1/* 2 * linux/include/linux/jbd2.h 3 * 4 * Written by Stephen C. Tweedie <sct@redhat.com> 5 * 6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved 7 * 8 * This file is part of the Linux kernel and is made available under 9 * the terms of the GNU General Public License, version 2, or at your 10 * option, any later version, incorporated herein by reference. 11 * 12 * Definitions for transaction data structures for the buffer cache 13 * filesystem journaling support. 14 */ 15 16#ifndef _LINUX_JBD2_H 17#define _LINUX_JBD2_H 18 19/* Allow this file to be included directly into e2fsprogs */ 20#ifndef __KERNEL__ 21#include "jfs_compat.h" 22#define JBD2_DEBUG 23#else 24 25#include <linux/types.h> 26#include <linux/buffer_head.h> 27#include <linux/journal-head.h> 28#include <linux/stddef.h> 29#include <linux/mutex.h> 30#include <linux/timer.h> 31#include <linux/slab.h> 32#include <linux/bit_spinlock.h> 33#include <crypto/hash.h> 34#endif 35 36#define journal_oom_retry 1 37 38/* 39 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds 40 * certain classes of error which can occur due to failed IOs. Under 41 * normal use we want ext4 to continue after such errors, because 42 * hardware _can_ fail, but for debugging purposes when running tests on 43 * known-good hardware we may want to trap these errors. 44 */ 45#undef JBD2_PARANOID_IOFAIL 46 47/* 48 * The default maximum commit age, in seconds. 49 */ 50#define JBD2_DEFAULT_MAX_COMMIT_AGE 5 51 52#ifdef CONFIG_JBD2_DEBUG 53/* 54 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal 55 * consistency checks. By default we don't do this unless 56 * CONFIG_JBD2_DEBUG is on. 57 */ 58#define JBD2_EXPENSIVE_CHECKING 59extern ushort jbd2_journal_enable_debug; 60void __jbd2_debug(int level, const char *file, const char *func, 61 unsigned int line, const char *fmt, ...); 62 63#define jbd_debug(n, fmt, a...) \ 64 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a) 65#else 66#define jbd_debug(n, fmt, a...) /**/ 67#endif 68 69extern void *jbd2_alloc(size_t size, gfp_t flags); 70extern void jbd2_free(void *ptr, size_t size); 71 72#define JBD2_MIN_JOURNAL_BLOCKS 1024 73 74#ifdef __KERNEL__ 75 76/** 77 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 78 * 79 * All filesystem modifications made by the process go 80 * through this handle. Recursive operations (such as quota operations) 81 * are gathered into a single update. 82 * 83 * The buffer credits field is used to account for journaled buffers 84 * being modified by the running process. To ensure that there is 85 * enough log space for all outstanding operations, we need to limit the 86 * number of outstanding buffers possible at any time. When the 87 * operation completes, any buffer credits not used are credited back to 88 * the transaction, so that at all times we know how many buffers the 89 * outstanding updates on a transaction might possibly touch. 90 * 91 * This is an opaque datatype. 92 **/ 93typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ 94 95 96/** 97 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 98 * 99 * journal_t is linked to from the fs superblock structure. 100 * 101 * We use the journal_t to keep track of all outstanding transaction 102 * activity on the filesystem, and to manage the state of the log 103 * writing process. 104 * 105 * This is an opaque datatype. 106 **/ 107typedef struct journal_s journal_t; /* Journal control structure */ 108#endif 109 110/* 111 * Internal structures used by the logging mechanism: 112 */ 113 114#define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 115 116/* 117 * On-disk structures 118 */ 119 120/* 121 * Descriptor block types: 122 */ 123 124#define JBD2_DESCRIPTOR_BLOCK 1 125#define JBD2_COMMIT_BLOCK 2 126#define JBD2_SUPERBLOCK_V1 3 127#define JBD2_SUPERBLOCK_V2 4 128#define JBD2_REVOKE_BLOCK 5 129 130/* 131 * Standard header for all descriptor blocks: 132 */ 133typedef struct journal_header_s 134{ 135 __be32 h_magic; 136 __be32 h_blocktype; 137 __be32 h_sequence; 138} journal_header_t; 139 140/* 141 * Checksum types. 142 */ 143#define JBD2_CRC32_CHKSUM 1 144#define JBD2_MD5_CHKSUM 2 145#define JBD2_SHA1_CHKSUM 3 146#define JBD2_CRC32C_CHKSUM 4 147 148#define JBD2_CRC32_CHKSUM_SIZE 4 149 150#define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 151/* 152 * Commit block header for storing transactional checksums: 153 * 154 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* 155 * fields are used to store a checksum of the descriptor and data blocks. 156 * 157 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum 158 * field is used to store crc32c(uuid+commit_block). Each journal metadata 159 * block gets its own checksum, and data block checksums are stored in 160 * journal_block_tag (in the descriptor). The other h_chksum* fields are 161 * not used. 162 * 163 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses 164 * journal_block_tag3_t to store a full 32-bit checksum. Everything else 165 * is the same as v2. 166 * 167 * Checksum v1, v2, and v3 are mutually exclusive features. 168 */ 169struct commit_header { 170 __be32 h_magic; 171 __be32 h_blocktype; 172 __be32 h_sequence; 173 unsigned char h_chksum_type; 174 unsigned char h_chksum_size; 175 unsigned char h_padding[2]; 176 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 177 __be64 h_commit_sec; 178 __be32 h_commit_nsec; 179}; 180 181/* 182 * The block tag: used to describe a single buffer in the journal. 