio->write) { 259 struct vdo *vdo = vdo_from_data_vio(data_vio); 260 261 if (vdo_is_read_only(vdo)) { 262 continue_data_vio_with_error(data_vio, VDO_READ_ONLY); 263 return; 264 } 265 } 266 267 data_vio->last_async_operation = VIO_ASYNC_OP_FIND_BLOCK_MAP_SLOT; 268 vdo_find_block_map_slot(data_vio); 269 } 270 271 static void acknowledge_data_vio(struct data_vio *data_vio) 272 { 273 struct vdo *vdo = vdo_from_data_vio(data_vio); 274 struct bio *bio = data_vio->user_bio; 275 int error = vdo_status_to_errno(data_vio->vio.completion.result); 276 277 if (bio == NULL) 278 return; 279 280 VDO_ASSERT_LOG_ONLY((data_vio->remaining_discard <= 281 (u32) (VDO_BLOCK_SIZE - data_vio->offset)), 282 "data_vio to acknowledge is not an incomplete discard"); 283 284 data_vio->user_bio = NULL; 285 vdo_count_bios(&vdo->stats.bios_acknowledged, bio); 286 if (data_vio->is_partial) 287 vdo_count_bios(&vdo->stats.bios_acknowledged_partial, bio); 288 289 bio->bi_status = errno_to_blk_status(error); 290 bio_endio(bio); 291 } 292 293 static void copy_to_bio(struct bio *bio, char *data_ptr) 294 { 295 struct bio_vec biovec; 296 struct bvec_iter iter; 297 298 bio_for_each_segment(biovec, bio, iter) { 299 memcpy_to_bvec(&biovec, data_ptr); 300 data_ptr += biovec.bv_len; 301 } 302 } 303 304 struct data_vio_compression_status get_data_vio_compression_status(struct data_vio *data_vio) 305 { 306 u32 packed = atomic_read(&data_vio->compression.status); 307 308 /* pairs with cmpxchg in set_data_vio_compression_status */ 309 smp_rmb(); 310 return (struct data_vio_compression_status) { 311 .stage = packed & COMPRESSION_STATUS_MASK, 312 .may_not_compress = ((packed & MAY_NOT_COMPRESS_MASK) != 0), 313 }; 314 } 315 316 /** 317 * pack_status() - Convert a data_vio_compression_status into a u32 which may be stored 318 * atomically. 319 * @status: The state to convert. 320 * 321 * Return: The compression state packed into a u32. 322 */ 323 static u32 __must_check pack_status(struct data_vio_compression_status status) 324 { 325 return status.stage | (status.may_not_compress ? MAY_NOT_COMPRESS_MASK : 0); 326 } 327 328 /** 329 * set_data_vio_compression_status() - Set the compression status of a data_vio. 330 * @state: The expected current status of the data_vio. 331 * @new_state: The status to set. 332 * 333 * Return: true if the new status was set, false if the data_vio's compression status did not 334 * match the expected state, and so was left unchanged. 335 */ 336 static bool __must_check 337 set_data_vio_compression_status(struct data_vio *data_vio, 338 struct data_vio_compression_status status, 339 struct data_vio_compression_status new_status) 340 { 341 u32 actual; 342 u32 expected = pack_status(status); 343 u32 replacement = pack_status(new_status); 344 345 /* 346 * Extra barriers because this was original developed using a CAS operation that implicitly 347 * had them. 348 */ 349 smp_mb__before_atomic(); 350 actual = atomic_cmpxchg(&data_vio->compression.status, expected, replacement); 351 /* same as before_atomic */ 352 smp_mb__after_atomic(); 353 return (expected == actual); 354 } 355 356 struct data_vio_compression_status advance_data_vio_compression_stage(struct data_vio *data_vio) 357 { 358 for (;;) { 359 struct data_vio_compression_status status = 360 get_data_vio_compression_status(data_vio); 361 struct data_vio_compression_status new_status = status; 362 363 if (status.stage == DATA_VIO_POST_PACKER) { 364 /* We're already in the last stage. */ 365 return status; 366 } 367 368 if (status.may_not_compress) { 369 /* 370 * Compression has been dis-allowed for this VIO, so skip the rest of the 371 * path and go to the end. 372 */ 373 new_status.stage = DATA_VIO_POST_PACKER; 374 } else { 375 /* Go to the next state. */ 376 new_status.stage++; 377 } 378 379 if (set_data_vio_compression_status(data_vio, status, new_status)) 380 return new_status; 381 382 /* Another thread changed the status out from under us so try again. */ 383 } 384 } 385 386 /** 387 * cancel_data_vio_compression() - Prevent this data_vio from being compressed or packed. 388 * 389 * Return: true if the data_vio is in the packer and the caller was the first caller to cancel it. 390 */ 391 bool cancel_data_vio_compression(struct data_vio *data_vio) 392 { 393 struct data_vio_compression_status status, new_status; 394 395 for (;;) { 396 status = get_data_vio_compression_status(data_vio); 397 if (status.may_not_compress || (status.stage == DATA_VIO_POST_PACKER)) { 398 /* This data_vio is already set up to not block in the packer. */ 399 break; 400 } 401 402 new_status.stage = status.stage; 403 new_status.may_not_compress = true; 404 405 if (set_data_vio_compression_status(data_vio, status, new_status)) 406 break; 407 } 408 409 return ((status.stage == DATA_VIO_PACKING) && !status.may_not_compress); 410 } 411 412 /** 413 * attempt_logical_block_lock() - Attempt to acquire the lock on a logical block. 414 * @completion: The data_vio for an external data request as a completion. 415 * 416 * This is the start of the path for all external requests. It is registered in launch_data_vio(). 417 */ 418 static void attempt_logical_block_lock(struct vdo_completion *completion) 419 { 420 struct data_vio *data_vio = as_data_vio(completion); 421 struct lbn_lock *lock = &data_vio->logical; 422 struct vdo *vdo = vdo_from_data_vio(data_vio); 423 struct data_vio *lock_holder; 424 int result; 425 426 assert_data_vio_in_logical_zone(data_vio); 427 428 if (data_vio->logical.lbn >= vdo->states.vdo.config.logical_blocks) { 429 continue_data_vio_with_error(data_vio, VDO_OUT_OF_RANGE); 430 return; 431 } 432 433 result = vdo_int_map_put(lock->zone->lbn_operations, lock->lbn, 434 data_vio, false, (void **) &lock_holder); 435 if (result != VDO_SUCCESS) { 436 continue_data_vio_with_error(data_vio, result); 437 return; 438 } 439 440 if (lock_holder == NULL) { 441 /* We got the lock */ 442 launch_locked_request(data_vio); 443 return; 444 } 445 446 result = VDO_ASSERT(lock_holder->logical.locked, "logical block lock held"); 447 if (result != VDO_SUCCESS) { 448 continue_data_vio_with_error(data_vio, result); 449 return; 450 } 451 452 /* 453 * If the new request is a pure read request (not read-modify-write) and the lock_holder is 454 * writing and has received an allocation, service the read request immediately by copying 455 * data from the lock_holder to avoid having to flush the write out of the packer just to 456 * prevent the read from waiting indefinitely. If the lock_holder does not yet have an 457 * allocation, prevent it from blocking in the packer and wait on it. This is necessary in 458 * order to prevent returning data that may not have actually been written. 459 */ 460 if (!data_vio->write && READ_ONCE(lock_holder->allocation_succeeded)) { 461 copy_to_bio(data_vio->user_bio, lock_holder->vio.data + data_vio->offset); 462 acknowledge_data_vio(data_vio); 463 complete_data_vio(completion); 464 return; 465 } 466 467 data_vio->last_async_operation = VIO_ASYNC_OP_ATTEMPT_LOGICAL_BLOCK_LOCK; 468 vdo_waitq_enqueue_waiter(&lock_holder->logical.waiters, &data_vio->waiter); 469 470 /* 471 * Prevent writes and read-modify-writes from blocking indefinitely on lock holders in the 472 * packer. 