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 __u32 s_padding[42]; 267 __be32 s_checksum; /* crc32c(superblock) */ 268 269 /* 0x0100 */ 270 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 271 /* 0x0400 */ 272 } journal_superblock_t; 273 274 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */ 275 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 276 ((j)->j_format_version >= 2 && \ 277 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 278 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 279 ((j)->j_format_version >= 2 && \ 280 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 281 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 282 ((j)->j_format_version >= 2 && \ 283 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 284 285 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 286 287 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 288 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 289 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 290 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 291 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 292 293 /* See "journal feature predicate functions" below */ 294 295 /* Features known to this kernel version: */ 296 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 297 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 298 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 299 JBD2_FEATURE_INCOMPAT_64BIT | \ 300 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 301 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 302 JBD2_FEATURE_INCOMPAT_CSUM_V3) 303 304 #ifdef __KERNEL__ 305 306 #include <linux/fs.h> 307 #include <linux/sched.h> 308 309 enum jbd_state_bits { 310 BH_JBD /* Has an attached ext3 journal_head */ 311 = BH_PrivateStart, 312 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 313 BH_Freed, /* Has been freed (truncated) */ 314 BH_Revoked, /* Has been revoked from the log */ 315 BH_RevokeValid, /* Revoked flag is valid */ 316 BH_JBDDirty, /* Is dirty but journaled */ 317 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 318 BH_Shadow, /* IO on shadow buffer is running */ 319 BH_Verified, /* Metadata block has been verified ok */ 320 BH_JBDPrivateStart, /* First bit available for private use by FS */ 321 }; 322 323 BUFFER_FNS(JBD, jbd) 324 BUFFER_FNS(JWrite, jwrite) 325 BUFFER_FNS(JBDDirty, jbddirty) 326 TAS_BUFFER_FNS(JBDDirty, jbddirty) 327 BUFFER_FNS(Revoked, revoked) 328 TAS_BUFFER_FNS(Revoked, revoked) 329 BUFFER_FNS(RevokeValid, revokevalid) 330 TAS_BUFFER_FNS(RevokeValid, revokevalid) 331 BUFFER_FNS(Freed, freed) 332 BUFFER_FNS(Shadow, shadow) 333 BUFFER_FNS(Verified, verified) 334 335 static inline struct buffer_head *jh2bh(struct journal_head *jh) 336 { 337 return jh->b_bh; 338 } 339 340 static inline struct journal_head *bh2jh(struct buffer_head *bh) 341 { 342 return bh->b_private; 343 } 344 345 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 346 { 347 bit_spin_lock(BH_JournalHead, &bh->b_state); 348 } 349 350 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 351 { 352 bit_spin_unlock(BH_JournalHead, &bh->b_state); 353 } 354 355 #define J_ASSERT(assert) BUG_ON(!(assert)) 356 357 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 358 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 359 360 #if defined(JBD2_PARANOID_IOFAIL) 361 #define J_EXPECT(expr, why...) J_ASSERT(expr) 362 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 363 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 364 #else 365 #define __journal_expect(expr, why...) \ 366 ({ \ 367 int val = (expr); \ 368 if (!val) { \ 369 printk(KERN_ERR \ 370 "JBD2 unexpected failure: %s: %s;\n", \ 371 __func__, #expr); \ 372 printk(KERN_ERR why "\n"); \ 373 } \ 374 val; \ 375 }) 376 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 377 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 378 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 379 #endif 380 381 /* Flags in jbd_inode->i_flags */ 382 #define __JI_COMMIT_RUNNING 0 383 #define __JI_WRITE_DATA 1 384 #define __JI_WAIT_DATA 2 385 386 /* 387 * Commit of the inode data in progress. We use this flag to protect us from 388 * concurrent deletion of inode. We cannot use reference to inode for this 389 * since we cannot afford doing last iput() on behalf of kjournald 390 */ 391 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 392 /* Write allocated dirty buffers in this inode before commit */ 393 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA) 394 /* Wait for outstanding data writes for this inode before commit */ 395 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA) 396 397 /** 398 * struct jbd_inode - The jbd_inode type is the structure linking inodes in 399 * ordered mode present in a transaction so that we can sync them during commit. 