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 72 #ifdef __KERNEL__ 73 74 /** 75 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 76 * 77 * All filesystem modifications made by the process go 78 * through this handle. Recursive operations (such as quota operations) 79 * are gathered into a single update. 80 * 81 * The buffer credits field is used to account for journaled buffers 82 * being modified by the running process. To ensure that there is 83 * enough log space for all outstanding operations, we need to limit the 84 * number of outstanding buffers possible at any time. When the 85 * operation completes, any buffer credits not used are credited back to 86 * the transaction, so that at all times we know how many buffers the 87 * outstanding updates on a transaction might possibly touch. 88 * 89 * This is an opaque datatype. 90 **/ 91 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ 92 93 94 /** 95 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 96 * 97 * journal_t is linked to from the fs superblock structure. 98 * 99 * We use the journal_t to keep track of all outstanding transaction 100 * activity on the filesystem, and to manage the state of the log 101 * writing process. 102 * 103 * This is an opaque datatype. 104 **/ 105 typedef struct journal_s journal_t; /* Journal control structure */ 106 #endif 107 108 /* 109 * Internal structures used by the logging mechanism: 110 */ 111 112 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 113 114 /* 115 * On-disk structures 116 */ 117 118 /* 119 * Descriptor block types: 120 */ 121 122 #define JBD2_DESCRIPTOR_BLOCK 1 123 #define JBD2_COMMIT_BLOCK 2 124 #define JBD2_SUPERBLOCK_V1 3 125 #define JBD2_SUPERBLOCK_V2 4 126 #define JBD2_REVOKE_BLOCK 5 127 128 /* 129 * Standard header for all descriptor blocks: 130 */ 131 typedef struct journal_header_s 132 { 133 __be32 h_magic; 134 __be32 h_blocktype; 135 __be32 h_sequence; 136 } journal_header_t; 137 138 /* 139 * Checksum types. 140 */ 141 #define JBD2_CRC32_CHKSUM 1 142 #define JBD2_MD5_CHKSUM 2 143 #define JBD2_SHA1_CHKSUM 3 144 #define JBD2_CRC32C_CHKSUM 4 145 146 #define JBD2_CRC32_CHKSUM_SIZE 4 147 148 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 149 /* 150 * Commit block header for storing transactional checksums: 151 * 152 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* 153 * fields are used to store a checksum of the descriptor and data blocks. 154 * 155 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum 156 * field is used to store crc32c(uuid+commit_block). Each journal metadata 157 * block gets its own checksum, and data block checksums are stored in 158 * journal_block_tag (in the descriptor). The other h_chksum* fields are 159 * not used. 160 * 161 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses 162 * journal_block_tag3_t to store a full 32-bit checksum. Everything else 163 * is the same as v2. 164 * 165 * Checksum v1, v2, and v3 are mutually exclusive features. 166 */ 167 struct commit_header { 168 __be32 h_magic; 169 __be32 h_blocktype; 170 __be32 h_sequence; 171 unsigned char h_chksum_type; 172 unsigned char h_chksum_size; 173 unsigned char h_padding[2]; 174 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 175 __be64 h_commit_sec; 176 __be32 h_commit_nsec; 177 }; 178 179 /* 180 * The block tag: used to describe a single buffer in the journal. 181 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 182 * raw struct shouldn't be used for pointer math or sizeof() - use 183 * journal_tag_bytes(journal) instead to compute this. 184 */ 185 typedef struct journal_block_tag3_s 186 { 187 __be32 t_blocknr; /* The on-disk block number */ 188 __be32 t_flags; /* See below */ 189 __be32 t_blocknr_high; /* most-significant high 32bits. */ 190 __be32 t_checksum; /* crc32c(uuid+seq+block) */ 191 } journal_block_tag3_t; 192 193 typedef struct journal_block_tag_s 194 { 195 __be32 t_blocknr; /* The on-disk block number */ 196 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ 197 __be16 t_flags; /* See below */ 198 __be32 t_blocknr_high; /* most-significant high 32bits. */ 199 } journal_block_tag_t; 200 201 /* Tail of descriptor or revoke block, for checksumming */ 202 struct jbd2_journal_block_tail { 203 __be32 t_checksum; /* crc32c(uuid+descr_block) */ 204 }; 205 206 /* 207 * The revoke descriptor: used on disk to describe a series of blocks to 208 * be revoked from the log 209 */ 210 typedef struct jbd2_journal_revoke_header_s 211 { 212 journal_header_t r_header; 213 __be32 r_count; /* Count of bytes used in the block */ 214 } jbd2_journal_revoke_header_t; 215 216 /* Definitions for the journal tag flags word: */ 217 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 218 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 219 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 220 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 221 222 223 /* 224 * The journal superblock. All fields are in big-endian byte order. 225 */ 226 typedef struct journal_superblock_s 227 { 228 /* 0x0000 */ 229 journal_header_t s_header; 230 231 /* 0x000C */ 232 /* Static information describing the journal */ 233 __be32 s_blocksize; /* journal device blocksize */ 234 __be32 s_maxlen; /* total blocks in journal file */ 235 __be32 s_first; /* first block of log information */ 236 237 /* 0x0018 */ 238 /* Dynamic information describing the current state of the log */ 239 __be32 s_sequence; /* first commit ID expected in log */ 240 __be32 s_start; /* blocknr of start of log */ 241 242 /* 0x0020 */ 243 /* Error value, as set by jbd2_journal_abort(). */ 244 __be32 s_errno; 245 246 /* 0x0024 */ 247 /* Remaining fields are only valid in a version-2 superblock */ 248 __be32 s_feature_compat; /* compatible feature set */ 249 __be32 s_feature_incompat; /* incompatible feature set */ 250 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 251 /* 0x0030 */ 252 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 253 254 /* 0x0040 */ 255 __be32 s_nr_users; /* Nr of filesystems sharing log */ 256 257 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 258 259 /* 0x0048 */ 260 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 261 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 262 263 /* 0x0050 */ 264 __u8 s_checksum_type; /* checksum type */ 265 __u8 s_padding2[3]; 266 /* 0x0054 */ 267 __be32 s_num_fc_blks; /* Number of fast commit blocks */ 268 /* 0x0058 */ 269 __u32 s_padding[41]; 270 __be32 s_checksum; /* crc32c(superblock) */ 271 272 /* 0x0100 */ 273 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 