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 #define jfs_debug jbd_debug 24 #else 25 26 #include <linux/types.h> 27 #include <linux/buffer_head.h> 28 #include <linux/journal-head.h> 29 #include <linux/stddef.h> 30 #include <linux/bit_spinlock.h> 31 #include <linux/mutex.h> 32 #include <linux/timer.h> 33 #include <linux/slab.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 59 extern u8 jbd2_journal_enable_debug; 60 61 #define jbd_debug(n, f, a...) \ 62 do { \ 63 if ((n) <= jbd2_journal_enable_debug) { \ 64 printk (KERN_DEBUG "(%s, %d): %s: ", \ 65 __FILE__, __LINE__, __func__); \ 66 printk (f, ## a); \ 67 } \ 68 } while (0) 69 #else 70 #define jbd_debug(f, a...) /**/ 71 #endif 72 73 extern void *jbd2_alloc(size_t size, gfp_t flags); 74 extern void jbd2_free(void *ptr, size_t size); 75 76 #define JBD2_MIN_JOURNAL_BLOCKS 1024 77 78 #ifdef __KERNEL__ 79 80 /** 81 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 82 * 83 * All filesystem modifications made by the process go 84 * through this handle. Recursive operations (such as quota operations) 85 * are gathered into a single update. 86 * 87 * The buffer credits field is used to account for journaled buffers 88 * being modified by the running process. To ensure that there is 89 * enough log space for all outstanding operations, we need to limit the 90 * number of outstanding buffers possible at any time. When the 91 * operation completes, any buffer credits not used are credited back to 92 * the transaction, so that at all times we know how many buffers the 93 * outstanding updates on a transaction might possibly touch. 94 * 95 * This is an opaque datatype. 96 **/ 97 typedef struct handle_s handle_t; /* Atomic operation type */ 98 99 100 /** 101 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 102 * 103 * journal_t is linked to from the fs superblock structure. 104 * 105 * We use the journal_t to keep track of all outstanding transaction 106 * activity on the filesystem, and to manage the state of the log 107 * writing process. 108 * 109 * This is an opaque datatype. 110 **/ 111 typedef struct journal_s journal_t; /* Journal control structure */ 112 #endif 113 114 /* 115 * Internal structures used by the logging mechanism: 116 */ 117 118 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 119 120 /* 121 * On-disk structures 122 */ 123 124 /* 125 * Descriptor block types: 126 */ 127 128 #define JBD2_DESCRIPTOR_BLOCK 1 129 #define JBD2_COMMIT_BLOCK 2 130 #define JBD2_SUPERBLOCK_V1 3 131 #define JBD2_SUPERBLOCK_V2 4 132 #define JBD2_REVOKE_BLOCK 5 133 134 /* 135 * Standard header for all descriptor blocks: 136 */ 137 typedef struct journal_header_s 138 { 139 __be32 h_magic; 140 __be32 h_blocktype; 141 __be32 h_sequence; 142 } journal_header_t; 143 144 /* 145 * Checksum types. 146 */ 147 #define JBD2_CRC32_CHKSUM 1 148 #define JBD2_MD5_CHKSUM 2 149 #define JBD2_SHA1_CHKSUM 3 150 151 #define JBD2_CRC32_CHKSUM_SIZE 4 152 153 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 154 /* 155 * Commit block header for storing transactional checksums: 156 */ 157 struct commit_header { 158 __be32 h_magic; 159 __be32 h_blocktype; 160 __be32 h_sequence; 161 unsigned char h_chksum_type; 162 unsigned char h_chksum_size; 163 unsigned char h_padding[2]; 164 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 165 __be64 h_commit_sec; 166 __be32 h_commit_nsec; 167 }; 168 169 /* 170 * The block tag: used to describe a single buffer in the journal. 171 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 172 * raw struct shouldn't be used for pointer math or sizeof() - use 173 * journal_tag_bytes(journal) instead to compute this. 174 */ 175 typedef struct journal_block_tag_s 176 { 177 __be32 t_blocknr; /* The on-disk block number */ 178 __be32 t_flags; /* See below */ 179 __be32 t_blocknr_high; /* most-significant high 32bits. */ 180 } journal_block_tag_t; 181 182 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high)) 183 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t)) 184 185 /* 186 * The revoke descriptor: used on disk to describe a series of blocks to 187 * be revoked from the log 188 */ 189 typedef struct jbd2_journal_revoke_header_s 190 { 191 journal_header_t r_header; 192 __be32 r_count; /* Count of bytes used in the block */ 193 } jbd2_journal_revoke_header_t; 194 195 196 /* Definitions for the journal tag flags word: */ 197 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 198 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 199 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 200 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 201 202 203 /* 204 * The journal superblock. All fields are in big-endian byte order. 