183 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 184 * raw struct shouldn't be used for pointer math or sizeof() - use 185 * journal_tag_bytes(journal) instead to compute this. 186 */ 187typedef struct journal_block_tag3_s 188{ 189 __be32 t_blocknr; /* The on-disk block number */ 190 __be32 t_flags; /* See below */ 191 __be32 t_blocknr_high; /* most-significant high 32bits. */ 192 __be32 t_checksum; /* crc32c(uuid+seq+block) */ 193} journal_block_tag3_t; 194 195typedef struct journal_block_tag_s 196{ 197 __be32 t_blocknr; /* The on-disk block number */ 198 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ 199 __be16 t_flags; /* See below */ 200 __be32 t_blocknr_high; /* most-significant high 32bits. */ 201} journal_block_tag_t; 202 203/* Tail of descriptor block, for checksumming */ 204struct jbd2_journal_block_tail { 205 __be32 t_checksum; /* crc32c(uuid+descr_block) */ 206}; 207 208/* 209 * The revoke descriptor: used on disk to describe a series of blocks to 210 * be revoked from the log 211 */ 212typedef struct jbd2_journal_revoke_header_s 213{ 214 journal_header_t r_header; 215 __be32 r_count; /* Count of bytes used in the block */ 216} jbd2_journal_revoke_header_t; 217 218/* Tail of revoke block, for checksumming */ 219struct jbd2_journal_revoke_tail { 220 __be32 r_checksum; /* crc32c(uuid+revoke_block) */ 221}; 222 223/* Definitions for the journal tag flags word: */ 224#define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 225#define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 226#define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 227#define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 228 229 230/* 231 * The journal superblock. All fields are in big-endian byte order. 232 */ 233typedef struct journal_superblock_s 234{ 235/* 0x0000 */ 236 journal_header_t s_header; 237 238/* 0x000C */ 239 /* Static information describing the journal */ 240 __be32 s_blocksize; /* journal device blocksize */ 241 __be32 s_maxlen; /* total blocks in journal file */ 242 __be32 s_first; /* first block of log information */ 243 244/* 0x0018 */ 245 /* Dynamic information describing the current state of the log */ 246 __be32 s_sequence; /* first commit ID expected in log */ 247 __be32 s_start; /* blocknr of start of log */ 248 249/* 0x0020 */ 250 /* Error value, as set by jbd2_journal_abort(). */ 251 __be32 s_errno; 252 253/* 0x0024 */ 254 /* Remaining fields are only valid in a version-2 superblock */ 255 __be32 s_feature_compat; /* compatible feature set */ 256 __be32 s_feature_incompat; /* incompatible feature set */ 257 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 258/* 0x0030 */ 259 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 260 261/* 0x0040 */ 262 __be32 s_nr_users; /* Nr of filesystems sharing log */ 263 264 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 265 266/* 0x0048 */ 267 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 268 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 269 270/* 0x0050 */ 271 __u8 s_checksum_type; /* checksum type */ 272 __u8 s_padding2[3]; 273 __u32 s_padding[42]; 274 __be32 s_checksum; /* crc32c(superblock) */ 275 276/* 0x0100 */ 277 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 278/* 0x0400 */ 279} journal_superblock_t; 280 281/* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */ 282#define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 283 ((j)->j_format_version >= 2 && \ 284 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 285#define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 286 ((j)->j_format_version >= 2 && \ 287 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 288#define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 289 ((j)->j_format_version >= 2 && \ 290 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 291 292#define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 293 294#define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 295#define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 296#define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 297#define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 298#define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 299 300/* See "journal feature predicate functions" below */ 301 302/* Features known to this kernel version: */ 303#define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 304#define JBD2_KNOWN_ROCOMPAT_FEATURES 0 305#define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 306 JBD2_FEATURE_INCOMPAT_64BIT | \ 307 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 308 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 309 JBD2_FEATURE_INCOMPAT_CSUM_V3) 310 311#ifdef __KERNEL__ 312 313#include <linux/fs.h> 314#include <linux/sched.h> 315 316enum jbd_state_bits { 317 BH_JBD /* Has an attached ext3 journal_head */ 318 = BH_PrivateStart, 319 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 320 BH_Freed, /* Has been freed (truncated) */ 321 BH_Revoked, /* Has been revoked from the log */ 322 BH_RevokeValid, /* Revoked flag is valid */ 323 BH_JBDDirty, /* Is dirty but journaled */ 324 BH_State, /* Pins most journal_head state */ 325 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 326 BH_Shadow, /* IO on shadow buffer is running */ 327 BH_Verified, /* Metadata block has been verified ok */ 328 BH_JBDPrivateStart, /* First bit available for private use by FS */ 329}; 330 331BUFFER_FNS(JBD, jbd) 332BUFFER_FNS(JWrite, jwrite) 333BUFFER_FNS(JBDDirty, jbddirty) 334TAS_BUFFER_FNS(JBDDirty, jbddirty) 335BUFFER_FNS(Revoked, revoked) 336TAS_BUFFER_FNS(Revoked, revoked) 337BUFFER_FNS(RevokeValid, revokevalid) 338TAS_BUFFER_FNS(RevokeValid, revokevalid) 339BUFFER_FNS(Freed, freed) 340BUFFER_FNS(Shadow, shadow) 341BUFFER_FNS(Verified, verified) 342 343static inline struct buffer_head *jh2bh(struct journal_head *jh) 344{ 345 return jh->b_bh; 346} 347 348static inline struct journal_head *bh2jh(struct buffer_head *bh) 349{ 350 return bh->b_private; 351} 352 353static inline void jbd_lock_bh_state(struct buffer_head *bh) 354{ 355 bit_spin_lock(BH_State, &bh->b_state); 356} 357 358static inline int jbd_trylock_bh_state(struct buffer_head *bh) 359{ 360 return bit_spin_trylock(BH_State, &bh->b_state); 361} 362 363static inline int jbd_is_locked_bh_state(struct buffer_head *bh) 364{ 365 return bit_spin_is_locked(BH_State, &bh->b_state); 366} 367 368static inline void jbd_unlock_bh_state(struct buffer_head *bh) 369{ 370 bit_spin_unlock(BH_State, &bh->b_state); 371} 372 373static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 374{ 375 bit_spin_lock(BH_JournalHead, &bh->b_state); 376} 377 378static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 379{ 380 bit_spin_unlock(BH_JournalHead, &bh->b_state); 381} 382 383#define J_ASSERT(assert) BUG_ON(!