274 /* 0x0400 */ 275 } journal_superblock_t; 276 277 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */ 278 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 279 ((j)->j_format_version >= 2 && \ 280 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 281 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 282 ((j)->j_format_version >= 2 && \ 283 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 284 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 285 ((j)->j_format_version >= 2 && \ 286 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 287 288 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 289 290 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 291 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 292 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 293 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 294 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 295 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT 0x00000020 296 297 /* See "journal feature predicate functions" below */ 298 299 /* Features known to this kernel version: */ 300 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 301 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 302 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 303 JBD2_FEATURE_INCOMPAT_64BIT | \ 304 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 305 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 306 JBD2_FEATURE_INCOMPAT_CSUM_V3 | \ 307 JBD2_FEATURE_INCOMPAT_FAST_COMMIT) 308 309 #ifdef __KERNEL__ 310 311 #include <linux/fs.h> 312 #include <linux/sched.h> 313 314 enum jbd_state_bits { 315 BH_JBD /* Has an attached ext3 journal_head */ 316 = BH_PrivateStart, 317 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 318 BH_Freed, /* Has been freed (truncated) */ 319 BH_Revoked, /* Has been revoked from the log */ 320 BH_RevokeValid, /* Revoked flag is valid */ 321 BH_JBDDirty, /* Is dirty but journaled */ 322 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 323 BH_Shadow, /* IO on shadow buffer is running */ 324 BH_Verified, /* Metadata block has been verified ok */ 325 BH_JBDPrivateStart, /* First bit available for private use by FS */ 326 }; 327 328 BUFFER_FNS(JBD, jbd) 329 BUFFER_FNS(JWrite, jwrite) 330 BUFFER_FNS(JBDDirty, jbddirty) 331 TAS_BUFFER_FNS(JBDDirty, jbddirty) 332 BUFFER_FNS(Revoked, revoked) 333 TAS_BUFFER_FNS(Revoked, revoked) 334 BUFFER_FNS(RevokeValid, revokevalid) 335 TAS_BUFFER_FNS(RevokeValid, revokevalid) 336 BUFFER_FNS(Freed, freed) 337 BUFFER_FNS(Shadow, shadow) 338 BUFFER_FNS(Verified, verified) 339 340 static inline struct buffer_head *jh2bh(struct journal_head *jh) 341 { 342 return jh->b_bh; 343 } 344 345 static inline struct journal_head *bh2jh(struct buffer_head *bh) 346 { 347 return bh->b_private; 348 } 349 350 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 351 { 352 bit_spin_lock(BH_JournalHead, &bh->b_state); 353 } 354 355 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 356 { 357 bit_spin_unlock(BH_JournalHead, &bh->b_state); 358 } 359 360 #define J_ASSERT(assert) BUG_ON(!(assert)) 361 362 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 363 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 364 365 #if defined(JBD2_PARANOID_IOFAIL) 366 #define J_EXPECT(expr, why...) J_ASSERT(expr) 367 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 368 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 369 #else 370 #define __journal_expect(expr, why...) \ 371 ({ \ 372 int val = (expr); \ 373 if (!val) { \ 374 printk(KERN_ERR \ 375 "JBD2 unexpected failure: %s: %s;\n", \ 376 __func__, #expr); \ 377 printk(KERN_ERR why "\n"); \ 378 } \ 379 val; \ 380 }) 381 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 382 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 383 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 384 #endif 385 386 /* Flags in jbd_inode->i_flags */ 387 #define __JI_COMMIT_RUNNING 0 388 #define __JI_WRITE_DATA 1 389 #define __JI_WAIT_DATA 2 390 391 /* 392 * Commit of the inode data in progress. We use this flag to protect us from 393 * concurrent deletion of inode. We cannot use reference to inode for this 394 * since we cannot afford doing last iput() on behalf of kjournald 395 */ 396 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 397 /* Write allocated dirty buffers in this inode before commit */ 398 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA) 399 /* Wait for outstanding data writes for this inode before commit */ 400 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA) 401 402 /** 403 * struct jbd_inode - The jbd_inode type is the structure linking inodes in 404 * ordered mode present in a transaction so that we can sync them during commit. 405 */ 406 struct jbd2_inode { 407 /** 408 * @i_transaction: 409 * 410 * Which transaction does this inode belong to? Either the running 411 * transaction or the committing one. [j_list_lock] 412 */ 413 transaction_t *i_transaction; 414 415 /** 416 * @i_next_transaction: 417 * 418 * Pointer to the running transaction modifying inode's data in case 419 * there is already a committing transaction touching it. [j_list_lock] 420 */ 421 transaction_t *i_next_transaction; 422 423 /** 424 * @i_list: List of inodes in the i_transaction [j_list_lock] 425 */ 426 struct list_head i_list; 427 428 /** 429 * @i_vfs_inode: 430 * 431 * VFS inode this inode belongs to [constant for lifetime of structure] 432 */ 433 struct inode *i_vfs_inode; 434 435 /** 436 * @i_flags: Flags of inode [j_list_lock] 437 */ 438 unsigned long i_flags; 439 440 /** 441 * @i_dirty_start: 442 * 443 * Offset in bytes where the dirty range for this inode starts. 444 * [j_list_lock] 445 */ 446 loff_t i_dirty_start; 447 448 /** 449 * @i_dirty_end: 450 * 451 * Inclusive offset in bytes where the dirty range for this inode 452 * ends. [j_list_lock] 453 */ 454 loff_t i_dirty_end; 455 }; 456 457 struct jbd2_revoke_table_s; 458 459 /** 460 * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete 461 * type associated with handle_t. 462 * @h_transaction: Which compound transaction is this update a part of? 463 * @h_journal: Which journal handle belongs to - used iff h_reserved set. 464 * @h_rsv_handle: Handle reserved for finishing the logical operation. 465 * @h_total_credits: Number of remaining buffers we are allowed to add to 466 * journal. These are dirty buffers and revoke descriptor blocks. 467 * @h_revoke_credits: Number of remaining revoke records available for handle 468 * @h_ref: Reference count on this handle. 