473 */ 474 if (lock_holder->write && cancel_data_vio_compression(lock_holder)) { 475 data_vio->compression.lock_holder = lock_holder; 476 launch_data_vio_packer_callback(data_vio, 477 vdo_remove_lock_holder_from_packer); 478 } 479 } 480 481 /** 482 * launch_data_vio() - (Re)initialize a data_vio to have a new logical block number, keeping the 483 * same parent and other state and send it on its way. 484 */ 485 static void launch_data_vio(struct data_vio *data_vio, logical_block_number_t lbn) 486 { 487 struct vdo_completion *completion = &data_vio->vio.completion; 488 489 /* 490 * Clearing the tree lock must happen before initializing the LBN lock, which also adds 491 * information to the tree lock. 492 */ 493 memset(&data_vio->tree_lock, 0, sizeof(data_vio->tree_lock)); 494 initialize_lbn_lock(data_vio, lbn); 495 INIT_LIST_HEAD(&data_vio->hash_lock_entry); 496 INIT_LIST_HEAD(&data_vio->write_entry); 497 498 memset(&data_vio->allocation, 0, sizeof(data_vio->allocation)); 499 500 data_vio->is_duplicate = false; 501 502 memset(&data_vio->record_name, 0, sizeof(data_vio->record_name)); 503 memset(&data_vio->duplicate, 0, sizeof(data_vio->duplicate)); 504 vdo_reset_completion(completion); 505 completion->error_handler = handle_data_vio_error; 506 set_data_vio_logical_callback(data_vio, attempt_logical_block_lock); 507 vdo_enqueue_completion(completion, VDO_DEFAULT_Q_MAP_BIO_PRIORITY); 508 } 509 510 static bool is_zero_block(char *block) 511 { 512 int i; 513 514 for (i = 0; i < VDO_BLOCK_SIZE; i += sizeof(u64)) { 515 if (*((u64 *) &block[i])) 516 return false; 517 } 518 519 return true; 520 } 521 522 static void copy_from_bio(struct bio *bio, char *data_ptr) 523 { 524 struct bio_vec biovec; 525 struct bvec_iter iter; 526 527 bio_for_each_segment(biovec, bio, iter) { 528 memcpy_from_bvec(data_ptr, &biovec); 529 data_ptr += biovec.bv_len; 530 } 531 } 532 533 static void launch_bio(struct vdo *vdo, struct data_vio *data_vio, struct bio *bio) 534 { 535 logical_block_number_t lbn; 536 /* 537 * Zero out the fields which don't need to be preserved (i.e. which are not pointers to 538 * separately allocated objects). 539 */ 540 memset(data_vio, 0, offsetof(struct data_vio, vio)); 541 memset(&data_vio->compression, 0, offsetof(struct compression_state, block)); 542 543 data_vio->user_bio = bio; 544 data_vio->offset = to_bytes(bio->bi_iter.bi_sector & VDO_SECTORS_PER_BLOCK_MASK); 545 data_vio->is_partial = (bio->bi_iter.bi_size < VDO_BLOCK_SIZE) || (data_vio->offset != 0); 546 547 /* 548 * Discards behave very differently than other requests when coming in from device-mapper. 549 * We have to be able to handle any size discards and various sector offsets within a 550 * block. 551 */ 552 if (bio_op(bio) == REQ_OP_DISCARD) { 553 data_vio->remaining_discard = bio->bi_iter.bi_size; 554 data_vio->write = true; 555 data_vio->is_discard = true; 556 if (data_vio->is_partial) { 557 vdo_count_bios(&vdo->stats.bios_in_partial, bio); 558 data_vio->read = true; 559 } 560 } else if (data_vio->is_partial) { 561 vdo_count_bios(&vdo->stats.bios_in_partial, bio); 562 data_vio->read = true; 563 if (bio_data_dir(bio) == WRITE) 564 data_vio->write = true; 565 } else if (bio_data_dir(bio) == READ) { 566 data_vio->read = true; 567 } else { 568 /* 569 * Copy the bio data to a char array so that we can continue to use the data after 570 * we acknowledge the bio. 571 */ 572 copy_from_bio(bio, data_vio->vio.data); 573 data_vio->is_zero = is_zero_block(data_vio->vio.data); 574 data_vio->write = true; 575 } 576 577 if (data_vio->user_bio->bi_opf & REQ_FUA) 578 data_vio->fua = true; 579 580 lbn = (bio->bi_iter.bi_sector - vdo->starting_sector_offset) / VDO_SECTORS_PER_BLOCK; 581 launch_data_vio(data_vio, lbn); 582 } 583 584 static void assign_data_vio(struct limiter *limiter, struct data_vio *data_vio) 585 { 586 struct bio *bio = bio_list_pop(limiter->permitted_waiters); 587 588 launch_bio(limiter->pool->completion.vdo, data_vio, bio); 589 limiter->wake_count++; 590 591 bio = bio_list_peek(limiter->permitted_waiters); 592 limiter->arrival = ((bio == NULL) ? U64_MAX : get_arrival_time(bio)); 593 } 594 595 static void assign_discard_permit(struct limiter *limiter) 596 { 597 struct bio *bio = bio_list_pop(&limiter->waiters); 598 599 if (limiter->arrival == U64_MAX) 600 limiter->arrival = get_arrival_time(bio); 601 602 bio_list_add(limiter->permitted_waiters, bio); 603 } 604 605 static void get_waiters(struct limiter *limiter) 606 { 607 bio_list_merge_init(&limiter->waiters, &limiter->new_waiters); 608 } 609 610 static inline struct data_vio *get_available_data_vio(struct data_vio_pool *pool) 611 { 612 struct data_vio *data_vio = 613 list_first_entry(&pool->available, struct data_vio, pool_entry); 614 615 list_del_init(&data_vio->pool_entry); 616 return data_vio; 617 } 618 619 static void assign_data_vio_to_waiter(struct limiter *limiter) 620 { 621 assign_data_vio(limiter, get_available_data_vio(limiter->pool)); 622 } 623 624 static void update_limiter(struct limiter *limiter) 625 { 626 struct bio_list *waiters = &limiter->waiters; 627 data_vio_count_t available = limiter->limit - limiter->busy; 628 629 VDO_ASSERT_LOG_ONLY((limiter->release_count <= limiter->busy), 630 "Release count %u is not more than busy count %u", 631 limiter->release_count, limiter->busy); 632 633 get_waiters(limiter); 634 for (; (limiter->release_count > 0) && !bio_list_empty(waiters); limiter->release_count--) 635 limiter->assigner(limiter); 636 637 if (limiter->release_count > 0) { 638 WRITE_ONCE(limiter->busy, limiter->busy - limiter->release_count); 639 limiter->release_count = 0; 640 return; 641 } 642 643 for (; (available > 0) && !bio_list_empty(waiters); available--) 644 limiter->assigner(limiter); 645 646 WRITE_ONCE(limiter->busy, limiter->limit - available); 647 if (limiter->max_busy < limiter->busy) 648 WRITE_ONCE(limiter->max_busy, limiter->busy); 649 } 650 651 /** 652 * schedule_releases() - Ensure that release processing is scheduled. 653 * 654 * If this call switches the state to processing, enqueue. Otherwise, some other thread has already 655 * done so. 656 */ 657 static void schedule_releases(struct data_vio_pool *pool) 658 { 659 /* Pairs with the barrier in process_release_callback(). */ 660 smp_mb__before_atomic(); 661 if (atomic_cmpxchg(&pool->processing, false, true)) 662 return; 663 664 pool->completion.requeue = true; 665 vdo_launch_completion_with_priority(&pool->completion, 666 CPU_Q_COMPLETE_VIO_PRIORITY); 667 } 668 669 static void reuse_or_release_resources(struct data_vio_pool *pool, 670 struct data_vio *data_vio, 671 struct list_head *returned) 672 { 673 if (data_vio->remaining_discard > 0) { 674 if (bio_list_empty(&pool->discard_limiter.waiters)) { 675 /* Return the data_vio's discard permit. */ 676 pool->discard_limiter.release_count++; 677 } else { 678 assign_discard_permit(&pool->discard_limiter); 679 } 680 } 681 682 if (pool->limiter.arrival < pool->discard_limiter.arrival) { 683 assign_data_vio(&pool->limiter, data_vio); 684 } else if (pool->discard_limiter.arrival < U64_MAX) { 685 assign_data_vio(&pool->discard_limiter, data_vio); 686 } else { 687 list_add(&data_vio->pool_entry, returned); 688 pool->limiter.release_count++; 689 } 690 } 691 692 /** 693 * process_release_callback() - Process a batch of data_vio releases. 694 * @completion: The pool with data_vios to release. 695 */ 696 static void process_release_callback(struct vdo_completion *completion) 697 { 698 struct data_vio_pool *pool = as_data_vio_pool(completion); 699 bool reschedule; 700 bool drained; 701 data_vio_count_t processed; 702 data_vio_count_t to_wake; 703 data_vio_count_t discards_to_wake; 704 LIST_HEAD(returned); 705 706 spin_lock(&pool->lock); 707 get_waiters(&pool->discard_limiter); 708 get_waiters(&pool->limiter); 709 spin_unlock(&pool->lock); 710 711 if (pool->limiter.arrival == U64_MAX) { 712 struct bio *bio = bio_list_peek(&pool->limiter.waiters); 713 714 if (bio != NULL) 715 pool->limiter.arrival = get_arrival_time(bio); 716 } 717 718 for (processed = 0; processed < DATA_VIO_RELEASE_BATCH_SIZE; processed++) { 719 struct data_vio *data_vio; 720 struct funnel_queue_entry *entry = vdo_funnel_queue_poll(pool->queue); 721 722 if (entry == NULL) 723 break; 724 725 data_vio = as_data_vio(container_of(entry, struct vdo_completion, 726 work_queue_entry_link)); 727 acknowledge_data_vio(data_vio); 728 reuse_or_release_resources(pool, data_vio, &returned); 729 } 730 731 spin_lock(&pool->lock); 732 /* 733 * There is a race where waiters could be added while we are in the unlocked section above. 