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 <crypto/hash.h> 31 #endif 32 33 #define journal_oom_retry 1 34 35 /* 36 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds 37 * certain classes of error which can occur due to failed IOs. Under 38 * normal use we want ext4 to continue after such errors, because 39 * hardware _can_ fail, but for debugging purposes when running tests on 40 * known-good hardware we may want to trap these errors. 41 */ 42 #undef JBD2_PARANOID_IOFAIL 43 44 /* 45 * The default maximum commit age, in seconds. 46 */ 47 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5 48 49 #ifdef CONFIG_JBD2_DEBUG 50 /* 51 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal 52 * consistency checks. By default we don't do this unless 53 * CONFIG_JBD2_DEBUG is on. 54 */ 55 #define JBD2_EXPENSIVE_CHECKING 56 extern ushort jbd2_journal_enable_debug; 57 void __jbd2_debug(int level, const char *file, const char *func, 58 unsigned int line, const char *fmt, ...); 59 60 #define jbd_debug(n, fmt, a...) \ 61 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a) 62 #else 63 #define jbd_debug(n, fmt, a...) /**/ 64 #endif 65 66 extern void *jbd2_alloc(size_t size, gfp_t flags); 67 extern void jbd2_free(void *ptr, size_t size); 68 69 #define JBD2_MIN_JOURNAL_BLOCKS 1024 70 71 #ifdef __KERNEL__ 72 73 /** 74 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 75 * 76 * All filesystem modifications made by the process go 77 * through this handle. Recursive operations (such as quota operations) 78 * are gathered into a single update. 79 * 80 * The buffer credits field is used to account for journaled buffers 81 * being modified by the running process. To ensure that there is 82 * enough log space for all outstanding operations, we need to limit the 83 * number of outstanding buffers possible at any time. When the 84 * operation completes, any buffer credits not used are credited back to 85 * the transaction, so that at all times we know how many buffers the 86 * outstanding updates on a transaction might possibly touch. 87 * 88 * This is an opaque datatype. 89 **/ 90 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ 91 92 93 /** 94 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 95 * 96 * journal_t is linked to from the fs superblock structure. 97 * 98 * We use the journal_t to keep track of all outstanding transaction 99 * activity on the filesystem, and to manage the state of the log 100 * writing process. 101 * 102 * This is an opaque datatype. 103 **/ 104 typedef struct journal_s journal_t; /* Journal control structure */ 105 #endif 106 107 /* 108 * Internal structures used by the logging mechanism: 109 */ 110 111 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 112 113 /* 114 * On-disk structures 115 */ 116 117 /* 118 * Descriptor block types: 119 */ 120 121 #define JBD2_DESCRIPTOR_BLOCK 1 122 #define JBD2_COMMIT_BLOCK 2 123 #define JBD2_SUPERBLOCK_V1 3 124 #define JBD2_SUPERBLOCK_V2 4 125 #define JBD2_REVOKE_BLOCK 5 126 127 /* 128 * Standard header for all descriptor blocks: 129 */ 130 typedef struct journal_header_s 131 { 132 __be32 h_magic; 133 __be32 h_blocktype; 134 __be32 h_sequence; 135 } journal_header_t; 136 137 /* 138 * Checksum types. 139 */ 140 #define JBD2_CRC32_CHKSUM 1 141 #define JBD2_MD5_CHKSUM 2 142 #define JBD2_SHA1_CHKSUM 3 143 #define JBD2_CRC32C_CHKSUM 4 144 145 #define JBD2_CRC32_CHKSUM_SIZE 4 146 147 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 148 /* 149 * Commit block header for storing transactional checksums: 150 * 151 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* 152 * fields are used to store a checksum of the descriptor and data blocks. 153 * 154 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum 155 * field is used to store crc32c(uuid+commit_block). Each journal metadata 156 * block gets its own checksum, and data block checksums are stored in 157 * journal_block_tag (in the descriptor). The other h_chksum* fields are 158 * not used. 159 * 160 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses 161 * journal_block_tag3_t to store a full 32-bit checksum. Everything else 162 * is the same as v2. 163 * 164 * Checksum v1, v2, and v3 are mutually exclusive features. 165 */ 166 struct commit_header { 167 __be32 h_magic; 168 __be32 h_blocktype; 169 __be32 h_sequence; 170 unsigned char h_chksum_type; 171 unsigned char h_chksum_size; 172 unsigned char h_padding[2]; 173 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 174 __be64 h_commit_sec; 175 __be32 h_commit_nsec; 176 }; 177 178 /* 179 * The block tag: used to describe a single buffer in the journal. 180 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 181 * raw struct shouldn't be used for pointer math or sizeof() - use 182 * journal_tag_bytes(journal) instead to compute this. 183 */ 184 typedef struct journal_block_tag3_s 185 { 186 __be32 t_blocknr; /* The on-disk block number */ 187 __be32 t_flags; /* See below */ 188 __be32 t_blocknr_high; /* most-significant high 32bits. */ 189 __be32 t_checksum; /* crc32c(uuid+seq+block) */ 190 } journal_block_tag3_t; 191 192 typedef struct journal_block_tag_s 193 { 194 __be32 t_blocknr; /* The on-disk block number */ 195 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ 196 __be16 t_flags; /* See below */ 197 __be32 t_blocknr_high; /* most-significant high 32bits. */ 198 } journal_block_tag_t; 199 200 /* Tail of descriptor or revoke block, for checksumming */ 201 struct jbd2_journal_block_tail { 202 __be32 t_checksum; /* crc32c(uuid+descr_block) */ 203 }; 204 205 /* 206 * The revoke descriptor: used on disk to describe a series of blocks to 207 * be revoked from the log 208 */ 209 typedef struct jbd2_journal_revoke_header_s 210 { 211 journal_header_t r_header; 212 __be32 r_count; /* Count of bytes used in the block */ 213 } jbd2_journal_revoke_header_t; 214 215 /* Definitions for the journal tag flags word: */ 216 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 217 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 218 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 219 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 220 221 222 /* 223 * The journal superblock. All fields are in big-endian byte order. 