400 */ 401 struct jbd2_inode { 402 /** 403 * @i_transaction: 404 * 405 * Which transaction does this inode belong to? Either the running 406 * transaction or the committing one. [j_list_lock] 407 */ 408 transaction_t *i_transaction; 409 410 /** 411 * @i_next_transaction: 412 * 413 * Pointer to the running transaction modifying inode's data in case 414 * there is already a committing transaction touching it. [j_list_lock] 415 */ 416 transaction_t *i_next_transaction; 417 418 /** 419 * @i_list: List of inodes in the i_transaction [j_list_lock] 420 */ 421 struct list_head i_list; 422 423 /** 424 * @i_vfs_inode: 425 * 426 * VFS inode this inode belongs to [constant for lifetime of structure] 427 */ 428 struct inode *i_vfs_inode; 429 430 /** 431 * @i_flags: Flags of inode [j_list_lock] 432 */ 433 unsigned long i_flags; 434 435 /** 436 * @i_dirty_start: 437 * 438 * Offset in bytes where the dirty range for this inode starts. 439 * [j_list_lock] 440 */ 441 loff_t i_dirty_start; 442 443 /** 444 * @i_dirty_end: 445 * 446 * Inclusive offset in bytes where the dirty range for this inode 447 * ends. [j_list_lock] 448 */ 449 loff_t i_dirty_end; 450 }; 451 452 struct jbd2_revoke_table_s; 453 454 /** 455 * struct handle_s - The handle_s type is the concrete type associated with 456 * handle_t. 457 * @h_transaction: Which compound transaction is this update a part of? 458 * @h_journal: Which journal handle belongs to - used iff h_reserved set. 459 * @h_rsv_handle: Handle reserved for finishing the logical operation. 460 * @h_total_credits: Number of remaining buffers we are allowed to add to 461 * journal. These are dirty buffers and revoke descriptor blocks. 462 * @h_revoke_credits: Number of remaining revoke records available for handle 463 * @h_ref: Reference count on this handle. 464 * @h_err: Field for caller's use to track errors through large fs operations. 465 * @h_sync: Flag for sync-on-close. 466 * @h_jdata: Flag to force data journaling. 467 * @h_reserved: Flag for handle for reserved credits. 468 * @h_aborted: Flag indicating fatal error on handle. 469 * @h_type: For handle statistics. 470 * @h_line_no: For handle statistics. 471 * @h_start_jiffies: Handle Start time. 472 * @h_requested_credits: Holds @h_total_credits after handle is started. 473 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started. 474 * @saved_alloc_context: Saved context while transaction is open. 475 **/ 476 477 /* Docbook can't yet cope with the bit fields, but will leave the documentation 478 * in so it can be fixed later. 479 */ 480 481 struct jbd2_journal_handle 482 { 483 union { 484 transaction_t *h_transaction; 485 /* Which journal handle belongs to - used iff h_reserved set */ 486 journal_t *h_journal; 487 }; 488 489 handle_t *h_rsv_handle; 490 int h_total_credits; 491 int h_revoke_credits; 492 int h_revoke_credits_requested; 493 int h_ref; 494 int h_err; 495 496 /* Flags [no locking] */ 497 unsigned int h_sync: 1; 498 unsigned int h_jdata: 1; 499 unsigned int h_reserved: 1; 500 unsigned int h_aborted: 1; 501 unsigned int h_type: 8; 502 unsigned int h_line_no: 16; 503 504 unsigned long h_start_jiffies; 505 unsigned int h_requested_credits; 506 507 unsigned int saved_alloc_context; 508 }; 509 510 511 /* 512 * Some stats for checkpoint phase 513 */ 514 struct transaction_chp_stats_s { 515 unsigned long cs_chp_time; 516 __u32 cs_forced_to_close; 517 __u32 cs_written; 518 __u32 cs_dropped; 519 }; 520 521 /* The transaction_t type is the guts of the journaling mechanism. It 522 * tracks a compound transaction through its various states: 523 * 524 * RUNNING: accepting new updates 525 * LOCKED: Updates still running but we don't accept new ones 526 * RUNDOWN: Updates are tidying up but have finished requesting 527 * new buffers to modify (state not used for now) 528 * FLUSH: All updates complete, but we are still writing to disk 529 * COMMIT: All data on disk, writing commit record 530 * FINISHED: We still have to keep the transaction for checkpointing. 531 * 532 * The transaction keeps track of all of the buffers modified by a 533 * running transaction, and all of the buffers committed but not yet 534 * flushed to home for finished transactions. 535 */ 536 537 /* 538 * Lock ranking: 539 * 540 * j_list_lock 541 * ->jbd_lock_bh_journal_head() (This is "innermost") 542 * 543 * j_state_lock 544 * ->b_state_lock 545 * 546 * b_state_lock 547 * ->j_list_lock 548 * 549 * j_state_lock 550 * ->t_handle_lock 551 * 552 * j_state_lock 553 * ->j_list_lock (journal_unmap_buffer) 554 * 555 */ 556 557 struct transaction_s 558 { 559 /* Pointer to the journal for this transaction. [no locking] */ 560 journal_t *t_journal; 561 562 /* Sequence number for this transaction [no locking] */ 563 tid_t t_tid; 564 565 /* 566 * Transaction's current state 567 * [no locking - only kjournald2 alters this] 568 * [j_list_lock] guards transition of a transaction into T_FINISHED 569 * state and subsequent call of __jbd2_journal_drop_transaction() 570 * FIXME: needs barriers 571 * KLUDGE: [use j_state_lock] 572 */ 573 enum { 574 T_RUNNING, 575 T_LOCKED, 576 T_SWITCH, 577 T_FLUSH, 578 T_COMMIT, 579 T_COMMIT_DFLUSH, 580 T_COMMIT_JFLUSH, 581 T_COMMIT_CALLBACK, 582 T_FINISHED 583 } t_state; 584 585 /* 586 * Where in the log does this transaction's commit start? [no locking] 587 */ 588 unsigned long t_log_start; 589 590 /* Number of buffers on the t_buffers list [j_list_lock] */ 591 int t_nr_buffers; 592 593 /* 594 * Doubly-linked circular list of all buffers reserved but not yet 595 * modified by this transaction [j_list_lock] 596 */ 597 struct journal_head *t_reserved_list; 598 599 /* 600 * Doubly-linked circular list of all metadata buffers