1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * linux/include/linux/jbd2.h 4 * 5 * Written by Stephen C. Tweedie <sct@redhat.com> 6 * 7 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved 8 * 9 * Definitions for transaction data structures for the buffer cache 10 * filesystem journaling support. 11 */ 12 13 #ifndef _LINUX_JBD2_H 14 #define _LINUX_JBD2_H 15 16 /* Allow this file to be included directly into e2fsprogs */ 17 #ifndef __KERNEL__ 18 #include "jfs_compat.h" 19 #define JBD2_DEBUG 20 #else 21 22 #include <linux/types.h> 23 #include <linux/buffer_head.h> 24 #include <linux/journal-head.h> 25 #include <linux/stddef.h> 26 #include <linux/mutex.h> 27 #include <linux/timer.h> 28 #include <linux/slab.h> 29 #include <linux/bit_spinlock.h> 30 #include <linux/blkdev.h> 31 #include <linux/crc32c.h> 32 #endif 33 34 #define journal_oom_retry 1 35 36 /* 37 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds 38 * certain classes of error which can occur due to failed IOs. Under 39 * normal use we want ext4 to continue after such errors, because 40 * hardware _can_ fail, but for debugging purposes when running tests on 41 * known-good hardware we may want to trap these errors. 42 */ 43 #undef JBD2_PARANOID_IOFAIL 44 45 /* 46 * The default maximum commit age, in seconds. 47 */ 48 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5 49 50 #ifdef CONFIG_JBD2_DEBUG 51 /* 52 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal 53 * consistency checks. By default we don't do this unless 54 * CONFIG_JBD2_DEBUG is on. 55 */ 56 #define JBD2_EXPENSIVE_CHECKING 57 void __jbd2_debug(int level, const char *file, const char *func, 58 unsigned int line, const char *fmt, ...); 59 60 #define jbd2_debug(n, fmt, a...) \ 61 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a) 62 #else 63 #define jbd2_debug(n, fmt, a...) no_printk(fmt, ##a) 64 #endif 65 66 extern void *jbd2_alloc(size_t size, gfp_t flags); 67 extern void jbd2_free(void *ptr, size_t size); 68 69 #define JBD2_MIN_JOURNAL_BLOCKS 1024 70 #define JBD2_DEFAULT_FAST_COMMIT_BLOCKS 256 71 72 #ifdef __KERNEL__ 73 74 /** 75 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 76 * 77 * All filesystem modifications made by the process go 78 * through this handle. Recursive operations (such as quota operations) 79 * are gathered into a single update. 80 * 81 * The buffer credits field is used to account for journaled buffers 82 * being modified by the running process. To ensure that there is 83 * enough log space for all outstanding operations, we need to limit the 84 * number of outstanding buffers possible at any time. When the 85 * operation completes, any buffer credits not used are credited back to 86 * the transaction, so that at all times we know how many buffers the 87 * outstanding updates on a transaction might possibly touch. 88 * 89 * This is an opaque datatype. 90 **/ 91 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ 92 93 94 /** 95 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 96 * 97 * journal_t is linked to from the fs superblock structure. 98 * 99 * We use the journal_t to keep track of all outstanding transaction 100 * activity on the filesystem, and to manage the state of the log 101 * writing process. 102 * 103 * This is an opaque datatype. 104 **/ 105 typedef struct journal_s journal_t; /* Journal control structure */ 106 #endif 107 108 /* 109 * Internal structures used by the logging mechanism: 110 */ 111 112 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 113 114 /* 115 * On-disk structures 116 */ 117 118 /* 119 * Descriptor block types: 120 */ 121 122 #define JBD2_DESCRIPTOR_BLOCK 1 123 #define JBD2_COMMIT_BLOCK 2 124 #define JBD2_SUPERBLOCK_V1 3 125 #define JBD2_SUPERBLOCK_V2 4 126 #define JBD2_REVOKE_BLOCK 5 127 128 /* 129 * Standard header for all descriptor blocks: 130 */ 131 typedef struct journal_header_s 132 { 133 __be32 h_magic; 134 __be32 h_blocktype; 135 __be32 h_sequence; 136 } journal_header_t; 137 138 /* 139 * Checksum types. 140 */ 141 #define JBD2_CRC32_CHKSUM 1 142 #define JBD2_MD5_CHKSUM 2 143 #define JBD2_SHA1_CHKSUM 3 144 #define JBD2_CRC32C_CHKSUM 4 145 146 #define JBD2_CRC32_CHKSUM_SIZE 4 147 148 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 149 /* 150 * Commit block header for storing transactional checksums: 151 * 152 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* 153 * fields are used to store a checksum of the descriptor and data blocks. 154 * 155 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum 156 * field is used to store crc32c(uuid+commit_block). Each journal metadata 157 * block gets its own checksum, and data block checksums are stored in 158 * journal_block_tag (in the descriptor). The other h_chksum* fields are 159 * not used. 160 * 161 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses 162 * journal_block_tag3_t to store a full 32-bit checksum. Everything else 163 * is the same as v2. 164 * 165 * Checksum v1, v2, and v3 are mutually exclusive features. 166 */ 167 struct commit_header { 168 __be32 h_magic; 169 __be32 h_blocktype; 170 __be32 h_sequence; 171 unsigned char h_chksum_type; 172 unsigned char h_chksum_size; 173 unsigned char h_padding[2]; 174 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 175 __be64 h_commit_sec; 176 __be32 h_commit_nsec; 177 }; 178 179 /* 180 * The block tag: used to describe a single buffer in the journal. 181 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 182 * raw struct shouldn't be used for pointer math or sizeof() - use 183 * journal_tag_bytes(journal) instead to compute this. 184 */ 185 typedef struct journal_block_tag3_s 186 { 187 __be32 t_blocknr; /* The on-disk block number */ 188 __be32 t_flags; /* See below */ 189 __be32 t_blocknr_high; /* most-significant high 32bits. */ 190 __be32 t_checksum; /* crc32c(uuid+seq+block) */ 191 } journal_block_tag3_t; 192 193 typedef struct journal_block_tag_s 194 { 195 __be32 t_blocknr; /* The on-disk block number */ 196 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ 197 __be16 t_flags; /* See below */ 198 __be32 t_blocknr_high; /* most-significant high 32bits. */ 199 } journal_block_tag_t; 200 201 /* Tail of descriptor or revoke block, for checksumming */ 202 struct jbd2_journal_block_tail { 203 __be32 t_checksum; /* crc32c(uuid+descr_block) */ 204 }; 205 206 /* 207 * The revoke descriptor: used on disk to describe a series of blocks to 208 * be revoked from the log 209 */ 210 typedef struct jbd2_journal_revoke_header_s 211 { 212 journal_header_t r_header; 213 __be32 r_count; /* Count of bytes used in the block */ 214 } jbd2_journal_revoke_header_t; 215 216 /* Definitions for the journal tag flags word: */ 217 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 218 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 219 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 220 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 221 222 223 /* 224 * The journal superblock. All fields are in big-endian byte order. 