(assert)) 384 385#define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 386#define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 387 388#if defined(JBD2_PARANOID_IOFAIL) 389#define J_EXPECT(expr, why...) J_ASSERT(expr) 390#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 391#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 392#else 393#define __journal_expect(expr, why...) \ 394 ({ \ 395 int val = (expr); \ 396 if (!val) { \ 397 printk(KERN_ERR \ 398 "JBD2 unexpected failure: %s: %s;\n", \ 399 __func__, #expr); \ 400 printk(KERN_ERR why "\n"); \ 401 } \ 402 val; \ 403 }) 404#define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 405#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 406#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 407#endif 408 409/* Flags in jbd_inode->i_flags */ 410#define __JI_COMMIT_RUNNING 0 411/* Commit of the inode data in progress. We use this flag to protect us from 412 * concurrent deletion of inode. We cannot use reference to inode for this 413 * since we cannot afford doing last iput() on behalf of kjournald 414 */ 415#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 416 417/** 418 * struct jbd_inode is the structure linking inodes in ordered mode 419 * present in a transaction so that we can sync them during commit. 420 */ 421struct jbd2_inode { 422 /* Which transaction does this inode belong to? Either the running 423 * transaction or the committing one. [j_list_lock] */ 424 transaction_t *i_transaction; 425 426 /* Pointer to the running transaction modifying inode's data in case 427 * there is already a committing transaction touching it. [j_list_lock] */ 428 transaction_t *i_next_transaction; 429 430 /* List of inodes in the i_transaction [j_list_lock] */ 431 struct list_head i_list; 432 433 /* VFS inode this inode belongs to [constant during the lifetime 434 * of the structure] */ 435 struct inode *i_vfs_inode; 436 437 /* Flags of inode [j_list_lock] */ 438 unsigned long i_flags; 439}; 440 441struct jbd2_revoke_table_s; 442 443/** 444 * struct handle_s - The handle_s type is the concrete type associated with 445 * handle_t. 446 * @h_transaction: Which compound transaction is this update a part of? 447 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. 448 * @h_ref: Reference count on this handle 449 * @h_err: Field for caller's use to track errors through large fs operations 450 * @h_sync: flag for sync-on-close 451 * @h_jdata: flag to force data journaling 452 * @h_aborted: flag indicating fatal error on handle 453 **/ 454 455/* Docbook can't yet cope with the bit fields, but will leave the documentation 456 * in so it can be fixed later. 457 */ 458 459struct jbd2_journal_handle 460{ 461 union { 462 /* Which compound transaction is this update a part of? */ 463 transaction_t *h_transaction; 464 /* Which journal handle belongs to - used iff h_reserved set */ 465 journal_t *h_journal; 466 }; 467 468 /* Handle reserved for finishing the logical operation */ 469 handle_t *h_rsv_handle; 470 471 /* Number of remaining buffers we are allowed to dirty: */ 472 int h_buffer_credits; 473 474 /* Reference count on this handle */ 475 int h_ref; 476 477 /* Field for caller's use to track errors through large fs */ 478 /* operations */ 479 int h_err; 480 481 /* Flags [no locking] */ 482 unsigned int h_sync: 1; /* sync-on-close */ 483 unsigned int h_jdata: 1; /* force data journaling */ 484 unsigned int h_reserved: 1; /* handle with reserved credits */ 485 unsigned int h_aborted: 1; /* fatal error on handle */ 486 unsigned int h_type: 8; /* for handle statistics */ 487 unsigned int h_line_no: 16; /* for handle statistics */ 488 489 unsigned long h_start_jiffies; 490 unsigned int h_requested_credits; 491 492#ifdef CONFIG_DEBUG_LOCK_ALLOC 493 struct lockdep_map h_lockdep_map; 494#endif 495}; 496 497 498/* 499 * Some stats for checkpoint phase 500 */ 501struct transaction_chp_stats_s { 502 unsigned long cs_chp_time; 503 __u32 cs_forced_to_close; 504 __u32 cs_written; 505 __u32 cs_dropped; 506}; 507 508/* The transaction_t type is the guts of the journaling mechanism. It 509 * tracks a compound transaction through its various states: 510 * 511 * RUNNING: accepting new updates 512 * LOCKED: Updates still running but we don't accept new ones 513 * RUNDOWN: Updates are tidying up but have finished requesting 514 * new buffers to modify (state not used for now) 515 * FLUSH: All updates complete, but we are still writing to disk 516 * COMMIT: All data on disk, writing commit record 517 * FINISHED: We still have to keep the transaction for checkpointing. 518 * 519 * The transaction keeps track of all of the buffers modified by a 520 * running transaction, and all of the buffers committed but not yet 521 * flushed to home for finished transactions. 522 */ 523 524/* 525 * Lock ranking: 526 * 527 * j_list_lock 528 * ->jbd_lock_bh_journal_head() (This is "innermost") 529 * 530 * j_state_lock 531 * ->jbd_lock_bh_state() 532 * 533 * jbd_lock_bh_state() 534 * ->j_list_lock 535 * 536 * j_state_lock 537 * ->t_handle_lock 538 * 539 * j_state_lock 540 * ->j_list_lock (journal_unmap_buffer) 541 * 542 */ 543 544struct transaction_s 545{ 546 /* Pointer to the journal for this transaction. [no locking] */ 547 journal_t *t_journal; 548 549 /* Sequence number for this transaction [no locking] */ 550 tid_t t_tid; 551 552 /* 553 * Transaction's current state 554 * [no locking - only kjournald2 alters this] 555 * [j_list_lock] guards transition of a transaction into T_FINISHED 556 * state and subsequent call of __jbd2_journal_drop_transaction() 557 * FIXME: needs barriers 558 * KLUDGE: [use j_state_lock] 559 */ 560 enum { 561 T_RUNNING, 562 T_LOCKED, 563 T_FLUSH, 564 T_COMMIT, 565 T_COMMIT_DFLUSH, 566 T_COMMIT_JFLUSH, 567 T_COMMIT_CALLBACK, 568 T_FINISHED 569 } t_state; 570 571 /* 572 * Where in the log does this transaction's commit start? [no locking] 573 */ 574 unsigned long t_log_start; 575 576 /* Number of buffers on the t_buffers list [j_list_lock] */ 577 int t_nr_buffers; 578 579 /* 580 * Doubly-linked circular list of all buffers reserved but not yet 581 * modified by this transaction [j_list_lock] 582 */ 583 struct journal_head *t_reserved_list; 584 585 /* 586 * Doubly-linked circular list of all metadata buffers owned by this 587 * transaction [j_list_lock] 588 */ 589 struct journal_head *t_buffers; 590 591 /* 592 * Doubly-linked circular list of all forget buffers (superseded 593 * buffers which we can un-checkpoint once this transaction commits) 594 * [j_list_lock] 595 */ 596 struct journal_head *t_forget; 597 598 /* 599 * Doubly-linked circular list of all buffers still to be flushed before 600 * this transaction can be checkpointed. [j_list_lock] 601 */ 602 struct journal_head *t_checkpoint_list; 603 604 /* 605 * Doubly-linked circular list of all buffers submitted for IO while 606 * checkpointing. [j_list_lock] 607 */ 608 struct journal_head *t_checkpoint_io_list; 609 610 /* 611 * Doubly-linked circular list of metadata buffers being shadowed by log 612 * IO. The IO buffers on the iobuf list and the shadow buffers on this 613 * list match each other one for one at all times. [j_list_lock] 614 */ 615 struct journal_head *t_shadow_list; 616 617 /* 618 * List of inodes whose data we've modified in data=ordered mode. 619 * [j_list_lock] 620 */ 621 struct list_head t_inode_list; 622 623 /* 624 * Protects info related to handles 625 */ 626 spinlock_t t_handle_lock; 627 628 /* 629 * Longest time some handle had to wait for running transaction 630 */ 631 unsigned long t_max_wait; 632 633 /* 634 * When transaction started 635 */ 636 unsigned long t_start; 637 638 /* 639 * When commit was requested 640 */ 641 unsigned long t_requested; 642 643 /* 644 * Checkpointing stats [j_checkpoint_sem] 645 */ 646 struct transaction_chp_stats_s t_chp_stats; 647 648 /* 649 * Number of outstanding updates running on this transaction 650 * [t_handle_lock] 651 */ 652 atomic_t t_updates; 653 654 /* 655 * Number of buffers reserved for use by all handles in this transaction 656 * handle but not yet modified. [t_handle_lock] 657 */ 658 atomic_t t_outstanding_credits; 659 660 /* 661 * Forward and backward links for the circular list of all transactions 662 * awaiting checkpoint. [j_list_lock] 663 */ 664 transaction_t *t_cpnext, *t_cpprev; 665 666 /* 667 * When will the transaction expire (become due for commit), in jiffies? 668 * [no locking] 669 */ 670 unsigned long t_expires; 671 672 /* 673 * When this transaction started, in nanoseconds [no locking] 674 */ 675 ktime_t t_start_time; 676 677 /* 678 * How many handles used this transaction? [t_handle_lock] 679 */ 680 atomic_t t_handle_count; 681 682 /* 683 * This transaction is being forced and some process is 684 * waiting for it to finish. 685 */ 686 unsigned int t_synchronous_commit:1; 687 688 /* Disk flush needs to be sent to fs partition [no locking] */ 689 int t_need_data_flush; 690 691 /* 692 * For use by the filesystem to store fs-specific data 693 * structures associated with the transaction 694 */ 695 struct list_head t_private_list; 696}; 697 698struct transaction_run_stats_s { 699 unsigned long rs_wait; 700 unsigned long rs_request_delay; 701 unsigned long rs_running; 702 unsigned long rs_locked; 703 unsigned long rs_flushing; 704 unsigned long rs_logging; 705 706 __u32 rs_handle_count; 707 __u32 rs_blocks; 708 __u32 rs_blocks_logged; 709}; 710 711struct transaction_stats_s { 712 unsigned long ts_tid; 713 unsigned long ts_requested; 714 struct transaction_run_stats_s run; 715}; 716 717static inline unsigned long 718jbd2_time_diff(unsigned long start, unsigned long end) 719{ 720 if (end >= start) 721 return end - start; 722 723 return end + (MAX_JIFFY_OFFSET - start); 724} 725 726#define JBD2_NR_BATCH 64 727 728/** 729 * struct journal_s - The journal_s type is the concrete type associated with 730 * journal_t. 731 * @j_flags: General journaling state flags 732 * @j_errno: Is there an outstanding uncleared error on the journal (from a 733 * prior abort)? 734 * @j_sb_buffer: First part of superblock buffer 735 * @j_superblock: Second part of superblock buffer 736 * @j_format_version: Version of the superblock format 737 * @j_state_lock: Protect the various scalars in the journal 738 * @j_barrier_count: Number of processes waiting to create a barrier lock 739 * @j_barrier: The barrier lock itself 740 * @j_running_transaction: The current running transaction.. 741 * @j_committing_transaction: the transaction we are pushing to disk 742 * @j_checkpoint_transactions: a linked circular list of all transactions 743 * waiting for checkpointing 744 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction 745 * to start committing, or for a barrier lock to be released 746 * @j_wait_done_commit: Wait queue for waiting for commit to complete 747 * @j_wait_commit: Wait queue to trigger commit 748 * @j_wait_updates: Wait queue to wait for updates to complete 749 * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop 750 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints 751 * @j_head: Journal head - identifies the first unused block in the journal 752 * @j_tail: Journal tail - identifies the oldest still-used block in the 753 * journal. 754 * @j_free: Journal free - how many free blocks are there in the journal? 755 * @j_first: The block number of the first usable block 756 * @j_last: The block number one beyond the last usable block 757 * @j_dev: Device where we store the journal 758 * @j_blocksize: blocksize for the location where we store the journal. 759 * @j_blk_offset: starting block offset for into the device where we store the 760 * journal 761 * @j_fs_dev: Device which holds the client fs. For internal journal this will 762 * be equal to j_dev 763 * @j_reserved_credits: Number of buffers reserved from the running transaction 764 * @j_maxlen: Total maximum capacity of the journal region on disk. 765 * @j_list_lock: Protects the buffer lists and internal buffer state. 766 * @j_inode: Optional inode where we store the journal. If present, all journal 767 * block numbers are mapped into this inode via bmap(). 768 * @j_tail_sequence: Sequence number of the oldest transaction in the log 769 * @j_transaction_sequence: Sequence number of the next transaction to grant 770 * @j_commit_sequence: Sequence number of the most recently committed 771 * transaction 772 * @j_commit_request: Sequence number of the most recent transaction wanting 773 * commit 774 * @j_uuid: Uuid of client object. 775 * @j_task: Pointer to the current commit thread for this journal 776 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a 777 * single compound commit transaction 778 * @j_commit_interval: What is the maximum transaction lifetime before we begin 779 * a commit? 