469 * @h_err: Field for caller's use to track errors through large fs operations. 470 * @h_sync: Flag for sync-on-close. 471 * @h_jdata: Flag to force data journaling. 472 * @h_reserved: Flag for handle for reserved credits. 473 * @h_aborted: Flag indicating fatal error on handle. 474 * @h_type: For handle statistics. 475 * @h_line_no: For handle statistics. 476 * @h_start_jiffies: Handle Start time. 477 * @h_requested_credits: Holds @h_total_credits after handle is started. 478 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started. 479 * @saved_alloc_context: Saved context while transaction is open. 480 **/ 481 482 /* Docbook can't yet cope with the bit fields, but will leave the documentation 483 * in so it can be fixed later. 484 */ 485 486 struct jbd2_journal_handle 487 { 488 union { 489 transaction_t *h_transaction; 490 /* Which journal handle belongs to - used iff h_reserved set */ 491 journal_t *h_journal; 492 }; 493 494 handle_t *h_rsv_handle; 495 int h_total_credits; 496 int h_revoke_credits; 497 int h_revoke_credits_requested; 498 int h_ref; 499 int h_err; 500 501 /* Flags [no locking] */ 502 unsigned int h_sync: 1; 503 unsigned int h_jdata: 1; 504 unsigned int h_reserved: 1; 505 unsigned int h_aborted: 1; 506 unsigned int h_type: 8; 507 unsigned int h_line_no: 16; 508 509 unsigned long h_start_jiffies; 510 unsigned int h_requested_credits; 511 512 unsigned int saved_alloc_context; 513 }; 514 515 516 /* 517 * Some stats for checkpoint phase 518 */ 519 struct transaction_chp_stats_s { 520 unsigned long cs_chp_time; 521 __u32 cs_forced_to_close; 522 __u32 cs_written; 523 __u32 cs_dropped; 524 }; 525 526 /* The transaction_t type is the guts of the journaling mechanism. It 527 * tracks a compound transaction through its various states: 528 * 529 * RUNNING: accepting new updates 530 * LOCKED: Updates still running but we don't accept new ones 531 * RUNDOWN: Updates are tidying up but have finished requesting 532 * new buffers to modify (state not used for now) 533 * FLUSH: All updates complete, but we are still writing to disk 534 * COMMIT: All data on disk, writing commit record 535 * FINISHED: We still have to keep the transaction for checkpointing. 536 * 537 * The transaction keeps track of all of the buffers modified by a 538 * running transaction, and all of the buffers committed but not yet 539 * flushed to home for finished transactions. 540 */ 541 542 /* 543 * Lock ranking: 544 * 545 * j_list_lock 546 * ->jbd_lock_bh_journal_head() (This is "innermost") 547 * 548 * j_state_lock 549 * ->b_state_lock 550 * 551 * b_state_lock 552 * ->j_list_lock 553 * 554 * j_state_lock 555 * ->t_handle_lock 556 * 557 * j_state_lock 558 * ->j_list_lock (journal_unmap_buffer) 559 * 560 */ 561 562 struct transaction_s 563 { 564 /* Pointer to the journal for this transaction. [no locking] */ 565 journal_t *t_journal; 566 567 /* Sequence number for this transaction [no locking] */ 568 tid_t t_tid; 569 570 /* 571 * Transaction's current state 572 * [no locking - only kjournald2 alters this] 573 * [j_list_lock] guards transition of a transaction into T_FINISHED 574 * state and subsequent call of __jbd2_journal_drop_transaction() 575 * FIXME: needs barriers 576 * KLUDGE: [use j_state_lock] 577 */ 578 enum { 579 T_RUNNING, 580 T_LOCKED, 581 T_SWITCH, 582 T_FLUSH, 583 T_COMMIT, 584 T_COMMIT_DFLUSH, 585 T_COMMIT_JFLUSH, 586 T_COMMIT_CALLBACK, 587 T_FINISHED 588 } t_state; 589 590 /* 591 * Where in the log does this transaction's commit start? [no locking] 592 */ 593 unsigned long t_log_start; 594 595 /* Number of buffers on the t_buffers list [j_list_lock] */ 596 int t_nr_buffers; 597 598 /* 599 * Doubly-linked circular list of all buffers reserved but not yet 600 * modified by this transaction [j_list_lock] 601 */ 602 struct journal_head *t_reserved_list; 603 604 /* 605 * Doubly-linked circular list of all metadata buffers owned by this 606 * transaction [j_list_lock] 607 */ 608 struct journal_head *t_buffers; 609 610 /* 611 * Doubly-linked circular list of all forget buffers (superseded 612 * buffers which we can un-checkpoint once this transaction commits) 613 * [j_list_lock] 614 */ 615 struct journal_head *t_forget; 616 617 /* 618 * Doubly-linked circular list of all buffers still to be flushed before 619 * this transaction can be checkpointed. [j_list_lock] 620 */ 621 struct journal_head *t_checkpoint_list; 622 623 /* 624 * Doubly-linked circular list of all buffers submitted for IO while 625 * checkpointing. [j_list_lock] 626 */ 627 struct journal_head *t_checkpoint_io_list; 628 629 /* 630 * Doubly-linked circular list of metadata buffers being shadowed by log 631 * IO. The IO buffers on the iobuf list and the shadow buffers on this 632 * list match each other one for one at all times. [j_list_lock] 633 */ 634 struct journal_head *t_shadow_list; 635 636 /* 637 * List of inodes associated with the transaction; e.g., ext4 uses 638 * this to track inodes in data=ordered and data=journal mode that 639 * need special handling on transaction commit; also used by ocfs2. 640 * [j_list_lock] 641 */ 642 struct list_head t_inode_list; 643 644 /* 645 * Protects info related to handles 646 */ 647 spinlock_t t_handle_lock; 648 649 /* 650 * Longest time some handle had to wait for running transaction 651 */ 652 unsigned long t_max_wait; 653 654 /* 655 * When transaction started 656 */ 657 unsigned long t_start; 658 659 /* 660 * When commit was requested 661 */ 662 unsigned long t_requested; 663 664 /* 665 * Checkpointing stats [j_checkpoint_sem] 666 */ 667 struct transaction_chp_stats_s t_chp_stats; 668 669 /* 670 * Number of outstanding updates running on this transaction 671 * [none] 672 */ 673 atomic_t t_updates; 674 675 /* 676 * Number of blocks reserved for this transaction in the journal. 677 * This is including all credits reserved when starting transaction 678 * handles as well as all journal descriptor blocks needed for this 679 * transaction. [none] 680 */ 681 atomic_t t_outstanding_credits; 682 683 /* 684 * Number of revoke records for this transaction added by already 685 * stopped handles. [none] 686 */ 687 atomic_t t_outstanding_revokes; 688 689 /* 690 * How many handles used this transaction? [none] 691 */ 692 atomic_t t_handle_count; 693 694 /* 695 * Forward and backward links for the circular list of all transactions 696 * awaiting checkpoint. [j_list_lock] 697 */ 698 transaction_t *t_cpnext, *t_cpprev; 699 700 /* 701 * When will the transaction expire (become due for commit), in jiffies? 