734 * Those waiters could not see the resources we are now about to release, so we assign 735 * those resources now as we have no guarantee of being rescheduled. This is handled in 736 * update_limiter(). 737 */ 738 update_limiter(&pool->discard_limiter); 739 list_splice(&returned, &pool->available); 740 update_limiter(&pool->limiter); 741 to_wake = pool->limiter.wake_count; 742 pool->limiter.wake_count = 0; 743 discards_to_wake = pool->discard_limiter.wake_count; 744 pool->discard_limiter.wake_count = 0; 745 746 atomic_set(&pool->processing, false); 747 /* Pairs with the barrier in schedule_releases(). */ 748 smp_mb(); 749 750 reschedule = !vdo_is_funnel_queue_empty(pool->queue); 751 drained = (!reschedule && 752 vdo_is_state_draining(&pool->state) && 753 check_for_drain_complete_locked(pool)); 754 spin_unlock(&pool->lock); 755 756 if (to_wake > 0) 757 wake_up_nr(&pool->limiter.blocked_threads, to_wake); 758 759 if (discards_to_wake > 0) 760 wake_up_nr(&pool->discard_limiter.blocked_threads, discards_to_wake); 761 762 if (reschedule) 763 schedule_releases(pool); 764 else if (drained) 765 vdo_finish_draining(&pool->state); 766 } 767 768 static void initialize_limiter(struct limiter *limiter, struct data_vio_pool *pool, 769 assigner_fn assigner, data_vio_count_t limit) 770 { 771 limiter->pool = pool; 772 limiter->assigner = assigner; 773 limiter->limit = limit; 774 limiter->arrival = U64_MAX; 775 init_waitqueue_head(&limiter->blocked_threads); 776 } 777 778 /** 779 * initialize_data_vio() - Allocate the components of a data_vio. 780 * 781 * The caller is responsible for cleaning up the data_vio on error. 782 * 783 * Return: VDO_SUCCESS or an error. 784 */ 785 static int initialize_data_vio(struct data_vio *data_vio, struct vdo *vdo) 786 { 787 struct bio *bio; 788 int result; 789 790 BUILD_BUG_ON(VDO_BLOCK_SIZE > PAGE_SIZE); 791 result = vdo_allocate_memory(VDO_BLOCK_SIZE, 0, "data_vio data", 792 &data_vio->vio.data); 793 if (result != VDO_SUCCESS) 794 return vdo_log_error_strerror(result, 795 "data_vio data allocation failure"); 796 797 result = vdo_allocate_memory(VDO_BLOCK_SIZE, 0, "compressed block", 798 &data_vio->compression.block); 799 if (result != VDO_SUCCESS) { 800 return vdo_log_error_strerror(result, 801 "data_vio compressed block allocation failure"); 802 } 803 804 result = vdo_allocate_memory(VDO_BLOCK_SIZE, 0, "vio scratch", 805 &data_vio->scratch_block); 806 if (result != VDO_SUCCESS) 807 return vdo_log_error_strerror(result, 808 "data_vio scratch allocation failure"); 809 810 result = vdo_create_bio(&bio); 811 if (result != VDO_SUCCESS) 812 return vdo_log_error_strerror(result, 813 "data_vio data bio allocation failure"); 814 815 vdo_initialize_completion(&data_vio->decrement_completion, vdo, 816 VDO_DECREMENT_COMPLETION); 817 initialize_vio(&data_vio->vio, bio, 1, VIO_TYPE_DATA, VIO_PRIORITY_DATA, vdo); 818 819 return VDO_SUCCESS; 820 } 821 822 static void destroy_data_vio(struct data_vio *data_vio) 823 { 824 if (data_vio == NULL) 825 return; 826 827 vdo_free_bio(vdo_forget(data_vio->vio.bio)); 828 vdo_free(vdo_forget(data_vio->vio.data)); 829 vdo_free(vdo_forget(data_vio->compression.block)); 830 vdo_free(vdo_forget(data_vio->scratch_block)); 831 } 832 833 /** 834 * make_data_vio_pool() - Initialize a data_vio pool. 835 * @vdo: The vdo to which the pool will belong. 836 * @pool_size: The number of data_vios in the pool. 837 * @discard_limit: The maximum number of data_vios which may be used for discards. 838 * @pool: A pointer to hold the newly allocated pool. 839 */ 840 int make_data_vio_pool(struct vdo *vdo, data_vio_count_t pool_size, 841 data_vio_count_t discard_limit, struct data_vio_pool **pool_ptr) 842 { 843 int result; 844 struct data_vio_pool *pool; 845 data_vio_count_t i; 846 847 result = vdo_allocate_extended(struct data_vio_pool, pool_size, struct data_vio, 848 __func__, &pool); 849 if (result != VDO_SUCCESS) 850 return result; 851 852 VDO_ASSERT_LOG_ONLY((discard_limit <= pool_size), 853 "discard limit does not exceed pool size"); 854 initialize_limiter(&pool->discard_limiter, pool, assign_discard_permit, 855 discard_limit); 856 pool->discard_limiter.permitted_waiters = &pool->permitted_discards; 857 initialize_limiter(&pool->limiter, pool, assign_data_vio_to_waiter, pool_size); 858 pool->limiter.permitted_waiters = &pool->limiter.waiters; 859 INIT_LIST_HEAD(&pool->available); 860 spin_lock_init(&pool->lock); 861 vdo_set_admin_state_code(&pool->state, VDO_ADMIN_STATE_NORMAL_OPERATION); 862 vdo_initialize_completion(&pool->completion, vdo, VDO_DATA_VIO_POOL_COMPLETION); 863 vdo_prepare_completion(&pool->completion, process_release_callback, 864 process_release_callback, vdo->thread_config.cpu_thread, 865 NULL); 866 867 result = vdo_make_funnel_queue(&pool->queue); 868 if (result != VDO_SUCCESS) { 869 free_data_vio_pool(vdo_forget(pool)); 870 return result; 871 } 872 873 for (i = 0; i < pool_size; i++) { 874 struct data_vio *data_vio = &pool->data_vios[i]; 875 876 result = initialize_data_vio(data_vio, vdo); 877 if (result != VDO_SUCCESS) { 878 destroy_data_vio(data_vio); 879 free_data_vio_pool(pool); 880 return result; 881 } 882 883 list_add(&data_vio->pool_entry, &pool->available); 884 } 885 886 *pool_ptr = pool; 887 return VDO_SUCCESS; 888 } 889 890 /** 891 * free_data_vio_pool() - Free a data_vio_pool and the data_vios in it. 892 * 893 * All data_vios must be returned to the pool before calling this function. 894 */ 895 void free_data_vio_pool(struct data_vio_pool *pool) 896 { 897 struct data_vio *data_vio, *tmp; 898 899 if (pool == NULL) 900 return; 901 902 /* 903 * Pairs with the barrier in process_release_callback(). Possibly not needed since it 904 * caters to an enqueue vs. free race. 905 */ 906 smp_mb(); 907 BUG_ON(atomic_read(&pool->processing)); 908 909 spin_lock(&pool->lock); 910 VDO_ASSERT_LOG_ONLY((pool->limiter.busy == 0), 911 "data_vio pool must not have %u busy entries when being freed", 912 pool->limiter.busy); 913 VDO_ASSERT_LOG_ONLY((bio_list_empty(&pool->limiter.waiters) && 914 bio_list_empty(&pool->limiter.new_waiters)), 915 "data_vio pool must not have threads waiting to read or write when being freed"); 916 VDO_ASSERT_LOG_ONLY((bio_list_empty(&pool->discard_limiter.waiters) && 917 bio_list_empty(&pool->discard_limiter.new_waiters)), 918 "data_vio pool must not have threads waiting to discard when being freed"); 919 spin_unlock(&pool->lock); 920 921 list_for_each_entry_safe(data_vio, tmp, &pool->available, pool_entry) { 922 list_del_init(&data_vio->pool_entry); 923 destroy_data_vio(data_vio); 924 } 925 926 vdo_free_funnel_queue(vdo_forget(pool->queue)); 927 vdo_free(pool); 928 } 929 930 static bool acquire_permit(struct limiter *limiter) 931 { 932 if (limiter->busy >= limiter->limit) 933 return false; 934 935 WRITE_ONCE(limiter->busy, limiter->busy + 1); 936 if (limiter->max_busy < limiter->busy) 937 WRITE_ONCE(limiter->max_busy, limiter->busy); 938 return true; 939 } 940 941 static void wait_permit(struct limiter *limiter, struct bio *bio) 942 __releases(&limiter->pool->lock) 943 { 944 DEFINE_WAIT(wait); 945 946 bio_list_add(&limiter->new_waiters, bio); 947 prepare_to_wait_exclusive(&limiter->blocked_threads, &wait, 948 TASK_UNINTERRUPTIBLE); 949 spin_unlock(&limiter->pool->lock); 950 io_schedule(); 951 finish_wait(&limiter->blocked_threads, &wait); 952 } 953 954 /** 955 * vdo_launch_bio() - Acquire a data_vio from the pool, assign the bio to it, and launch it. 956 * 957 * This will block if data_vios or discard permits are not available. 