780 * @j_commit_timer: The timer used to wakeup the commit thread 781 * @j_revoke_lock: Protect the revoke table 782 * @j_revoke: The revoke table - maintains the list of revoked blocks in the 783 * current transaction. 784 * @j_revoke_table: alternate revoke tables for j_revoke 785 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction 786 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the 787 * number that will fit in j_blocksize 788 * @j_last_sync_writer: most recent pid which did a synchronous write 789 * @j_history: Buffer storing the transactions statistics history 790 * @j_history_max: Maximum number of transactions in the statistics history 791 * @j_history_cur: Current number of transactions in the statistics history 792 * @j_history_lock: Protect the transactions statistics history 793 * @j_proc_entry: procfs entry for the jbd statistics directory 794 * @j_stats: Overall statistics 795 * @j_private: An opaque pointer to fs-private information. 796 */ 797 798struct journal_s 799{ 800 /* General journaling state flags [j_state_lock] */ 801 unsigned long j_flags; 802 803 /* 804 * Is there an outstanding uncleared error on the journal (from a prior 805 * abort)? [j_state_lock] 806 */ 807 int j_errno; 808 809 /* The superblock buffer */ 810 struct buffer_head *j_sb_buffer; 811 journal_superblock_t *j_superblock; 812 813 /* Version of the superblock format */ 814 int j_format_version; 815 816 /* 817 * Protect the various scalars in the journal 818 */ 819 rwlock_t j_state_lock; 820 821 /* 822 * Number of processes waiting to create a barrier lock [j_state_lock] 823 */ 824 int j_barrier_count; 825 826 /* The barrier lock itself */ 827 struct mutex j_barrier; 828 829 /* 830 * Transactions: The current running transaction... 831 * [j_state_lock] [caller holding open handle] 832 */ 833 transaction_t *j_running_transaction; 834 835 /* 836 * the transaction we are pushing to disk 837 * [j_state_lock] [caller holding open handle] 838 */ 839 transaction_t *j_committing_transaction; 840 841 /* 842 * ... and a linked circular list of all transactions waiting for 843 * checkpointing. [j_list_lock] 844 */ 845 transaction_t *j_checkpoint_transactions; 846 847 /* 848 * Wait queue for waiting for a locked transaction to start committing, 849 * or for a barrier lock to be released 850 */ 851 wait_queue_head_t j_wait_transaction_locked; 852 853 /* Wait queue for waiting for commit to complete */ 854 wait_queue_head_t j_wait_done_commit; 855 856 /* Wait queue to trigger commit */ 857 wait_queue_head_t j_wait_commit; 858 859 /* Wait queue to wait for updates to complete */ 860 wait_queue_head_t j_wait_updates; 861 862 /* Wait queue to wait for reserved buffer credits to drop */ 863 wait_queue_head_t j_wait_reserved; 864 865 /* Semaphore for locking against concurrent checkpoints */ 866 struct mutex j_checkpoint_mutex; 867 868 /* 869 * List of buffer heads used by the checkpoint routine. This 870 * was moved from jbd2_log_do_checkpoint() to reduce stack 871 * usage. Access to this array is controlled by the 872 * j_checkpoint_mutex. [j_checkpoint_mutex] 873 */ 874 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 875 876 /* 877 * Journal head: identifies the first unused block in the journal. 878 * [j_state_lock] 879 */ 880 unsigned long j_head; 881 882 /* 883 * Journal tail: identifies the oldest still-used block in the journal. 884 * [j_state_lock] 885 */ 886 unsigned long j_tail; 887 888 /* 889 * Journal free: how many free blocks are there in the journal? 890 * [j_state_lock] 891 */ 892 unsigned long j_free; 893 894 /* 895 * Journal start and end: the block numbers of the first usable block 896 * and one beyond the last usable block in the journal. [j_state_lock] 897 */ 898 unsigned long j_first; 899 unsigned long j_last; 900 901 /* 902 * Device, blocksize and starting block offset for the location where we 903 * store the journal. 904 */ 905 struct block_device *j_dev; 906 int j_blocksize; 907 unsigned long long j_blk_offset; 908 char j_devname[BDEVNAME_SIZE+24]; 909 910 /* 911 * Device which holds the client fs. For internal journal this will be 912 * equal to j_dev. 913 */ 914 struct block_device *j_fs_dev; 915 916 /* Total maximum capacity of the journal region on disk. */ 917 unsigned int j_maxlen; 918 919 /* Number of buffers reserved from the running transaction */ 920 atomic_t j_reserved_credits; 921 922 /* 923 * Protects the buffer lists and internal buffer state. 924 */ 925 spinlock_t j_list_lock; 926 927 /* Optional inode where we store the journal. If present, all */ 928 /* journal block numbers are mapped into this inode via */ 929 /* bmap(). */ 930 struct inode *j_inode; 931 932 /* 933 * Sequence number of the oldest transaction in the log [j_state_lock] 934 */ 935 tid_t j_tail_sequence; 936 937 /* 938 * Sequence number of the next transaction to grant [j_state_lock] 939 */ 940 tid_t j_transaction_sequence; 941 942 /* 943 * Sequence number of the most recently committed transaction 944 * [j_state_lock]. 945 */ 946 tid_t j_commit_sequence; 947 948 /* 949 * Sequence number of the most recent transaction wanting commit 950 * [j_state_lock] 951 */ 952 tid_t j_commit_request; 953 954 /* 955 * Journal uuid: identifies the object (filesystem, LVM volume etc) 956 * backed by this journal. This will eventually be replaced by an array 957 * of uuids, allowing us to index multiple devices within a single 958 * journal and to perform atomic updates across them. 959 */ 960 __u8 j_uuid[16]; 961 962 /* Pointer to the current commit thread for this journal */ 963 struct task_struct *j_task; 964 965 /* 966 * Maximum number of metadata buffers to allow in a single compound 967 * commit transaction 968 */ 969 int j_max_transaction_buffers; 970 971 /* 972 * What is the maximum transaction lifetime before we begin a commit? 