205 */ 206 typedef struct journal_superblock_s 207 { 208 /* 0x0000 */ 209 journal_header_t s_header; 210 211 /* 0x000C */ 212 /* Static information describing the journal */ 213 __be32 s_blocksize; /* journal device blocksize */ 214 __be32 s_maxlen; /* total blocks in journal file */ 215 __be32 s_first; /* first block of log information */ 216 217 /* 0x0018 */ 218 /* Dynamic information describing the current state of the log */ 219 __be32 s_sequence; /* first commit ID expected in log */ 220 __be32 s_start; /* blocknr of start of log */ 221 222 /* 0x0020 */ 223 /* Error value, as set by jbd2_journal_abort(). */ 224 __be32 s_errno; 225 226 /* 0x0024 */ 227 /* Remaining fields are only valid in a version-2 superblock */ 228 __be32 s_feature_compat; /* compatible feature set */ 229 __be32 s_feature_incompat; /* incompatible feature set */ 230 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 231 /* 0x0030 */ 232 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 233 234 /* 0x0040 */ 235 __be32 s_nr_users; /* Nr of filesystems sharing log */ 236 237 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 238 239 /* 0x0048 */ 240 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 241 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 242 243 /* 0x0050 */ 244 __u32 s_padding[44]; 245 246 /* 0x0100 */ 247 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 248 /* 0x0400 */ 249 } journal_superblock_t; 250 251 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 252 ((j)->j_format_version >= 2 && \ 253 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 254 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 255 ((j)->j_format_version >= 2 && \ 256 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 257 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 258 ((j)->j_format_version >= 2 && \ 259 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 260 261 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 262 263 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 264 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 265 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 266 267 /* Features known to this kernel version: */ 268 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 269 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 270 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 271 JBD2_FEATURE_INCOMPAT_64BIT | \ 272 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT) 273 274 #ifdef __KERNEL__ 275 276 #include <linux/fs.h> 277 #include <linux/sched.h> 278 279 #define J_ASSERT(assert) BUG_ON(!(assert)) 280 281 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 282 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 283 284 #if defined(JBD2_PARANOID_IOFAIL) 285 #define J_EXPECT(expr, why...) J_ASSERT(expr) 286 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 287 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 288 #else 289 #define __journal_expect(expr, why...) \ 290 ({ \ 291 int val = (expr); \ 292 if (!val) { \ 293 printk(KERN_ERR \ 294 "JBD2 unexpected failure: %s: %s;\n", \ 295 __func__, #expr); \ 296 printk(KERN_ERR why "\n"); \ 297 } \ 298 val; \ 299 }) 300 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 301 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 302 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 303 #endif 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_State, /* Pins most journal_head state */ 314 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 315 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */ 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) 328 329 static inline struct buffer_head *jh2bh(struct journal_head *jh) 330 { 331 return jh->b_bh; 332 } 333 334 static inline struct journal_head *bh2jh(struct buffer_head *bh) 335 { 336 return bh->b_private; 337 } 338 339 static inline void jbd_lock_bh_state(struct buffer_head *bh) 340 { 341 bit_spin_lock(BH_State, &bh->b_state); 342 } 343 344 static inline int jbd_trylock_bh_state(struct buffer_head *bh) 345 { 346 return bit_spin_trylock(BH_State, &bh->b_state); 347 } 348 349 static inline int jbd_is_locked_bh_state(struct buffer_head *bh) 350 { 351 return bit_spin_is_locked(BH_State, &bh->b_state); 352 } 353 354 static inline void jbd_unlock_bh_state(struct buffer_head *bh) 355 { 356 bit_spin_unlock(BH_State, &bh->b_state); 357 } 358 359 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 360 { 361 bit_spin_lock(BH_JournalHead, &bh->b_state); 362 } 363 364 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 365 { 366 bit_spin_unlock(BH_JournalHead, &bh->b_state); 367 } 368 369 /* Flags in jbd_inode->i_flags */ 370 #define __JI_COMMIT_RUNNING 0 371 /* Commit of the inode data in progress. We use this flag to protect us from 372 * concurrent deletion of inode. We cannot use reference to inode for this 373 * since we cannot afford doing last iput() on behalf of kjournald 374 */ 375 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 376 377 /** 378 * struct jbd_inode is the structure linking inodes in ordered mode 379 * present in a transaction so that we can sync them during commit. 