702 * [no locking] 703 */ 704 unsigned long t_expires; 705 706 /* 707 * When this transaction started, in nanoseconds [no locking] 708 */ 709 ktime_t t_start_time; 710 711 /* 712 * This transaction is being forced and some process is 713 * waiting for it to finish. 714 */ 715 unsigned int t_synchronous_commit:1; 716 717 /* Disk flush needs to be sent to fs partition [no locking] */ 718 int t_need_data_flush; 719 720 /* 721 * For use by the filesystem to store fs-specific data 722 * structures associated with the transaction 723 */ 724 struct list_head t_private_list; 725 }; 726 727 struct transaction_run_stats_s { 728 unsigned long rs_wait; 729 unsigned long rs_request_delay; 730 unsigned long rs_running; 731 unsigned long rs_locked; 732 unsigned long rs_flushing; 733 unsigned long rs_logging; 734 735 __u32 rs_handle_count; 736 __u32 rs_blocks; 737 __u32 rs_blocks_logged; 738 }; 739 740 struct transaction_stats_s { 741 unsigned long ts_tid; 742 unsigned long ts_requested; 743 struct transaction_run_stats_s run; 744 }; 745 746 static inline unsigned long 747 jbd2_time_diff(unsigned long start, unsigned long end) 748 { 749 if (end >= start) 750 return end - start; 751 752 return end + (MAX_JIFFY_OFFSET - start); 753 } 754 755 #define JBD2_NR_BATCH 64 756 757 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY}; 758 759 #define JBD2_FC_REPLAY_STOP 0 760 #define JBD2_FC_REPLAY_CONTINUE 1 761 762 /** 763 * struct journal_s - The journal_s type is the concrete type associated with 764 * journal_t. 765 */ 766 struct journal_s 767 { 768 /** 769 * @j_flags: General journaling state flags [j_state_lock] 770 */ 771 unsigned long j_flags; 772 773 /** 774 * @j_errno: 775 * 776 * Is there an outstanding uncleared error on the journal (from a prior 777 * abort)? [j_state_lock] 778 */ 779 int j_errno; 780 781 /** 782 * @j_abort_mutex: Lock the whole aborting procedure. 783 */ 784 struct mutex j_abort_mutex; 785 786 /** 787 * @j_sb_buffer: The first part of the superblock buffer. 788 */ 789 struct buffer_head *j_sb_buffer; 790 791 /** 792 * @j_superblock: The second part of the superblock buffer. 793 */ 794 journal_superblock_t *j_superblock; 795 796 /** 797 * @j_format_version: Version of the superblock format. 798 */ 799 int j_format_version; 800 801 /** 802 * @j_state_lock: Protect the various scalars in the journal. 803 */ 804 rwlock_t j_state_lock; 805 806 /** 807 * @j_barrier_count: 808 * 809 * Number of processes waiting to create a barrier lock [j_state_lock] 810 */ 811 int j_barrier_count; 812 813 /** 814 * @j_barrier: The barrier lock itself. 815 */ 816 struct mutex j_barrier; 817 818 /** 819 * @j_running_transaction: 820 * 821 * Transactions: The current running transaction... 822 * [j_state_lock] [caller holding open handle] 823 */ 824 transaction_t *j_running_transaction; 825 826 /** 827 * @j_committing_transaction: 828 * 829 * the transaction we are pushing to disk 830 * [j_state_lock] [caller holding open handle] 831 */ 832 transaction_t *j_committing_transaction; 833 834 /** 835 * @j_checkpoint_transactions: 836 * 837 * ... and a linked circular list of all transactions waiting for 838 * checkpointing. [j_list_lock] 839 */ 840 transaction_t *j_checkpoint_transactions; 841 842 /** 843 * @j_wait_transaction_locked: 844 * 845 * Wait queue for waiting for a locked transaction to start committing, 846 * or for a barrier lock to be released. 847 */ 848 wait_queue_head_t j_wait_transaction_locked; 849 850 /** 851 * @j_wait_done_commit: Wait queue for waiting for commit to complete. 852 */ 853 wait_queue_head_t j_wait_done_commit; 854 855 /** 856 * @j_wait_commit: Wait queue to trigger commit. 857 */ 858 wait_queue_head_t j_wait_commit; 859 860 /** 861 * @j_wait_updates: Wait queue to wait for updates to complete. 862 */ 863 wait_queue_head_t j_wait_updates; 864 865 /** 866 * @j_wait_reserved: 867 * 868 * Wait queue to wait for reserved buffer credits to drop. 869 */ 870 wait_queue_head_t j_wait_reserved; 871 872 /** 873 * @j_fc_wait: 874 * 875 * Wait queue to wait for completion of async fast commits. 876 */ 877 wait_queue_head_t j_fc_wait; 878 879 /** 880 * @j_checkpoint_mutex: 881 * 882 * Semaphore for locking against concurrent checkpoints. 883 */ 884 struct mutex j_checkpoint_mutex; 885 886 /** 887 * @j_chkpt_bhs: 888 * 889 * List of buffer heads used by the checkpoint routine. This 890 * was moved from jbd2_log_do_checkpoint() to reduce stack 891 * usage. Access to this array is controlled by the 892 * @j_checkpoint_mutex. [j_checkpoint_mutex] 893 */ 894 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 895 896 /** 897 * @j_head: 898 * 899 * Journal head: identifies the first unused block in the journal. 900 * [j_state_lock] 901 */ 902 unsigned long j_head; 903 904 /** 905 * @j_tail: 906 * 907 * Journal tail: identifies the oldest still-used block in the journal. 908 * [j_state_lock] 909 */ 910 unsigned long j_tail; 911 912 /** 913 * @j_free: 914 * 915 * Journal free: how many free blocks are there in the journal? 916 * [j_state_lock] 917 */ 918 unsigned long j_free; 919 920 /** 921 * @j_first: 922 * 923 * The block number of the first usable block in the journal 924 * [j_state_lock]. 925 */ 926 unsigned long j_first; 927 928 /** 929 * @j_last: 930 * 931 * The block number one beyond the last usable block in the journal 932 * [j_state_lock]. 933 */ 934 unsigned long j_last; 935 936 /** 937 * @j_fc_first: 938 * 939 * The block number of the first fast commit block in the journal 940 * [j_state_lock]. 941 */ 942 unsigned long j_fc_first; 943 944 /** 945 * @j_fc_off: 946 * 947 * Number of fast commit blocks currently allocated. 948 * [j_state_lock]. 949 */ 950 unsigned long j_fc_off; 951 952 /** 953 * @j_fc_last: 954 * 955 * The block number one beyond the last fast commit block in the journal 956 * [j_state_lock]. 957 */ 958 unsigned long j_fc_last; 959 960 /** 961 * @j_dev: Device where we store the journal. 962 */ 963 struct block_device *j_dev; 964 965 /** 966 * @j_blocksize: Block size for the location where we store the journal. 967 */ 968 int j_blocksize; 969 970 /** 971 * @j_blk_offset: 972 * 973 * Starting block offset into the device where we store the journal. 974 */ 975 unsigned long long j_blk_offset; 976 977 /** 978 * @j_devname: Journal device name. 979 */ 980 char j_devname[BDEVNAME_SIZE+24]; 981 982 /** 983 * @j_fs_dev: 984 * 985 * Device which holds the client fs. For internal journal this will be 986 * equal to j_dev. 987 */ 988 struct block_device *j_fs_dev; 989 990 /** 991 * @j_maxlen: Total maximum capacity of the journal region on disk. 992 */ 993 unsigned int j_maxlen; 994 995 /** 996 * @j_reserved_credits: 997 * 998 * Number of buffers reserved from the running transaction. 