225 */ 226 typedef struct journal_superblock_s 227 { 228 /* 0x0000 */ 229 journal_header_t s_header; 230 231 /* 0x000C */ 232 /* Static information describing the journal */ 233 __be32 s_blocksize; /* journal device blocksize */ 234 __be32 s_maxlen; /* total blocks in journal file */ 235 __be32 s_first; /* first block of log information */ 236 237 /* 0x0018 */ 238 /* Dynamic information describing the current state of the log */ 239 __be32 s_sequence; /* first commit ID expected in log */ 240 __be32 s_start; /* blocknr of start of log */ 241 242 /* 0x0020 */ 243 /* Error value, as set by jbd2_journal_abort(). */ 244 __be32 s_errno; 245 246 /* 0x0024 */ 247 /* Remaining fields are only valid in a version-2 superblock */ 248 __be32 s_feature_compat; /* compatible feature set */ 249 __be32 s_feature_incompat; /* incompatible feature set */ 250 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 251 /* 0x0030 */ 252 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 253 254 /* 0x0040 */ 255 __be32 s_nr_users; /* Nr of filesystems sharing log */ 256 257 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 258 259 /* 0x0048 */ 260 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 261 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 262 263 /* 0x0050 */ 264 __u8 s_checksum_type; /* checksum type */ 265 __u8 s_padding2[3]; 266 /* 0x0054 */ 267 __be32 s_num_fc_blks; /* Number of fast commit blocks */ 268 __be32 s_head; /* blocknr of head of log, only uptodate 269 * while the filesystem is clean */ 270 /* 0x005C */ 271 __u32 s_padding[40]; 272 __be32 s_checksum; /* crc32c(superblock) */ 273 274 /* 0x0100 */ 275 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 276 /* 0x0400 */ 277 } journal_superblock_t; 278 279 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 280 281 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 282 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 283 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 284 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 285 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 286 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT 0x00000020 287 288 /* See "journal feature predicate functions" below */ 289 290 /* Features known to this kernel version: */ 291 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 292 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 293 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 294 JBD2_FEATURE_INCOMPAT_64BIT | \ 295 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 296 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 297 JBD2_FEATURE_INCOMPAT_CSUM_V3 | \ 298 JBD2_FEATURE_INCOMPAT_FAST_COMMIT) 299 300 #ifdef __KERNEL__ 301 302 #include <linux/fs.h> 303 #include <linux/sched.h> 304 305 enum jbd_state_bits { 306 BH_JBD /* Has an attached ext3 journal_head */ 307 = BH_PrivateStart, 308 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 309 BH_Freed, /* Has been freed (truncated) */ 310 BH_Revoked, /* Has been revoked from the log */ 311 BH_RevokeValid, /* Revoked flag is valid */ 312 BH_JBDDirty, /* Is dirty but journaled */ 313 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 314 BH_Shadow, /* IO on shadow buffer is running */ 315 BH_Verified, /* Metadata block has been verified ok */ 316 BH_JBDPrivateStart, /* First bit available for private use by FS */ 317 }; 318 319 BUFFER_FNS(JBD, jbd) 320 BUFFER_FNS(JWrite, jwrite) 321 BUFFER_FNS(JBDDirty, jbddirty) 322 TAS_BUFFER_FNS(JBDDirty, jbddirty) 323 BUFFER_FNS(Revoked, revoked) 324 TAS_BUFFER_FNS(Revoked, revoked) 325 BUFFER_FNS(RevokeValid, revokevalid) 326 TAS_BUFFER_FNS(RevokeValid, revokevalid) 327 BUFFER_FNS(Freed, freed) 328 BUFFER_FNS(Shadow, shadow) 329 BUFFER_FNS(Verified, verified) 330 331 static inline struct buffer_head *jh2bh(struct journal_head *jh) 332 { 333 return jh->b_bh; 334 } 335 336 static inline struct journal_head *bh2jh(struct buffer_head *bh) 337 { 338 return bh->b_private; 339 } 340 341 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 342 { 343 bit_spin_lock(BH_JournalHead, &bh->b_state); 344 } 345 346 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 347 { 348 bit_spin_unlock(BH_JournalHead, &bh->b_state); 349 } 350 351 #define J_ASSERT(assert) BUG_ON(!(assert)) 352 353 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 354 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 355 356 #if defined(JBD2_PARANOID_IOFAIL) 357 #define J_EXPECT(expr, why...) J_ASSERT(expr) 358 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 359 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 360 #else 361 #define __journal_expect(expr, why...) \ 362 ({ \ 363 int val = (expr); \ 364 if (!val) { \ 365 printk(KERN_ERR \ 366 "JBD2 unexpected failure: %s: %s;\n", \ 367 __func__, #expr); \ 368 printk(KERN_ERR why "\n"); \ 369 } \ 370 val; \ 371 }) 372 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 373 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 374 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 375 #endif 376 377 /* Flags in jbd_inode->i_flags */ 378 #define __JI_COMMIT_RUNNING 0 379 #define __JI_WRITE_DATA 1 380 #define __JI_WAIT_DATA 2 381 382 /* 383 * Commit of the inode data in progress. We use this flag to protect us from 384 * concurrent deletion of inode. We cannot use reference to inode for this 385 * since we cannot afford doing last iput() on behalf of kjournald 386 */ 387 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 388 /* Write allocated dirty buffers in this inode before commit */ 389 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA) 390 /* Wait for outstanding data writes for this inode before commit */ 391 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA) 392 393 /** 394 * struct jbd2_inode - The jbd_inode type is the structure linking inodes in 395 * ordered mode present in a transaction so that we can sync them during commit. 396 */ 397 struct jbd2_inode { 398 /** 399 * @i_transaction: 400 * 401 * Which transaction does this inode belong to? Either the running 402 * transaction or the committing one. [j_list_lock] 403 */ 404 transaction_t *i_transaction; 405 406 /** 407 * @i_next_transaction: 408 * 409 * Pointer to the running transaction modifying inode's data in case 410 * there is already a committing transaction touching it. [j_list_lock] 411 */ 412 transaction_t *i_next_transaction; 413 414 /** 415 * @i_list: List of inodes in the i_transaction [j_list_lock] 416 */ 417 struct list_head i_list; 418 419 /** 420 * @i_vfs_inode: 421 * 422 * VFS inode this inode belongs to [constant for lifetime of structure] 423 */ 424 struct inode *i_vfs_inode; 425 426 /** 427 * @i_flags: Flags of inode [j_list_lock] 428 */ 429 unsigned long i_flags; 430 431 /** 432 * @i_dirty_start_page: 433 * 434 * Dirty range start in PAGE_SIZE units. 435 * 436 * The dirty range is empty if @i_dirty_start_page is greater than or 437 * equal to @i_dirty_end_page. 438 * 439 * [j_list_lock] 440 */ 441 pgoff_t i_dirty_start_page; 442 443 /** 444 * @i_dirty_end_page: 445 * 446 * Dirty range end in PAGE_SIZE units (exclusive). 447 * 448 * [j_list_lock] 449 */ 450 pgoff_t i_dirty_end_page; 451 }; 452 453 /* 454 * Lockless readers treat start_page >= end_page as an empty range. 