1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * linux/include/linux/jbd2.h 4 * 5 * Written by Stephen C. Tweedie <sct@redhat.com> 6 * 7 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved 8 * 9 * Definitions for transaction data structures for the buffer cache 10 * filesystem journaling support. 11 */ 12 13 #ifndef _LINUX_JBD2_H 14 #define _LINUX_JBD2_H 15 16 /* Allow this file to be included directly into e2fsprogs */ 17 #ifndef __KERNEL__ 18 #include "jfs_compat.h" 19 #define JBD2_DEBUG 20 #else 21 22 #include <linux/types.h> 23 #include <linux/buffer_head.h> 24 #include <linux/journal-head.h> 25 #include <linux/stddef.h> 26 #include <linux/mutex.h> 27 #include <linux/timer.h> 28 #include <linux/slab.h> 29 #include <linux/bit_spinlock.h> 30 #include <linux/blkdev.h> 31 #include <crypto/hash.h> 32 #endif 33 34 #define journal_oom_retry 1 35 36 /* 37 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds 38 * certain classes of error which can occur due to failed IOs. Under 39 * normal use we want ext4 to continue after such errors, because 40 * hardware _can_ fail, but for debugging purposes when running tests on 41 * known-good hardware we may want to trap these errors. 42 */ 43 #undef JBD2_PARANOID_IOFAIL 44 45 /* 46 * The default maximum commit age, in seconds. 47 */ 48 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5 49 50 #ifdef CONFIG_JBD2_DEBUG 51 /* 52 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal 53 * consistency checks. By default we don't do this unless 54 * CONFIG_JBD2_DEBUG is on. 55 */ 56 #define JBD2_EXPENSIVE_CHECKING 57 extern ushort jbd2_journal_enable_debug; 58 void __jbd2_debug(int level, const char *file, const char *func, 59 unsigned int line, const char *fmt, ...); 60 61 #define jbd_debug(n, fmt, a...) \ 62 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a) 63 #else 64 #define jbd_debug(n, fmt, a...) /**/ 65 #endif 66 67 extern void *jbd2_alloc(size_t size, gfp_t flags); 68 extern void jbd2_free(void *ptr, size_t size); 69 70 #define JBD2_MIN_JOURNAL_BLOCKS 1024 71 #define JBD2_DEFAULT_FAST_COMMIT_BLOCKS 256 72 73 #ifdef __KERNEL__ 74 75 /** 76 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 77 * 78 * All filesystem modifications made by the process go 79 * through this handle. Recursive operations (such as quota operations) 80 * are gathered into a single update. 81 * 82 * The buffer credits field is used to account for journaled buffers 83 * being modified by the running process. To ensure that there is 84 * enough log space for all outstanding operations, we need to limit the 85 * number of outstanding buffers possible at any time. When the 86 * operation completes, any buffer credits not used are credited back to 87 * the transaction, so that at all times we know how many buffers the 88 * outstanding updates on a transaction might possibly touch. 89 * 90 * This is an opaque datatype. 91 **/ 92 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ 93 94 95 /** 96 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 97 * 98 * journal_t is linked to from the fs superblock structure. 99 * 100 * We use the journal_t to keep track of all outstanding transaction 101 * activity on the filesystem, and to manage the state of the log 102 * writing process. 103 * 104 * This is an opaque datatype. 105 **/ 106 typedef struct journal_s journal_t; /* Journal control structure */ 107 #endif 108 109 /* 110 * Internal structures used by the logging mechanism: 111 */ 112 113 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 114 115 /* 116 * On-disk structures 117 */ 118 119 /* 120 * Descriptor block types: 121 */ 122 123 #define JBD2_DESCRIPTOR_BLOCK 1 124 #define JBD2_COMMIT_BLOCK 2 125 #define JBD2_SUPERBLOCK_V1 3 126 #define JBD2_SUPERBLOCK_V2 4 127 #define JBD2_REVOKE_BLOCK 5 128 129 /* 130 * Standard header for all descriptor blocks: 131 */ 132 typedef struct journal_header_s 133 { 134 __be32 h_magic; 135 __be32 h_blocktype; 136 __be32 h_sequence; 137 } journal_header_t; 138 139 /* 140 * Checksum types. 141 */ 142 #define JBD2_CRC32_CHKSUM 1 143 #define JBD2_MD5_CHKSUM 2 144 #define JBD2_SHA1_CHKSUM 3 145 #define JBD2_CRC32C_CHKSUM 4 146 147 #define JBD2_CRC32_CHKSUM_SIZE 4 148 149 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 150 /* 151 * Commit block header for storing transactional checksums: 152 * 153 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* 154 * fields are used to store a checksum of the descriptor and data blocks. 155 * 156 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum 157 * field is used to store crc32c(uuid+commit_block). Each journal metadata 158 * block gets its own checksum, and data block checksums are stored in 159 * journal_block_tag (in the descriptor). The other h_chksum* fields are 160 * not used. 161 * 162 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses 163 * journal_block_tag3_t to store a full 32-bit checksum. Everything else 164 * is the same as v2. 165 * 166 * Checksum v1, v2, and v3 are mutually exclusive features. 167 */ 168 struct commit_header { 169 __be32 h_magic; 170 __be32 h_blocktype; 171 __be32 h_sequence; 172 unsigned char h_chksum_type; 173 unsigned char h_chksum_size; 174 unsigned char h_padding[2]; 175 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 176 __be64 h_commit_sec; 177 __be32 h_commit_nsec; 178 }; 179 180 /* 181 * The block tag: used to describe a single buffer in the journal. 182 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 183 * raw struct shouldn't be used for pointer math or sizeof() - use 184 * journal_tag_bytes(journal) instead to compute this. 185 */ 186 typedef struct journal_block_tag3_s 187 { 188 __be32 t_blocknr; /* The on-disk block number */ 189 __be32 t_flags; /* See below */ 190 __be32 t_blocknr_high; /* most-significant high 32bits. */ 191 __be32 t_checksum; /* crc32c(uuid+seq+block) */ 192 } journal_block_tag3_t; 193 194 typedef struct journal_block_tag_s 195 { 196 __be32 t_blocknr; /* The on-disk block number */ 197 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ 198 __be16 t_flags; /* See below */ 199 __be32 t_blocknr_high; /* most-significant high 32bits. */ 200 } journal_block_tag_t; 201 202 /* Tail of descriptor or revoke block, for checksumming */ 203 struct jbd2_journal_block_tail { 204 __be32 t_checksum; /* crc32c(uuid+descr_block) */ 205 }; 206 207 /* 208 * The revoke descriptor: used on disk to describe a series of blocks to 209 * be revoked from the log 210 */ 211 typedef struct jbd2_journal_revoke_header_s 212 { 213 journal_header_t r_header; 214 __be32 r_count; /* Count of bytes used in the block */ 215 } jbd2_journal_revoke_header_t; 216 217 /* Definitions for the journal tag flags word: */ 218 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 219 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 220 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 221 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 222 223 224 /* 225 * The journal superblock. All fields are in big-endian byte order. 