973 */ 974 unsigned long j_commit_interval; 975 976 /* The timer used to wakeup the commit thread: */ 977 struct timer_list j_commit_timer; 978 979 /* 980 * The revoke table: maintains the list of revoked blocks in the 981 * current transaction. [j_revoke_lock] 982 */ 983 spinlock_t j_revoke_lock; 984 struct jbd2_revoke_table_s *j_revoke; 985 struct jbd2_revoke_table_s *j_revoke_table[2]; 986 987 /* 988 * array of bhs for jbd2_journal_commit_transaction 989 */ 990 struct buffer_head **j_wbuf; 991 int j_wbufsize; 992 993 /* 994 * this is the pid of hte last person to run a synchronous operation 995 * through the journal 996 */ 997 pid_t j_last_sync_writer; 998 999 /* 1000 * the average amount of time in nanoseconds it takes to commit a 1001 * transaction to disk. [j_state_lock] 1002 */ 1003 u64 j_average_commit_time; 1004 1005 /* 1006 * minimum and maximum times that we should wait for 1007 * additional filesystem operations to get batched into a 1008 * synchronous handle in microseconds 1009 */ 1010 u32 j_min_batch_time; 1011 u32 j_max_batch_time; 1012 1013 /* This function is called when a transaction is closed */ 1014 void (*j_commit_callback)(journal_t *, 1015 transaction_t *); 1016 1017 /* 1018 * Journal statistics 1019 */ 1020 spinlock_t j_history_lock; 1021 struct proc_dir_entry *j_proc_entry; 1022 struct transaction_stats_s j_stats; 1023 1024 /* Failed journal commit ID */ 1025 unsigned int j_failed_commit; 1026 1027 /* 1028 * An opaque pointer to fs-private information. ext3 puts its 1029 * superblock pointer here 1030 */ 1031 void *j_private; 1032 1033 /* Reference to checksum algorithm driver via cryptoapi */ 1034 struct crypto_shash *j_chksum_driver; 1035 1036 /* Precomputed journal UUID checksum for seeding other checksums */ 1037 __u32 j_csum_seed; 1038}; 1039 1040/* journal feature predicate functions */ 1041#define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1042static inline bool jbd2_has_feature_##name(journal_t *j) \ 1043{ \ 1044 return ((j)->j_format_version >= 2 && \ 1045 ((j)->j_superblock->s_feature_compat & \ 1046 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1047} \ 1048static inline void jbd2_set_feature_##name(journal_t *j) \ 1049{ \ 1050 (j)->j_superblock->s_feature_compat |= \ 1051 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1052} \ 1053static inline void jbd2_clear_feature_##name(journal_t *j) \ 1054{ \ 1055 (j)->j_superblock->s_feature_compat &= \ 1056 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1057} 1058 1059#define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1060static inline bool jbd2_has_feature_##name(journal_t *j) \ 1061{ \ 1062 return ((j)->j_format_version >= 2 && \ 1063 ((j)->j_superblock->s_feature_ro_compat & \ 1064 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1065} \ 1066static inline void jbd2_set_feature_##name(journal_t *j) \ 1067{ \ 1068 (j)->j_superblock->s_feature_ro_compat |= \ 1069 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1070} \ 1071static inline void jbd2_clear_feature_##name(journal_t *j) \ 1072{ \ 1073 (j)->j_superblock->s_feature_ro_compat &= \ 1074 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1075} 1076 1077#define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1078static inline bool jbd2_has_feature_##name(journal_t *j) \ 1079{ \ 1080 return ((j)->j_format_version >= 2 && \ 1081 ((j)->j_superblock->s_feature_incompat & \ 1082 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1083} \ 1084static inline void jbd2_set_feature_##name(journal_t *j) \ 1085{ \ 1086 (j)->j_superblock->s_feature_incompat |= \ 1087 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1088} \ 1089static inline void jbd2_clear_feature_##name(journal_t *j) \ 1090{ \ 1091 (j)->j_superblock->s_feature_incompat &= \ 1092 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1093} 1094 1095JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1096 1097JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1098JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1099JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1100JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1101JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1102 1103/* 1104 * Journal flag definitions 1105 */ 1106#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1107#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1108#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1109#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1110#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1111#define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1112#define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1113 * data write error in ordered 1114 * mode */ 1115#define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */ 1116 1117/* 1118 * Function declarations for the journaling transaction and buffer 1119 * management 1120 */ 1121 1122/* Filing buffers */ 1123extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1124extern void __jbd2_journal_refile_buffer(struct journal_head *); 1125extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1126extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1127extern void __journal_free_buffer(struct journal_head *bh); 1128extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1129extern void __journal_clean_data_list(transaction_t *transaction); 1130static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1131{ 1132 list_add_tail(&bh->b_assoc_buffers, head); 1133} 1134static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1135{ 1136 list_del_init(&bh->b_assoc_buffers); 1137} 1138 1139/* Log buffer allocation */ 1140struct buffer_head *jbd2_journal_get_descriptor_buffer(journal_t *journal); 1141int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1142int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1143 unsigned long *block); 1144int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1145void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1146 1147/* Commit management */ 1148extern void jbd2_journal_commit_transaction(journal_t *); 1149 1150/* Checkpoint list management */ 1151void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1152int __jbd2_journal_remove_checkpoint(struct journal_head *); 1153void jbd2_journal_destroy_checkpoint(journal_t *journal); 1154void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1155 1156 1157/* 1158 * Triggers 1159 */ 1160 1161struct jbd2_buffer_trigger_type { 1162 /* 1163 * Fired a the moment data to write to the journal are known to be 1164 * stable - so either at the moment b_frozen_data is created or just 1165 * before a buffer is written to the journal. mapped_data is a mapped 1166 * buffer that is the frozen data for commit. 