455 * Writers publish a new non-empty range by storing i_dirty_end_page before 456 * i_dirty_start_page. 457 */ 458 static inline bool jbd2_jinode_get_dirty_range(const struct jbd2_inode *jinode, 459 loff_t *start, loff_t *end) 460 { 461 pgoff_t start_page = READ_ONCE(jinode->i_dirty_start_page); 462 pgoff_t end_page = READ_ONCE(jinode->i_dirty_end_page); 463 464 if (start_page >= end_page) 465 return false; 466 467 *start = (loff_t)start_page << PAGE_SHIFT; 468 *end = ((loff_t)end_page << PAGE_SHIFT) - 1; 469 return true; 470 } 471 472 struct jbd2_revoke_table_s; 473 474 /** 475 * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete 476 * type associated with handle_t. 477 * @h_transaction: Which compound transaction is this update a part of? 478 * @h_journal: Which journal handle belongs to - used iff h_reserved set. 479 * @h_rsv_handle: Handle reserved for finishing the logical operation. 480 * @h_total_credits: Number of remaining buffers we are allowed to add to 481 * journal. These are dirty buffers and revoke descriptor blocks. 482 * @h_revoke_credits: Number of remaining revoke records available for handle 483 * @h_ref: Reference count on this handle. 484 * @h_err: Field for caller's use to track errors through large fs operations. 485 * @h_sync: Flag for sync-on-close. 486 * @h_reserved: Flag for handle for reserved credits. 487 * @h_aborted: Flag indicating fatal error on handle. 488 * @h_type: For handle statistics. 489 * @h_line_no: For handle statistics. 490 * @h_start_jiffies: Handle Start time. 491 * @h_requested_credits: Holds @h_total_credits after handle is started. 492 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started. 493 * @saved_alloc_context: Saved context while transaction is open. 494 **/ 495 496 /* Docbook can't yet cope with the bit fields, but will leave the documentation 497 * in so it can be fixed later. 498 */ 499 500 struct jbd2_journal_handle 501 { 502 union { 503 transaction_t *h_transaction; 504 /* Which journal handle belongs to - used iff h_reserved set */ 505 journal_t *h_journal; 506 }; 507 508 handle_t *h_rsv_handle; 509 int h_total_credits; 510 int h_revoke_credits; 511 int h_revoke_credits_requested; 512 int h_ref; 513 int h_err; 514 515 /* Flags [no locking] */ 516 unsigned int h_sync: 1; 517 unsigned int h_reserved: 1; 518 unsigned int h_aborted: 1; 519 unsigned int h_type: 8; 520 unsigned int h_line_no: 16; 521 522 unsigned long h_start_jiffies; 523 unsigned int h_requested_credits; 524 525 unsigned int saved_alloc_context; 526 }; 527 528 529 /* 530 * Some stats for checkpoint phase 531 */ 532 struct transaction_chp_stats_s { 533 unsigned long cs_chp_time; 534 __u32 cs_forced_to_close; 535 __u32 cs_written; 536 __u32 cs_dropped; 537 }; 538 539 /* The transaction_t type is the guts of the journaling mechanism. It 540 * tracks a compound transaction through its various states: 541 * 542 * RUNNING: accepting new updates 543 * LOCKED: Updates still running but we don't accept new ones 544 * RUNDOWN: Updates are tidying up but have finished requesting 545 * new buffers to modify (state not used for now) 546 * FLUSH: All updates complete, but we are still writing to disk 547 * COMMIT: All data on disk, writing commit record 548 * FINISHED: We still have to keep the transaction for checkpointing. 549 * 550 * The transaction keeps track of all of the buffers modified by a 551 * running transaction, and all of the buffers committed but not yet 552 * flushed to home for finished transactions. 553 * (Locking Documentation improved by LockDoc) 554 */ 555 556 /* 557 * Lock ranking: 558 * 559 * j_list_lock 560 * ->jbd_lock_bh_journal_head() (This is "innermost") 561 * 562 * j_state_lock 563 * ->b_state_lock 564 * 565 * b_state_lock 566 * ->j_list_lock 567 * 568 * j_state_lock 569 * ->j_list_lock (journal_unmap_buffer) 570 * 571 */ 572 573 struct transaction_s 574 { 575 /* Pointer to the journal for this transaction. [no locking] */ 576 journal_t *t_journal; 577 578 /* Sequence number for this transaction [no locking] */ 579 tid_t t_tid; 580 581 /* 582 * Transaction's current state 583 * [no locking - only kjournald2 alters this] 584 * [j_list_lock] guards transition of a transaction into T_FINISHED 585 * state and subsequent call of __jbd2_journal_drop_transaction() 586 * FIXME: needs barriers 587 * KLUDGE: [use j_state_lock] 588 */ 589 enum { 590 T_RUNNING, 591 T_LOCKED, 592 T_SWITCH, 593 T_FLUSH, 594 T_COMMIT, 595 T_COMMIT_DFLUSH, 596 T_COMMIT_JFLUSH, 597 T_COMMIT_CALLBACK, 598 T_FINISHED 599 } t_state; 600 601 /* 602 * Where in the log does this transaction's commit start? [no locking] 603 */ 604 unsigned long t_log_start; 605 606 /* 607 * Number of buffers on the t_buffers list [j_list_lock, no locks 608 * needed for jbd2 thread] 609 */ 610 int t_nr_buffers; 611 612 /* 613 * Doubly-linked circular list of all buffers reserved but not yet 614 * modified by this transaction [j_list_lock, no locks needed fo 615 * jbd2 thread] 616 */ 617 struct journal_head *t_reserved_list; 618 619 /* 620 * Doubly-linked circular list of all metadata buffers owned by this 621 * transaction [j_list_lock, no locks needed for jbd2 thread] 622 */ 623 struct journal_head *t_buffers; 624 625 /* 626 * Doubly-linked circular list of all forget buffers (superseded 627 * buffers which we can un-checkpoint once this transaction commits) 628 * [j_list_lock] 629 */ 630 struct journal_head *t_forget; 631 632 /* 633 * Doubly-linked circular list of all buffers still to be flushed before 634 * this transaction can be checkpointed. [j_list_lock] 635 */ 636 struct journal_head *t_checkpoint_list; 637 638 /* 639 * Doubly-linked circular list of metadata buffers being 640 * shadowed by log IO. The IO buffers on the iobuf list and 641 * the shadow buffers on this list match each other one for 642 * one at all times. [j_list_lock, no locks needed for jbd2 643 * thread] 644 */ 645 struct journal_head *t_shadow_list; 646 647 /* 648 * List of inodes associated with the transaction; e.g., ext4 uses 649 * this to track inodes in data=ordered and data=journal mode that 650 * need special handling on transaction commit; also used by ocfs2. 651 * [j_list_lock] 652 */ 653 struct list_head t_inode_list; 654 655 /* 656 * Longest time some handle had to wait for running transaction 657 */ 658 unsigned long t_max_wait; 659 660 /* 661 * When transaction started 662 */ 663 unsigned long t_start; 664 665 /* 666 * When commit was requested [j_state_lock] 667 */ 668 unsigned long t_requested; 669 670 /* 671 * Checkpointing stats [j_list_lock] 672 */ 673 struct transaction_chp_stats_s t_chp_stats; 674 675 /* 676 * Number of outstanding updates running on this transaction 677 * [none] 678 */ 679 atomic_t t_updates; 680 681 /* 682 * Number of blocks reserved for this transaction in the journal. 