226 */ 227 typedef struct journal_superblock_s 228 { 229 /* 0x0000 */ 230 journal_header_t s_header; 231 232 /* 0x000C */ 233 /* Static information describing the journal */ 234 __be32 s_blocksize; /* journal device blocksize */ 235 __be32 s_maxlen; /* total blocks in journal file */ 236 __be32 s_first; /* first block of log information */ 237 238 /* 0x0018 */ 239 /* Dynamic information describing the current state of the log */ 240 __be32 s_sequence; /* first commit ID expected in log */ 241 __be32 s_start; /* blocknr of start of log */ 242 243 /* 0x0020 */ 244 /* Error value, as set by jbd2_journal_abort(). */ 245 __be32 s_errno; 246 247 /* 0x0024 */ 248 /* Remaining fields are only valid in a version-2 superblock */ 249 __be32 s_feature_compat; /* compatible feature set */ 250 __be32 s_feature_incompat; /* incompatible feature set */ 251 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 252 /* 0x0030 */ 253 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 254 255 /* 0x0040 */ 256 __be32 s_nr_users; /* Nr of filesystems sharing log */ 257 258 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 259 260 /* 0x0048 */ 261 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 262 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 263 264 /* 0x0050 */ 265 __u8 s_checksum_type; /* checksum type */ 266 __u8 s_padding2[3]; 267 /* 0x0054 */ 268 __be32 s_num_fc_blks; /* Number of fast commit blocks */ 269 /* 0x0058 */ 270 __u32 s_padding[41]; 271 __be32 s_checksum; /* crc32c(superblock) */ 272 273 /* 0x0100 */ 274 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 275 /* 0x0400 */ 276 } journal_superblock_t; 277 278 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */ 279 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 280 ((j)->j_format_version >= 2 && \ 281 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 282 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 283 ((j)->j_format_version >= 2 && \ 284 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 285 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 286 ((j)->j_format_version >= 2 && \ 287 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 288 289 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 290 291 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 292 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 293 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 294 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 295 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 296 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT 0x00000020 297 298 /* See "journal feature predicate functions" below */ 299 300 /* Features known to this kernel version: */ 301 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 302 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 303 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 304 JBD2_FEATURE_INCOMPAT_64BIT | \ 305 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 306 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 307 JBD2_FEATURE_INCOMPAT_CSUM_V3 | \ 308 JBD2_FEATURE_INCOMPAT_FAST_COMMIT) 309 310 #ifdef __KERNEL__ 311 312 #include <linux/fs.h> 313 #include <linux/sched.h> 314 315 enum jbd_state_bits { 316 BH_JBD /* Has an attached ext3 journal_head */ 317 = BH_PrivateStart, 318 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 319 BH_Freed, /* Has been freed (truncated) */ 320 BH_Revoked, /* Has been revoked from the log */ 321 BH_RevokeValid, /* Revoked flag is valid */ 322 BH_JBDDirty, /* Is dirty but journaled */ 323 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 324 BH_Shadow, /* IO on shadow buffer is running */ 325 BH_Verified, /* Metadata block has been verified ok */ 326 BH_JBDPrivateStart, /* First bit available for private use by FS */ 327 }; 328 329 BUFFER_FNS(JBD, jbd) 330 BUFFER_FNS(JWrite, jwrite) 331 BUFFER_FNS(JBDDirty, jbddirty) 332 TAS_BUFFER_FNS(JBDDirty, jbddirty) 333 BUFFER_FNS(Revoked, revoked) 334 TAS_BUFFER_FNS(Revoked, revoked) 335 BUFFER_FNS(RevokeValid, revokevalid) 336 TAS_BUFFER_FNS(RevokeValid, revokevalid) 337 BUFFER_FNS(Freed, freed) 338 BUFFER_FNS(Shadow, shadow) 339 BUFFER_FNS(Verified, verified) 340 341 static inline struct buffer_head *jh2bh(struct journal_head *jh) 342 { 343 return jh->b_bh; 344 } 345 346 static inline struct journal_head *bh2jh(struct buffer_head *bh) 347 { 348 return bh->b_private; 349 } 350 351 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 352 { 353 bit_spin_lock(BH_JournalHead, &bh->b_state); 354 } 355 356 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 357 { 358 bit_spin_unlock(BH_JournalHead, &bh->b_state); 359 } 360 361 #define J_ASSERT(assert) BUG_ON(!(assert)) 362 363 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 364 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 365 366 #if defined(JBD2_PARANOID_IOFAIL) 367 #define J_EXPECT(expr, why...) J_ASSERT(expr) 368 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 369 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 370 #else 371 #define __journal_expect(expr, why...) \ 372 ({ \ 373 int val = (expr); \ 374 if (!val) { \ 375 printk(KERN_ERR \ 376 "JBD2 unexpected failure: %s: %s;\n", \ 377 __func__, #expr); \ 378 printk(KERN_ERR why "\n"); \ 379 } \ 380 val; \ 381 }) 382 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 383 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 384 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 385 #endif 386 387 /* Flags in jbd_inode->i_flags */ 388 #define __JI_COMMIT_RUNNING 0 389 #define __JI_WRITE_DATA 1 390 #define __JI_WAIT_DATA 2 391 392 /* 393 * Commit of the inode data in progress. We use this flag to protect us from 394 * concurrent deletion of inode. We cannot use reference to inode for this 395 * since we cannot afford doing last iput() on behalf of kjournald 396 */ 397 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 398 /* Write allocated dirty buffers in this inode before commit */ 399 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA) 400 /* Wait for outstanding data writes for this inode before commit */ 401 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA) 402 403 /** 404 * struct jbd2_inode - The jbd_inode type is the structure linking inodes in 405 * ordered mode present in a transaction so that we can sync them during commit. 