958 */ 959 void vdo_launch_bio(struct data_vio_pool *pool, struct bio *bio) 960 { 961 struct data_vio *data_vio; 962 963 VDO_ASSERT_LOG_ONLY(!vdo_is_state_quiescent(&pool->state), 964 "data_vio_pool not quiescent on acquire"); 965 966 bio->bi_private = (void *) jiffies; 967 spin_lock(&pool->lock); 968 if ((bio_op(bio) == REQ_OP_DISCARD) && 969 !acquire_permit(&pool->discard_limiter)) { 970 wait_permit(&pool->discard_limiter, bio); 971 return; 972 } 973 974 if (!acquire_permit(&pool->limiter)) { 975 wait_permit(&pool->limiter, bio); 976 return; 977 } 978 979 data_vio = get_available_data_vio(pool); 980 spin_unlock(&pool->lock); 981 launch_bio(pool->completion.vdo, data_vio, bio); 982 } 983 984 /* Implements vdo_admin_initiator_fn. */ 985 static void initiate_drain(struct admin_state *state) 986 { 987 bool drained; 988 struct data_vio_pool *pool = container_of(state, struct data_vio_pool, state); 989 990 spin_lock(&pool->lock); 991 drained = check_for_drain_complete_locked(pool); 992 spin_unlock(&pool->lock); 993 994 if (drained) 995 vdo_finish_draining(state); 996 } 997 998 static void assert_on_vdo_cpu_thread(const struct vdo *vdo, const char *name) 999 { 1000 VDO_ASSERT_LOG_ONLY((vdo_get_callback_thread_id() == vdo->thread_config.cpu_thread), 1001 "%s called on cpu thread", name); 1002 } 1003 1004 /** 1005 * drain_data_vio_pool() - Wait asynchronously for all data_vios to be returned to the pool. 1006 * @completion: The completion to notify when the pool has drained. 1007 */ 1008 void drain_data_vio_pool(struct data_vio_pool *pool, struct vdo_completion *completion) 1009 { 1010 assert_on_vdo_cpu_thread(completion->vdo, __func__); 1011 vdo_start_draining(&pool->state, VDO_ADMIN_STATE_SUSPENDING, completion, 1012 initiate_drain); 1013 } 1014 1015 /** 1016 * resume_data_vio_pool() - Resume a data_vio pool. 1017 * @completion: The completion to notify when the pool has resumed. 1018 */ 1019 void resume_data_vio_pool(struct data_vio_pool *pool, struct vdo_completion *completion) 1020 { 1021 assert_on_vdo_cpu_thread(completion->vdo, __func__); 1022 vdo_continue_completion(completion, vdo_resume_if_quiescent(&pool->state)); 1023 } 1024 1025 static void dump_limiter(const char *name, struct limiter *limiter) 1026 { 1027 vdo_log_info("%s: %u of %u busy (max %u), %s", name, limiter->busy, 1028 limiter->limit, limiter->max_busy, 1029 ((bio_list_empty(&limiter->waiters) && 1030 bio_list_empty(&limiter->new_waiters)) ? 1031 "no waiters" : "has waiters")); 1032 } 1033 1034 /** 1035 * dump_data_vio_pool() - Dump a data_vio pool to the log. 1036 * @dump_vios: Whether to dump the details of each busy data_vio as well. 1037 */ 1038 void dump_data_vio_pool(struct data_vio_pool *pool, bool dump_vios) 1039 { 1040 /* 1041 * In order that syslog can empty its buffer, sleep after 35 elements for 4ms (till the 1042 * second clock tick). These numbers were picked based on experiments with lab machines. 1043 */ 1044 static const int ELEMENTS_PER_BATCH = 35; 1045 static const int SLEEP_FOR_SYSLOG = 4000; 1046 1047 if (pool == NULL) 1048 return; 1049 1050 spin_lock(&pool->lock); 1051 dump_limiter("data_vios", &pool->limiter); 1052 dump_limiter("discard permits", &pool->discard_limiter); 1053 if (dump_vios) { 1054 int i; 1055 int dumped = 0; 1056 1057 for (i = 0; i < pool->limiter.limit; i++) { 1058 struct data_vio *data_vio = &pool->data_vios[i]; 1059 1060 if (!list_empty(&data_vio->pool_entry)) 1061 continue; 1062 1063 dump_data_vio(data_vio); 1064 if (++dumped >= ELEMENTS_PER_BATCH) { 1065 spin_unlock(&pool->lock); 1066 dumped = 0; 1067 fsleep(SLEEP_FOR_SYSLOG); 1068 spin_lock(&pool->lock); 1069 } 1070 } 1071 } 1072 1073 spin_unlock(&pool->lock); 1074 } 1075 1076 data_vio_count_t get_data_vio_pool_active_discards(struct data_vio_pool *pool) 1077 { 1078 return READ_ONCE(pool->discard_limiter.busy); 1079 } 1080 1081 data_vio_count_t get_data_vio_pool_discard_limit(struct data_vio_pool *pool) 1082 { 1083 return READ_ONCE(pool->discard_limiter.limit); 1084 } 1085 1086 data_vio_count_t get_data_vio_pool_maximum_discards(struct data_vio_pool *pool) 1087 { 1088 return READ_ONCE(pool->discard_limiter.max_busy); 1089 } 1090 1091 int set_data_vio_pool_discard_limit(struct data_vio_pool *pool, data_vio_count_t limit) 1092 { 1093 if (get_data_vio_pool_request_limit(pool) < limit) { 1094 // The discard limit may not be higher than the data_vio limit. 1095 return -EINVAL; 1096 } 1097 1098 spin_lock(&pool->lock); 1099 pool->discard_limiter.limit = limit; 1100 spin_unlock(&pool->lock); 1101 1102 return VDO_SUCCESS; 1103 } 1104 1105 data_vio_count_t get_data_vio_pool_active_requests(struct data_vio_pool *pool) 1106 { 1107 return READ_ONCE(pool->limiter.busy); 1108 } 1109 1110 data_vio_count_t get_data_vio_pool_request_limit(struct data_vio_pool *pool) 1111 { 1112 return READ_ONCE(pool->limiter.limit); 1113 } 1114 1115 data_vio_count_t get_data_vio_pool_maximum_requests(struct data_vio_pool *pool) 1116 { 1117 return READ_ONCE(pool->limiter.max_busy); 1118 } 1119 1120 static void update_data_vio_error_stats(struct data_vio *data_vio) 1121 { 1122 u8 index = 0; 1123 static const char * const operations[] = { 1124 [0] = "empty", 1125 [1] = "read", 1126 [2] = "write", 1127 [3] = "read-modify-write", 1128 [5] = "read+fua", 1129 [6] = "write+fua", 1130 [7] = "read-modify-write+fua", 1131 }; 1132 1133 if (data_vio->read) 1134 index = 1; 1135 1136 if (data_vio->write) 1137 index += 2; 1138 1139 if (data_vio->fua) 1140 index += 4; 1141 1142 update_vio_error_stats(&data_vio->vio, 1143 "Completing %s vio for LBN %llu with error after %s", 1144 operations[index], 1145 (unsigned long long) data_vio->logical.lbn, 1146 get_data_vio_operation_name(data_vio)); 1147 } 1148 1149 static void perform_cleanup_stage(struct data_vio *data_vio, 1150 enum data_vio_cleanup_stage stage); 1151 1152 /** 1153 * release_allocated_lock() - Release the PBN lock and/or the reference on the allocated block at 1154 * the end of processing a data_vio. 1155 */ 1156 static void release_allocated_lock(struct vdo_completion *completion) 1157 { 1158 struct data_vio *data_vio = as_data_vio(completion); 1159 1160 assert_data_vio_in_allocated_zone(data_vio); 1161 release_data_vio_allocation_lock(data_vio, false); 1162 perform_cleanup_stage(data_vio, VIO_RELEASE_RECOVERY_LOCKS); 1163 } 1164 1165 /** release_lock() - Release an uncontended LBN lock. */ 1166 static void release_lock(struct data_vio *data_vio, struct lbn_lock *lock) 1167 { 1168 struct int_map *lock_map = lock->zone->lbn_operations; 1169 struct data_vio *lock_holder; 1170 1171 if (!lock->locked) { 1172 /* The lock is not locked, so it had better not be registered in the lock map. */ 1173 struct data_vio *lock_holder = vdo_int_map_get(lock_map, lock->lbn); 1174 1175 VDO_ASSERT_LOG_ONLY((data_vio != lock_holder), 1176 "no logical block lock held for block %llu", 1177 (unsigned long long) lock->lbn); 1178 return; 1179 } 1180 1181 /* Release the lock by removing the lock from the map. */ 1182 lock_holder = vdo_int_map_remove(lock_map, lock->lbn); 1183 VDO_ASSERT_LOG_ONLY((data_vio == lock_holder), 1184 "logical block lock mismatch for block %llu", 1185 (unsigned long long) lock->lbn); 1186 lock->locked = false; 1187 } 1188 1189 /** transfer_lock() - Transfer a contended LBN lock to the eldest waiter. */ 1190 static void transfer_lock(struct data_vio *data_vio, struct lbn_lock *lock) 1191 { 1192 struct data_vio *lock_holder, *next_lock_holder; 1193 int result; 1194 1195 VDO_ASSERT_LOG_ONLY(lock->locked, "lbn_lock with waiters is not locked"); 1196 1197 /* Another data_vio is waiting for the lock, transfer it in a single lock map operation. */ 1198 next_lock_holder = 1199 vdo_waiter_as_data_vio(vdo_waitq_dequeue_waiter(&lock->waiters)); 1200 1201 /* Transfer the remaining lock waiters to the next lock holder. */ 1202 vdo_waitq_transfer_all_waiters(&lock->waiters, 1203 &next_lock_holder->logical.waiters); 1204 1205 result = vdo_int_map_put(lock->zone->lbn_operations, lock->lbn, 1206 next_lock_holder, true, (void **) &lock_holder); 1207 if (result != VDO_SUCCESS) { 1208 continue_data_vio_with_error(next_lock_holder, result); 1209 return; 1210 } 1211 1212 VDO_ASSERT_LOG_ONLY((lock_holder == data_vio), 1213 "logical block lock mismatch for block %llu", 1214 (unsigned long long) lock->lbn); 1215 lock->locked = false; 1216 1217 /* 1218 * If there are still waiters, other data_vios must be trying to get the lock we just 1219 * transferred. We must ensure that the new lock holder doesn't block in the packer. 