683 * This is including all credits reserved when starting transaction 684 * handles as well as all journal descriptor blocks needed for this 685 * transaction. [none] 686 */ 687 atomic_t t_outstanding_credits; 688 689 /* 690 * Number of revoke records for this transaction added by already 691 * stopped handles. [none] 692 */ 693 atomic_t t_outstanding_revokes; 694 695 /* 696 * How many handles used this transaction? [none] 697 */ 698 atomic_t t_handle_count; 699 700 /* 701 * Forward and backward links for the circular list of all transactions 702 * awaiting checkpoint. [j_list_lock] 703 */ 704 transaction_t *t_cpnext, *t_cpprev; 705 706 /* 707 * When will the transaction expire (become due for commit), in jiffies? 708 * [no locking] 709 */ 710 unsigned long t_expires; 711 712 /* 713 * When this transaction started, in nanoseconds [no locking] 714 */ 715 ktime_t t_start_time; 716 717 /* 718 * This transaction is being forced and some process is 719 * waiting for it to finish. 720 */ 721 unsigned int t_synchronous_commit:1; 722 723 /* Disk flush needs to be sent to fs partition [no locking] */ 724 int t_need_data_flush; 725 }; 726 727 struct transaction_run_stats_s { 728 unsigned long rs_wait; 729 unsigned long rs_request_delay; 730 unsigned long rs_running; 731 unsigned long rs_locked; 732 unsigned long rs_flushing; 733 unsigned long rs_logging; 734 735 __u32 rs_handle_count; 736 __u32 rs_blocks; 737 __u32 rs_blocks_logged; 738 }; 739 740 struct transaction_stats_s { 741 unsigned long ts_tid; 742 unsigned long ts_requested; 743 struct transaction_run_stats_s run; 744 }; 745 746 static inline unsigned long 747 jbd2_time_diff(unsigned long start, unsigned long end) 748 { 749 if (end >= start) 750 return end - start; 751 752 return end + (MAX_JIFFY_OFFSET - start); 753 } 754 755 #define JBD2_NR_BATCH 64 756 757 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY}; 758 759 #define JBD2_FC_REPLAY_STOP 0 760 #define JBD2_FC_REPLAY_CONTINUE 1 761 762 /** 763 * struct journal_s - The journal_s type is the concrete type associated with 764 * journal_t. 765 */ 766 struct journal_s 767 { 768 /** 769 * @j_flags: General journaling state flags [j_state_lock, 770 * no lock for quick racy checks] 771 */ 772 unsigned long j_flags; 773 774 /** 775 * @j_errno: 776 * 777 * Is there an outstanding uncleared error on the journal (from a prior 778 * abort)? [j_state_lock] 779 */ 780 int j_errno; 781 782 /** 783 * @j_abort_mutex: Lock the whole aborting procedure. 784 */ 785 struct mutex j_abort_mutex; 786 787 /** 788 * @j_sb_buffer: The first part of the superblock buffer. 789 */ 790 struct buffer_head *j_sb_buffer; 791 792 /** 793 * @j_superblock: The second part of the superblock buffer. 794 */ 795 journal_superblock_t *j_superblock; 796 797 /** 798 * @j_state_lock: Protect the various scalars in the journal. 799 */ 800 rwlock_t j_state_lock; 801 802 /** 803 * @j_barrier_count: 804 * 805 * Number of processes waiting to create a barrier lock [j_state_lock, 806 * no lock for quick racy checks] 807 */ 808 int j_barrier_count; 809 810 /** 811 * @j_barrier: The barrier lock itself. 812 */ 813 struct mutex j_barrier; 814 815 /** 816 * @j_running_transaction: 817 * 818 * Transactions: The current running transaction... 819 * [j_state_lock, no lock for quick racy checks] [caller holding 820 * open handle] 821 */ 822 transaction_t *j_running_transaction; 823 824 /** 825 * @j_committing_transaction: 826 * 827 * the transaction we are pushing to disk 828 * [j_state_lock] [caller holding open handle] 829 */ 830 transaction_t *j_committing_transaction; 831 832 /** 833 * @j_checkpoint_transactions: 834 * 835 * ... and a linked circular list of all transactions waiting for 836 * checkpointing. [j_list_lock] 837 */ 838 transaction_t *j_checkpoint_transactions; 839 840 /** 841 * @j_wait_transaction_locked: 842 * 843 * Wait queue for waiting for a locked transaction to start committing, 844 * or for a barrier lock to be released. 845 */ 846 wait_queue_head_t j_wait_transaction_locked; 847 848 /** 849 * @j_wait_done_commit: Wait queue for waiting for commit to complete. 850 */ 851 wait_queue_head_t j_wait_done_commit; 852 853 /** 854 * @j_wait_commit: Wait queue to trigger commit. 855 */ 856 wait_queue_head_t j_wait_commit; 857 858 /** 859 * @j_wait_updates: Wait queue to wait for updates to complete. 860 */ 861 wait_queue_head_t j_wait_updates; 862 863 /** 864 * @j_wait_reserved: 865 * 866 * Wait queue to wait for reserved buffer credits to drop. 867 */ 868 wait_queue_head_t j_wait_reserved; 869 870 /** 871 * @j_fc_wait: 872 * 873 * Wait queue to wait for completion of async fast commits. 874 */ 875 wait_queue_head_t j_fc_wait; 876 877 /** 878 * @j_checkpoint_mutex: 879 * 880 * Semaphore for locking against concurrent checkpoints. 881 */ 882 struct mutex j_checkpoint_mutex; 883 884 /** 885 * @j_chkpt_bhs: 886 * 887 * List of buffer heads used by the checkpoint routine. This 888 * was moved from jbd2_log_do_checkpoint() to reduce stack 889 * usage. Access to this array is controlled by the 890 * @j_checkpoint_mutex. [j_checkpoint_mutex] 891 */ 892 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 893 894 /** 895 * @j_shrinker: 896 * 897 * Journal head shrinker, reclaim buffer's journal head which 898 * has been written back. 899 */ 900 struct shrinker *j_shrinker; 901 902 /** 903 * @j_checkpoint_jh_count: 904 * 905 * Number of journal buffers on the checkpoint list. [j_list_lock] 906 */ 907 struct percpu_counter j_checkpoint_jh_count; 908 909 /** 910 * @j_shrink_transaction: 911 * 912 * Record next transaction will shrink on the checkpoint list. 913 * [j_list_lock] 914 */ 915 transaction_t *j_shrink_transaction; 916 917 /** 918 * @j_head: 919 * 920 * Journal head: identifies the first unused block in the journal. 921 * [j_state_lock] 922 */ 923 unsigned long j_head; 924 925 /** 926 * @j_tail: 927 * 928 * Journal tail: identifies the oldest still-used block in the journal. 929 * [j_state_lock] 930 */ 931 unsigned long j_tail; 932 933 /** 934 * @j_free: 935 * 936 * Journal free: how many free blocks are there in the journal? 937 * [j_state_lock] 938 */ 939 unsigned long j_free; 940 941 /** 942 * @j_first: 943 * 944 * The block number of the first usable block in the journal 945 * [j_state_lock]. 946 */ 947 unsigned long j_first; 948 949 /** 950 * @j_last: 951 * 952 * The block number one beyond the last usable block in the journal 953 * [j_state_lock]. 954 */ 955 unsigned long j_last; 956 957 /** 958 * @j_fc_first: 959 * 960 * The block number of the first fast commit block in the journal 961 * [j_state_lock]. 962 */ 963 unsigned long j_fc_first; 964 965 /** 966 * @j_fc_off: 967 * 968 * Number of fast commit blocks currently allocated. Accessed only 969 * during fast commit. Currently only process can do fast commit, so 970 * this field is not protected by any lock. 971 */ 972 unsigned long j_fc_off; 973 974 /** 975 * @j_fc_last: 976 * 977 * The block number one beyond the last fast commit block in the journal 978 * [j_state_lock]. 979 */ 980 unsigned long j_fc_last; 981 982 /** 983 * @j_dev: Device where we store the journal. 984 */ 985 struct block_device *j_dev; 986 987 /** 988 * @j_blocksize: Block size for the location where we store the journal. 989 */ 990 int j_blocksize; 991 992 /** 993 * @j_blk_offset: 994 * 995 * Starting block offset into the device where we store the journal. 