406 */ 407 struct jbd2_inode { 408 /** 409 * @i_transaction: 410 * 411 * Which transaction does this inode belong to? Either the running 412 * transaction or the committing one. [j_list_lock] 413 */ 414 transaction_t *i_transaction; 415 416 /** 417 * @i_next_transaction: 418 * 419 * Pointer to the running transaction modifying inode's data in case 420 * there is already a committing transaction touching it. [j_list_lock] 421 */ 422 transaction_t *i_next_transaction; 423 424 /** 425 * @i_list: List of inodes in the i_transaction [j_list_lock] 426 */ 427 struct list_head i_list; 428 429 /** 430 * @i_vfs_inode: 431 * 432 * VFS inode this inode belongs to [constant for lifetime of structure] 433 */ 434 struct inode *i_vfs_inode; 435 436 /** 437 * @i_flags: Flags of inode [j_list_lock] 438 */ 439 unsigned long i_flags; 440 441 /** 442 * @i_dirty_start: 443 * 444 * Offset in bytes where the dirty range for this inode starts. 445 * [j_list_lock] 446 */ 447 loff_t i_dirty_start; 448 449 /** 450 * @i_dirty_end: 451 * 452 * Inclusive offset in bytes where the dirty range for this inode 453 * ends. [j_list_lock] 454 */ 455 loff_t i_dirty_end; 456 }; 457 458 struct jbd2_revoke_table_s; 459 460 /** 461 * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete 462 * type associated with handle_t. 463 * @h_transaction: Which compound transaction is this update a part of? 464 * @h_journal: Which journal handle belongs to - used iff h_reserved set. 465 * @h_rsv_handle: Handle reserved for finishing the logical operation. 466 * @h_total_credits: Number of remaining buffers we are allowed to add to 467 * journal. These are dirty buffers and revoke descriptor blocks. 468 * @h_revoke_credits: Number of remaining revoke records available for handle 469 * @h_ref: Reference count on this handle. 470 * @h_err: Field for caller's use to track errors through large fs operations. 471 * @h_sync: Flag for sync-on-close. 472 * @h_jdata: Flag to force data journaling. 473 * @h_reserved: Flag for handle for reserved credits. 474 * @h_aborted: Flag indicating fatal error on handle. 475 * @h_type: For handle statistics. 476 * @h_line_no: For handle statistics. 477 * @h_start_jiffies: Handle Start time. 478 * @h_requested_credits: Holds @h_total_credits after handle is started. 479 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started. 480 * @saved_alloc_context: Saved context while transaction is open. 481 **/ 482 483 /* Docbook can't yet cope with the bit fields, but will leave the documentation 484 * in so it can be fixed later. 485 */ 486 487 struct jbd2_journal_handle 488 { 489 union { 490 transaction_t *h_transaction; 491 /* Which journal handle belongs to - used iff h_reserved set */ 492 journal_t *h_journal; 493 }; 494 495 handle_t *h_rsv_handle; 496 int h_total_credits; 497 int h_revoke_credits; 498 int h_revoke_credits_requested; 499 int h_ref; 500 int h_err; 501 502 /* Flags [no locking] */ 503 unsigned int h_sync: 1; 504 unsigned int h_jdata: 1; 505 unsigned int h_reserved: 1; 506 unsigned int h_aborted: 1; 507 unsigned int h_type: 8; 508 unsigned int h_line_no: 16; 509 510 unsigned long h_start_jiffies; 511 unsigned int h_requested_credits; 512 513 unsigned int saved_alloc_context; 514 }; 515 516 517 /* 518 * Some stats for checkpoint phase 519 */ 520 struct transaction_chp_stats_s { 521 unsigned long cs_chp_time; 522 __u32 cs_forced_to_close; 523 __u32 cs_written; 524 __u32 cs_dropped; 525 }; 526 527 /* The transaction_t type is the guts of the journaling mechanism. It 528 * tracks a compound transaction through its various states: 529 * 530 * RUNNING: accepting new updates 531 * LOCKED: Updates still running but we don't accept new ones 532 * RUNDOWN: Updates are tidying up but have finished requesting 533 * new buffers to modify (state not used for now) 534 * FLUSH: All updates complete, but we are still writing to disk 535 * COMMIT: All data on disk, writing commit record 536 * FINISHED: We still have to keep the transaction for checkpointing. 537 * 538 * The transaction keeps track of all of the buffers modified by a 539 * running transaction, and all of the buffers committed but not yet 540 * flushed to home for finished transactions. 541 * (Locking Documentation improved by LockDoc) 542 */ 543 544 /* 545 * Lock ranking: 546 * 547 * j_list_lock 548 * ->jbd_lock_bh_journal_head() (This is "innermost") 549 * 550 * j_state_lock 551 * ->b_state_lock 552 * 553 * b_state_lock 554 * ->j_list_lock 555 * 556 * j_state_lock 557 * ->t_handle_lock 558 * 559 * j_state_lock 560 * ->j_list_lock (journal_unmap_buffer) 561 * 562 */ 563 564 struct transaction_s 565 { 566 /* Pointer to the journal for this transaction. [no locking] */ 567 journal_t *t_journal; 568 569 /* Sequence number for this transaction [no locking] */ 570 tid_t t_tid; 571 572 /* 573 * Transaction's current state 574 * [no locking - only kjournald2 alters this] 575 * [j_list_lock] guards transition of a transaction into T_FINISHED 576 * state and subsequent call of __jbd2_journal_drop_transaction() 577 * FIXME: needs barriers 578 * KLUDGE: [use j_state_lock] 579 */ 580 enum { 581 T_RUNNING, 582 T_LOCKED, 583 T_SWITCH, 584 T_FLUSH, 585 T_COMMIT, 586 T_COMMIT_DFLUSH, 587 T_COMMIT_JFLUSH, 588 T_COMMIT_CALLBACK, 589 T_FINISHED 590 } t_state; 591 592 /* 593 * Where in the log does this transaction's commit start? [no locking] 594 */ 595 unsigned long t_log_start; 596 597 /* Number of buffers on the t_buffers list [j_list_lock] */ 598 int t_nr_buffers; 599 600 /* 601 * Doubly-linked circular list of all buffers reserved but not yet 602 * modified by this transaction [j_list_lock] 603 */ 604 struct journal_head *t_reserved_list; 605 606 /* 607 * Doubly-linked circular list of all metadata buffers owned by this 608 * transaction [j_list_lock] 609 */ 610 struct journal_head *t_buffers; 611 612 /* 613 * Doubly-linked circular list of all forget buffers (superseded 614 * buffers which we can un-checkpoint once this transaction commits) 615 * [j_list_lock] 616 */ 617 struct journal_head *t_forget; 618 619 /* 620 * Doubly-linked circular list of all buffers still to be flushed before 621 * this transaction can be checkpointed. [j_list_lock] 622 */ 623 struct journal_head *t_checkpoint_list; 624 625 /* 626 * Doubly-linked circular list of all buffers submitted for IO while 627 * checkpointing. [j_list_lock] 628 */ 629 struct journal_head *t_checkpoint_io_list; 630 631 /* 632 * Doubly-linked circular list of metadata buffers being shadowed by log 633 * IO. The IO buffers on the iobuf list and the shadow buffers on this 634 * list match each other one for one at all times. [j_list_lock] 635 */ 636 struct journal_head *t_shadow_list; 637 638 /* 639 * List of inodes associated with the transaction; e.g., ext4 uses 640 * this to track inodes in data=ordered and data=journal mode that 641 * need special handling on transaction commit; also used by ocfs2. 