1220 */ 1221 if (vdo_waitq_has_waiters(&next_lock_holder->logical.waiters)) 1222 cancel_data_vio_compression(next_lock_holder); 1223 1224 /* 1225 * Avoid stack overflow on lock transfer. 1226 * FIXME: this is only an issue in the 1 thread config. 1227 */ 1228 next_lock_holder->vio.completion.requeue = true; 1229 launch_locked_request(next_lock_holder); 1230 } 1231 1232 /** 1233 * release_logical_lock() - Release the logical block lock and flush generation lock at the end of 1234 * processing a data_vio. 1235 */ 1236 static void release_logical_lock(struct vdo_completion *completion) 1237 { 1238 struct data_vio *data_vio = as_data_vio(completion); 1239 struct lbn_lock *lock = &data_vio->logical; 1240 1241 assert_data_vio_in_logical_zone(data_vio); 1242 1243 if (vdo_waitq_has_waiters(&lock->waiters)) 1244 transfer_lock(data_vio, lock); 1245 else 1246 release_lock(data_vio, lock); 1247 1248 vdo_release_flush_generation_lock(data_vio); 1249 perform_cleanup_stage(data_vio, VIO_CLEANUP_DONE); 1250 } 1251 1252 /** clean_hash_lock() - Release the hash lock at the end of processing a data_vio. */ 1253 static void clean_hash_lock(struct vdo_completion *completion) 1254 { 1255 struct data_vio *data_vio = as_data_vio(completion); 1256 1257 assert_data_vio_in_hash_zone(data_vio); 1258 if (completion->result != VDO_SUCCESS) { 1259 vdo_clean_failed_hash_lock(data_vio); 1260 return; 1261 } 1262 1263 vdo_release_hash_lock(data_vio); 1264 perform_cleanup_stage(data_vio, VIO_RELEASE_LOGICAL); 1265 } 1266 1267 /** 1268 * finish_cleanup() - Make some assertions about a data_vio which has finished cleaning up. 1269 * 1270 * If it is part of a multi-block discard, starts on the next block, otherwise, returns it to the 1271 * pool. 1272 */ 1273 static void finish_cleanup(struct data_vio *data_vio) 1274 { 1275 struct vdo_completion *completion = &data_vio->vio.completion; 1276 1277 VDO_ASSERT_LOG_ONLY(data_vio->allocation.lock == NULL, 1278 "complete data_vio has no allocation lock"); 1279 VDO_ASSERT_LOG_ONLY(data_vio->hash_lock == NULL, 1280 "complete data_vio has no hash lock"); 1281 if ((data_vio->remaining_discard <= VDO_BLOCK_SIZE) || 1282 (completion->result != VDO_SUCCESS)) { 1283 struct data_vio_pool *pool = completion->vdo->data_vio_pool; 1284 1285 vdo_funnel_queue_put(pool->queue, &completion->work_queue_entry_link); 1286 schedule_releases(pool); 1287 return; 1288 } 1289 1290 data_vio->remaining_discard -= min_t(u32, data_vio->remaining_discard, 1291 VDO_BLOCK_SIZE - data_vio->offset); 1292 data_vio->is_partial = (data_vio->remaining_discard < VDO_BLOCK_SIZE); 1293 data_vio->read = data_vio->is_partial; 1294 data_vio->offset = 0; 1295 completion->requeue = true; 1296 launch_data_vio(data_vio, data_vio->logical.lbn + 1); 1297 } 1298 1299 /** perform_cleanup_stage() - Perform the next step in the process of cleaning up a data_vio. */ 1300 static void perform_cleanup_stage(struct data_vio *data_vio, 1301 enum data_vio_cleanup_stage stage) 1302 { 1303 struct vdo *vdo = vdo_from_data_vio(data_vio); 1304 1305 switch (stage) { 1306 case VIO_RELEASE_HASH_LOCK: 1307 if (data_vio->hash_lock != NULL) { 1308 launch_data_vio_hash_zone_callback(data_vio, clean_hash_lock); 1309 return; 1310 } 1311 fallthrough; 1312 1313 case VIO_RELEASE_ALLOCATED: 1314 if (data_vio_has_allocation(data_vio)) { 1315 launch_data_vio_allocated_zone_callback(data_vio, 1316 release_allocated_lock); 1317 return; 1318 } 1319 fallthrough; 1320 1321 case VIO_RELEASE_RECOVERY_LOCKS: 1322 if ((data_vio->recovery_sequence_number > 0) && 1323 (READ_ONCE(vdo->read_only_notifier.read_only_error) == VDO_SUCCESS) && 1324 (data_vio->vio.completion.result != VDO_READ_ONLY)) 1325 vdo_log_warning("VDO not read-only when cleaning data_vio with RJ lock"); 1326 fallthrough; 1327 1328 case VIO_RELEASE_LOGICAL: 1329 launch_data_vio_logical_callback(data_vio, release_logical_lock); 1330 return; 1331 1332 default: 1333 finish_cleanup(data_vio); 1334 } 1335 } 1336 1337 void complete_data_vio(struct vdo_completion *completion) 1338 { 1339 struct data_vio *data_vio = as_data_vio(completion); 1340 1341 completion->error_handler = NULL; 1342 data_vio->last_async_operation = VIO_ASYNC_OP_CLEANUP; 1343 perform_cleanup_stage(data_vio, 1344 (data_vio->write ? VIO_CLEANUP_START : VIO_RELEASE_LOGICAL)); 1345 } 1346 1347 static void enter_read_only_mode(struct vdo_completion *completion) 1348 { 1349 if (vdo_is_read_only(completion->vdo)) 1350 return; 1351 1352 if (completion->result != VDO_READ_ONLY) { 1353 struct data_vio *data_vio = as_data_vio(completion); 1354 1355 vdo_log_error_strerror(completion->result, 1356 "Preparing to enter read-only mode: data_vio for LBN %llu (becoming mapped to %llu, previously mapped to %llu, allocated %llu) is completing with a fatal error after operation %s", 1357 (unsigned long long) data_vio->logical.lbn, 1358 (unsigned long long) data_vio->new_mapped.pbn, 1359 (unsigned long long) data_vio->mapped.pbn, 1360 (unsigned long long) data_vio->allocation.pbn, 1361 get_data_vio_operation_name(data_vio)); 1362 } 1363 1364 vdo_enter_read_only_mode(completion->vdo, completion->result); 1365 } 1366 1367 void handle_data_vio_error(struct vdo_completion *completion) 1368 { 1369 struct data_vio *data_vio = as_data_vio(completion); 1370 1371 if ((completion->result == VDO_READ_ONLY) || (data_vio->user_bio == NULL)) 1372 enter_read_only_mode(completion); 1373 1374 update_data_vio_error_stats(data_vio); 1375 complete_data_vio(completion); 1376 } 1377 1378 /** 1379 * get_data_vio_operation_name() - Get the name of the last asynchronous operation performed on a 1380 * data_vio. 1381 */ 1382 const char *get_data_vio_operation_name(struct data_vio *data_vio) 1383 { 1384 BUILD_BUG_ON((MAX_VIO_ASYNC_OPERATION_NUMBER - MIN_VIO_ASYNC_OPERATION_NUMBER) != 1385 ARRAY_SIZE(ASYNC_OPERATION_NAMES)); 1386 1387 return ((data_vio->last_async_operation < MAX_VIO_ASYNC_OPERATION_NUMBER) ? 1388 ASYNC_OPERATION_NAMES[data_vio->last_async_operation] : 1389 "unknown async operation"); 1390 } 1391 1392 /** 1393 * data_vio_allocate_data_block() - Allocate a data block. 1394 * 1395 * @write_lock_type: The type of write lock to obtain on the block. 1396 * @callback: The callback which will attempt an allocation in the current zone and continue if it 1397 * succeeds. 1398 * @error_handler: The handler for errors while allocating. 1399 */ 1400 void data_vio_allocate_data_block(struct data_vio *data_vio, 1401 enum pbn_lock_type write_lock_type, 1402 vdo_action_fn callback, vdo_action_fn error_handler) 1403 { 1404 struct allocation *allocation = &data_vio->allocation; 1405 1406 VDO_ASSERT_LOG_ONLY((allocation->pbn == VDO_ZERO_BLOCK), 1407 "data_vio does not have an allocation"); 1408 allocation->write_lock_type = write_lock_type; 1409 allocation->zone = vdo_get_next_allocation_zone(data_vio->logical.zone); 1410 allocation->first_allocation_zone = allocation->zone->zone_number; 1411 1412 data_vio->vio.completion.error_handler = error_handler; 1413 launch_data_vio_allocated_zone_callback(data_vio, callback); 1414 } 1415 1416 /** 1417 * release_data_vio_allocation_lock() - Release the PBN lock on a data_vio's allocated block. 