owned by this 601 * transaction [j_list_lock] 602 */ 603 struct journal_head *t_buffers; 604 605 /* 606 * Doubly-linked circular list of all forget buffers (superseded 607 * buffers which we can un-checkpoint once this transaction commits) 608 * [j_list_lock] 609 */ 610 struct journal_head *t_forget; 611 612 /* 613 * Doubly-linked circular list of all buffers still to be flushed before 614 * this transaction can be checkpointed. [j_list_lock] 615 */ 616 struct journal_head *t_checkpoint_list; 617 618 /* 619 * Doubly-linked circular list of all buffers submitted for IO while 620 * checkpointing. [j_list_lock] 621 */ 622 struct journal_head *t_checkpoint_io_list; 623 624 /* 625 * Doubly-linked circular list of metadata buffers being shadowed by log 626 * IO. The IO buffers on the iobuf list and the shadow buffers on this 627 * list match each other one for one at all times. [j_list_lock] 628 */ 629 struct journal_head *t_shadow_list; 630 631 /* 632 * List of inodes whose data we've modified in data=ordered mode. 633 * [j_list_lock] 634 */ 635 struct list_head t_inode_list; 636 637 /* 638 * Protects info related to handles 639 */ 640 spinlock_t t_handle_lock; 641 642 /* 643 * Longest time some handle had to wait for running transaction 644 */ 645 unsigned long t_max_wait; 646 647 /* 648 * When transaction started 649 */ 650 unsigned long t_start; 651 652 /* 653 * When commit was requested 654 */ 655 unsigned long t_requested; 656 657 /* 658 * Checkpointing stats [j_checkpoint_sem] 659 */ 660 struct transaction_chp_stats_s t_chp_stats; 661 662 /* 663 * Number of outstanding updates running on this transaction 664 * [none] 665 */ 666 atomic_t t_updates; 667 668 /* 669 * Number of blocks reserved for this transaction in the journal. 670 * This is including all credits reserved when starting transaction 671 * handles as well as all journal descriptor blocks needed for this 672 * transaction. [none] 673 */ 674 atomic_t t_outstanding_credits; 675 676 /* 677 * Number of revoke records for this transaction added by already 678 * stopped handles. [none] 679 */ 680 atomic_t t_outstanding_revokes; 681 682 /* 683 * How many handles used this transaction? [none] 684 */ 685 atomic_t t_handle_count; 686 687 /* 688 * Forward and backward links for the circular list of all transactions 689 * awaiting checkpoint. [j_list_lock] 690 */ 691 transaction_t *t_cpnext, *t_cpprev; 692 693 /* 694 * When will the transaction expire (become due for commit), in jiffies? 695 * [no locking] 696 */ 697 unsigned long t_expires; 698 699 /* 700 * When this transaction started, in nanoseconds [no locking] 701 */ 702 ktime_t t_start_time; 703 704 /* 705 * This transaction is being forced and some process is 706 * waiting for it to finish. 707 */ 708 unsigned int t_synchronous_commit:1; 709 710 /* Disk flush needs to be sent to fs partition [no locking] */ 711 int t_need_data_flush; 712 713 /* 714 * For use by the filesystem to store fs-specific data 715 * structures associated with the transaction 716 */ 717 struct list_head t_private_list; 718 }; 719 720 struct transaction_run_stats_s { 721 unsigned long rs_wait; 722 unsigned long rs_request_delay; 723 unsigned long rs_running; 724 unsigned long rs_locked; 725 unsigned long rs_flushing; 726 unsigned long rs_logging; 727 728 __u32 rs_handle_count; 729 __u32 rs_blocks; 730 __u32 rs_blocks_logged; 731 }; 732 733 struct transaction_stats_s { 734 unsigned long ts_tid; 735 unsigned long ts_requested; 736 struct transaction_run_stats_s run; 737 }; 738 739 static inline unsigned long 740 jbd2_time_diff(unsigned long start, unsigned long end) 741 { 742 if (end >= start) 743 return end - start; 744 745 return end + (MAX_JIFFY_OFFSET - start); 746 } 747 748 #define JBD2_NR_BATCH 64 749 750 /** 751 * struct journal_s - The journal_s type is the concrete type associated with 752 * journal_t. 753 */ 754 struct journal_s 755 { 756 /** 757 * @j_flags: General journaling state flags [j_state_lock] 758 */ 759 unsigned long j_flags; 760 761 /** 762 * @j_errno: 763 * 764 * Is there an outstanding uncleared error on the journal (from a prior 765 * abort)? [j_state_lock] 766 */ 767 int j_errno; 768 769 /** 770 * @j_abort_mutex: Lock the whole aborting procedure. 771 */ 772 struct mutex j_abort_mutex; 773 774 /** 775 * @j_sb_buffer: The first part of the superblock buffer. 776 */ 777 struct buffer_head *j_sb_buffer; 778 779 /** 780 * @j_superblock: The second part of the superblock buffer. 781 */ 782 journal_superblock_t *j_superblock; 783 784 /** 785 * @j_format_version: Version of the superblock format. 786 */ 787 int j_format_version; 788 789 /** 790 * @j_state_lock: Protect the various scalars in the journal. 791 */ 792 rwlock_t j_state_lock; 793 794 /** 795 * @j_barrier_count: 796 * 797 * Number of processes waiting to create a barrier lock [j_state_lock] 798 */ 799 int j_barrier_count; 800 801 /** 802 * @j_barrier: The barrier lock itself. 803 */ 804 struct mutex j_barrier; 805 806 /** 807 * @j_running_transaction: 808 * 809 * Transactions: The current running transaction... 