999 */ 1000 atomic_t j_reserved_credits; 1001 1002 /** 1003 * @j_list_lock: Protects the buffer lists and internal buffer state. 1004 */ 1005 spinlock_t j_list_lock; 1006 1007 /** 1008 * @j_inode: 1009 * 1010 * Optional inode where we store the journal. If present, all 1011 * journal block numbers are mapped into this inode via bmap(). 1012 */ 1013 struct inode *j_inode; 1014 1015 /** 1016 * @j_tail_sequence: 1017 * 1018 * Sequence number of the oldest transaction in the log [j_state_lock] 1019 */ 1020 tid_t j_tail_sequence; 1021 1022 /** 1023 * @j_transaction_sequence: 1024 * 1025 * Sequence number of the next transaction to grant [j_state_lock] 1026 */ 1027 tid_t j_transaction_sequence; 1028 1029 /** 1030 * @j_commit_sequence: 1031 * 1032 * Sequence number of the most recently committed transaction 1033 * [j_state_lock]. 1034 */ 1035 tid_t j_commit_sequence; 1036 1037 /** 1038 * @j_commit_request: 1039 * 1040 * Sequence number of the most recent transaction wanting commit 1041 * [j_state_lock] 1042 */ 1043 tid_t j_commit_request; 1044 1045 /** 1046 * @j_uuid: 1047 * 1048 * Journal uuid: identifies the object (filesystem, LVM volume etc) 1049 * backed by this journal. This will eventually be replaced by an array 1050 * of uuids, allowing us to index multiple devices within a single 1051 * journal and to perform atomic updates across them. 1052 */ 1053 __u8 j_uuid[16]; 1054 1055 /** 1056 * @j_task: Pointer to the current commit thread for this journal. 1057 */ 1058 struct task_struct *j_task; 1059 1060 /** 1061 * @j_max_transaction_buffers: 1062 * 1063 * Maximum number of metadata buffers to allow in a single compound 1064 * commit transaction. 1065 */ 1066 int j_max_transaction_buffers; 1067 1068 /** 1069 * @j_revoke_records_per_block: 1070 * 1071 * Number of revoke records that fit in one descriptor block. 1072 */ 1073 int j_revoke_records_per_block; 1074 1075 /** 1076 * @j_commit_interval: 1077 * 1078 * What is the maximum transaction lifetime before we begin a commit? 1079 */ 1080 unsigned long j_commit_interval; 1081 1082 /** 1083 * @j_commit_timer: The timer used to wakeup the commit thread. 1084 */ 1085 struct timer_list j_commit_timer; 1086 1087 /** 1088 * @j_revoke_lock: Protect the revoke table. 1089 */ 1090 spinlock_t j_revoke_lock; 1091 1092 /** 1093 * @j_revoke: 1094 * 1095 * The revoke table - maintains the list of revoked blocks in the 1096 * current transaction. 1097 */ 1098 struct jbd2_revoke_table_s *j_revoke; 1099 1100 /** 1101 * @j_revoke_table: Alternate revoke tables for j_revoke. 1102 */ 1103 struct jbd2_revoke_table_s *j_revoke_table[2]; 1104 1105 /** 1106 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction. 1107 */ 1108 struct buffer_head **j_wbuf; 1109 1110 /** 1111 * @j_fc_wbuf: Array of fast commit bhs for 1112 * jbd2_journal_commit_transaction. 1113 */ 1114 struct buffer_head **j_fc_wbuf; 1115 1116 /** 1117 * @j_wbufsize: 1118 * 1119 * Size of @j_wbuf array. 1120 */ 1121 int j_wbufsize; 1122 1123 /** 1124 * @j_fc_wbufsize: 1125 * 1126 * Size of @j_fc_wbuf array. 1127 */ 1128 int j_fc_wbufsize; 1129 1130 /** 1131 * @j_last_sync_writer: 1132 * 1133 * The pid of the last person to run a synchronous operation 1134 * through the journal. 1135 */ 1136 pid_t j_last_sync_writer; 1137 1138 /** 1139 * @j_average_commit_time: 1140 * 1141 * The average amount of time in nanoseconds it takes to commit a 1142 * transaction to disk. [j_state_lock] 1143 */ 1144 u64 j_average_commit_time; 1145 1146 /** 1147 * @j_min_batch_time: 1148 * 1149 * Minimum time that we should wait for additional filesystem operations 1150 * to get batched into a synchronous handle in microseconds. 1151 */ 1152 u32 j_min_batch_time; 1153 1154 /** 1155 * @j_max_batch_time: 1156 * 1157 * Maximum time that we should wait for additional filesystem operations 1158 * to get batched into a synchronous handle in microseconds. 1159 */ 1160 u32 j_max_batch_time; 1161 1162 /** 1163 * @j_commit_callback: 1164 * 1165 * This function is called when a transaction is closed. 1166 */ 1167 void (*j_commit_callback)(journal_t *, 1168 transaction_t *); 1169 1170 /** 1171 * @j_submit_inode_data_buffers: 1172 * 1173 * This function is called for all inodes associated with the 1174 * committing transaction marked with JI_WRITE_DATA flag 1175 * before we start to write out the transaction to the journal. 1176 */ 1177 int (*j_submit_inode_data_buffers) 1178 (struct jbd2_inode *); 1179 1180 /** 1181 * @j_finish_inode_data_buffers: 1182 * 1183 * This function is called for all inodes associated with the 1184 * committing transaction marked with JI_WAIT_DATA flag 1185 * after we have written the transaction to the journal 1186 * but before we write out the commit block. 1187 */ 1188 int (*j_finish_inode_data_buffers) 1189 (struct jbd2_inode *); 1190 1191 /* 1192 * Journal statistics 1193 */ 1194 1195 /** 1196 * @j_history_lock: Protect the transactions statistics history. 1197 */ 1198 spinlock_t j_history_lock; 1199 1200 /** 1201 * @j_proc_entry: procfs entry for the jbd statistics directory. 1202 */ 1203 struct proc_dir_entry *j_proc_entry; 1204 1205 /** 1206 * @j_stats: Overall statistics. 1207 */ 1208 struct transaction_stats_s j_stats; 1209 1210 /** 1211 * @j_failed_commit: Failed journal commit ID. 1212 */ 1213 unsigned int j_failed_commit; 1214 1215 /** 1216 * @j_private: 1217 * 1218 * An opaque pointer to fs-private information. ext3 puts its 1219 * superblock pointer here. 1220 */ 1221 void *j_private; 1222 1223 /** 1224 * @j_chksum_driver: 1225 * 1226 * Reference to checksum algorithm driver via cryptoapi. 1227 */ 1228 struct crypto_shash *j_chksum_driver; 1229 1230 /** 1231 * @j_csum_seed: 1232 * 1233 * Precomputed journal UUID checksum for seeding other checksums. 1234 */ 1235 __u32 j_csum_seed; 1236 1237 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1238 /** 1239 * @j_trans_commit_map: 1240 * 1241 * Lockdep entity to track transaction commit dependencies. Handles 1242 * hold this "lock" for read, when we wait for commit, we acquire the 1243 * "lock" for writing. This matches the properties of jbd2 journalling 1244 * where the running transaction has to wait for all handles to be 1245 * dropped to commit that transaction and also acquiring a handle may 1246 * require transaction commit to finish. 