1167 */ 1168 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1169 struct buffer_head *bh, void *mapped_data, 1170 size_t size); 1171 1172 /* 1173 * Fired during journal abort for dirty buffers that will not be 1174 * committed. 1175 */ 1176 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1177 struct buffer_head *bh); 1178}; 1179 1180extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1181 void *mapped_data, 1182 struct jbd2_buffer_trigger_type *triggers); 1183extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1184 struct jbd2_buffer_trigger_type *triggers); 1185 1186/* Buffer IO */ 1187extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1188 struct journal_head *jh_in, 1189 struct buffer_head **bh_out, 1190 sector_t blocknr); 1191 1192/* Transaction locking */ 1193extern void __wait_on_journal (journal_t *); 1194 1195/* Transaction cache support */ 1196extern void jbd2_journal_destroy_transaction_cache(void); 1197extern int jbd2_journal_init_transaction_cache(void); 1198extern void jbd2_journal_free_transaction(transaction_t *); 1199 1200/* 1201 * Journal locking. 1202 * 1203 * We need to lock the journal during transaction state changes so that nobody 1204 * ever tries to take a handle on the running transaction while we are in the 1205 * middle of moving it to the commit phase. j_state_lock does this. 1206 * 1207 * Note that the locking is completely interrupt unsafe. We never touch 1208 * journal structures from interrupts. 1209 */ 1210 1211static inline handle_t *journal_current_handle(void) 1212{ 1213 return current->journal_info; 1214} 1215 1216/* The journaling code user interface: 1217 * 1218 * Create and destroy handles 1219 * Register buffer modifications against the current transaction. 1220 */ 1221 1222extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1223extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1224 gfp_t gfp_mask, unsigned int type, 1225 unsigned int line_no); 1226extern int jbd2_journal_restart(handle_t *, int nblocks); 1227extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask); 1228extern int jbd2_journal_start_reserved(handle_t *handle, 1229 unsigned int type, unsigned int line_no); 1230extern void jbd2_journal_free_reserved(handle_t *handle); 1231extern int jbd2_journal_extend (handle_t *, int nblocks); 1232extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1233extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1234extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1235void jbd2_journal_set_triggers(struct buffer_head *, 1236 struct jbd2_buffer_trigger_type *type); 1237extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1238extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1239extern void journal_sync_buffer (struct buffer_head *); 1240extern int jbd2_journal_invalidatepage(journal_t *, 1241 struct page *, unsigned int, unsigned int); 1242extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); 1243extern int jbd2_journal_stop(handle_t *); 1244extern int jbd2_journal_flush (journal_t *); 1245extern void jbd2_journal_lock_updates (journal_t *); 1246extern void jbd2_journal_unlock_updates (journal_t *); 1247 1248extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1249 struct block_device *fs_dev, 1250 unsigned long long start, int len, int bsize); 1251extern journal_t * jbd2_journal_init_inode (struct inode *); 1252extern int jbd2_journal_update_format (journal_t *); 1253extern int jbd2_journal_check_used_features 1254 (journal_t *, unsigned long, unsigned long, unsigned long); 1255extern int jbd2_journal_check_available_features 1256 (journal_t *, unsigned long, unsigned long, unsigned long); 1257extern int jbd2_journal_set_features 1258 (journal_t *, unsigned long, unsigned long, unsigned long); 1259extern void jbd2_journal_clear_features 1260 (journal_t *, unsigned long, unsigned long, unsigned long); 1261extern int jbd2_journal_load (journal_t *journal); 1262extern int jbd2_journal_destroy (journal_t *); 1263extern int jbd2_journal_recover (journal_t *journal); 1264extern int jbd2_journal_wipe (journal_t *, int); 1265extern int jbd2_journal_skip_recovery (journal_t *); 1266extern void jbd2_journal_update_sb_errno(journal_t *); 1267extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1268 unsigned long, int); 1269extern void __jbd2_journal_abort_hard (journal_t *); 1270extern void jbd2_journal_abort (journal_t *, int); 1271extern int jbd2_journal_errno (journal_t *); 1272extern void jbd2_journal_ack_err (journal_t *); 1273extern int jbd2_journal_clear_err (journal_t *); 1274extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1275extern int jbd2_journal_force_commit(journal_t *); 1276extern int jbd2_journal_force_commit_nested(journal_t *); 1277extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode); 1278extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1279 struct jbd2_inode *inode, loff_t new_size); 1280extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1281extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1282 1283/* 1284 * journal_head management 1285 */ 1286struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1287struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1288void jbd2_journal_put_journal_head(struct journal_head *jh); 1289 1290/* 1291 * handle management 1292 */ 1293extern struct kmem_cache *jbd2_handle_cache; 1294 1295static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1296{ 1297 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1298} 1299 1300static inline void jbd2_free_handle(handle_t *handle) 1301{ 1302 kmem_cache_free(jbd2_handle_cache, handle); 1303} 1304 1305/* 1306 * jbd2_inode management (optional, for those file systems that want to use 1307 * dynamically allocated jbd2_inode structures) 1308 */ 1309extern struct kmem_cache *jbd2_inode_cache; 1310 1311static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1312{ 1313 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1314} 1315 1316static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1317{ 1318 kmem_cache_free(jbd2_inode_cache, jinode); 1319} 1320 1321/* Primary revoke support */ 1322#define JOURNAL_REVOKE_DEFAULT_HASH 256 1323extern int jbd2_journal_init_revoke(journal_t *, int); 1324extern void jbd2_journal_destroy_revoke_caches(void); 1325extern int jbd2_journal_init_revoke_caches(void); 1326 1327extern void jbd2_journal_destroy_revoke(journal_t *); 1328extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1329extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1330extern void jbd2_journal_write_revoke_records(journal_t *journal, 1331 transaction_t *transaction, 1332 struct list_head *log_bufs, 1333 int write_op); 1334 1335/* Recovery revoke support */ 1336extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1337extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1338extern void jbd2_journal_clear_revoke(journal_t *); 1339extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1340extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1341 1342/* 1343 * The log thread user interface: 1344 * 1345 * Request space in the current transaction, and force transaction commit 1346 * transitions on demand. 