380 */ 381 struct jbd2_inode { 382 /* Which transaction does this inode belong to? Either the running 383 * transaction or the committing one. [j_list_lock] */ 384 transaction_t *i_transaction; 385 386 /* Pointer to the running transaction modifying inode's data in case 387 * there is already a committing transaction touching it. [j_list_lock] */ 388 transaction_t *i_next_transaction; 389 390 /* List of inodes in the i_transaction [j_list_lock] */ 391 struct list_head i_list; 392 393 /* VFS inode this inode belongs to [constant during the lifetime 394 * of the structure] */ 395 struct inode *i_vfs_inode; 396 397 /* Flags of inode [j_list_lock] */ 398 unsigned int i_flags; 399 }; 400 401 struct jbd2_revoke_table_s; 402 403 /** 404 * struct handle_s - The handle_s type is the concrete type associated with 405 * handle_t. 406 * @h_transaction: Which compound transaction is this update a part of? 407 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. 408 * @h_ref: Reference count on this handle 409 * @h_err: Field for caller's use to track errors through large fs operations 410 * @h_sync: flag for sync-on-close 411 * @h_jdata: flag to force data journaling 412 * @h_aborted: flag indicating fatal error on handle 413 **/ 414 415 /* Docbook can't yet cope with the bit fields, but will leave the documentation 416 * in so it can be fixed later. 417 */ 418 419 struct handle_s 420 { 421 /* Which compound transaction is this update a part of? */ 422 transaction_t *h_transaction; 423 424 /* Number of remaining buffers we are allowed to dirty: */ 425 int h_buffer_credits; 426 427 /* Reference count on this handle */ 428 int h_ref; 429 430 /* Field for caller's use to track errors through large fs */ 431 /* operations */ 432 int h_err; 433 434 /* Flags [no locking] */ 435 unsigned int h_sync: 1; /* sync-on-close */ 436 unsigned int h_jdata: 1; /* force data journaling */ 437 unsigned int h_aborted: 1; /* fatal error on handle */ 438 439 #ifdef CONFIG_DEBUG_LOCK_ALLOC 440 struct lockdep_map h_lockdep_map; 441 #endif 442 }; 443 444 445 /* 446 * Some stats for checkpoint phase 447 */ 448 struct transaction_chp_stats_s { 449 unsigned long cs_chp_time; 450 __u32 cs_forced_to_close; 451 __u32 cs_written; 452 __u32 cs_dropped; 453 }; 454 455 /* The transaction_t type is the guts of the journaling mechanism. It 456 * tracks a compound transaction through its various states: 457 * 458 * RUNNING: accepting new updates 459 * LOCKED: Updates still running but we don't accept new ones 460 * RUNDOWN: Updates are tidying up but have finished requesting 461 * new buffers to modify (state not used for now) 462 * FLUSH: All updates complete, but we are still writing to disk 463 * COMMIT: All data on disk, writing commit record 464 * FINISHED: We still have to keep the transaction for checkpointing. 465 * 466 * The transaction keeps track of all of the buffers modified by a 467 * running transaction, and all of the buffers committed but not yet 468 * flushed to home for finished transactions. 469 */ 470 471 /* 472 * Lock ranking: 473 * 474 * j_list_lock 475 * ->jbd_lock_bh_journal_head() (This is "innermost") 476 * 477 * j_state_lock 478 * ->jbd_lock_bh_state() 479 * 480 * jbd_lock_bh_state() 481 * ->j_list_lock 482 * 483 * j_state_lock 484 * ->t_handle_lock 485 * 486 * j_state_lock 487 * ->j_list_lock (journal_unmap_buffer) 488 * 489 */ 490 491 struct transaction_s 492 { 493 /* Pointer to the journal for this transaction. [no locking] */ 494 journal_t *t_journal; 495 496 /* Sequence number for this transaction [no locking] */ 497 tid_t t_tid; 498 499 /* 500 * Transaction's current state 501 * [no locking - only kjournald2 alters this] 502 * [j_list_lock] guards transition of a transaction into T_FINISHED 503 * state and subsequent call of __jbd2_journal_drop_transaction() 504 * FIXME: needs barriers 505 * KLUDGE: [use j_state_lock] 506 */ 507 enum { 508 T_RUNNING, 509 T_LOCKED, 510 T_RUNDOWN, 511 T_FLUSH, 512 T_COMMIT, 513 T_FINISHED 514 } t_state; 515 516 /* 517 * Where in the log does this transaction's commit start? [no locking] 518 */ 519 unsigned long t_log_start; 520 521 /* Number of buffers on the t_buffers list [j_list_lock] */ 522 int t_nr_buffers; 523 524 /* 525 * Doubly-linked circular list of all buffers reserved but not yet 526 * modified by this transaction [j_list_lock] 527 */ 528 struct journal_head *t_reserved_list; 529 530 /* 531 * Doubly-linked circular list of all metadata buffers owned by this 532 * transaction [j_list_lock] 533 */ 534 struct journal_head *t_buffers; 535 536 /* 537 * Doubly-linked circular list of all forget buffers (superseded 538 * buffers which we can un-checkpoint once this transaction commits) 539 * [j_list_lock] 540 */ 541 struct journal_head *t_forget; 542 543 /* 544 * Doubly-linked circular list of all buffers still to be flushed before 545 * this transaction can be checkpointed. [j_list_lock] 546 */ 547 struct journal_head *t_checkpoint_list; 548 549 /* 550 * Doubly-linked circular list of all buffers submitted for IO while 551 * checkpointing. [j_list_lock] 552 */ 553 struct journal_head *t_checkpoint_io_list; 554 555 /* 556 * Doubly-linked circular list of temporary buffers currently undergoing 557 * IO in the log [j_list_lock] 558 */ 559 struct journal_head *t_iobuf_list; 