1247 */ 1248 struct lockdep_map j_trans_commit_map; 1249 #endif 1250 1251 /** 1252 * @j_fc_cleanup_callback: 1253 * 1254 * Clean-up after fast commit or full commit. JBD2 calls this function 1255 * after every commit operation. 1256 */ 1257 void (*j_fc_cleanup_callback)(struct journal_s *journal, int); 1258 1259 /** 1260 * @j_fc_replay_callback: 1261 * 1262 * File-system specific function that performs replay of a fast 1263 * commit. JBD2 calls this function for each fast commit block found in 1264 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE 1265 * to indicate that the block was processed correctly and more fast 1266 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP 1267 * indicates the end of replay (no more blocks remaining). A negative 1268 * return value indicates error. 1269 */ 1270 int (*j_fc_replay_callback)(struct journal_s *journal, 1271 struct buffer_head *bh, 1272 enum passtype pass, int off, 1273 tid_t expected_commit_id); 1274 }; 1275 1276 #define jbd2_might_wait_for_commit(j) \ 1277 do { \ 1278 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \ 1279 rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \ 1280 } while (0) 1281 1282 /* journal feature predicate functions */ 1283 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1284 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1285 { \ 1286 return ((j)->j_format_version >= 2 && \ 1287 ((j)->j_superblock->s_feature_compat & \ 1288 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1289 } \ 1290 static inline void jbd2_set_feature_##name(journal_t *j) \ 1291 { \ 1292 (j)->j_superblock->s_feature_compat |= \ 1293 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1294 } \ 1295 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1296 { \ 1297 (j)->j_superblock->s_feature_compat &= \ 1298 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1299 } 1300 1301 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1302 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1303 { \ 1304 return ((j)->j_format_version >= 2 && \ 1305 ((j)->j_superblock->s_feature_ro_compat & \ 1306 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1307 } \ 1308 static inline void jbd2_set_feature_##name(journal_t *j) \ 1309 { \ 1310 (j)->j_superblock->s_feature_ro_compat |= \ 1311 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1312 } \ 1313 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1314 { \ 1315 (j)->j_superblock->s_feature_ro_compat &= \ 1316 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1317 } 1318 1319 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1320 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1321 { \ 1322 return ((j)->j_format_version >= 2 && \ 1323 ((j)->j_superblock->s_feature_incompat & \ 1324 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1325 } \ 1326 static inline void jbd2_set_feature_##name(journal_t *j) \ 1327 { \ 1328 (j)->j_superblock->s_feature_incompat |= \ 1329 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1330 } \ 1331 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1332 { \ 1333 (j)->j_superblock->s_feature_incompat &= \ 1334 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1335 } 1336 1337 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1338 1339 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1340 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1341 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1342 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1343 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1344 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit, FAST_COMMIT) 1345 1346 /* 1347 * Journal flag definitions 1348 */ 1349 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1350 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1351 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1352 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1353 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1354 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1355 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1356 * data write error in ordered 1357 * mode */ 1358 #define JBD2_FAST_COMMIT_ONGOING 0x100 /* Fast commit is ongoing */ 1359 #define JBD2_FULL_COMMIT_ONGOING 0x200 /* Full commit is ongoing */ 1360 1361 /* 1362 * Function declarations for the journaling transaction and buffer 1363 * management 1364 */ 1365 1366 /* Filing buffers */ 1367 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1368 extern bool __jbd2_journal_refile_buffer(struct journal_head *); 1369 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1370 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1371 extern void __journal_free_buffer(struct journal_head *bh); 1372 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1373 extern void __journal_clean_data_list(transaction_t *transaction); 1374 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1375 { 1376 list_add_tail(&bh->b_assoc_buffers, head); 1377 } 1378 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1379 { 1380 list_del_init(&bh->b_assoc_buffers); 1381 } 1382 1383 /* Log buffer allocation */ 1384 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1385 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1386 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1387 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1388 unsigned long *block); 1389 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1390 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1391 1392 /* Commit management */ 1393 extern void jbd2_journal_commit_transaction(journal_t *); 1394 1395 /* Checkpoint list management */ 1396 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1397 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1398 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1399 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1400 1401 1402 /* 1403 * Triggers 1404 */ 1405 1406 struct jbd2_buffer_trigger_type { 1407 /* 1408 * Fired a the moment data to write to the journal are known to be 1409 * stable - so either at the moment b_frozen_data is created or just 1410 * before a buffer is written to the journal. mapped_data is a mapped 1411 * buffer that is the frozen data for commit. 1412 */ 1413 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1414 struct buffer_head *bh, void *mapped_data, 1415 size_t size); 1416 1417 /* 1418 * Fired during journal abort for dirty buffers that will not be 1419 * committed. 