996 */ 997 unsigned long long j_blk_offset; 998 999 /** 1000 * @j_devname: Journal device name. 1001 */ 1002 char j_devname[BDEVNAME_SIZE+24]; 1003 1004 /** 1005 * @j_fs_dev: 1006 * 1007 * Device which holds the client fs. For internal journal this will be 1008 * equal to j_dev. 1009 */ 1010 struct block_device *j_fs_dev; 1011 1012 /** 1013 * @j_fs_dev_wb_err: 1014 * 1015 * Records the errseq of the client fs's backing block device. 1016 */ 1017 errseq_t j_fs_dev_wb_err; 1018 1019 /** 1020 * @j_total_len: Total maximum capacity of the journal region on disk. 1021 */ 1022 unsigned int j_total_len; 1023 1024 /** 1025 * @j_reserved_credits: 1026 * 1027 * Number of buffers reserved from the running transaction. 1028 */ 1029 atomic_t j_reserved_credits; 1030 1031 /** 1032 * @j_list_lock: Protects the buffer lists and internal buffer state. 1033 */ 1034 spinlock_t j_list_lock; 1035 1036 /** 1037 * @j_inode: 1038 * 1039 * Optional inode where we store the journal. If present, all 1040 * journal block numbers are mapped into this inode via bmap(). 1041 */ 1042 struct inode *j_inode; 1043 1044 /** 1045 * @j_tail_sequence: 1046 * 1047 * Sequence number of the oldest transaction in the log [j_state_lock] 1048 */ 1049 tid_t j_tail_sequence; 1050 1051 /** 1052 * @j_transaction_sequence: 1053 * 1054 * Sequence number of the next transaction to grant [j_state_lock] 1055 */ 1056 tid_t j_transaction_sequence; 1057 1058 /** 1059 * @j_commit_sequence: 1060 * 1061 * Sequence number of the most recently committed transaction 1062 * [j_state_lock, no lock for quick racy checks] 1063 */ 1064 tid_t j_commit_sequence; 1065 1066 /** 1067 * @j_commit_request: 1068 * 1069 * Sequence number of the most recent transaction wanting commit 1070 * [j_state_lock, no lock for quick racy checks] 1071 */ 1072 tid_t j_commit_request; 1073 1074 /** 1075 * @j_uuid: 1076 * 1077 * Journal uuid: identifies the object (filesystem, LVM volume etc) 1078 * backed by this journal. This will eventually be replaced by an array 1079 * of uuids, allowing us to index multiple devices within a single 1080 * journal and to perform atomic updates across them. 1081 */ 1082 __u8 j_uuid[16]; 1083 1084 /** 1085 * @j_task: Pointer to the current commit thread for this journal. 1086 */ 1087 struct task_struct *j_task; 1088 1089 /** 1090 * @j_max_transaction_buffers: 1091 * 1092 * Maximum number of metadata buffers to allow in a single compound 1093 * commit transaction. 1094 */ 1095 int j_max_transaction_buffers; 1096 1097 /** 1098 * @j_revoke_records_per_block: 1099 * 1100 * Number of revoke records that fit in one descriptor block. 1101 */ 1102 int j_revoke_records_per_block; 1103 1104 /** 1105 * @j_transaction_overhead_buffers: 1106 * 1107 * Number of blocks each transaction needs for its own bookkeeping 1108 */ 1109 int j_transaction_overhead_buffers; 1110 1111 /** 1112 * @j_commit_interval: 1113 * 1114 * What is the maximum transaction lifetime before we begin a commit? 1115 */ 1116 unsigned long j_commit_interval; 1117 1118 /** 1119 * @j_commit_timer: The timer used to wakeup the commit thread. 1120 */ 1121 struct timer_list j_commit_timer; 1122 1123 /** 1124 * @j_revoke_lock: Protect the revoke table. 1125 */ 1126 spinlock_t j_revoke_lock; 1127 1128 /** 1129 * @j_revoke: 1130 * 1131 * The revoke table - maintains the list of revoked blocks in the 1132 * current transaction. 1133 */ 1134 struct jbd2_revoke_table_s *j_revoke; 1135 1136 /** 1137 * @j_revoke_table: Alternate revoke tables for j_revoke. 1138 */ 1139 struct jbd2_revoke_table_s *j_revoke_table[2]; 1140 1141 /** 1142 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction. 1143 */ 1144 struct buffer_head **j_wbuf; 1145 1146 /** 1147 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only 1148 * during a fast commit. Currently only process can do fast commit, so 1149 * this field is not protected by any lock. 1150 */ 1151 struct buffer_head **j_fc_wbuf; 1152 1153 /** 1154 * @j_wbufsize: 1155 * 1156 * Size of @j_wbuf array. 1157 */ 1158 int j_wbufsize; 1159 1160 /** 1161 * @j_fc_wbufsize: 1162 * 1163 * Size of @j_fc_wbuf array. 1164 */ 1165 int j_fc_wbufsize; 1166 1167 /** 1168 * @j_last_sync_writer: 1169 * 1170 * The pid of the last person to run a synchronous operation 1171 * through the journal. 1172 */ 1173 pid_t j_last_sync_writer; 1174 1175 /** 1176 * @j_average_commit_time: 1177 * 1178 * The average amount of time in nanoseconds it takes to commit a 1179 * transaction to disk. [j_state_lock] 1180 */ 1181 u64 j_average_commit_time; 1182 1183 /** 1184 * @j_min_batch_time: 1185 * 1186 * Minimum time that we should wait for additional filesystem operations 1187 * to get batched into a synchronous handle in microseconds. 1188 */ 1189 u32 j_min_batch_time; 1190 1191 /** 1192 * @j_max_batch_time: 1193 * 1194 * Maximum time that we should wait for additional filesystem operations 1195 * to get batched into a synchronous handle in microseconds. 1196 */ 1197 u32 j_max_batch_time; 1198 1199 /** 1200 * @j_commit_callback: 1201 * 1202 * This function is called when a transaction is closed. 1203 */ 1204 void (*j_commit_callback)(journal_t *, 1205 transaction_t *); 1206 1207 /** 1208 * @j_submit_inode_data_buffers: 1209 * 1210 * This function is called for all inodes associated with the 1211 * committing transaction marked with JI_WRITE_DATA flag 1212 * before we start to write out the transaction to the journal. 1213 */ 1214 int (*j_submit_inode_data_buffers) 1215 (struct jbd2_inode *); 1216 1217 /** 1218 * @j_finish_inode_data_buffers: 1219 * 1220 * This function is called for all inodes associated with the 1221 * committing transaction marked with JI_WAIT_DATA flag 1222 * after we have written the transaction to the journal 1223 * but before we write out the commit block. 1224 */ 1225 int (*j_finish_inode_data_buffers) 1226 (struct jbd2_inode *); 1227 1228 /* 1229 * Journal statistics 1230 */ 1231 1232 /** 1233 * @j_history_lock: Protect the transactions statistics history. 1234 */ 1235 spinlock_t j_history_lock; 1236 1237 /** 1238 * @j_proc_entry: procfs entry for the jbd statistics directory. 1239 */ 1240 struct proc_dir_entry *j_proc_entry; 1241 1242 /** 1243 * @j_stats: Overall statistics. 1244 */ 1245 struct transaction_stats_s j_stats; 1246 1247 /** 1248 * @j_failed_commit: Failed journal commit ID. 1249 */ 1250 unsigned int j_failed_commit; 1251 1252 /** 1253 * @j_private: 1254 * 1255 * An opaque pointer to fs-private information. ext3 puts its 1256 * superblock pointer here. 1257 */ 1258 void *j_private; 1259 1260 /** 1261 * @j_csum_seed: 1262 * 1263 * Precomputed journal UUID checksum for seeding other checksums. 1264 */ 1265 __u32 j_csum_seed; 1266 1267 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1268 /** 1269 * @j_trans_commit_map: 1270 * 1271 * Lockdep entity to track transaction commit dependencies. Handles 1272 * hold this "lock" for read, when we wait for commit, we acquire the 1273 * "lock" for writing. This matches the properties of jbd2 journalling 1274 * where the running transaction has to wait for all handles to be 1275 * dropped to commit that transaction and also acquiring a handle may 1276 * require transaction commit to finish. 