642 * [j_list_lock] 643 */ 644 struct list_head t_inode_list; 645 646 /* 647 * Protects info related to handles 648 */ 649 spinlock_t t_handle_lock; 650 651 /* 652 * Longest time some handle had to wait for running transaction 653 */ 654 unsigned long t_max_wait; 655 656 /* 657 * When transaction started 658 */ 659 unsigned long t_start; 660 661 /* 662 * When commit was requested [j_state_lock] 663 */ 664 unsigned long t_requested; 665 666 /* 667 * Checkpointing stats [j_list_lock] 668 */ 669 struct transaction_chp_stats_s t_chp_stats; 670 671 /* 672 * Number of outstanding updates running on this transaction 673 * [none] 674 */ 675 atomic_t t_updates; 676 677 /* 678 * Number of blocks reserved for this transaction in the journal. 679 * This is including all credits reserved when starting transaction 680 * handles as well as all journal descriptor blocks needed for this 681 * transaction. [none] 682 */ 683 atomic_t t_outstanding_credits; 684 685 /* 686 * Number of revoke records for this transaction added by already 687 * stopped handles. [none] 688 */ 689 atomic_t t_outstanding_revokes; 690 691 /* 692 * How many handles used this transaction? [none] 693 */ 694 atomic_t t_handle_count; 695 696 /* 697 * Forward and backward links for the circular list of all transactions 698 * awaiting checkpoint. [j_list_lock] 699 */ 700 transaction_t *t_cpnext, *t_cpprev; 701 702 /* 703 * When will the transaction expire (become due for commit), in jiffies? 704 * [no locking] 705 */ 706 unsigned long t_expires; 707 708 /* 709 * When this transaction started, in nanoseconds [no locking] 710 */ 711 ktime_t t_start_time; 712 713 /* 714 * This transaction is being forced and some process is 715 * waiting for it to finish. 716 */ 717 unsigned int t_synchronous_commit:1; 718 719 /* Disk flush needs to be sent to fs partition [no locking] */ 720 int t_need_data_flush; 721 722 /* 723 * For use by the filesystem to store fs-specific data 724 * structures associated with the transaction 725 */ 726 struct list_head t_private_list; 727 }; 728 729 struct transaction_run_stats_s { 730 unsigned long rs_wait; 731 unsigned long rs_request_delay; 732 unsigned long rs_running; 733 unsigned long rs_locked; 734 unsigned long rs_flushing; 735 unsigned long rs_logging; 736 737 __u32 rs_handle_count; 738 __u32 rs_blocks; 739 __u32 rs_blocks_logged; 740 }; 741 742 struct transaction_stats_s { 743 unsigned long ts_tid; 744 unsigned long ts_requested; 745 struct transaction_run_stats_s run; 746 }; 747 748 static inline unsigned long 749 jbd2_time_diff(unsigned long start, unsigned long end) 750 { 751 if (end >= start) 752 return end - start; 753 754 return end + (MAX_JIFFY_OFFSET - start); 755 } 756 757 #define JBD2_NR_BATCH 64 758 759 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY}; 760 761 #define JBD2_FC_REPLAY_STOP 0 762 #define JBD2_FC_REPLAY_CONTINUE 1 763 764 /** 765 * struct journal_s - The journal_s type is the concrete type associated with 766 * journal_t. 767 */ 768 struct journal_s 769 { 770 /** 771 * @j_flags: General journaling state flags [j_state_lock] 772 */ 773 unsigned long j_flags; 774 775 /** 776 * @j_errno: 777 * 778 * Is there an outstanding uncleared error on the journal (from a prior 779 * abort)? [j_state_lock] 780 */ 781 int j_errno; 782 783 /** 784 * @j_abort_mutex: Lock the whole aborting procedure. 785 */ 786 struct mutex j_abort_mutex; 787 788 /** 789 * @j_sb_buffer: The first part of the superblock buffer. 790 */ 791 struct buffer_head *j_sb_buffer; 792 793 /** 794 * @j_superblock: The second part of the superblock buffer. 795 */ 796 journal_superblock_t *j_superblock; 797 798 /** 799 * @j_format_version: Version of the superblock format. 800 */ 801 int j_format_version; 802 803 /** 804 * @j_state_lock: Protect the various scalars in the journal. 805 */ 806 rwlock_t j_state_lock; 807 808 /** 809 * @j_barrier_count: 810 * 811 * Number of processes waiting to create a barrier lock [j_state_lock] 812 */ 813 int j_barrier_count; 814 815 /** 816 * @j_barrier: The barrier lock itself. 817 */ 818 struct mutex j_barrier; 819 820 /** 821 * @j_running_transaction: 822 * 823 * Transactions: The current running transaction... 824 * [j_state_lock] [caller holding open handle] 825 */ 826 transaction_t *j_running_transaction; 827 828 /** 829 * @j_committing_transaction: 830 * 831 * the transaction we are pushing to disk 832 * [j_state_lock] [caller holding open handle] 833 */ 834 transaction_t *j_committing_transaction; 835 836 /** 837 * @j_checkpoint_transactions: 838 * 839 * ... and a linked circular list of all transactions waiting for 840 * checkpointing. [j_list_lock] 841 */ 842 transaction_t *j_checkpoint_transactions; 843 844 /** 845 * @j_wait_transaction_locked: 846 * 847 * Wait queue for waiting for a locked transaction to start committing, 848 * or for a barrier lock to be released. 849 */ 850 wait_queue_head_t j_wait_transaction_locked; 851 852 /** 853 * @j_wait_done_commit: Wait queue for waiting for commit to complete. 854 */ 855 wait_queue_head_t j_wait_done_commit; 856 857 /** 858 * @j_wait_commit: Wait queue to trigger commit. 859 */ 860 wait_queue_head_t j_wait_commit; 861 862 /** 863 * @j_wait_updates: Wait queue to wait for updates to complete. 864 */ 865 wait_queue_head_t j_wait_updates; 866 867 /** 868 * @j_wait_reserved: 869 * 870 * Wait queue to wait for reserved buffer credits to drop. 871 */ 872 wait_queue_head_t j_wait_reserved; 873 874 /** 875 * @j_fc_wait: 876 * 877 * Wait queue to wait for completion of async fast commits. 878 */ 879 wait_queue_head_t j_fc_wait; 880 881 /** 882 * @j_checkpoint_mutex: 883 * 884 * Semaphore for locking against concurrent checkpoints. 885 */ 886 struct mutex j_checkpoint_mutex; 887 888 /** 889 * @j_chkpt_bhs: 890 * 891 * List of buffer heads used by the checkpoint routine. This 892 * was moved from jbd2_log_do_checkpoint() to reduce stack 893 * usage. Access to this array is controlled by the 894 * @j_checkpoint_mutex. [j_checkpoint_mutex] 895 */ 896 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 897 898 /** 899 * @j_head: 900 * 901 * Journal head: identifies the first unused block in the journal. 902 * [j_state_lock] 903 */ 904 unsigned long j_head; 905 906 /** 907 * @j_tail: 908 * 909 * Journal tail: identifies the oldest still-used block in the journal. 910 * [j_state_lock] 911 */ 912 unsigned long j_tail; 913 914 /** 915 * @j_free: 916 * 917 * Journal free: how many free blocks are there in the journal? 918 * [j_state_lock] 919 */ 920 unsigned long j_free; 921 922 /** 923 * @j_first: 924 * 925 * The block number of the first usable block in the journal 926 * [j_state_lock]. 927 */ 928 unsigned long j_first; 929 930 /** 931 * @j_last: 932 * 933 * The block number one beyond the last usable block in the journal 934 * [j_state_lock]. 935 */ 936 unsigned long j_last; 937 938 /** 939 * @j_fc_first: 940 * 941 * The block number of the first fast commit block in the journal 942 * [j_state_lock]. 943 */ 944 unsigned long j_fc_first; 945 946 /** 947 * @j_fc_off: 948 * 949 * Number of fast commit blocks currently allocated. Accessed only 950 * during fast commit. Currently only process can do fast commit, so 951 * this field is not protected by any lock. 952 */ 953 unsigned long j_fc_off; 954 955 /** 956 * @j_fc_last: 957 * 958 * The block number one beyond the last fast commit block in the journal 959 * [j_state_lock]. 