810 * [j_state_lock] [caller holding open handle] 811 */ 812 transaction_t *j_running_transaction; 813 814 /** 815 * @j_committing_transaction: 816 * 817 * the transaction we are pushing to disk 818 * [j_state_lock] [caller holding open handle] 819 */ 820 transaction_t *j_committing_transaction; 821 822 /** 823 * @j_checkpoint_transactions: 824 * 825 * ... and a linked circular list of all transactions waiting for 826 * checkpointing. [j_list_lock] 827 */ 828 transaction_t *j_checkpoint_transactions; 829 830 /** 831 * @j_wait_transaction_locked: 832 * 833 * Wait queue for waiting for a locked transaction to start committing, 834 * or for a barrier lock to be released. 835 */ 836 wait_queue_head_t j_wait_transaction_locked; 837 838 /** 839 * @j_wait_done_commit: Wait queue for waiting for commit to complete. 840 */ 841 wait_queue_head_t j_wait_done_commit; 842 843 /** 844 * @j_wait_commit: Wait queue to trigger commit. 845 */ 846 wait_queue_head_t j_wait_commit; 847 848 /** 849 * @j_wait_updates: Wait queue to wait for updates to complete. 850 */ 851 wait_queue_head_t j_wait_updates; 852 853 /** 854 * @j_wait_reserved: 855 * 856 * Wait queue to wait for reserved buffer credits to drop. 857 */ 858 wait_queue_head_t j_wait_reserved; 859 860 /** 861 * @j_checkpoint_mutex: 862 * 863 * Semaphore for locking against concurrent checkpoints. 864 */ 865 struct mutex j_checkpoint_mutex; 866 867 /** 868 * @j_chkpt_bhs: 869 * 870 * List of buffer heads used by the checkpoint routine. This 871 * was moved from jbd2_log_do_checkpoint() to reduce stack 872 * usage. Access to this array is controlled by the 873 * @j_checkpoint_mutex. [j_checkpoint_mutex] 874 */ 875 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 876 877 /** 878 * @j_head: 879 * 880 * Journal head: identifies the first unused block in the journal. 881 * [j_state_lock] 882 */ 883 unsigned long j_head; 884 885 /** 886 * @j_tail: 887 * 888 * Journal tail: identifies the oldest still-used block in the journal. 889 * [j_state_lock] 890 */ 891 unsigned long j_tail; 892 893 /** 894 * @j_free: 895 * 896 * Journal free: how many free blocks are there in the journal? 897 * [j_state_lock] 898 */ 899 unsigned long j_free; 900 901 /** 902 * @j_first: 903 * 904 * The block number of the first usable block in the journal 905 * [j_state_lock]. 906 */ 907 unsigned long j_first; 908 909 /** 910 * @j_last: 911 * 912 * The block number one beyond the last usable block in the journal 913 * [j_state_lock]. 914 */ 915 unsigned long j_last; 916 917 /** 918 * @j_dev: Device where we store the journal. 919 */ 920 struct block_device *j_dev; 921 922 /** 923 * @j_blocksize: Block size for the location where we store the journal. 924 */ 925 int j_blocksize; 926 927 /** 928 * @j_blk_offset: 929 * 930 * Starting block offset into the device where we store the journal. 931 */ 932 unsigned long long j_blk_offset; 933 934 /** 935 * @j_devname: Journal device name. 936 */ 937 char j_devname[BDEVNAME_SIZE+24]; 938 939 /** 940 * @j_fs_dev: 941 * 942 * Device which holds the client fs. For internal journal this will be 943 * equal to j_dev. 944 */ 945 struct block_device *j_fs_dev; 946 947 /** 948 * @j_maxlen: Total maximum capacity of the journal region on disk. 949 */ 950 unsigned int j_maxlen; 951 952 /** 953 * @j_reserved_credits: 954 * 955 * Number of buffers reserved from the running transaction. 956 */ 957 atomic_t j_reserved_credits; 958 959 /** 960 * @j_list_lock: Protects the buffer lists and internal buffer state. 961 */ 962 spinlock_t j_list_lock; 963 964 /** 965 * @j_inode: 966 * 967 * Optional inode where we store the journal. If present, all 968 * journal block numbers are mapped into this inode via bmap(). 969 */ 970 struct inode *j_inode; 971 972 /** 973 * @j_tail_sequence: 974 * 975 * Sequence number of the oldest transaction in the log [j_state_lock] 976 */ 977 tid_t j_tail_sequence; 978 979 /** 980 * @j_transaction_sequence: 981 * 982 * Sequence number of the next transaction to grant [j_state_lock] 983 */ 984 tid_t j_transaction_sequence; 985 986 /** 987 * @j_commit_sequence: 988 * 989 * Sequence number of the most recently committed transaction 990 * [j_state_lock]. 991 */ 992 tid_t j_commit_sequence; 993 994 /** 995 * @j_commit_request: 996 * 997 * Sequence number of the most recent transaction wanting commit 998 * [j_state_lock] 999 */ 1000 tid_t j_commit_request; 1001 1002 /** 1003 * @j_uuid: 1004 * 1005 * Journal uuid: identifies the object (filesystem, LVM volume etc) 1006 * backed by this journal. This will eventually be replaced by an array 1007 * of uuids, allowing us to index multiple devices within a single 1008 * journal and to perform atomic updates across them. 1009 */ 1010 __u8 j_uuid[16]; 1011 1012 /** 1013 * @j_task: Pointer to the current commit thread for this journal. 1014 */ 1015 struct task_struct *j_task; 1016 1017 /** 1018 * @j_max_transaction_buffers: 1019 * 1020 * Maximum number of metadata buffers to allow in a single compound 1021 * commit transaction. 1022 */ 1023 int j_max_transaction_buffers; 1024 1025 /** 1026 * @j_revoke_records_per_block: 1027 * 1028 * Number of revoke records that fit in one descriptor block. 1029 */ 1030 int j_revoke_records_per_block; 1031 1032 /** 1033 * @j_commit_interval: 1034 * 1035 * What is the maximum transaction lifetime before we begin a commit? 1036 */ 1037 unsigned long j_commit_interval; 1038 1039 /** 1040 * @j_commit_timer: The timer used to wakeup the commit thread. 