560 561 /* 562 * Doubly-linked circular list of metadata buffers being shadowed by log 563 * IO. The IO buffers on the iobuf list and the shadow buffers on this 564 * list match each other one for one at all times. [j_list_lock] 565 */ 566 struct journal_head *t_shadow_list; 567 568 /* 569 * Doubly-linked circular list of control buffers being written to the 570 * log. [j_list_lock] 571 */ 572 struct journal_head *t_log_list; 573 574 /* 575 * List of inodes whose data we've modified in data=ordered mode. 576 * [j_list_lock] 577 */ 578 struct list_head t_inode_list; 579 580 /* 581 * Protects info related to handles 582 */ 583 spinlock_t t_handle_lock; 584 585 /* 586 * Longest time some handle had to wait for running transaction 587 */ 588 unsigned long t_max_wait; 589 590 /* 591 * When transaction started 592 */ 593 unsigned long t_start; 594 595 /* 596 * Checkpointing stats [j_checkpoint_sem] 597 */ 598 struct transaction_chp_stats_s t_chp_stats; 599 600 /* 601 * Number of outstanding updates running on this transaction 602 * [t_handle_lock] 603 */ 604 atomic_t t_updates; 605 606 /* 607 * Number of buffers reserved for use by all handles in this transaction 608 * handle but not yet modified. [t_handle_lock] 609 */ 610 atomic_t t_outstanding_credits; 611 612 /* 613 * Forward and backward links for the circular list of all transactions 614 * awaiting checkpoint. [j_list_lock] 615 */ 616 transaction_t *t_cpnext, *t_cpprev; 617 618 /* 619 * When will the transaction expire (become due for commit), in jiffies? 620 * [no locking] 621 */ 622 unsigned long t_expires; 623 624 /* 625 * When this transaction started, in nanoseconds [no locking] 626 */ 627 ktime_t t_start_time; 628 629 /* 630 * How many handles used this transaction? [t_handle_lock] 631 */ 632 atomic_t t_handle_count; 633 634 /* 635 * This transaction is being forced and some process is 636 * waiting for it to finish. 637 */ 638 unsigned int t_synchronous_commit:1; 639 unsigned int t_flushed_data_blocks:1; 640 641 /* 642 * For use by the filesystem to store fs-specific data 643 * structures associated with the transaction 644 */ 645 struct list_head t_private_list; 646 }; 647 648 struct transaction_run_stats_s { 649 unsigned long rs_wait; 650 unsigned long rs_running; 651 unsigned long rs_locked; 652 unsigned long rs_flushing; 653 unsigned long rs_logging; 654 655 __u32 rs_handle_count; 656 __u32 rs_blocks; 657 __u32 rs_blocks_logged; 658 }; 659 660 struct transaction_stats_s { 661 unsigned long ts_tid; 662 struct transaction_run_stats_s run; 663 }; 664 665 static inline unsigned long 666 jbd2_time_diff(unsigned long start, unsigned long end) 667 { 668 if (end >= start) 669 return end - start; 670 671 return end + (MAX_JIFFY_OFFSET - start); 672 } 673 674 #define JBD2_NR_BATCH 64 675 676 /** 677 * struct journal_s - The journal_s type is the concrete type associated with 678 * journal_t. 679 * @j_flags: General journaling state flags 680 * @j_errno: Is there an outstanding uncleared error on the journal (from a 681 * prior abort)? 682 * @j_sb_buffer: First part of superblock buffer 683 * @j_superblock: Second part of superblock buffer 684 * @j_format_version: Version of the superblock format 685 * @j_state_lock: Protect the various scalars in the journal 686 * @j_barrier_count: Number of processes waiting to create a barrier lock 687 * @j_barrier: The barrier lock itself 688 * @j_running_transaction: The current running transaction.. 689 * @j_committing_transaction: the transaction we are pushing to disk 690 * @j_checkpoint_transactions: a linked circular list of all transactions 691 * waiting for checkpointing 692 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction 693 * to start committing, or for a barrier lock to be released 694 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete 695 * @j_wait_done_commit: Wait queue for waiting for commit to complete 696 * @j_wait_checkpoint: Wait queue to trigger checkpointing 697 * @j_wait_commit: Wait queue to trigger commit 698 * @j_wait_updates: Wait queue to wait for updates to complete 699 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints 700 * @j_head: Journal head - identifies the first unused block in the journal 701 * @j_tail: Journal tail - identifies the oldest still-used block in the 702 * journal. 703 * @j_free: Journal free - how many free blocks are there in the journal? 704 * @j_first: The block number of the first usable block 705 * @j_last: The block number one beyond the last usable block 706 * @j_dev: Device where we store the journal 707 * @j_blocksize: blocksize for the location where we store the journal. 708 * @j_blk_offset: starting block offset for into the device where we store the 709 * journal 710 * @j_fs_dev: Device which holds the client fs. For internal journal this will 711 * be equal to j_dev 712 * @j_maxlen: Total maximum capacity of the journal region on disk. 713 * @j_list_lock: Protects the buffer lists and internal buffer state. 