1418 * @reset: If true, the allocation will be reset (i.e. any allocated pbn will be forgotten). 1419 * 1420 * If the reference to the locked block is still provisional, it will be released as well. 1421 */ 1422 void release_data_vio_allocation_lock(struct data_vio *data_vio, bool reset) 1423 { 1424 struct allocation *allocation = &data_vio->allocation; 1425 physical_block_number_t locked_pbn = allocation->pbn; 1426 1427 assert_data_vio_in_allocated_zone(data_vio); 1428 1429 if (reset || vdo_pbn_lock_has_provisional_reference(allocation->lock)) 1430 allocation->pbn = VDO_ZERO_BLOCK; 1431 1432 vdo_release_physical_zone_pbn_lock(allocation->zone, locked_pbn, 1433 vdo_forget(allocation->lock)); 1434 } 1435 1436 /** 1437 * uncompress_data_vio() - Uncompress the data a data_vio has just read. 1438 * @mapping_state: The mapping state indicating which fragment to decompress. 1439 * @buffer: The buffer to receive the uncompressed data. 1440 */ 1441 int uncompress_data_vio(struct data_vio *data_vio, 1442 enum block_mapping_state mapping_state, char *buffer) 1443 { 1444 int size; 1445 u16 fragment_offset, fragment_size; 1446 struct compressed_block *block = data_vio->compression.block; 1447 int result = vdo_get_compressed_block_fragment(mapping_state, block, 1448 &fragment_offset, &fragment_size); 1449 1450 if (result != VDO_SUCCESS) { 1451 vdo_log_debug("%s: compressed fragment error %d", __func__, result); 1452 return result; 1453 } 1454 1455 size = LZ4_decompress_safe((block->data + fragment_offset), buffer, 1456 fragment_size, VDO_BLOCK_SIZE); 1457 if (size != VDO_BLOCK_SIZE) { 1458 vdo_log_debug("%s: lz4 error", __func__); 1459 return VDO_INVALID_FRAGMENT; 1460 } 1461 1462 return VDO_SUCCESS; 1463 } 1464 1465 /** 1466 * modify_for_partial_write() - Do the modify-write part of a read-modify-write cycle. 1467 * @completion: The data_vio which has just finished its read. 1468 * 1469 * This callback is registered in read_block(). 1470 */ 1471 static void modify_for_partial_write(struct vdo_completion *completion) 1472 { 1473 struct data_vio *data_vio = as_data_vio(completion); 1474 char *data = data_vio->vio.data; 1475 struct bio *bio = data_vio->user_bio; 1476 1477 assert_data_vio_on_cpu_thread(data_vio); 1478 1479 if (bio_op(bio) == REQ_OP_DISCARD) { 1480 memset(data + data_vio->offset, '\0', min_t(u32, 1481 data_vio->remaining_discard, 1482 VDO_BLOCK_SIZE - data_vio->offset)); 1483 } else { 1484 copy_from_bio(bio, data + data_vio->offset); 1485 } 1486 1487 data_vio->is_zero = is_zero_block(data); 1488 data_vio->read = false; 1489 launch_data_vio_logical_callback(data_vio, 1490 continue_data_vio_with_block_map_slot); 1491 } 1492 1493 static void complete_read(struct vdo_completion *completion) 1494 { 1495 struct data_vio *data_vio = as_data_vio(completion); 1496 char *data = data_vio->vio.data; 1497 bool compressed = vdo_is_state_compressed(data_vio->mapped.state); 1498 1499 assert_data_vio_on_cpu_thread(data_vio); 1500 1501 if (compressed) { 1502 int result = uncompress_data_vio(data_vio, data_vio->mapped.state, data); 1503 1504 if (result != VDO_SUCCESS) { 1505 continue_data_vio_with_error(data_vio, result); 1506 return; 1507 } 1508 } 1509 1510 if (data_vio->write) { 1511 modify_for_partial_write(completion); 1512 return; 1513 } 1514 1515 if (compressed || data_vio->is_partial) 1516 copy_to_bio(data_vio->user_bio, data + data_vio->offset); 1517 1518 acknowledge_data_vio(data_vio); 1519 complete_data_vio(completion); 1520 } 1521 1522 static void read_endio(struct bio *bio) 1523 { 1524 struct data_vio *data_vio = vio_as_data_vio(bio->bi_private); 1525 int result = blk_status_to_errno(bio->bi_status); 1526 1527 vdo_count_completed_bios(bio); 1528 if (result != VDO_SUCCESS) { 1529 continue_data_vio_with_error(data_vio, result); 1530 return; 1531 } 1532 1533 launch_data_vio_cpu_callback(data_vio, complete_read, 1534 CPU_Q_COMPLETE_READ_PRIORITY); 1535 } 1536 1537 static void complete_zero_read(struct vdo_completion *completion) 1538 { 1539 struct data_vio *data_vio = as_data_vio(completion); 1540 1541 assert_data_vio_on_cpu_thread(data_vio); 1542 1543 if (data_vio->is_partial) { 1544 memset(data_vio->vio.data, 0, VDO_BLOCK_SIZE); 1545 if (data_vio->write) { 1546 modify_for_partial_write(completion); 1547 return; 1548 } 1549 } else { 1550 zero_fill_bio(data_vio->user_bio); 1551 } 1552 1553 complete_read(completion); 1554 } 1555 1556 /** 1557 * read_block() - Read a block asynchronously. 1558 * 1559 * This is the callback registered in read_block_mapping(). 1560 */ 1561 static void read_block(struct vdo_completion *completion) 1562 { 1563 struct data_vio *data_vio = as_data_vio(completion); 1564 struct vio *vio = as_vio(completion); 1565 int result = VDO_SUCCESS; 1566 1567 if (data_vio->mapped.pbn == VDO_ZERO_BLOCK) { 1568 launch_data_vio_cpu_callback(data_vio, complete_zero_read, 1569 CPU_Q_COMPLETE_VIO_PRIORITY); 1570 return; 1571 } 1572 1573 data_vio->last_async_operation = VIO_ASYNC_OP_READ_DATA_VIO; 1574 if (vdo_is_state_compressed(data_vio->mapped.state)) { 1575 result = vio_reset_bio(vio, (char *) data_vio->compression.block, 1576 read_endio, REQ_OP_READ, data_vio->mapped.pbn); 1577 } else { 1578 blk_opf_t opf = ((data_vio->user_bio->bi_opf & PASSTHROUGH_FLAGS) | REQ_OP_READ); 1579 1580 if (data_vio->is_partial) { 1581 result = vio_reset_bio(vio, vio->data, read_endio, opf, 1582 data_vio->mapped.pbn); 1583 } else { 1584 /* A full 4k read. Use the incoming bio to avoid having to copy the data */ 1585 bio_reset(vio->bio, vio->bio->bi_bdev, opf); 1586 bio_init_clone(data_vio->user_bio->bi_bdev, vio->bio, 1587 data_vio->user_bio, GFP_KERNEL); 1588 1589 /* Copy over the original bio iovec and opflags. */ 1590 vdo_set_bio_properties(vio->bio, vio, read_endio, opf, 1591 data_vio->mapped.pbn); 1592 } 1593 } 1594 1595 if (result != VDO_SUCCESS) { 1596 continue_data_vio_with_error(data_vio, result); 1597 return; 1598 } 1599 1600 vdo_submit_data_vio(data_vio); 1601 } 1602 1603 static inline struct data_vio * 1604 reference_count_update_completion_as_data_vio(struct vdo_completion *completion) 1605 { 1606 if (completion->type == VIO_COMPLETION) 1607 return as_data_vio(completion); 1608 1609 return container_of(completion, struct data_vio, decrement_completion); 1610 } 1611 1612 /** 1613 * update_block_map() - Rendezvous of the data_vio and decrement completions after each has 1614 * made its reference updates. Handle any error from either, or proceed 1615 * to updating the block map. 1616 * @completion: The completion of the write in progress. 