1277 */ 1278 struct lockdep_map j_trans_commit_map; 1279 #endif 1280 /** 1281 * @jbd2_trans_commit_key: 1282 * 1283 * "struct lock_class_key" for @j_trans_commit_map 1284 */ 1285 struct lock_class_key jbd2_trans_commit_key; 1286 1287 /** 1288 * @j_fc_cleanup_callback: 1289 * 1290 * Clean-up after fast commit or full commit. JBD2 calls this function 1291 * after every commit operation. 1292 */ 1293 void (*j_fc_cleanup_callback)(struct journal_s *journal, int full, tid_t tid); 1294 1295 /** 1296 * @j_fc_replay_callback: 1297 * 1298 * File-system specific function that performs replay of a fast 1299 * commit. JBD2 calls this function for each fast commit block found in 1300 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE 1301 * to indicate that the block was processed correctly and more fast 1302 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP 1303 * indicates the end of replay (no more blocks remaining). A negative 1304 * return value indicates error. 1305 */ 1306 int (*j_fc_replay_callback)(struct journal_s *journal, 1307 struct buffer_head *bh, 1308 enum passtype pass, int off, 1309 tid_t expected_commit_id); 1310 1311 /** 1312 * @j_bmap: 1313 * 1314 * Bmap function that should be used instead of the generic 1315 * VFS bmap function. 1316 */ 1317 int (*j_bmap)(struct journal_s *journal, sector_t *block); 1318 }; 1319 1320 #define jbd2_might_wait_for_commit(j) \ 1321 do { \ 1322 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \ 1323 rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \ 1324 } while (0) 1325 1326 /* 1327 * We can support any known requested features iff the 1328 * superblock is not in version 1. Otherwise we fail to support any 1329 * extended sb features. 1330 */ 1331 static inline bool jbd2_format_support_feature(journal_t *j) 1332 { 1333 return j->j_superblock->s_header.h_blocktype != 1334 cpu_to_be32(JBD2_SUPERBLOCK_V1); 1335 } 1336 1337 /* journal feature predicate functions */ 1338 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1339 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1340 { \ 1341 return (jbd2_format_support_feature(j) && \ 1342 ((j)->j_superblock->s_feature_compat & \ 1343 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1344 } \ 1345 static inline void jbd2_set_feature_##name(journal_t *j) \ 1346 { \ 1347 (j)->j_superblock->s_feature_compat |= \ 1348 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1349 } \ 1350 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1351 { \ 1352 (j)->j_superblock->s_feature_compat &= \ 1353 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1354 } 1355 1356 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1357 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1358 { \ 1359 return (jbd2_format_support_feature(j) && \ 1360 ((j)->j_superblock->s_feature_ro_compat & \ 1361 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1362 } \ 1363 static inline void jbd2_set_feature_##name(journal_t *j) \ 1364 { \ 1365 (j)->j_superblock->s_feature_ro_compat |= \ 1366 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1367 } \ 1368 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1369 { \ 1370 (j)->j_superblock->s_feature_ro_compat &= \ 1371 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1372 } 1373 1374 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1375 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1376 { \ 1377 return (jbd2_format_support_feature(j) && \ 1378 ((j)->j_superblock->s_feature_incompat & \ 1379 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1380 } \ 1381 static inline void jbd2_set_feature_##name(journal_t *j) \ 1382 { \ 1383 (j)->j_superblock->s_feature_incompat |= \ 1384 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1385 } \ 1386 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1387 { \ 1388 (j)->j_superblock->s_feature_incompat &= \ 1389 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1390 } 1391 1392 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1393 1394 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1395 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1396 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1397 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1398 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1399 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit, FAST_COMMIT) 1400 1401 /* Journal high priority write IO operation flags */ 1402 #define JBD2_JOURNAL_REQ_FLAGS (REQ_META | REQ_SYNC | REQ_IDLE) 1403 1404 /* 1405 * Journal flag definitions 1406 */ 1407 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1408 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1409 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1410 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1411 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1412 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1413 #define JBD2_CYCLE_RECORD 0x080 /* Journal cycled record log on 1414 * clean and empty filesystem 1415 * logging area */ 1416 #define JBD2_FAST_COMMIT_ONGOING 0x100 /* Fast commit is ongoing */ 1417 #define JBD2_FULL_COMMIT_ONGOING 0x200 /* Full commit is ongoing */ 1418 #define JBD2_JOURNAL_FLUSH_DISCARD 0x0001 1419 #define JBD2_JOURNAL_FLUSH_ZEROOUT 0x0002 1420 #define JBD2_JOURNAL_FLUSH_VALID (JBD2_JOURNAL_FLUSH_DISCARD | \ 1421 JBD2_JOURNAL_FLUSH_ZEROOUT) 1422 1423 /* 1424 * Function declarations for the journaling transaction and buffer 1425 * management 1426 */ 1427 1428 /* Filing buffers */ 1429 extern bool __jbd2_journal_refile_buffer(struct journal_head *); 1430 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1431 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1432 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1433 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1434 { 1435 list_add_tail(&bh->b_assoc_buffers, head); 1436 } 1437 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1438 { 1439 list_del_init(&bh->b_assoc_buffers); 1440 } 1441 1442 /* Log buffer allocation */ 1443 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1444 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1445 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1446 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1447 unsigned long *block); 1448 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1449 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1450 1451 /* Commit management */ 1452 extern void jbd2_journal_commit_transaction(journal_t *); 1453 1454 /* Checkpoint list management */ 1455 enum jbd2_shrink_type {JBD2_SHRINK_DESTROY, JBD2_SHRINK_BUSY_STOP, JBD2_SHRINK_BUSY_SKIP}; 1456 1457 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, enum jbd2_shrink_type type); 1458 unsigned long jbd2_journal_shrink_checkpoint_list(journal_t *journal, unsigned long *nr_to_scan); 1459 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1460 int jbd2_journal_try_remove_checkpoint(struct journal_head *jh); 1461 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1462 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1463 1464 1465 /* 1466 * Triggers 1467 */ 1468 1469 struct jbd2_buffer_trigger_type { 1470 /* 1471 * Fired a the moment data to write to the journal are known to be 1472 * stable - so either at the moment b_frozen_data is created or just 1473 * before a buffer is written to the journal. mapped_data is a mapped 1474 * buffer that is the frozen data for commit. 