1041 */ 1042 struct timer_list j_commit_timer; 1043 1044 /** 1045 * @j_revoke_lock: Protect the revoke table. 1046 */ 1047 spinlock_t j_revoke_lock; 1048 1049 /** 1050 * @j_revoke: 1051 * 1052 * The revoke table - maintains the list of revoked blocks in the 1053 * current transaction. 1054 */ 1055 struct jbd2_revoke_table_s *j_revoke; 1056 1057 /** 1058 * @j_revoke_table: Alternate revoke tables for j_revoke. 1059 */ 1060 struct jbd2_revoke_table_s *j_revoke_table[2]; 1061 1062 /** 1063 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction. 1064 */ 1065 struct buffer_head **j_wbuf; 1066 1067 /** 1068 * @j_wbufsize: 1069 * 1070 * Size of @j_wbuf array. 1071 */ 1072 int j_wbufsize; 1073 1074 /** 1075 * @j_last_sync_writer: 1076 * 1077 * The pid of the last person to run a synchronous operation 1078 * through the journal. 1079 */ 1080 pid_t j_last_sync_writer; 1081 1082 /** 1083 * @j_average_commit_time: 1084 * 1085 * The average amount of time in nanoseconds it takes to commit a 1086 * transaction to disk. [j_state_lock] 1087 */ 1088 u64 j_average_commit_time; 1089 1090 /** 1091 * @j_min_batch_time: 1092 * 1093 * Minimum time that we should wait for additional filesystem operations 1094 * to get batched into a synchronous handle in microseconds. 1095 */ 1096 u32 j_min_batch_time; 1097 1098 /** 1099 * @j_max_batch_time: 1100 * 1101 * Maximum time that we should wait for additional filesystem operations 1102 * to get batched into a synchronous handle in microseconds. 1103 */ 1104 u32 j_max_batch_time; 1105 1106 /** 1107 * @j_commit_callback: 1108 * 1109 * This function is called when a transaction is closed. 1110 */ 1111 void (*j_commit_callback)(journal_t *, 1112 transaction_t *); 1113 1114 /* 1115 * Journal statistics 1116 */ 1117 1118 /** 1119 * @j_history_lock: Protect the transactions statistics history. 1120 */ 1121 spinlock_t j_history_lock; 1122 1123 /** 1124 * @j_proc_entry: procfs entry for the jbd statistics directory. 1125 */ 1126 struct proc_dir_entry *j_proc_entry; 1127 1128 /** 1129 * @j_stats: Overall statistics. 1130 */ 1131 struct transaction_stats_s j_stats; 1132 1133 /** 1134 * @j_failed_commit: Failed journal commit ID. 1135 */ 1136 unsigned int j_failed_commit; 1137 1138 /** 1139 * @j_private: 1140 * 1141 * An opaque pointer to fs-private information. ext3 puts its 1142 * superblock pointer here. 1143 */ 1144 void *j_private; 1145 1146 /** 1147 * @j_chksum_driver: 1148 * 1149 * Reference to checksum algorithm driver via cryptoapi. 1150 */ 1151 struct crypto_shash *j_chksum_driver; 1152 1153 /** 1154 * @j_csum_seed: 1155 * 1156 * Precomputed journal UUID checksum for seeding other checksums. 1157 */ 1158 __u32 j_csum_seed; 1159 1160 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1161 /** 1162 * @j_trans_commit_map: 1163 * 1164 * Lockdep entity to track transaction commit dependencies. Handles 1165 * hold this "lock" for read, when we wait for commit, we acquire the 1166 * "lock" for writing. This matches the properties of jbd2 journalling 1167 * where the running transaction has to wait for all handles to be 1168 * dropped to commit that transaction and also acquiring a handle may 1169 * require transaction commit to finish. 1170 */ 1171 struct lockdep_map j_trans_commit_map; 1172 #endif 1173 }; 1174 1175 #define jbd2_might_wait_for_commit(j) \ 1176 do { \ 1177 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \ 1178 rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \ 1179 } while (0) 1180 1181 /* journal feature predicate functions */ 1182 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1183 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1184 { \ 1185 return ((j)->j_format_version >= 2 && \ 1186 ((j)->j_superblock->s_feature_compat & \ 1187 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1188 } \ 1189 static inline void jbd2_set_feature_##name(journal_t *j) \ 1190 { \ 1191 (j)->j_superblock->s_feature_compat |= \ 1192 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1193 } \ 1194 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1195 { \ 1196 (j)->j_superblock->s_feature_compat &= \ 1197 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1198 } 1199 1200 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1201 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1202 { \ 1203 return ((j)->j_format_version >= 2 && \ 1204 ((j)->j_superblock->s_feature_ro_compat & \ 1205 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1206 } \ 1207 static inline void jbd2_set_feature_##name(journal_t *j) \ 1208 { \ 1209 (j)->j_superblock->s_feature_ro_compat |= \ 1210 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1211 } \ 1212 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1213 { \ 1214 (j)->j_superblock->s_feature_ro_compat &= \ 1215 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1216 } 1217 1218 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1219 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1220 { \ 1221 return ((j)->j_format_version >= 2 && \ 1222 ((j)->j_superblock->s_feature_incompat & \ 1223 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1224 } \ 1225 static inline void jbd2_set_feature_##name(journal_t *j) \ 1226 { \ 1227 (j)->j_superblock->s_feature_incompat |= \ 1228 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1229 } \ 1230 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1231 { \ 1232 (j)->j_superblock->s_feature_incompat &= \ 1233 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1234 } 1235 1236 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1237 1238 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1239 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1240 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1241 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1242 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1243 1244 /* 1245 * Journal flag definitions 1246 */ 1247 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1248 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1249 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1250 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1251 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1252 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1253 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1254 * data write error in ordered 1255 * mode */ 1256 1257 /* 1258 * Function declarations for the journaling transaction and buffer 1259 * management 1260 */ 1261 1262 /* Filing buffers */ 1263 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1264 extern bool __jbd2_journal_refile_buffer(struct journal_head *); 1265 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1266 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1267 extern void __journal_free_buffer(struct journal_head *bh); 1268 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1269 extern void __journal_clean_data_list(transaction_t *transaction); 1270 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1271 { 1272 list_add_tail(&bh->b_assoc_buffers, head); 1273 } 1274 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1275 { 1276 list_del_init(&bh->b_assoc_buffers); 1277 } 1278 1279 /* Log buffer allocation */ 1280 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1281 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1282 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1283 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1284 unsigned long *block); 1285 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1286 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1287 1288 /* Commit management */ 1289 extern void jbd2_journal_commit_transaction(journal_t *); 1290 1291 /* Checkpoint list management */ 1292 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1293 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1294 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1295 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1296 1297 1298 /* 1299 * Triggers 1300 */ 1301 1302 struct jbd2_buffer_trigger_type { 1303 /* 1304 * Fired a the moment data to write to the journal are known to be 1305 * stable - so either at the moment b_frozen_data is created or just 1306 * before a buffer is written to the journal. mapped_data is a mapped 1307 * buffer that is the frozen data for commit. 1308 */ 1309 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1310 struct buffer_head *bh, void *mapped_data, 1311 size_t size); 1312 1313 /* 1314 * Fired during journal abort for dirty buffers that will not be 1315 * committed. 1316 */ 1317 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1318 struct buffer_head *bh); 1319 }; 1320 1321 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1322 void *mapped_data, 1323 struct jbd2_buffer_trigger_type *triggers); 1324 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1325 struct jbd2_buffer_trigger_type *triggers); 1326 1327 /* Buffer IO */ 1328 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1329 struct journal_head *jh_in, 1330 struct buffer_head **bh_out, 1331 sector_t blocknr); 1332 1333 /* Transaction locking */ 1334 extern void __wait_on_journal (journal_t *); 1335 1336 /* Transaction cache support */ 1337 extern void jbd2_journal_destroy_transaction_cache(void); 1338 extern int __init jbd2_journal_init_transaction_cache(void); 1339 extern void jbd2_journal_free_transaction(transaction_t *); 1340 1341 /* 1342 * Journal locking. 1343 * 1344 * We need to lock the journal during transaction state changes so that nobody 1345 * ever tries to take a handle on the running transaction while we are in the 1346 * middle of moving it to the commit phase. j_state_lock does this. 1347 * 1348 * Note that the locking is completely interrupt unsafe. We never touch 1349 * journal structures from interrupts. 1350 */ 1351 1352 static inline handle_t *journal_current_handle(void) 1353 { 1354 return current->journal_info; 1355 } 1356 1357 /* The journaling code user interface: 1358 * 1359 * Create and destroy handles 1360 * Register buffer modifications against the current transaction. 