960 */ 961 unsigned long j_fc_last; 962 963 /** 964 * @j_dev: Device where we store the journal. 965 */ 966 struct block_device *j_dev; 967 968 /** 969 * @j_blocksize: Block size for the location where we store the journal. 970 */ 971 int j_blocksize; 972 973 /** 974 * @j_blk_offset: 975 * 976 * Starting block offset into the device where we store the journal. 977 */ 978 unsigned long long j_blk_offset; 979 980 /** 981 * @j_devname: Journal device name. 982 */ 983 char j_devname[BDEVNAME_SIZE+24]; 984 985 /** 986 * @j_fs_dev: 987 * 988 * Device which holds the client fs. For internal journal this will be 989 * equal to j_dev. 990 */ 991 struct block_device *j_fs_dev; 992 993 /** 994 * @j_total_len: Total maximum capacity of the journal region on disk. 995 */ 996 unsigned int j_total_len; 997 998 /** 999 * @j_reserved_credits: 1000 * 1001 * Number of buffers reserved from the running transaction. 1002 */ 1003 atomic_t j_reserved_credits; 1004 1005 /** 1006 * @j_list_lock: Protects the buffer lists and internal buffer state. 1007 */ 1008 spinlock_t j_list_lock; 1009 1010 /** 1011 * @j_inode: 1012 * 1013 * Optional inode where we store the journal. If present, all 1014 * journal block numbers are mapped into this inode via bmap(). 1015 */ 1016 struct inode *j_inode; 1017 1018 /** 1019 * @j_tail_sequence: 1020 * 1021 * Sequence number of the oldest transaction in the log [j_state_lock] 1022 */ 1023 tid_t j_tail_sequence; 1024 1025 /** 1026 * @j_transaction_sequence: 1027 * 1028 * Sequence number of the next transaction to grant [j_state_lock] 1029 */ 1030 tid_t j_transaction_sequence; 1031 1032 /** 1033 * @j_commit_sequence: 1034 * 1035 * Sequence number of the most recently committed transaction 1036 * [j_state_lock]. 1037 */ 1038 tid_t j_commit_sequence; 1039 1040 /** 1041 * @j_commit_request: 1042 * 1043 * Sequence number of the most recent transaction wanting commit 1044 * [j_state_lock] 1045 */ 1046 tid_t j_commit_request; 1047 1048 /** 1049 * @j_uuid: 1050 * 1051 * Journal uuid: identifies the object (filesystem, LVM volume etc) 1052 * backed by this journal. This will eventually be replaced by an array 1053 * of uuids, allowing us to index multiple devices within a single 1054 * journal and to perform atomic updates across them. 1055 */ 1056 __u8 j_uuid[16]; 1057 1058 /** 1059 * @j_task: Pointer to the current commit thread for this journal. 1060 */ 1061 struct task_struct *j_task; 1062 1063 /** 1064 * @j_max_transaction_buffers: 1065 * 1066 * Maximum number of metadata buffers to allow in a single compound 1067 * commit transaction. 1068 */ 1069 int j_max_transaction_buffers; 1070 1071 /** 1072 * @j_revoke_records_per_block: 1073 * 1074 * Number of revoke records that fit in one descriptor block. 1075 */ 1076 int j_revoke_records_per_block; 1077 1078 /** 1079 * @j_commit_interval: 1080 * 1081 * What is the maximum transaction lifetime before we begin a commit? 1082 */ 1083 unsigned long j_commit_interval; 1084 1085 /** 1086 * @j_commit_timer: The timer used to wakeup the commit thread. 1087 */ 1088 struct timer_list j_commit_timer; 1089 1090 /** 1091 * @j_revoke_lock: Protect the revoke table. 1092 */ 1093 spinlock_t j_revoke_lock; 1094 1095 /** 1096 * @j_revoke: 1097 * 1098 * The revoke table - maintains the list of revoked blocks in the 1099 * current transaction. 1100 */ 1101 struct jbd2_revoke_table_s *j_revoke; 1102 1103 /** 1104 * @j_revoke_table: Alternate revoke tables for j_revoke. 1105 */ 1106 struct jbd2_revoke_table_s *j_revoke_table[2]; 1107 1108 /** 1109 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction. 1110 */ 1111 struct buffer_head **j_wbuf; 1112 1113 /** 1114 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only 1115 * during a fast commit. Currently only process can do fast commit, so 1116 * this field is not protected by any lock. 1117 */ 1118 struct buffer_head **j_fc_wbuf; 1119 1120 /** 1121 * @j_wbufsize: 1122 * 1123 * Size of @j_wbuf array. 1124 */ 1125 int j_wbufsize; 1126 1127 /** 1128 * @j_fc_wbufsize: 1129 * 1130 * Size of @j_fc_wbuf array. 1131 */ 1132 int j_fc_wbufsize; 1133 1134 /** 1135 * @j_last_sync_writer: 1136 * 1137 * The pid of the last person to run a synchronous operation 1138 * through the journal. 1139 */ 1140 pid_t j_last_sync_writer; 1141 1142 /** 1143 * @j_average_commit_time: 1144 * 1145 * The average amount of time in nanoseconds it takes to commit a 1146 * transaction to disk. [j_state_lock] 1147 */ 1148 u64 j_average_commit_time; 1149 1150 /** 1151 * @j_min_batch_time: 1152 * 1153 * Minimum time that we should wait for additional filesystem operations 1154 * to get batched into a synchronous handle in microseconds. 1155 */ 1156 u32 j_min_batch_time; 1157 1158 /** 1159 * @j_max_batch_time: 1160 * 1161 * Maximum time that we should wait for additional filesystem operations 1162 * to get batched into a synchronous handle in microseconds. 1163 */ 1164 u32 j_max_batch_time; 1165 1166 /** 1167 * @j_commit_callback: 1168 * 1169 * This function is called when a transaction is closed. 1170 */ 1171 void (*j_commit_callback)(journal_t *, 1172 transaction_t *); 1173 1174 /** 1175 * @j_submit_inode_data_buffers: 1176 * 1177 * This function is called for all inodes associated with the 1178 * committing transaction marked with JI_WRITE_DATA flag 1179 * before we start to write out the transaction to the journal. 1180 */ 1181 int (*j_submit_inode_data_buffers) 1182 (struct jbd2_inode *); 1183 1184 /** 1185 * @j_finish_inode_data_buffers: 1186 * 1187 * This function is called for all inodes associated with the 1188 * committing transaction marked with JI_WAIT_DATA flag 1189 * after we have written the transaction to the journal 1190 * but before we write out the commit block. 1191 */ 1192 int (*j_finish_inode_data_buffers) 1193 (struct jbd2_inode *); 1194 1195 /* 1196 * Journal statistics 1197 */ 1198 1199 /** 1200 * @j_history_lock: Protect the transactions statistics history. 1201 */ 1202 spinlock_t j_history_lock; 1203 1204 /** 1205 * @j_proc_entry: procfs entry for the jbd statistics directory. 1206 */ 1207 struct proc_dir_entry *j_proc_entry; 1208 1209 /** 1210 * @j_stats: Overall statistics. 1211 */ 1212 struct transaction_stats_s j_stats; 1213 1214 /** 1215 * @j_failed_commit: Failed journal commit ID. 1216 */ 1217 unsigned int j_failed_commit; 1218 1219 /** 1220 * @j_private: 1221 * 1222 * An opaque pointer to fs-private information. ext3 puts its 1223 * superblock pointer here. 1224 */ 1225 void *j_private; 1226 1227 /** 1228 * @j_chksum_driver: 1229 * 1230 * Reference to checksum algorithm driver via cryptoapi. 1231 */ 1232 struct crypto_shash *j_chksum_driver; 1233 1234 /** 1235 * @j_csum_seed: 1236 * 1237 * Precomputed journal UUID checksum for seeding other checksums. 