1475 */ 1476 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1477 struct buffer_head *bh, void *mapped_data, 1478 size_t size); 1479 1480 /* 1481 * Fired during journal abort for dirty buffers that will not be 1482 * committed. 1483 */ 1484 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1485 struct buffer_head *bh); 1486 }; 1487 1488 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1489 void *mapped_data, 1490 struct jbd2_buffer_trigger_type *triggers); 1491 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1492 struct jbd2_buffer_trigger_type *triggers); 1493 1494 /* Buffer IO */ 1495 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1496 struct journal_head *jh_in, 1497 struct buffer_head **bh_out, 1498 sector_t blocknr); 1499 1500 /* Transaction cache support */ 1501 extern void jbd2_journal_destroy_transaction_cache(void); 1502 extern int __init jbd2_journal_init_transaction_cache(void); 1503 extern void jbd2_journal_free_transaction(transaction_t *); 1504 1505 /* 1506 * Journal locking. 1507 * 1508 * We need to lock the journal during transaction state changes so that nobody 1509 * ever tries to take a handle on the running transaction while we are in the 1510 * middle of moving it to the commit phase. j_state_lock does this. 1511 * 1512 * Note that the locking is completely interrupt unsafe. We never touch 1513 * journal structures from interrupts. 1514 */ 1515 1516 static inline handle_t *journal_current_handle(void) 1517 { 1518 return current->journal_info; 1519 } 1520 1521 /* The journaling code user interface: 1522 * 1523 * Create and destroy handles 1524 * Register buffer modifications against the current transaction. 1525 */ 1526 1527 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1528 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1529 int revoke_records, gfp_t gfp_mask, 1530 unsigned int type, unsigned int line_no); 1531 extern int jbd2_journal_restart(handle_t *, int nblocks); 1532 extern int jbd2__journal_restart(handle_t *, int nblocks, 1533 int revoke_records, gfp_t gfp_mask); 1534 extern int jbd2_journal_start_reserved(handle_t *handle, 1535 unsigned int type, unsigned int line_no); 1536 extern void jbd2_journal_free_reserved(handle_t *handle); 1537 extern int jbd2_journal_extend(handle_t *handle, int nblocks, 1538 int revoke_records); 1539 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1540 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1541 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1542 void jbd2_journal_set_triggers(struct buffer_head *, 1543 struct jbd2_buffer_trigger_type *type); 1544 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1545 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1546 int jbd2_journal_invalidate_folio(journal_t *, struct folio *, 1547 size_t offset, size_t length); 1548 bool jbd2_journal_try_to_free_buffers(journal_t *journal, struct folio *folio); 1549 extern int jbd2_journal_stop(handle_t *); 1550 extern int jbd2_journal_flush(journal_t *journal, unsigned int flags); 1551 extern void jbd2_journal_lock_updates (journal_t *); 1552 extern void jbd2_journal_unlock_updates (journal_t *); 1553 1554 void jbd2_journal_wait_updates(journal_t *); 1555 1556 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1557 struct block_device *fs_dev, 1558 unsigned long long start, int len, int bsize); 1559 extern journal_t * jbd2_journal_init_inode (struct inode *); 1560 extern int jbd2_journal_update_format (journal_t *); 1561 extern int jbd2_journal_check_used_features 1562 (journal_t *, unsigned long, unsigned long, unsigned long); 1563 extern int jbd2_journal_check_available_features 1564 (journal_t *, unsigned long, unsigned long, unsigned long); 1565 extern int jbd2_journal_set_features 1566 (journal_t *, unsigned long, unsigned long, unsigned long); 1567 extern void jbd2_journal_clear_features 1568 (journal_t *, unsigned long, unsigned long, unsigned long); 1569 extern int jbd2_journal_load (journal_t *journal); 1570 extern int jbd2_journal_destroy (journal_t *); 1571 extern int jbd2_journal_recover (journal_t *journal); 1572 extern int jbd2_journal_wipe (journal_t *, int); 1573 extern int jbd2_journal_skip_recovery (journal_t *); 1574 extern void jbd2_journal_update_sb_errno(journal_t *); 1575 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1576 unsigned long, blk_opf_t); 1577 extern void jbd2_journal_abort (journal_t *, int); 1578 extern int jbd2_journal_errno (journal_t *); 1579 extern void jbd2_journal_ack_err (journal_t *); 1580 extern int jbd2_journal_clear_err (journal_t *); 1581 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1582 extern int jbd2_journal_force_commit(journal_t *); 1583 extern int jbd2_journal_force_commit_nested(journal_t *); 1584 extern int jbd2_journal_inode_ranged_write(handle_t *handle, 1585 struct jbd2_inode *inode, loff_t start_byte, 1586 loff_t length); 1587 extern int jbd2_journal_inode_ranged_wait(handle_t *handle, 1588 struct jbd2_inode *inode, loff_t start_byte, 1589 loff_t length); 1590 extern int jbd2_journal_finish_inode_data_buffers( 1591 struct jbd2_inode *jinode); 1592 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1593 struct jbd2_inode *inode, loff_t new_size); 1594 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1595 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1596 1597 /* 1598 * journal_head management 1599 */ 1600 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1601 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1602 void jbd2_journal_put_journal_head(struct journal_head *jh); 1603 1604 /* 1605 * handle management 1606 */ 1607 extern struct kmem_cache *jbd2_handle_cache; 1608 1609 /* 1610 * This specialized allocator has to be a macro for its allocations to be 1611 * accounted separately (to have a separate alloc_tag). The typecast is 1612 * intentional to enforce typesafety. 1613 */ 1614 #define jbd2_alloc_handle(_gfp_flags) \ 1615 ((handle_t *)kmem_cache_zalloc(jbd2_handle_cache, _gfp_flags)) 1616 1617 static inline void jbd2_free_handle(handle_t *handle) 1618 { 1619 kmem_cache_free(jbd2_handle_cache, handle); 1620 } 1621 1622 /* 1623 * jbd2_inode management (optional, for those file systems that want to use 1624 * dynamically allocated jbd2_inode structures) 1625 */ 1626 extern struct kmem_cache *jbd2_inode_cache; 1627 1628 /* 1629 * This specialized allocator has to be a macro for its allocations to be 1630 * accounted separately (to have a separate alloc_tag). The typecast is 1631 * intentional to enforce typesafety. 