714 * @j_inode: Optional inode where we store the journal. If present, all journal 715 * block numbers are mapped into this inode via bmap(). 716 * @j_tail_sequence: Sequence number of the oldest transaction in the log 717 * @j_transaction_sequence: Sequence number of the next transaction to grant 718 * @j_commit_sequence: Sequence number of the most recently committed 719 * transaction 720 * @j_commit_request: Sequence number of the most recent transaction wanting 721 * commit 722 * @j_uuid: Uuid of client object. 723 * @j_task: Pointer to the current commit thread for this journal 724 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a 725 * single compound commit transaction 726 * @j_commit_interval: What is the maximum transaction lifetime before we begin 727 * a commit? 728 * @j_commit_timer: The timer used to wakeup the commit thread 729 * @j_revoke_lock: Protect the revoke table 730 * @j_revoke: The revoke table - maintains the list of revoked blocks in the 731 * current transaction. 732 * @j_revoke_table: alternate revoke tables for j_revoke 733 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction 734 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the 735 * number that will fit in j_blocksize 736 * @j_last_sync_writer: most recent pid which did a synchronous write 737 * @j_history: Buffer storing the transactions statistics history 738 * @j_history_max: Maximum number of transactions in the statistics history 739 * @j_history_cur: Current number of transactions in the statistics history 740 * @j_history_lock: Protect the transactions statistics history 741 * @j_proc_entry: procfs entry for the jbd statistics directory 742 * @j_stats: Overall statistics 743 * @j_private: An opaque pointer to fs-private information. 744 */ 745 746 struct journal_s 747 { 748 /* General journaling state flags [j_state_lock] */ 749 unsigned long j_flags; 750 751 /* 752 * Is there an outstanding uncleared error on the journal (from a prior 753 * abort)? [j_state_lock] 754 */ 755 int j_errno; 756 757 /* The superblock buffer */ 758 struct buffer_head *j_sb_buffer; 759 journal_superblock_t *j_superblock; 760 761 /* Version of the superblock format */ 762 int j_format_version; 763 764 /* 765 * Protect the various scalars in the journal 766 */ 767 rwlock_t j_state_lock; 768 769 /* 770 * Number of processes waiting to create a barrier lock [j_state_lock] 771 */ 772 int j_barrier_count; 773 774 /* The barrier lock itself */ 775 struct mutex j_barrier; 776 777 /* 778 * Transactions: The current running transaction... 779 * [j_state_lock] [caller holding open handle] 780 */ 781 transaction_t *j_running_transaction; 782 783 /* 784 * the transaction we are pushing to disk 785 * [j_state_lock] [caller holding open handle] 786 */ 787 transaction_t *j_committing_transaction; 788 789 /* 790 * ... and a linked circular list of all transactions waiting for 791 * checkpointing. [j_list_lock] 792 */ 793 transaction_t *j_checkpoint_transactions; 794 795 /* 796 * Wait queue for waiting for a locked transaction to start committing, 797 * or for a barrier lock to be released 798 */ 799 wait_queue_head_t j_wait_transaction_locked; 800 801 /* Wait queue for waiting for checkpointing to complete */ 802 wait_queue_head_t j_wait_logspace; 803 804 /* Wait queue for waiting for commit to complete */ 805 wait_queue_head_t j_wait_done_commit; 806 807 /* Wait queue to trigger checkpointing */ 808 wait_queue_head_t j_wait_checkpoint; 809 810 /* Wait queue to trigger commit */ 811 wait_queue_head_t j_wait_commit; 812 813 /* Wait queue to wait for updates to complete */ 814 wait_queue_head_t j_wait_updates; 815 816 /* Semaphore for locking against concurrent checkpoints */ 817 struct mutex j_checkpoint_mutex; 818 819 /* 820 * List of buffer heads used by the checkpoint routine. This 821 * was moved from jbd2_log_do_checkpoint() to reduce stack 822 * usage. Access to this array is controlled by the 823 * j_checkpoint_mutex. [j_checkpoint_mutex] 824 */ 825 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 826 827 /* 828 * Journal head: identifies the first unused block in the journal. 829 * [j_state_lock] 830 */ 831 unsigned long j_head; 832 833 /* 834 * Journal tail: identifies the oldest still-used block in the journal. 835 * [j_state_lock] 836 */ 837 unsigned long j_tail; 838 839 /* 840 * Journal free: how many free blocks are there in the journal? 841 * [j_state_lock] 842 */ 843 unsigned long j_free; 844 845 /* 846 * Journal start and end: the block numbers of the first usable block 847 * and one beyond the last usable block in the journal. [j_state_lock] 848 */ 849 unsigned long j_first; 850 unsigned long j_last; 851 852 /* 853 * Device, blocksize and starting block offset for the location where we 854 * store the journal. 855 */ 856 struct block_device *j_dev; 857 int j_blocksize; 858 unsigned long long j_blk_offset; 859 char j_devname[BDEVNAME_SIZE+24]; 860 861 /* 862 * Device which holds the client fs. For internal journal this will be 863 * equal to j_dev. 864 */ 865 struct block_device *j_fs_dev; 866 867 /* Total maximum capacity of the journal region on disk. */ 868 unsigned int j_maxlen; 869 870 /* 871 * Protects the buffer lists and internal buffer state. 