1617 */ 1618 static void update_block_map(struct vdo_completion *completion) 1619 { 1620 struct data_vio *data_vio = reference_count_update_completion_as_data_vio(completion); 1621 1622 assert_data_vio_in_logical_zone(data_vio); 1623 1624 if (!data_vio->first_reference_operation_complete) { 1625 /* Rendezvous, we're first */ 1626 data_vio->first_reference_operation_complete = true; 1627 return; 1628 } 1629 1630 completion = &data_vio->vio.completion; 1631 vdo_set_completion_result(completion, data_vio->decrement_completion.result); 1632 if (completion->result != VDO_SUCCESS) { 1633 handle_data_vio_error(completion); 1634 return; 1635 } 1636 1637 completion->error_handler = handle_data_vio_error; 1638 if (data_vio->hash_lock != NULL) 1639 set_data_vio_hash_zone_callback(data_vio, vdo_continue_hash_lock); 1640 else 1641 completion->callback = complete_data_vio; 1642 1643 data_vio->last_async_operation = VIO_ASYNC_OP_PUT_MAPPED_BLOCK; 1644 vdo_put_mapped_block(data_vio); 1645 } 1646 1647 static void decrement_reference_count(struct vdo_completion *completion) 1648 { 1649 struct data_vio *data_vio = container_of(completion, struct data_vio, 1650 decrement_completion); 1651 1652 assert_data_vio_in_mapped_zone(data_vio); 1653 1654 vdo_set_completion_callback(completion, update_block_map, 1655 data_vio->logical.zone->thread_id); 1656 completion->error_handler = update_block_map; 1657 vdo_modify_reference_count(completion, &data_vio->decrement_updater); 1658 } 1659 1660 static void increment_reference_count(struct vdo_completion *completion) 1661 { 1662 struct data_vio *data_vio = as_data_vio(completion); 1663 1664 assert_data_vio_in_new_mapped_zone(data_vio); 1665 1666 if (data_vio->downgrade_allocation_lock) { 1667 /* 1668 * Now that the data has been written, it's safe to deduplicate against the 1669 * block. Downgrade the allocation lock to a read lock so it can be used later by 1670 * the hash lock. This is done here since it needs to happen sometime before we 1671 * return to the hash zone, and we are currently on the correct thread. For 1672 * compressed blocks, the downgrade will have already been done. 1673 */ 1674 vdo_downgrade_pbn_write_lock(data_vio->allocation.lock, false); 1675 } 1676 1677 set_data_vio_logical_callback(data_vio, update_block_map); 1678 completion->error_handler = update_block_map; 1679 vdo_modify_reference_count(completion, &data_vio->increment_updater); 1680 } 1681 1682 /** journal_remapping() - Add a recovery journal entry for a data remapping. */ 1683 static void journal_remapping(struct vdo_completion *completion) 1684 { 1685 struct data_vio *data_vio = as_data_vio(completion); 1686 1687 assert_data_vio_in_journal_zone(data_vio); 1688 1689 data_vio->decrement_updater.operation = VDO_JOURNAL_DATA_REMAPPING; 1690 data_vio->decrement_updater.zpbn = data_vio->mapped; 1691 if (data_vio->new_mapped.pbn == VDO_ZERO_BLOCK) { 1692 data_vio->first_reference_operation_complete = true; 1693 if (data_vio->mapped.pbn == VDO_ZERO_BLOCK) 1694 set_data_vio_logical_callback(data_vio, update_block_map); 1695 } else { 1696 set_data_vio_new_mapped_zone_callback(data_vio, 1697 increment_reference_count); 1698 } 1699 1700 if (data_vio->mapped.pbn == VDO_ZERO_BLOCK) { 1701 data_vio->first_reference_operation_complete = true; 1702 } else { 1703 vdo_set_completion_callback(&data_vio->decrement_completion, 1704 decrement_reference_count, 1705 data_vio->mapped.zone->thread_id); 1706 } 1707 1708 data_vio->last_async_operation = VIO_ASYNC_OP_JOURNAL_REMAPPING; 1709 vdo_add_recovery_journal_entry(completion->vdo->recovery_journal, data_vio); 1710 } 1711 1712 /** 1713 * read_old_block_mapping() - Get the previous PBN/LBN mapping of an in-progress write. 1714 * 1715 * Gets the previous PBN mapped to this LBN from the block map, so as to make an appropriate 1716 * journal entry referencing the removal of this LBN->PBN mapping. 1717 */ 1718 static void read_old_block_mapping(struct vdo_completion *completion) 1719 { 1720 struct data_vio *data_vio = as_data_vio(completion); 1721 1722 assert_data_vio_in_logical_zone(data_vio); 1723 1724 data_vio->last_async_operation = VIO_ASYNC_OP_GET_MAPPED_BLOCK_FOR_WRITE; 1725 set_data_vio_journal_callback(data_vio, journal_remapping); 1726 vdo_get_mapped_block(data_vio); 1727 } 1728 1729 void update_metadata_for_data_vio_write(struct data_vio *data_vio, struct pbn_lock *lock) 1730 { 1731 data_vio->increment_updater = (struct reference_updater) { 1732 .operation = VDO_JOURNAL_DATA_REMAPPING, 1733 .increment = true, 1734 .zpbn = data_vio->new_mapped, 1735 .lock = lock, 1736 }; 1737 1738 launch_data_vio_logical_callback(data_vio, read_old_block_mapping); 1739 } 1740 1741 /** 1742 * pack_compressed_data() - Attempt to pack the compressed data_vio into a block. 1743 * 1744 * This is the callback registered in launch_compress_data_vio(). 1745 */ 1746 static void pack_compressed_data(struct vdo_completion *completion) 1747 { 1748 struct data_vio *data_vio = as_data_vio(completion); 1749 1750 assert_data_vio_in_packer_zone(data_vio); 1751 1752 if (!vdo_get_compressing(vdo_from_data_vio(data_vio)) || 1753 get_data_vio_compression_status(data_vio).may_not_compress) { 1754 write_data_vio(data_vio); 1755 return; 1756 } 1757 1758 data_vio->last_async_operation = VIO_ASYNC_OP_ATTEMPT_PACKING; 1759 vdo_attempt_packing(data_vio); 1760 } 1761 1762 /** 1763 * compress_data_vio() - Do the actual work of compressing the data on a CPU queue. 1764 * 1765 * This callback is registered in launch_compress_data_vio(). 1766 */ 1767 static void compress_data_vio(struct vdo_completion *completion) 1768 { 1769 struct data_vio *data_vio = as_data_vio(completion); 1770 int size; 1771 1772 assert_data_vio_on_cpu_thread(data_vio); 1773 1774 /* 1775 * By putting the compressed data at the start of the compressed block data field, we won't 1776 * need to copy it if this data_vio becomes a compressed write agent. 1777 */ 1778 size = LZ4_compress_default(data_vio->vio.data, 1779 data_vio->compression.block->data, VDO_BLOCK_SIZE, 1780 VDO_MAX_COMPRESSED_FRAGMENT_SIZE, 1781 (char *) vdo_get_work_queue_private_data()); 1782 if ((size > 0) && (size < VDO_COMPRESSED_BLOCK_DATA_SIZE)) { 1783 data_vio->compression.size = size; 1784 launch_data_vio_packer_callback(data_vio, pack_compressed_data); 1785 return; 1786 } 1787 1788 write_data_vio(data_vio); 1789 } 1790 1791 /** 1792 * launch_compress_data_vio() - Continue a write by attempting to compress the data. 1793 * 1794 * This is a re-entry point to vio_write used by hash locks. 1795 */ 1796 void launch_compress_data_vio(struct data_vio *data_vio) 1797 { 1798 VDO_ASSERT_LOG_ONLY(!data_vio->is_duplicate, "compressing a non-duplicate block"); 1799 VDO_ASSERT_LOG_ONLY(data_vio->hash_lock != NULL, 1800 "data_vio to compress has a hash_lock"); 1801 VDO_ASSERT_LOG_ONLY(data_vio_has_allocation(data_vio), 1802 "data_vio to compress has an allocation"); 1803 1804 /* 1805 * There are 4 reasons why a data_vio which has reached this point will not be eligible for 1806 * compression: 1807 * 1808 * 1) Since data_vios can block indefinitely in the packer, it would be bad to do so if the 1809 * write request also requests FUA. 1810 * 1811 * 2) A data_vio should not be compressed when compression is disabled for the vdo. 1812 * 1813 * 3) A data_vio could be doing a partial write on behalf of a larger discard which has not 1814 * yet been acknowledged and hence blocking in the packer would be bad. 1815 * 1816 * 4) Some other data_vio may be waiting on this data_vio in which case blocking in the 1817 * packer would also be bad. 1818 */ 1819 if (data_vio->fua || 1820 !vdo_get_compressing(vdo_from_data_vio(data_vio)) || 1821 ((data_vio->user_bio != NULL) && (bio_op(data_vio->user_bio) == REQ_OP_DISCARD)) || 1822 (advance_data_vio_compression_stage(data_vio).stage != DATA_VIO_COMPRESSING)) { 1823 write_data_vio(data_vio); 1824 return; 1825 } 1826 1827 data_vio->last_async_operation = VIO_ASYNC_OP_COMPRESS_DATA_VIO; 1828 launch_data_vio_cpu_callback(data_vio, compress_data_vio, 1829 CPU_Q_COMPRESS_BLOCK_PRIORITY); 1830 } 1831 1832 /** 1833 * hash_data_vio() - Hash the data in a data_vio and set the hash zone (which also flags the record 1834 * name as set). 1835 1836 * This callback is registered in prepare_for_dedupe(). 