1238 */ 1239 __u32 j_csum_seed; 1240 1241 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1242 /** 1243 * @j_trans_commit_map: 1244 * 1245 * Lockdep entity to track transaction commit dependencies. Handles 1246 * hold this "lock" for read, when we wait for commit, we acquire the 1247 * "lock" for writing. This matches the properties of jbd2 journalling 1248 * where the running transaction has to wait for all handles to be 1249 * dropped to commit that transaction and also acquiring a handle may 1250 * require transaction commit to finish. 1251 */ 1252 struct lockdep_map j_trans_commit_map; 1253 #endif 1254 1255 /** 1256 * @j_fc_cleanup_callback: 1257 * 1258 * Clean-up after fast commit or full commit. JBD2 calls this function 1259 * after every commit operation. 1260 */ 1261 void (*j_fc_cleanup_callback)(struct journal_s *journal, int); 1262 1263 /** 1264 * @j_fc_replay_callback: 1265 * 1266 * File-system specific function that performs replay of a fast 1267 * commit. JBD2 calls this function for each fast commit block found in 1268 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE 1269 * to indicate that the block was processed correctly and more fast 1270 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP 1271 * indicates the end of replay (no more blocks remaining). A negative 1272 * return value indicates error. 1273 */ 1274 int (*j_fc_replay_callback)(struct journal_s *journal, 1275 struct buffer_head *bh, 1276 enum passtype pass, int off, 1277 tid_t expected_commit_id); 1278 }; 1279 1280 #define jbd2_might_wait_for_commit(j) \ 1281 do { \ 1282 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \ 1283 rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \ 1284 } while (0) 1285 1286 /* journal feature predicate functions */ 1287 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1288 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1289 { \ 1290 return ((j)->j_format_version >= 2 && \ 1291 ((j)->j_superblock->s_feature_compat & \ 1292 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1293 } \ 1294 static inline void jbd2_set_feature_##name(journal_t *j) \ 1295 { \ 1296 (j)->j_superblock->s_feature_compat |= \ 1297 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1298 } \ 1299 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1300 { \ 1301 (j)->j_superblock->s_feature_compat &= \ 1302 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1303 } 1304 1305 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1306 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1307 { \ 1308 return ((j)->j_format_version >= 2 && \ 1309 ((j)->j_superblock->s_feature_ro_compat & \ 1310 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1311 } \ 1312 static inline void jbd2_set_feature_##name(journal_t *j) \ 1313 { \ 1314 (j)->j_superblock->s_feature_ro_compat |= \ 1315 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1316 } \ 1317 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1318 { \ 1319 (j)->j_superblock->s_feature_ro_compat &= \ 1320 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1321 } 1322 1323 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1324 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1325 { \ 1326 return ((j)->j_format_version >= 2 && \ 1327 ((j)->j_superblock->s_feature_incompat & \ 1328 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1329 } \ 1330 static inline void jbd2_set_feature_##name(journal_t *j) \ 1331 { \ 1332 (j)->j_superblock->s_feature_incompat |= \ 1333 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1334 } \ 1335 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1336 { \ 1337 (j)->j_superblock->s_feature_incompat &= \ 1338 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1339 } 1340 1341 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1342 1343 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1344 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1345 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1346 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1347 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1348 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit, FAST_COMMIT) 1349 1350 /* 1351 * Journal flag definitions 1352 */ 1353 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1354 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1355 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1356 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1357 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1358 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1359 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1360 * data write error in ordered 1361 * mode */ 1362 #define JBD2_FAST_COMMIT_ONGOING 0x100 /* Fast commit is ongoing */ 1363 #define JBD2_FULL_COMMIT_ONGOING 0x200 /* Full commit is ongoing */ 1364 1365 /* 1366 * Function declarations for the journaling transaction and buffer 1367 * management 1368 */ 1369 1370 /* Filing buffers */ 1371 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1372 extern bool __jbd2_journal_refile_buffer(struct journal_head *); 1373 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1374 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1375 extern void __journal_free_buffer(struct journal_head *bh); 1376 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1377 extern void __journal_clean_data_list(transaction_t *transaction); 1378 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1379 { 1380 list_add_tail(&bh->b_assoc_buffers, head); 1381 } 1382 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1383 { 1384 list_del_init(&bh->b_assoc_buffers); 1385 } 1386 1387 /* Log buffer allocation */ 1388 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1389 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1390 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1391 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1392 unsigned long *block); 1393 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1394 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1395 1396 /* Commit management */ 1397 extern void jbd2_journal_commit_transaction(journal_t *); 1398 1399 /* Checkpoint list management */ 1400 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1401 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1402 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1403 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1404 1405 1406 /* 1407 * Triggers 1408 */ 1409 1410 struct jbd2_buffer_trigger_type { 1411 /* 1412 * Fired a the moment data to write to the journal are known to be 1413 * stable - so either at the moment b_frozen_data is created or just 1414 * before a buffer is written to the journal. mapped_data is a mapped 1415 * buffer that is the frozen data for commit. 