1361 */ 1362 1363 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1364 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1365 int revoke_records, gfp_t gfp_mask, 1366 unsigned int type, unsigned int line_no); 1367 extern int jbd2_journal_restart(handle_t *, int nblocks); 1368 extern int jbd2__journal_restart(handle_t *, int nblocks, 1369 int revoke_records, gfp_t gfp_mask); 1370 extern int jbd2_journal_start_reserved(handle_t *handle, 1371 unsigned int type, unsigned int line_no); 1372 extern void jbd2_journal_free_reserved(handle_t *handle); 1373 extern int jbd2_journal_extend(handle_t *handle, int nblocks, 1374 int revoke_records); 1375 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1376 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1377 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1378 void jbd2_journal_set_triggers(struct buffer_head *, 1379 struct jbd2_buffer_trigger_type *type); 1380 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1381 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1382 extern int jbd2_journal_invalidatepage(journal_t *, 1383 struct page *, unsigned int, unsigned int); 1384 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); 1385 extern int jbd2_journal_stop(handle_t *); 1386 extern int jbd2_journal_flush (journal_t *); 1387 extern void jbd2_journal_lock_updates (journal_t *); 1388 extern void jbd2_journal_unlock_updates (journal_t *); 1389 1390 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1391 struct block_device *fs_dev, 1392 unsigned long long start, int len, int bsize); 1393 extern journal_t * jbd2_journal_init_inode (struct inode *); 1394 extern int jbd2_journal_update_format (journal_t *); 1395 extern int jbd2_journal_check_used_features 1396 (journal_t *, unsigned long, unsigned long, unsigned long); 1397 extern int jbd2_journal_check_available_features 1398 (journal_t *, unsigned long, unsigned long, unsigned long); 1399 extern int jbd2_journal_set_features 1400 (journal_t *, unsigned long, unsigned long, unsigned long); 1401 extern void jbd2_journal_clear_features 1402 (journal_t *, unsigned long, unsigned long, unsigned long); 1403 extern int jbd2_journal_load (journal_t *journal); 1404 extern int jbd2_journal_destroy (journal_t *); 1405 extern int jbd2_journal_recover (journal_t *journal); 1406 extern int jbd2_journal_wipe (journal_t *, int); 1407 extern int jbd2_journal_skip_recovery (journal_t *); 1408 extern void jbd2_journal_update_sb_errno(journal_t *); 1409 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1410 unsigned long, int); 1411 extern void jbd2_journal_abort (journal_t *, int); 1412 extern int jbd2_journal_errno (journal_t *); 1413 extern void jbd2_journal_ack_err (journal_t *); 1414 extern int jbd2_journal_clear_err (journal_t *); 1415 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1416 extern int jbd2_journal_force_commit(journal_t *); 1417 extern int jbd2_journal_force_commit_nested(journal_t *); 1418 extern int jbd2_journal_inode_ranged_write(handle_t *handle, 1419 struct jbd2_inode *inode, loff_t start_byte, 1420 loff_t length); 1421 extern int jbd2_journal_inode_ranged_wait(handle_t *handle, 1422 struct jbd2_inode *inode, loff_t start_byte, 1423 loff_t length); 1424 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1425 struct jbd2_inode *inode, loff_t new_size); 1426 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1427 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1428 1429 /* 1430 * journal_head management 1431 */ 1432 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1433 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1434 void jbd2_journal_put_journal_head(struct journal_head *jh); 1435 1436 /* 1437 * handle management 1438 */ 1439 extern struct kmem_cache *jbd2_handle_cache; 1440 1441 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1442 { 1443 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1444 } 1445 1446 static inline void jbd2_free_handle(handle_t *handle) 1447 { 1448 kmem_cache_free(jbd2_handle_cache, handle); 1449 } 1450 1451 /* 1452 * jbd2_inode management (optional, for those file systems that want to use 1453 * dynamically allocated jbd2_inode structures) 1454 */ 1455 extern struct kmem_cache *jbd2_inode_cache; 1456 1457 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1458 { 1459 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1460 } 1461 1462 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1463 { 1464 kmem_cache_free(jbd2_inode_cache, jinode); 1465 } 1466 1467 /* Primary revoke support */ 1468 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1469 extern int jbd2_journal_init_revoke(journal_t *, int); 1470 extern void jbd2_journal_destroy_revoke_record_cache(void); 1471 extern void jbd2_journal_destroy_revoke_table_cache(void); 1472 extern int __init jbd2_journal_init_revoke_record_cache(void); 1473 extern int __init jbd2_journal_init_revoke_table_cache(void); 1474 1475 extern void jbd2_journal_destroy_revoke(journal_t *); 1476 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1477 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1478 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1479 struct list_head *log_bufs); 1480 1481 /* Recovery revoke support */ 1482 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1483 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1484 extern void jbd2_journal_clear_revoke(journal_t *); 1485 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1486 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1487 1488 /* 1489 * The log thread user interface: 1490 * 1491 * Request space in the current transaction, and force transaction commit 1492 * transitions on demand. 