872 */ 873 spinlock_t j_list_lock; 874 875 /* Optional inode where we store the journal. If present, all */ 876 /* journal block numbers are mapped into this inode via */ 877 /* bmap(). */ 878 struct inode *j_inode; 879 880 /* 881 * Sequence number of the oldest transaction in the log [j_state_lock] 882 */ 883 tid_t j_tail_sequence; 884 885 /* 886 * Sequence number of the next transaction to grant [j_state_lock] 887 */ 888 tid_t j_transaction_sequence; 889 890 /* 891 * Sequence number of the most recently committed transaction 892 * [j_state_lock]. 893 */ 894 tid_t j_commit_sequence; 895 896 /* 897 * Sequence number of the most recent transaction wanting commit 898 * [j_state_lock] 899 */ 900 tid_t j_commit_request; 901 902 /* 903 * Journal uuid: identifies the object (filesystem, LVM volume etc) 904 * backed by this journal. This will eventually be replaced by an array 905 * of uuids, allowing us to index multiple devices within a single 906 * journal and to perform atomic updates across them. 907 */ 908 __u8 j_uuid[16]; 909 910 /* Pointer to the current commit thread for this journal */ 911 struct task_struct *j_task; 912 913 /* 914 * Maximum number of metadata buffers to allow in a single compound 915 * commit transaction 916 */ 917 int j_max_transaction_buffers; 918 919 /* 920 * What is the maximum transaction lifetime before we begin a commit? 921 */ 922 unsigned long j_commit_interval; 923 924 /* The timer used to wakeup the commit thread: */ 925 struct timer_list j_commit_timer; 926 927 /* 928 * The revoke table: maintains the list of revoked blocks in the 929 * current transaction. [j_revoke_lock] 930 */ 931 spinlock_t j_revoke_lock; 932 struct jbd2_revoke_table_s *j_revoke; 933 struct jbd2_revoke_table_s *j_revoke_table[2]; 934 935 /* 936 * array of bhs for jbd2_journal_commit_transaction 937 */ 938 struct buffer_head **j_wbuf; 939 int j_wbufsize; 940 941 /* 942 * this is the pid of hte last person to run a synchronous operation 943 * through the journal 944 */ 945 pid_t j_last_sync_writer; 946 947 /* 948 * the average amount of time in nanoseconds it takes to commit a 949 * transaction to disk. [j_state_lock] 950 */ 951 u64 j_average_commit_time; 952 953 /* 954 * minimum and maximum times that we should wait for 955 * additional filesystem operations to get batched into a 956 * synchronous handle in microseconds 957 */ 958 u32 j_min_batch_time; 959 u32 j_max_batch_time; 960 961 /* This function is called when a transaction is closed */ 962 void (*j_commit_callback)(journal_t *, 963 transaction_t *); 964 965 /* 966 * Journal statistics 967 */ 968 spinlock_t j_history_lock; 969 struct proc_dir_entry *j_proc_entry; 970 struct transaction_stats_s j_stats; 971 972 /* Failed journal commit ID */ 973 unsigned int j_failed_commit; 974 975 /* 976 * An opaque pointer to fs-private information. ext3 puts its 977 * superblock pointer here 978 */ 979 void *j_private; 980 }; 981 982 /* 983 * Journal flag definitions 984 */ 985 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 986 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 987 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 988 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 989 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 990 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 991 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 992 * data write error in ordered 993 * mode */ 994 995 /* 996 * Function declarations for the journaling transaction and buffer 997 * management 998 */ 999 1000 /* Filing buffers */ 1001 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1002 extern void __jbd2_journal_unfile_buffer(struct journal_head *); 1003 extern void __jbd2_journal_refile_buffer(struct journal_head *); 1004 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1005 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1006 extern void __journal_free_buffer(struct journal_head *bh); 1007 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1008 extern void __journal_clean_data_list(transaction_t *transaction); 1009 1010 /* Log buffer allocation */ 1011 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *); 1012 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1013 1014 /* Commit management */ 1015 extern void jbd2_journal_commit_transaction(journal_t *); 1016 1017 /* Checkpoint list management */ 1018 int __jbd2_journal_clean_checkpoint_list(journal_t *journal); 1019 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1020 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1021 1022 1023 /* 1024 * Triggers 1025 */ 1026 1027 struct jbd2_buffer_trigger_type { 1028 /* 1029 * Fired a the moment data to write to the journal are known to be 1030 * stable - so either at the moment b_frozen_data is created or just 1031 * before a buffer is written to the journal. mapped_data is a mapped 1032 * buffer that is the frozen data for commit. 