1420 */ 1421 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1422 struct buffer_head *bh); 1423 }; 1424 1425 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1426 void *mapped_data, 1427 struct jbd2_buffer_trigger_type *triggers); 1428 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1429 struct jbd2_buffer_trigger_type *triggers); 1430 1431 /* Buffer IO */ 1432 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1433 struct journal_head *jh_in, 1434 struct buffer_head **bh_out, 1435 sector_t blocknr); 1436 1437 /* Transaction locking */ 1438 extern void __wait_on_journal (journal_t *); 1439 1440 /* Transaction cache support */ 1441 extern void jbd2_journal_destroy_transaction_cache(void); 1442 extern int __init jbd2_journal_init_transaction_cache(void); 1443 extern void jbd2_journal_free_transaction(transaction_t *); 1444 1445 /* 1446 * Journal locking. 1447 * 1448 * We need to lock the journal during transaction state changes so that nobody 1449 * ever tries to take a handle on the running transaction while we are in the 1450 * middle of moving it to the commit phase. j_state_lock does this. 1451 * 1452 * Note that the locking is completely interrupt unsafe. We never touch 1453 * journal structures from interrupts. 1454 */ 1455 1456 static inline handle_t *journal_current_handle(void) 1457 { 1458 return current->journal_info; 1459 } 1460 1461 /* The journaling code user interface: 1462 * 1463 * Create and destroy handles 1464 * Register buffer modifications against the current transaction. 1465 */ 1466 1467 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1468 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1469 int revoke_records, gfp_t gfp_mask, 1470 unsigned int type, unsigned int line_no); 1471 extern int jbd2_journal_restart(handle_t *, int nblocks); 1472 extern int jbd2__journal_restart(handle_t *, int nblocks, 1473 int revoke_records, gfp_t gfp_mask); 1474 extern int jbd2_journal_start_reserved(handle_t *handle, 1475 unsigned int type, unsigned int line_no); 1476 extern void jbd2_journal_free_reserved(handle_t *handle); 1477 extern int jbd2_journal_extend(handle_t *handle, int nblocks, 1478 int revoke_records); 1479 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1480 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1481 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1482 void jbd2_journal_set_triggers(struct buffer_head *, 1483 struct jbd2_buffer_trigger_type *type); 1484 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1485 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1486 extern int jbd2_journal_invalidatepage(journal_t *, 1487 struct page *, unsigned int, unsigned int); 1488 extern int jbd2_journal_try_to_free_buffers(journal_t *journal, struct page *page); 1489 extern int jbd2_journal_stop(handle_t *); 1490 extern int jbd2_journal_flush (journal_t *); 1491 extern void jbd2_journal_lock_updates (journal_t *); 1492 extern void jbd2_journal_unlock_updates (journal_t *); 1493 1494 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1495 struct block_device *fs_dev, 1496 unsigned long long start, int len, int bsize); 1497 extern journal_t * jbd2_journal_init_inode (struct inode *); 1498 extern int jbd2_journal_update_format (journal_t *); 1499 extern int jbd2_journal_check_used_features 1500 (journal_t *, unsigned long, unsigned long, unsigned long); 1501 extern int jbd2_journal_check_available_features 1502 (journal_t *, unsigned long, unsigned long, unsigned long); 1503 extern int jbd2_journal_set_features 1504 (journal_t *, unsigned long, unsigned long, unsigned long); 1505 extern void jbd2_journal_clear_features 1506 (journal_t *, unsigned long, unsigned long, unsigned long); 1507 extern int jbd2_journal_load (journal_t *journal); 1508 extern int jbd2_journal_destroy (journal_t *); 1509 extern int jbd2_journal_recover (journal_t *journal); 1510 extern int jbd2_journal_wipe (journal_t *, int); 1511 extern int jbd2_journal_skip_recovery (journal_t *); 1512 extern void jbd2_journal_update_sb_errno(journal_t *); 1513 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1514 unsigned long, int); 1515 extern void jbd2_journal_abort (journal_t *, int); 1516 extern int jbd2_journal_errno (journal_t *); 1517 extern void jbd2_journal_ack_err (journal_t *); 1518 extern int jbd2_journal_clear_err (journal_t *); 1519 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1520 extern int jbd2_journal_force_commit(journal_t *); 1521 extern int jbd2_journal_force_commit_nested(journal_t *); 1522 extern int jbd2_journal_inode_ranged_write(handle_t *handle, 1523 struct jbd2_inode *inode, loff_t start_byte, 1524 loff_t length); 1525 extern int jbd2_journal_inode_ranged_wait(handle_t *handle, 1526 struct jbd2_inode *inode, loff_t start_byte, 1527 loff_t length); 1528 extern int jbd2_journal_submit_inode_data_buffers( 1529 struct jbd2_inode *jinode); 1530 extern int jbd2_journal_finish_inode_data_buffers( 1531 struct jbd2_inode *jinode); 1532 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1533 struct jbd2_inode *inode, loff_t new_size); 1534 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1535 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1536 1537 /* 1538 * journal_head management 1539 */ 1540 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1541 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1542 void jbd2_journal_put_journal_head(struct journal_head *jh); 1543 1544 /* 1545 * handle management 1546 */ 1547 extern struct kmem_cache *jbd2_handle_cache; 1548 1549 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1550 { 1551 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1552 } 1553 1554 static inline void jbd2_free_handle(handle_t *handle) 1555 { 1556 kmem_cache_free(jbd2_handle_cache, handle); 1557 } 1558 1559 /* 1560 * jbd2_inode management (optional, for those file systems that want to use 1561 * dynamically allocated jbd2_inode structures) 1562 */ 1563 extern struct kmem_cache *jbd2_inode_cache; 1564 1565 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1566 { 1567 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1568 } 1569 1570 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1571 { 1572 kmem_cache_free(jbd2_inode_cache, jinode); 1573 } 1574 1575 /* Primary revoke support */ 1576 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1577 extern int jbd2_journal_init_revoke(journal_t *, int); 1578 extern void jbd2_journal_destroy_revoke_record_cache(void); 1579 extern void jbd2_journal_destroy_revoke_table_cache(void); 1580 extern int __init jbd2_journal_init_revoke_record_cache(void); 1581 extern int __init jbd2_journal_init_revoke_table_cache(void); 1582 1583 extern void jbd2_journal_destroy_revoke(journal_t *); 1584 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1585 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1586 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1587 struct list_head *log_bufs); 1588 1589 /* Recovery revoke support */ 1590 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1591 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1592 extern void jbd2_journal_clear_revoke(journal_t *); 1593 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1594 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1595 1596 /* 1597 * The log thread user interface: 1598 * 1599 * Request space in the current transaction, and force transaction commit 1600 * transitions on demand. 