1493 */ 1494 1495 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1496 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1497 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1498 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1499 int jbd2_transaction_committed(journal_t *journal, tid_t tid); 1500 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1501 int jbd2_log_do_checkpoint(journal_t *journal); 1502 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1503 1504 void __jbd2_log_wait_for_space(journal_t *journal); 1505 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1506 extern int jbd2_cleanup_journal_tail(journal_t *); 1507 1508 /* 1509 * is_journal_abort 1510 * 1511 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1512 * bit, when set, indicates that we have had a fatal error somewhere, 1513 * either inside the journaling layer or indicated to us by the client 1514 * (eg. ext3), and that we and should not commit any further 1515 * transactions. 1516 */ 1517 1518 static inline int is_journal_aborted(journal_t *journal) 1519 { 1520 return journal->j_flags & JBD2_ABORT; 1521 } 1522 1523 static inline int is_handle_aborted(handle_t *handle) 1524 { 1525 if (handle->h_aborted || !handle->h_transaction) 1526 return 1; 1527 return is_journal_aborted(handle->h_transaction->t_journal); 1528 } 1529 1530 static inline void jbd2_journal_abort_handle(handle_t *handle) 1531 { 1532 handle->h_aborted = 1; 1533 } 1534 1535 #endif /* __KERNEL__ */ 1536 1537 /* Comparison functions for transaction IDs: perform comparisons using 1538 * modulo arithmetic so that they work over sequence number wraps. */ 1539 1540 static inline int tid_gt(tid_t x, tid_t y) 1541 { 1542 int difference = (x - y); 1543 return (difference > 0); 1544 } 1545 1546 static inline int tid_geq(tid_t x, tid_t y) 1547 { 1548 int difference = (x - y); 1549 return (difference >= 0); 1550 } 1551 1552 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1553 extern size_t journal_tag_bytes(journal_t *journal); 1554 1555 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1556 { 1557 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1558 } 1559 1560 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1561 { 1562 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1563 journal->j_chksum_driver == NULL); 1564 1565 return journal->j_chksum_driver != NULL; 1566 } 1567 1568 /* 1569 * Return number of free blocks in the log. Must be called under j_state_lock. 1570 */ 1571 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1572 { 1573 /* Allow for rounding errors */ 1574 long free = journal->j_free - 32; 1575 1576 if (journal->j_committing_transaction) { 1577 free -= atomic_read(&journal-> 1578 j_committing_transaction->t_outstanding_credits); 1579 } 1580 return max_t(long, free, 0); 1581 } 1582 1583 /* 1584 * Definitions which augment the buffer_head layer 1585 */ 1586 1587 /* journaling buffer types */ 1588 #define BJ_None 0 /* Not journaled */ 1589 #define BJ_Metadata 1 /* Normal journaled metadata */ 1590 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1591 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1592 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1593 #define BJ_Types 5 1594 1595 extern int jbd_blocks_per_page(struct inode *inode); 1596 1597 /* JBD uses a CRC32 checksum */ 1598 #define JBD_MAX_CHECKSUM_SIZE 4 1599 1600 static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1601 const void *address, unsigned int length) 1602 { 1603 struct { 1604 struct shash_desc shash; 1605 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1606 } desc; 1607 int err; 1608 1609 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1610 JBD_MAX_CHECKSUM_SIZE); 1611 1612 desc.shash.tfm = journal->j_chksum_driver; 1613 *(u32 *)desc.ctx = crc; 1614 1615 err = crypto_shash_update(&desc.shash, address, length); 1616 BUG_ON(err); 1617 1618 return *(u32 *)desc.ctx; 1619 } 1620 1621 /* Return most recent uncommitted transaction */ 1622 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1623 { 1624 tid_t tid; 1625 1626 read_lock(&journal->j_state_lock); 1627 tid = journal->j_commit_request; 1628 if (journal->j_running_transaction) 1629 tid = journal->j_running_transaction->t_tid; 1630 read_unlock(&journal->j_state_lock); 1631 return tid; 1632 } 1633 1634 static inline int jbd2_handle_buffer_credits(handle_t *handle) 1635 { 1636 journal_t *journal; 1637 1638 if (!handle->h_reserved) 1639 journal = handle->h_transaction->t_journal; 1640 else 1641 journal = handle->h_journal; 1642 1643 return handle->h_total_credits - 1644 DIV_ROUND_UP(handle->h_revoke_credits_requested, 1645 journal->j_revoke_records_per_block); 1646 } 1647 1648 #ifdef __KERNEL__ 1649 1650 #define buffer_trace_init(bh) do {} while (0) 1651 #define print_buffer_fields(bh) do {} while (0) 1652 #define print_buffer_trace(bh) do {} while (0) 1653 #define BUFFER_TRACE(bh, info) do {} while (0) 1654 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1655 #define JBUFFER_TRACE(jh, info) do {} while (0) 1656 1657 #endif /* __KERNEL__ */ 1658 1659 #define EFSBADCRC EBADMSG /* Bad CRC detected */ 1660 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1661 1662 #endif /* _LINUX_JBD2_H */ 1663