1416 */ 1417 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1418 struct buffer_head *bh, void *mapped_data, 1419 size_t size); 1420 1421 /* 1422 * Fired during journal abort for dirty buffers that will not be 1423 * committed. 1424 */ 1425 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1426 struct buffer_head *bh); 1427 }; 1428 1429 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1430 void *mapped_data, 1431 struct jbd2_buffer_trigger_type *triggers); 1432 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1433 struct jbd2_buffer_trigger_type *triggers); 1434 1435 /* Buffer IO */ 1436 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1437 struct journal_head *jh_in, 1438 struct buffer_head **bh_out, 1439 sector_t blocknr); 1440 1441 /* Transaction locking */ 1442 extern void __wait_on_journal (journal_t *); 1443 1444 /* Transaction cache support */ 1445 extern void jbd2_journal_destroy_transaction_cache(void); 1446 extern int __init jbd2_journal_init_transaction_cache(void); 1447 extern void jbd2_journal_free_transaction(transaction_t *); 1448 1449 /* 1450 * Journal locking. 1451 * 1452 * We need to lock the journal during transaction state changes so that nobody 1453 * ever tries to take a handle on the running transaction while we are in the 1454 * middle of moving it to the commit phase. j_state_lock does this. 1455 * 1456 * Note that the locking is completely interrupt unsafe. We never touch 1457 * journal structures from interrupts. 1458 */ 1459 1460 static inline handle_t *journal_current_handle(void) 1461 { 1462 return current->journal_info; 1463 } 1464 1465 /* The journaling code user interface: 1466 * 1467 * Create and destroy handles 1468 * Register buffer modifications against the current transaction. 1469 */ 1470 1471 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1472 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1473 int revoke_records, gfp_t gfp_mask, 1474 unsigned int type, unsigned int line_no); 1475 extern int jbd2_journal_restart(handle_t *, int nblocks); 1476 extern int jbd2__journal_restart(handle_t *, int nblocks, 1477 int revoke_records, gfp_t gfp_mask); 1478 extern int jbd2_journal_start_reserved(handle_t *handle, 1479 unsigned int type, unsigned int line_no); 1480 extern void jbd2_journal_free_reserved(handle_t *handle); 1481 extern int jbd2_journal_extend(handle_t *handle, int nblocks, 1482 int revoke_records); 1483 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1484 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1485 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1486 void jbd2_journal_set_triggers(struct buffer_head *, 1487 struct jbd2_buffer_trigger_type *type); 1488 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1489 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1490 extern int jbd2_journal_invalidatepage(journal_t *, 1491 struct page *, unsigned int, unsigned int); 1492 extern int jbd2_journal_try_to_free_buffers(journal_t *journal, struct page *page); 1493 extern int jbd2_journal_stop(handle_t *); 1494 extern int jbd2_journal_flush (journal_t *); 1495 extern void jbd2_journal_lock_updates (journal_t *); 1496 extern void jbd2_journal_unlock_updates (journal_t *); 1497 1498 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1499 struct block_device *fs_dev, 1500 unsigned long long start, int len, int bsize); 1501 extern journal_t * jbd2_journal_init_inode (struct inode *); 1502 extern int jbd2_journal_update_format (journal_t *); 1503 extern int jbd2_journal_check_used_features 1504 (journal_t *, unsigned long, unsigned long, unsigned long); 1505 extern int jbd2_journal_check_available_features 1506 (journal_t *, unsigned long, unsigned long, unsigned long); 1507 extern int jbd2_journal_set_features 1508 (journal_t *, unsigned long, unsigned long, unsigned long); 1509 extern void jbd2_journal_clear_features 1510 (journal_t *, unsigned long, unsigned long, unsigned long); 1511 extern int jbd2_journal_load (journal_t *journal); 1512 extern int jbd2_journal_destroy (journal_t *); 1513 extern int jbd2_journal_recover (journal_t *journal); 1514 extern int jbd2_journal_wipe (journal_t *, int); 1515 extern int jbd2_journal_skip_recovery (journal_t *); 1516 extern void jbd2_journal_update_sb_errno(journal_t *); 1517 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1518 unsigned long, int); 1519 extern void jbd2_journal_abort (journal_t *, int); 1520 extern int jbd2_journal_errno (journal_t *); 1521 extern void jbd2_journal_ack_err (journal_t *); 1522 extern int jbd2_journal_clear_err (journal_t *); 1523 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1524 extern int jbd2_journal_force_commit(journal_t *); 1525 extern int jbd2_journal_force_commit_nested(journal_t *); 1526 extern int jbd2_journal_inode_ranged_write(handle_t *handle, 1527 struct jbd2_inode *inode, loff_t start_byte, 1528 loff_t length); 1529 extern int jbd2_journal_inode_ranged_wait(handle_t *handle, 1530 struct jbd2_inode *inode, loff_t start_byte, 1531 loff_t length); 1532 extern int jbd2_journal_submit_inode_data_buffers( 1533 struct jbd2_inode *jinode); 1534 extern int jbd2_journal_finish_inode_data_buffers( 1535 struct jbd2_inode *jinode); 1536 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1537 struct jbd2_inode *inode, loff_t new_size); 1538 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1539 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1540 1541 /* 1542 * journal_head management 1543 */ 1544 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1545 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1546 void jbd2_journal_put_journal_head(struct journal_head *jh); 1547 1548 /* 1549 * handle management 1550 */ 1551 extern struct kmem_cache *jbd2_handle_cache; 1552 1553 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1554 { 1555 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1556 } 1557 1558 static inline void jbd2_free_handle(handle_t *handle) 1559 { 1560 kmem_cache_free(jbd2_handle_cache, handle); 1561 } 1562 1563 /* 1564 * jbd2_inode management (optional, for those file systems that want to use 1565 * dynamically allocated jbd2_inode structures) 1566 */ 1567 extern struct kmem_cache *jbd2_inode_cache; 1568 1569 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1570 { 1571 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1572 } 1573 1574 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1575 { 1576 kmem_cache_free(jbd2_inode_cache, jinode); 1577 } 1578 1579 /* Primary revoke support */ 1580 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1581 extern int jbd2_journal_init_revoke(journal_t *, int); 1582 extern void jbd2_journal_destroy_revoke_record_cache(void); 1583 extern void jbd2_journal_destroy_revoke_table_cache(void); 1584 extern int __init jbd2_journal_init_revoke_record_cache(void); 1585 extern int __init jbd2_journal_init_revoke_table_cache(void); 1586 1587 extern void jbd2_journal_destroy_revoke(journal_t *); 1588 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1589 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1590 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1591 struct list_head *log_bufs); 1592 1593 /* Recovery revoke support */ 1594 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1595 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1596 extern void jbd2_journal_clear_revoke(journal_t *); 1597 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1598 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1599 1600 /* 1601 * The log thread user interface: 1602 * 1603 * Request space in the current transaction, and force transaction commit 1604 * transitions on demand. 