1601 */ 1602 1603 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1604 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1605 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1606 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1607 int jbd2_transaction_committed(journal_t *journal, tid_t tid); 1608 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1609 int jbd2_log_do_checkpoint(journal_t *journal); 1610 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1611 1612 void __jbd2_log_wait_for_space(journal_t *journal); 1613 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1614 extern int jbd2_cleanup_journal_tail(journal_t *); 1615 1616 /* Fast commit related APIs */ 1617 int jbd2_fc_init(journal_t *journal, int num_fc_blks); 1618 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid); 1619 int jbd2_fc_end_commit(journal_t *journal); 1620 int jbd2_fc_end_commit_fallback(journal_t *journal, tid_t tid); 1621 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out); 1622 int jbd2_submit_inode_data(struct jbd2_inode *jinode); 1623 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode); 1624 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks); 1625 int jbd2_fc_release_bufs(journal_t *journal); 1626 1627 /* 1628 * is_journal_abort 1629 * 1630 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1631 * bit, when set, indicates that we have had a fatal error somewhere, 1632 * either inside the journaling layer or indicated to us by the client 1633 * (eg. ext3), and that we and should not commit any further 1634 * transactions. 1635 */ 1636 1637 static inline int is_journal_aborted(journal_t *journal) 1638 { 1639 return journal->j_flags & JBD2_ABORT; 1640 } 1641 1642 static inline int is_handle_aborted(handle_t *handle) 1643 { 1644 if (handle->h_aborted || !handle->h_transaction) 1645 return 1; 1646 return is_journal_aborted(handle->h_transaction->t_journal); 1647 } 1648 1649 static inline void jbd2_journal_abort_handle(handle_t *handle) 1650 { 1651 handle->h_aborted = 1; 1652 } 1653 1654 #endif /* __KERNEL__ */ 1655 1656 /* Comparison functions for transaction IDs: perform comparisons using 1657 * modulo arithmetic so that they work over sequence number wraps. */ 1658 1659 static inline int tid_gt(tid_t x, tid_t y) 1660 { 1661 int difference = (x - y); 1662 return (difference > 0); 1663 } 1664 1665 static inline int tid_geq(tid_t x, tid_t y) 1666 { 1667 int difference = (x - y); 1668 return (difference >= 0); 1669 } 1670 1671 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1672 extern size_t journal_tag_bytes(journal_t *journal); 1673 1674 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1675 { 1676 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1677 } 1678 1679 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1680 { 1681 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1682 journal->j_chksum_driver == NULL); 1683 1684 return journal->j_chksum_driver != NULL; 1685 } 1686 1687 /* 1688 * Return number of free blocks in the log. Must be called under j_state_lock. 1689 */ 1690 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1691 { 1692 /* Allow for rounding errors */ 1693 long free = journal->j_free - 32; 1694 1695 if (journal->j_committing_transaction) { 1696 free -= atomic_read(&journal-> 1697 j_committing_transaction->t_outstanding_credits); 1698 } 1699 return max_t(long, free, 0); 1700 } 1701 1702 /* 1703 * Definitions which augment the buffer_head layer 1704 */ 1705 1706 /* journaling buffer types */ 1707 #define BJ_None 0 /* Not journaled */ 1708 #define BJ_Metadata 1 /* Normal journaled metadata */ 1709 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1710 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1711 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1712 #define BJ_Types 5 1713 1714 extern int jbd_blocks_per_page(struct inode *inode); 1715 1716 /* JBD uses a CRC32 checksum */ 1717 #define JBD_MAX_CHECKSUM_SIZE 4 1718 1719 static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1720 const void *address, unsigned int length) 1721 { 1722 struct { 1723 struct shash_desc shash; 1724 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1725 } desc; 1726 int err; 1727 1728 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1729 JBD_MAX_CHECKSUM_SIZE); 1730 1731 desc.shash.tfm = journal->j_chksum_driver; 1732 *(u32 *)desc.ctx = crc; 1733 1734 err = crypto_shash_update(&desc.shash, address, length); 1735 BUG_ON(err); 1736 1737 return *(u32 *)desc.ctx; 1738 } 1739 1740 /* Return most recent uncommitted transaction */ 1741 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1742 { 1743 tid_t tid; 1744 1745 read_lock(&journal->j_state_lock); 1746 tid = journal->j_commit_request; 1747 if (journal->j_running_transaction) 1748 tid = journal->j_running_transaction->t_tid; 1749 read_unlock(&journal->j_state_lock); 1750 return tid; 1751 } 1752 1753 static inline int jbd2_handle_buffer_credits(handle_t *handle) 1754 { 1755 journal_t *journal; 1756 1757 if (!handle->h_reserved) 1758 journal = handle->h_transaction->t_journal; 1759 else 1760 journal = handle->h_journal; 1761 1762 return handle->h_total_credits - 1763 DIV_ROUND_UP(handle->h_revoke_credits_requested, 1764 journal->j_revoke_records_per_block); 1765 } 1766 1767 #ifdef __KERNEL__ 1768 1769 #define buffer_trace_init(bh) do {} while (0) 1770 #define print_buffer_fields(bh) do {} while (0) 1771 #define print_buffer_trace(bh) do {} while (0) 1772 #define BUFFER_TRACE(bh, info) do {} while (0) 1773 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1774 #define JBUFFER_TRACE(jh, info) do {} while (0) 1775 1776 #endif /* __KERNEL__ */ 1777 1778 #define EFSBADCRC EBADMSG /* Bad CRC detected */ 1779 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1780 1781 #endif /* _LINUX_JBD2_H */ 1782