1033 */ 1034 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1035 struct buffer_head *bh, void *mapped_data, 1036 size_t size); 1037 1038 /* 1039 * Fired during journal abort for dirty buffers that will not be 1040 * committed. 1041 */ 1042 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1043 struct buffer_head *bh); 1044 }; 1045 1046 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1047 void *mapped_data, 1048 struct jbd2_buffer_trigger_type *triggers); 1049 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1050 struct jbd2_buffer_trigger_type *triggers); 1051 1052 /* Buffer IO */ 1053 extern int 1054 jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1055 struct journal_head *jh_in, 1056 struct journal_head **jh_out, 1057 unsigned long long blocknr); 1058 1059 /* Transaction locking */ 1060 extern void __wait_on_journal (journal_t *); 1061 1062 /* 1063 * Journal locking. 1064 * 1065 * We need to lock the journal during transaction state changes so that nobody 1066 * ever tries to take a handle on the running transaction while we are in the 1067 * middle of moving it to the commit phase. j_state_lock does this. 1068 * 1069 * Note that the locking is completely interrupt unsafe. We never touch 1070 * journal structures from interrupts. 1071 */ 1072 1073 static inline handle_t *journal_current_handle(void) 1074 { 1075 return current->journal_info; 1076 } 1077 1078 /* The journaling code user interface: 1079 * 1080 * Create and destroy handles 1081 * Register buffer modifications against the current transaction. 1082 */ 1083 1084 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1085 extern handle_t *jbd2__journal_start(journal_t *, int nblocks, int gfp_mask); 1086 extern int jbd2_journal_restart(handle_t *, int nblocks); 1087 extern int jbd2__journal_restart(handle_t *, int nblocks, int gfp_mask); 1088 extern int jbd2_journal_extend (handle_t *, int nblocks); 1089 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1090 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1091 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1092 void jbd2_journal_set_triggers(struct buffer_head *, 1093 struct jbd2_buffer_trigger_type *type); 1094 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1095 extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *); 1096 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1097 extern void journal_sync_buffer (struct buffer_head *); 1098 extern void jbd2_journal_invalidatepage(journal_t *, 1099 struct page *, unsigned long); 1100 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); 1101 extern int jbd2_journal_stop(handle_t *); 1102 extern int jbd2_journal_flush (journal_t *); 1103 extern void jbd2_journal_lock_updates (journal_t *); 1104 extern void jbd2_journal_unlock_updates (journal_t *); 1105 1106 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1107 struct block_device *fs_dev, 1108 unsigned long long start, int len, int bsize); 1109 extern journal_t * jbd2_journal_init_inode (struct inode *); 1110 extern int jbd2_journal_update_format (journal_t *); 1111 extern int jbd2_journal_check_used_features 1112 (journal_t *, unsigned long, unsigned long, unsigned long); 1113 extern int jbd2_journal_check_available_features 1114 (journal_t *, unsigned long, unsigned long, unsigned long); 1115 extern int jbd2_journal_set_features 1116 (journal_t *, unsigned long, unsigned long, unsigned long); 1117 extern void jbd2_journal_clear_features 1118 (journal_t *, unsigned long, unsigned long, unsigned long); 1119 extern int jbd2_journal_load (journal_t *journal); 1120 extern int jbd2_journal_destroy (journal_t *); 1121 extern int jbd2_journal_recover (journal_t *journal); 1122 extern int jbd2_journal_wipe (journal_t *, int); 1123 extern int jbd2_journal_skip_recovery (journal_t *); 1124 extern void jbd2_journal_update_superblock (journal_t *, int); 1125 extern void __jbd2_journal_abort_hard (journal_t *); 1126 extern void jbd2_journal_abort (journal_t *, int); 1127 extern int jbd2_journal_errno (journal_t *); 1128 extern void jbd2_journal_ack_err (journal_t *); 1129 extern int jbd2_journal_clear_err (journal_t *); 1130 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1131 extern int jbd2_journal_force_commit(journal_t *); 1132 extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode); 1133 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1134 struct jbd2_inode *inode, loff_t new_size); 1135 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1136 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1137 1138 /* 1139 * journal_head management 1140 */ 1141 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1142 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1143 void jbd2_journal_remove_journal_head(struct buffer_head *bh); 1144 void jbd2_journal_put_journal_head(struct journal_head *jh); 1145 1146 /* 1147 * handle management 1148 */ 1149 extern struct kmem_cache *jbd2_handle_cache; 1150 1151 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1152 { 1153 