1605 */ 1606 1607 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1608 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1609 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1610 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1611 int jbd2_transaction_committed(journal_t *journal, tid_t tid); 1612 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1613 int jbd2_log_do_checkpoint(journal_t *journal); 1614 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1615 1616 void __jbd2_log_wait_for_space(journal_t *journal); 1617 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1618 extern int jbd2_cleanup_journal_tail(journal_t *); 1619 1620 /* Fast commit related APIs */ 1621 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid); 1622 int jbd2_fc_end_commit(journal_t *journal); 1623 int jbd2_fc_end_commit_fallback(journal_t *journal); 1624 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out); 1625 int jbd2_submit_inode_data(struct jbd2_inode *jinode); 1626 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode); 1627 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks); 1628 int jbd2_fc_release_bufs(journal_t *journal); 1629 1630 static inline int jbd2_journal_get_max_txn_bufs(journal_t *journal) 1631 { 1632 return (journal->j_total_len - journal->j_fc_wbufsize) / 4; 1633 } 1634 1635 /* 1636 * is_journal_abort 1637 * 1638 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1639 * bit, when set, indicates that we have had a fatal error somewhere, 1640 * either inside the journaling layer or indicated to us by the client 1641 * (eg. ext3), and that we and should not commit any further 1642 * transactions. 1643 */ 1644 1645 static inline int is_journal_aborted(journal_t *journal) 1646 { 1647 return journal->j_flags & JBD2_ABORT; 1648 } 1649 1650 static inline int is_handle_aborted(handle_t *handle) 1651 { 1652 if (handle->h_aborted || !handle->h_transaction) 1653 return 1; 1654 return is_journal_aborted(handle->h_transaction->t_journal); 1655 } 1656 1657 static inline void jbd2_journal_abort_handle(handle_t *handle) 1658 { 1659 handle->h_aborted = 1; 1660 } 1661 1662 #endif /* __KERNEL__ */ 1663 1664 /* Comparison functions for transaction IDs: perform comparisons using 1665 * modulo arithmetic so that they work over sequence number wraps. */ 1666 1667 static inline int tid_gt(tid_t x, tid_t y) 1668 { 1669 int difference = (x - y); 1670 return (difference > 0); 1671 } 1672 1673 static inline int tid_geq(tid_t x, tid_t y) 1674 { 1675 int difference = (x - y); 1676 return (difference >= 0); 1677 } 1678 1679 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1680 extern size_t journal_tag_bytes(journal_t *journal); 1681 1682 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1683 { 1684 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1685 } 1686 1687 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1688 { 1689 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1690 journal->j_chksum_driver == NULL); 1691 1692 return journal->j_chksum_driver != NULL; 1693 } 1694 1695 static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb) 1696 { 1697 int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks); 1698 1699 return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS; 1700 } 1701 1702 /* 1703 * Return number of free blocks in the log. Must be called under j_state_lock. 1704 */ 1705 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1706 { 1707 /* Allow for rounding errors */ 1708 long free = journal->j_free - 32; 1709 1710 if (journal->j_committing_transaction) { 1711 free -= atomic_read(&journal-> 1712 j_committing_transaction->t_outstanding_credits); 1713 } 1714 return max_t(long, free, 0); 1715 } 1716 1717 /* 1718 * Definitions which augment the buffer_head layer 1719 */ 1720 1721 /* journaling buffer types */ 1722 #define BJ_None 0 /* Not journaled */ 1723 #define BJ_Metadata 1 /* Normal journaled metadata */ 1724 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1725 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1726 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1727 #define BJ_Types 5 1728 1729 extern int jbd_blocks_per_page(struct inode *inode); 1730 1731 /* JBD uses a CRC32 checksum */ 1732 #define JBD_MAX_CHECKSUM_SIZE 4 1733 1734 static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1735 const void *address, unsigned int length) 1736 { 1737 struct { 1738 struct shash_desc shash; 1739 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1740 } desc; 1741 int err; 1742 1743 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1744 JBD_MAX_CHECKSUM_SIZE); 1745 1746 desc.shash.tfm = journal->j_chksum_driver; 1747 *(u32 *)desc.ctx = crc; 1748 1749 err = crypto_shash_update(&desc.shash, address, length); 1750 BUG_ON(err); 1751 1752 return *(u32 *)desc.ctx; 1753 } 1754 1755 /* Return most recent uncommitted transaction */ 1756 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1757 { 1758 tid_t tid; 1759 1760 read_lock(&journal->j_state_lock); 1761 tid = journal->j_commit_request; 1762 if (journal->j_running_transaction) 1763 tid = journal->j_running_transaction->t_tid; 1764 read_unlock(&journal->j_state_lock); 1765 return tid; 1766 } 1767 1768 static inline int jbd2_handle_buffer_credits(handle_t *handle) 1769 { 1770 journal_t *journal; 1771 1772 if (!handle->h_reserved) 1773 journal = handle->h_transaction->t_journal; 1774 else 1775 journal = handle->h_journal; 1776 1777 return handle->h_total_credits - 1778 DIV_ROUND_UP(handle->h_revoke_credits_requested, 1779 journal->j_revoke_records_per_block); 1780 } 1781 1782 #ifdef __KERNEL__ 1783 1784 #define buffer_trace_init(bh) do {} while (0) 1785 #define print_buffer_fields(bh) do {} while (0) 1786 #define print_buffer_trace(bh) do {} while (0) 1787 #define BUFFER_TRACE(bh, info) do {} while (0) 1788 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1789 #define JBUFFER_TRACE(jh, info) do {} while (0) 1790 1791 #endif /* __KERNEL__ */ 1792 1793 #define EFSBADCRC EBADMSG /* Bad CRC detected */ 1794 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1795 1796 #endif /* _LINUX_JBD2_H */ 1797