1347 */ 1348 1349int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1350int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1351int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1352int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1353int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1354int jbd2_log_do_checkpoint(journal_t *journal); 1355int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1356 1357void __jbd2_log_wait_for_space(journal_t *journal); 1358extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1359extern int jbd2_cleanup_journal_tail(journal_t *); 1360 1361/* 1362 * is_journal_abort 1363 * 1364 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1365 * bit, when set, indicates that we have had a fatal error somewhere, 1366 * either inside the journaling layer or indicated to us by the client 1367 * (eg. ext3), and that we and should not commit any further 1368 * transactions. 1369 */ 1370 1371static inline int is_journal_aborted(journal_t *journal) 1372{ 1373 return journal->j_flags & JBD2_ABORT; 1374} 1375 1376static inline int is_handle_aborted(handle_t *handle) 1377{ 1378 if (handle->h_aborted || !handle->h_transaction) 1379 return 1; 1380 return is_journal_aborted(handle->h_transaction->t_journal); 1381} 1382 1383static inline void jbd2_journal_abort_handle(handle_t *handle) 1384{ 1385 handle->h_aborted = 1; 1386} 1387 1388#endif /* __KERNEL__ */ 1389 1390/* Comparison functions for transaction IDs: perform comparisons using 1391 * modulo arithmetic so that they work over sequence number wraps. */ 1392 1393static inline int tid_gt(tid_t x, tid_t y) 1394{ 1395 int difference = (x - y); 1396 return (difference > 0); 1397} 1398 1399static inline int tid_geq(tid_t x, tid_t y) 1400{ 1401 int difference = (x - y); 1402 return (difference >= 0); 1403} 1404 1405extern int jbd2_journal_blocks_per_page(struct inode *inode); 1406extern size_t journal_tag_bytes(journal_t *journal); 1407 1408static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1409{ 1410 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1411} 1412 1413static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1414{ 1415 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1416 journal->j_chksum_driver == NULL); 1417 1418 return journal->j_chksum_driver != NULL; 1419} 1420 1421/* 1422 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for 1423 * transaction control blocks. 1424 */ 1425#define JBD2_CONTROL_BLOCKS_SHIFT 5 1426 1427/* 1428 * Return the minimum number of blocks which must be free in the journal 1429 * before a new transaction may be started. Must be called under j_state_lock. 1430 */ 1431static inline int jbd2_space_needed(journal_t *journal) 1432{ 1433 int nblocks = journal->j_max_transaction_buffers; 1434 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT); 1435} 1436 1437/* 1438 * Return number of free blocks in the log. Must be called under j_state_lock. 1439 */ 1440static inline unsigned long jbd2_log_space_left(journal_t *journal) 1441{ 1442 /* Allow for rounding errors */ 1443 unsigned long free = journal->j_free - 32; 1444 1445 if (journal->j_committing_transaction) { 1446 unsigned long committing = atomic_read(&journal-> 1447 j_committing_transaction->t_outstanding_credits); 1448 1449 /* Transaction + control blocks */ 1450 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT); 1451 } 1452 return free; 1453} 1454 1455/* 1456 * Definitions which augment the buffer_head layer 1457 */ 1458 1459/* journaling buffer types */ 1460#define BJ_None 0 /* Not journaled */ 1461#define BJ_Metadata 1 /* Normal journaled metadata */ 1462#define BJ_Forget 2 /* Buffer superseded by this transaction */ 1463#define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1464#define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1465#define BJ_Types 5 1466 1467extern int jbd_blocks_per_page(struct inode *inode); 1468 1469/* JBD uses a CRC32 checksum */ 1470#define JBD_MAX_CHECKSUM_SIZE 4 1471 1472static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1473 const void *address, unsigned int length) 1474{ 1475 struct { 1476 struct shash_desc shash; 1477 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1478 } desc; 1479 int err; 1480 1481 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1482 JBD_MAX_CHECKSUM_SIZE); 1483 1484 desc.shash.tfm = journal->j_chksum_driver; 1485 desc.shash.flags = 0; 1486 *(u32 *)desc.ctx = crc; 1487 1488 err = crypto_shash_update(&desc.shash, address, length); 1489 BUG_ON(err); 1490 1491 return *(u32 *)desc.ctx; 1492} 1493 1494/* Return most recent uncommitted transaction */ 1495static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1496{ 1497 tid_t tid; 1498 1499 read_lock(&journal->j_state_lock); 1500 tid = journal->j_commit_request; 1501 if (journal->j_running_transaction) 1502 tid = journal->j_running_transaction->t_tid; 1503 read_unlock(&journal->j_state_lock); 1504 return tid; 1505} 1506 1507#ifdef __KERNEL__ 1508 1509#define buffer_trace_init(bh) do {} while (0) 1510#define print_buffer_fields(bh) do {} while (0) 1511#define print_buffer_trace(bh) do {} while (0) 1512#define BUFFER_TRACE(bh, info) do {} while (0) 1513#define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1514#define JBUFFER_TRACE(jh, info) do {} while (0) 1515 1516#endif /* __KERNEL__ */ 1517 1518#define EFSBADCRC EBADMSG /* Bad CRC detected */ 1519#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1520 1521#endif /* _LINUX_JBD2_H */ 1522