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags); 1154 } 1155 1156 static inline void jbd2_free_handle(handle_t *handle) 1157 { 1158 kmem_cache_free(jbd2_handle_cache, handle); 1159 } 1160 1161 /* Primary revoke support */ 1162 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1163 extern int jbd2_journal_init_revoke(journal_t *, int); 1164 extern void jbd2_journal_destroy_revoke_caches(void); 1165 extern int jbd2_journal_init_revoke_caches(void); 1166 1167 extern void jbd2_journal_destroy_revoke(journal_t *); 1168 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1169 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1170 extern void jbd2_journal_write_revoke_records(journal_t *, 1171 transaction_t *, int); 1172 1173 /* Recovery revoke support */ 1174 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1175 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1176 extern void jbd2_journal_clear_revoke(journal_t *); 1177 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1178 1179 /* 1180 * The log thread user interface: 1181 * 1182 * Request space in the current transaction, and force transaction commit 1183 * transitions on demand. 1184 */ 1185 1186 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */ 1187 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1188 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1189 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1190 int jbd2_journal_force_commit_nested(journal_t *journal); 1191 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1192 int jbd2_log_do_checkpoint(journal_t *journal); 1193 1194 void __jbd2_log_wait_for_space(journal_t *journal); 1195 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1196 extern int jbd2_cleanup_journal_tail(journal_t *); 1197 1198 /* Debugging code only: */ 1199 1200 #define jbd_ENOSYS() \ 1201 do { \ 1202 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \ 1203 current->state = TASK_UNINTERRUPTIBLE; \ 1204 schedule(); \ 1205 } while (1) 1206 1207 /* 1208 * is_journal_abort 1209 * 1210 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1211 * bit, when set, indicates that we have had a fatal error somewhere, 1212 * either inside the journaling layer or indicated to us by the client 1213 * (eg. ext3), and that we and should not commit any further 1214 * transactions. 1215 */ 1216 1217 static inline int is_journal_aborted(journal_t *journal) 1218 { 1219 return journal->j_flags & JBD2_ABORT; 1220 } 1221 1222 static inline int is_handle_aborted(handle_t *handle) 1223 { 1224 if (handle->h_aborted) 1225 return 1; 1226 return is_journal_aborted(handle->h_transaction->t_journal); 1227 } 1228 1229 static inline void jbd2_journal_abort_handle(handle_t *handle) 1230 { 1231 handle->h_aborted = 1; 1232 } 1233 1234 #endif /* __KERNEL__ */ 1235 1236 /* Comparison functions for transaction IDs: perform comparisons using 1237 * modulo arithmetic so that they work over sequence number wraps. */ 1238 1239 static inline int tid_gt(tid_t x, tid_t y) 1240 { 1241 int difference = (x - y); 1242 return (difference > 0); 1243 } 1244 1245 static inline int tid_geq(tid_t x, tid_t y) 1246 { 1247 int difference = (x - y); 1248 return (difference >= 0); 1249 } 1250 1251 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1252 extern size_t journal_tag_bytes(journal_t *journal); 1253 1254 /* 1255 * Return the minimum number of blocks which must be free in the journal 1256 * before a new transaction may be started. Must be called under j_state_lock. 1257 */ 1258 static inline int jbd_space_needed(journal_t *journal) 1259 { 1260 int nblocks = journal->j_max_transaction_buffers; 1261 if (journal->j_committing_transaction) 1262 nblocks += atomic_read(&journal->j_committing_transaction-> 1263 t_outstanding_credits); 1264 return nblocks; 1265 } 1266 1267 /* 1268 * Definitions which augment the buffer_head layer 1269 */ 1270 1271 /* journaling buffer types */ 1272 #define BJ_None 0 /* Not journaled */ 1273 #define BJ_Metadata 1 /* Normal journaled metadata */ 1274 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1275 #define BJ_IO 3 /* Buffer is for temporary IO use */ 1276 #define BJ_Shadow 4 /* Buffer contents being shadowed to the log */ 1277 #define BJ_LogCtl 5 /* Buffer contains log descriptors */ 1278 #define BJ_Reserved 6 /* Buffer is reserved for access by journal */ 1279 #define BJ_Types 7 1280 1281 extern int jbd_blocks_per_page(struct inode *inode); 1282 1283 #ifdef __KERNEL__ 1284 1285 #define buffer_trace_init(bh) do {} while (0) 1286 #define print_buffer_fields(bh) do {} while (0) 1287 #define print_buffer_trace(bh) do {} while (0) 1288 #define BUFFER_TRACE(bh, info) do {} while (0) 1289 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1290 #define JBUFFER_TRACE(jh, info) do {} while (0) 1291 1292 /* 1293 * jbd2_dev_to_name is a utility function used by the jbd2 and ext4 1294 * tracing infrastructure to map a dev_t to a device name. 1295 */ 1296 extern const char *jbd2_dev_to_name(dev_t device); 1297 1298 #endif /* __KERNEL__ */ 1299 1300 #endif /* _LINUX_JBD2_H */ 1301