1837 */ 1838 static void hash_data_vio(struct vdo_completion *completion) 1839 { 1840 struct data_vio *data_vio = as_data_vio(completion); 1841 1842 assert_data_vio_on_cpu_thread(data_vio); 1843 VDO_ASSERT_LOG_ONLY(!data_vio->is_zero, "zero blocks should not be hashed"); 1844 1845 murmurhash3_128(data_vio->vio.data, VDO_BLOCK_SIZE, 0x62ea60be, 1846 &data_vio->record_name); 1847 1848 data_vio->hash_zone = vdo_select_hash_zone(vdo_from_data_vio(data_vio)->hash_zones, 1849 &data_vio->record_name); 1850 data_vio->last_async_operation = VIO_ASYNC_OP_ACQUIRE_VDO_HASH_LOCK; 1851 launch_data_vio_hash_zone_callback(data_vio, vdo_acquire_hash_lock); 1852 } 1853 1854 /** prepare_for_dedupe() - Prepare for the dedupe path after attempting to get an allocation. */ 1855 static void prepare_for_dedupe(struct data_vio *data_vio) 1856 { 1857 /* We don't care what thread we are on. */ 1858 VDO_ASSERT_LOG_ONLY(!data_vio->is_zero, "must not prepare to dedupe zero blocks"); 1859 1860 /* 1861 * Before we can dedupe, we need to know the record name, so the first 1862 * step is to hash the block data. 1863 */ 1864 data_vio->last_async_operation = VIO_ASYNC_OP_HASH_DATA_VIO; 1865 launch_data_vio_cpu_callback(data_vio, hash_data_vio, CPU_Q_HASH_BLOCK_PRIORITY); 1866 } 1867 1868 /** 1869 * write_bio_finished() - This is the bio_end_io function registered in write_block() to be called 1870 * when a data_vio's write to the underlying storage has completed. 1871 */ 1872 static void write_bio_finished(struct bio *bio) 1873 { 1874 struct data_vio *data_vio = vio_as_data_vio((struct vio *) bio->bi_private); 1875 1876 vdo_count_completed_bios(bio); 1877 vdo_set_completion_result(&data_vio->vio.completion, 1878 blk_status_to_errno(bio->bi_status)); 1879 data_vio->downgrade_allocation_lock = true; 1880 update_metadata_for_data_vio_write(data_vio, data_vio->allocation.lock); 1881 } 1882 1883 /** write_data_vio() - Write a data block to storage without compression. */ 1884 void write_data_vio(struct data_vio *data_vio) 1885 { 1886 struct data_vio_compression_status status, new_status; 1887 int result; 1888 1889 if (!data_vio_has_allocation(data_vio)) { 1890 /* 1891 * There was no space to write this block and we failed to deduplicate or compress 1892 * it. 1893 */ 1894 continue_data_vio_with_error(data_vio, VDO_NO_SPACE); 1895 return; 1896 } 1897 1898 new_status = (struct data_vio_compression_status) { 1899 .stage = DATA_VIO_POST_PACKER, 1900 .may_not_compress = true, 1901 }; 1902 1903 do { 1904 status = get_data_vio_compression_status(data_vio); 1905 } while ((status.stage != DATA_VIO_POST_PACKER) && 1906 !set_data_vio_compression_status(data_vio, status, new_status)); 1907 1908 /* Write the data from the data block buffer. */ 1909 result = vio_reset_bio(&data_vio->vio, data_vio->vio.data, 1910 write_bio_finished, REQ_OP_WRITE, 1911 data_vio->allocation.pbn); 1912 if (result != VDO_SUCCESS) { 1913 continue_data_vio_with_error(data_vio, result); 1914 return; 1915 } 1916 1917 data_vio->last_async_operation = VIO_ASYNC_OP_WRITE_DATA_VIO; 1918 vdo_submit_data_vio(data_vio); 1919 } 1920 1921 /** 1922 * acknowledge_write_callback() - Acknowledge a write to the requestor. 1923 * 1924 * This callback is registered in allocate_block() and continue_write_with_block_map_slot(). 1925 */ 1926 static void acknowledge_write_callback(struct vdo_completion *completion) 1927 { 1928 struct data_vio *data_vio = as_data_vio(completion); 1929 struct vdo *vdo = completion->vdo; 1930 1931 VDO_ASSERT_LOG_ONLY((!vdo_uses_bio_ack_queue(vdo) || 1932 (vdo_get_callback_thread_id() == vdo->thread_config.bio_ack_thread)), 1933 "%s() called on bio ack queue", __func__); 1934 VDO_ASSERT_LOG_ONLY(data_vio_has_flush_generation_lock(data_vio), 1935 "write VIO to be acknowledged has a flush generation lock"); 1936 acknowledge_data_vio(data_vio); 1937 if (data_vio->new_mapped.pbn == VDO_ZERO_BLOCK) { 1938 /* This is a zero write or discard */ 1939 update_metadata_for_data_vio_write(data_vio, NULL); 1940 return; 1941 } 1942 1943 prepare_for_dedupe(data_vio); 1944 } 1945 1946 /** 1947 * allocate_block() - Attempt to allocate a block in the current allocation zone. 1948 * 1949 * This callback is registered in continue_write_with_block_map_slot(). 1950 */ 1951 static void allocate_block(struct vdo_completion *completion) 1952 { 1953 struct data_vio *data_vio = as_data_vio(completion); 1954 1955 assert_data_vio_in_allocated_zone(data_vio); 1956 1957 if (!vdo_allocate_block_in_zone(data_vio)) 1958 return; 1959 1960 completion->error_handler = handle_data_vio_error; 1961 WRITE_ONCE(data_vio->allocation_succeeded, true); 1962 data_vio->new_mapped = (struct zoned_pbn) { 1963 .zone = data_vio->allocation.zone, 1964 .pbn = data_vio->allocation.pbn, 1965 .state = VDO_MAPPING_STATE_UNCOMPRESSED, 1966 }; 1967 1968 if (data_vio->fua) { 1969 prepare_for_dedupe(data_vio); 1970 return; 1971 } 1972 1973 data_vio->last_async_operation = VIO_ASYNC_OP_ACKNOWLEDGE_WRITE; 1974 launch_data_vio_on_bio_ack_queue(data_vio, acknowledge_write_callback); 1975 } 1976 1977 /** 1978 * handle_allocation_error() - Handle an error attempting to allocate a block. 1979 * 1980 * This error handler is registered in continue_write_with_block_map_slot(). 1981 */ 1982 static void handle_allocation_error(struct vdo_completion *completion) 1983 { 1984 struct data_vio *data_vio = as_data_vio(completion); 1985 1986 if (completion->result == VDO_NO_SPACE) { 1987 /* We failed to get an allocation, but we can try to dedupe. */ 1988 vdo_reset_completion(completion); 1989 completion->error_handler = handle_data_vio_error; 1990 prepare_for_dedupe(data_vio); 1991 return; 1992 } 1993 1994 /* We got a "real" error, not just a failure to allocate, so fail the request. */ 1995 handle_data_vio_error(completion); 1996 } 1997 1998 static int assert_is_discard(struct data_vio *data_vio) 1999 { 2000 int result = VDO_ASSERT(data_vio->is_discard, 2001 "data_vio with no block map page is a discard"); 2002 2003 return ((result == VDO_SUCCESS) ? result : VDO_READ_ONLY); 2004 } 2005 2006 /** 2007 * continue_data_vio_with_block_map_slot() - Read the data_vio's mapping from the block map. 2008 * 2009 * This callback is registered in launch_read_data_vio(). 2010 */ 2011 void continue_data_vio_with_block_map_slot(struct vdo_completion *completion) 2012 { 2013 struct data_vio *data_vio = as_data_vio(completion); 2014 2015 assert_data_vio_in_logical_zone(data_vio); 2016 if (data_vio->read) { 2017 set_data_vio_logical_callback(data_vio, read_block); 2018 data_vio->last_async_operation = VIO_ASYNC_OP_GET_MAPPED_BLOCK_FOR_READ; 2019 vdo_get_mapped_block(data_vio); 2020 return; 2021 } 2022 2023 vdo_acquire_flush_generation_lock(data_vio); 2024 2025 if (data_vio->tree_lock.tree_slots[0].block_map_slot.pbn == VDO_ZERO_BLOCK) { 2026 /* 2027 * This is a discard for a block on a block map page which has not been allocated, so 2028 * there's nothing more we need to do. 2029 */ 2030 completion->callback = complete_data_vio; 2031 continue_data_vio_with_error(data_vio, assert_is_discard(data_vio)); 2032 return; 2033 } 2034 2035 /* 2036 * We need an allocation if this is neither a full-block discard nor a 2037 * full-block zero write. 2038 */ 2039 if (!data_vio->is_zero && (!data_vio->is_discard || data_vio->is_partial)) { 2040 data_vio_allocate_data_block(data_vio, VIO_WRITE_LOCK, allocate_block, 2041 handle_allocation_error); 2042 return; 2043 } 2044 2045 2046 /* 2047 * We don't need to write any data, so skip allocation and just update the block map and 2048 * reference counts (via the journal). 2049 */ 2050 data_vio->new_mapped.pbn = VDO_ZERO_BLOCK; 2051 if (data_vio->is_zero) 2052 data_vio->new_mapped.state = VDO_MAPPING_STATE_UNCOMPRESSED; 2053 2054 if (data_vio->remaining_discard > VDO_BLOCK_SIZE) { 2055 /* This is not the final block of a discard so we can't acknowledge it yet. */ 2056 update_metadata_for_data_vio_write(data_vio, NULL); 2057 return; 2058 } 2059 2060 data_vio->last_async_operation = VIO_ASYNC_OP_ACKNOWLEDGE_WRITE; 2061 launch_data_vio_on_bio_ack_queue(data_vio, acknowledge_write_callback); 2062 } 2063