1632 */ 1633 #define jbd2_alloc_inode(_gfp_flags) \ 1634 ((struct jbd2_inode *)kmem_cache_alloc(jbd2_inode_cache, _gfp_flags)) 1635 1636 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1637 { 1638 kmem_cache_free(jbd2_inode_cache, jinode); 1639 } 1640 1641 /* Primary revoke support */ 1642 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1643 extern int jbd2_journal_init_revoke(journal_t *, int); 1644 extern void jbd2_journal_destroy_revoke_record_cache(void); 1645 extern void jbd2_journal_destroy_revoke_table_cache(void); 1646 extern int __init jbd2_journal_init_revoke_record_cache(void); 1647 extern int __init jbd2_journal_init_revoke_table_cache(void); 1648 struct jbd2_revoke_table_s *jbd2_journal_init_revoke_table(int hash_size); 1649 void jbd2_journal_destroy_revoke_table(struct jbd2_revoke_table_s *table); 1650 1651 extern void jbd2_journal_destroy_revoke(journal_t *); 1652 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1653 extern void jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1654 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1655 struct list_head *log_bufs); 1656 1657 /* Recovery revoke support */ 1658 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1659 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1660 extern void jbd2_journal_clear_revoke(journal_t *); 1661 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1662 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1663 1664 /* 1665 * The log thread user interface: 1666 * 1667 * Request space in the current transaction, and force transaction commit 1668 * transitions on demand. 1669 */ 1670 1671 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1672 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1673 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1674 int jbd2_transaction_committed(journal_t *journal, tid_t tid); 1675 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1676 int jbd2_log_do_checkpoint(journal_t *journal); 1677 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1678 1679 void __jbd2_log_wait_for_space(journal_t *journal); 1680 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1681 extern int jbd2_cleanup_journal_tail(journal_t *); 1682 1683 /* Fast commit related APIs */ 1684 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid); 1685 int jbd2_fc_end_commit(journal_t *journal); 1686 int jbd2_fc_end_commit_fallback(journal_t *journal); 1687 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out); 1688 int jbd2_submit_inode_data(journal_t *journal, struct jbd2_inode *jinode); 1689 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode); 1690 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks); 1691 void jbd2_fc_release_bufs(journal_t *journal); 1692 1693 /* 1694 * is_journal_abort 1695 * 1696 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1697 * bit, when set, indicates that we have had a fatal error somewhere, 1698 * either inside the journaling layer or indicated to us by the client 1699 * (eg. ext3), and that we and should not commit any further 1700 * transactions. 1701 */ 1702 1703 static inline int is_journal_aborted(journal_t *journal) 1704 { 1705 return journal->j_flags & JBD2_ABORT; 1706 } 1707 1708 static inline int is_handle_aborted(handle_t *handle) 1709 { 1710 if (handle->h_aborted || !handle->h_transaction) 1711 return 1; 1712 return is_journal_aborted(handle->h_transaction->t_journal); 1713 } 1714 1715 static inline void jbd2_journal_abort_handle(handle_t *handle) 1716 { 1717 handle->h_aborted = 1; 1718 } 1719 1720 static inline void jbd2_init_fs_dev_write_error(journal_t *journal) 1721 { 1722 struct address_space *mapping = journal->j_fs_dev->bd_mapping; 1723 1724 /* 1725 * Save the original wb_err value of client fs's bdev mapping which 1726 * could be used to detect the client fs's metadata async write error. 1727 */ 1728 errseq_check_and_advance(&mapping->wb_err, &journal->j_fs_dev_wb_err); 1729 } 1730 1731 static inline int jbd2_check_fs_dev_write_error(journal_t *journal) 1732 { 1733 struct address_space *mapping = journal->j_fs_dev->bd_mapping; 1734 1735 return errseq_check(&mapping->wb_err, 1736 READ_ONCE(journal->j_fs_dev_wb_err)); 1737 } 1738 1739 #endif /* __KERNEL__ */ 1740 1741 /* Comparison functions for transaction IDs: perform comparisons using 1742 * modulo arithmetic so that they work over sequence number wraps. */ 1743 1744 static inline int tid_gt(tid_t x, tid_t y) 1745 { 1746 int difference = (x - y); 1747 return (difference > 0); 1748 } 1749 1750 static inline int tid_geq(tid_t x, tid_t y) 1751 { 1752 int difference = (x - y); 1753 return (difference >= 0); 1754 } 1755 1756 extern int jbd2_journal_blocks_per_folio(struct inode *inode); 1757 extern size_t journal_tag_bytes(journal_t *journal); 1758 1759 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1760 { 1761 return jbd2_has_feature_csum2(journal) || 1762 jbd2_has_feature_csum3(journal); 1763 } 1764 1765 static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb) 1766 { 1767 int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks); 1768 1769 return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS; 1770 } 1771 1772 /* 1773 * Return number of free blocks in the log. Must be called under j_state_lock. 1774 */ 1775 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1776 { 1777 /* Allow for rounding errors */ 1778 long free = journal->j_free - 32; 1779 1780 if (journal->j_committing_transaction) { 1781 free -= atomic_read(&journal-> 1782 j_committing_transaction->t_outstanding_credits); 1783 } 1784 return max_t(long, free, 0); 1785 } 1786 1787 /* 1788 * Definitions which augment the buffer_head layer 1789 */ 1790 1791 /* journaling buffer types */ 1792 #define BJ_None 0 /* Not journaled */ 1793 #define BJ_Metadata 1 /* Normal journaled metadata */ 1794 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1795 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1796 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1797 #define BJ_Types 5 1798 1799 static inline u32 jbd2_chksum(u32 crc, const void *address, unsigned int length) 1800 { 1801 return crc32c(crc, address, length); 1802 } 1803 1804 /* Return most recent uncommitted transaction */ 1805 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1806 { 1807 tid_t tid; 1808 1809 read_lock(&journal->j_state_lock); 1810 tid = journal->j_commit_request; 1811 if (journal->j_running_transaction) 1812 tid = journal->j_running_transaction->t_tid; 1813 read_unlock(&journal->j_state_lock); 1814 return tid; 1815 } 1816 1817 static inline int jbd2_handle_buffer_credits(handle_t *handle) 1818 { 1819 journal_t *journal; 1820 1821 if (!handle->h_reserved) 1822 journal = handle->h_transaction->t_journal; 1823 else 1824 journal = handle->h_journal; 1825 1826 return handle->h_total_credits - 1827 DIV_ROUND_UP(handle->h_revoke_credits_requested, 1828 journal->j_revoke_records_per_block); 1829 } 1830 1831 #ifdef __KERNEL__ 1832 1833 #define buffer_trace_init(bh) do {} while (0) 1834 #define print_buffer_fields(bh) do {} while (0) 1835 #define print_buffer_trace(bh) do {} while (0) 1836 #define BUFFER_TRACE(bh, info) do {} while (0) 1837 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1838 #define JBUFFER_TRACE(jh, info) do {} while (0) 1839 1840 #endif /* __KERNEL__ */ 1841 1842 #endif /* _LINUX_JBD2_H */ 1843