224 */ 225 typedef struct journal_superblock_s 226 { 227 /* 0x0000 */ 228 journal_header_t s_header; 229 230 /* 0x000C */ 231 /* Static information describing the journal */ 232 __be32 s_blocksize; /* journal device blocksize */ 233 __be32 s_maxlen; /* total blocks in journal file */ 234 __be32 s_first; /* first block of log information */ 235 236 /* 0x0018 */ 237 /* Dynamic information describing the current state of the log */ 238 __be32 s_sequence; /* first commit ID expected in log */ 239 __be32 s_start; /* blocknr of start of log */ 240 241 /* 0x0020 */ 242 /* Error value, as set by jbd2_journal_abort(). */ 243 __be32 s_errno; 244 245 /* 0x0024 */ 246 /* Remaining fields are only valid in a version-2 superblock */ 247 __be32 s_feature_compat; /* compatible feature set */ 248 __be32 s_feature_incompat; /* incompatible feature set */ 249 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 250 /* 0x0030 */ 251 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 252 253 /* 0x0040 */ 254 __be32 s_nr_users; /* Nr of filesystems sharing log */ 255 256 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 257 258 /* 0x0048 */ 259 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 260 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 261 262 /* 0x0050 */ 263 __u8 s_checksum_type; /* checksum type */ 264 __u8 s_padding2[3]; 265 __u32 s_padding[42]; 266 __be32 s_checksum; /* crc32c(superblock) */ 267 268 /* 0x0100 */ 269 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 270 /* 0x0400 */ 271 } journal_superblock_t; 272 273 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */ 274 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 275 ((j)->j_format_version >= 2 && \ 276 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 277 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 278 ((j)->j_format_version >= 2 && \ 279 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 280 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 281 ((j)->j_format_version >= 2 && \ 282 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 283 284 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 285 286 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 287 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 288 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 289 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 290 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 291 292 /* See "journal feature predicate functions" below */ 293 294 /* Features known to this kernel version: */ 295 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 296 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 297 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 298 JBD2_FEATURE_INCOMPAT_64BIT | \ 299 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 300 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 301 JBD2_FEATURE_INCOMPAT_CSUM_V3) 302 303 #ifdef __KERNEL__ 304 305 #include <linux/fs.h> 306 #include <linux/sched.h> 307 308 enum jbd_state_bits { 309 BH_JBD /* Has an attached ext3 journal_head */ 310 = BH_PrivateStart, 311 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 312 BH_Freed, /* Has been freed (truncated) */ 313 BH_Revoked, /* Has been revoked from the log */ 314 BH_RevokeValid, /* Revoked flag is valid */ 315 BH_JBDDirty, /* Is dirty but journaled */ 316 BH_State, /* Pins most journal_head state */ 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_state(struct buffer_head *bh) 346 { 347 bit_spin_lock(BH_State, &bh->b_state); 348 } 349 350 static inline int jbd_trylock_bh_state(struct buffer_head *bh) 351 { 352 return bit_spin_trylock(BH_State, &bh->b_state); 353 } 354 355 static inline int jbd_is_locked_bh_state(struct buffer_head *bh) 356 { 357 return bit_spin_is_locked(BH_State, &bh->b_state); 358 } 359 360 static inline void jbd_unlock_bh_state(struct buffer_head *bh) 361 { 362 bit_spin_unlock(BH_State, &bh->b_state); 363 } 364 365 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 366 { 367 bit_spin_lock(BH_JournalHead, &bh->b_state); 368 } 369 370 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 371 { 372 bit_spin_unlock(BH_JournalHead, &bh->b_state); 373 } 374 375 #define J_ASSERT(assert) BUG_ON(!(assert)) 376 377 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 378 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 379 380 #if defined(JBD2_PARANOID_IOFAIL) 381 #define J_EXPECT(expr, why...) J_ASSERT(expr) 382 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 383 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 384 #else 385 #define __journal_expect(expr, why...) \ 386 ({ \ 387 int val = (expr); \ 388 if (!val) { \ 389 printk(KERN_ERR \ 390 "JBD2 unexpected failure: %s: %s;\n", \ 391 __func__, #expr); \ 392 printk(KERN_ERR why "\n"); \ 393 } \ 394 val; \ 395 }) 396 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 397 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 398 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 399 #endif 400 401 /* Flags in jbd_inode->i_flags */ 402 #define __JI_COMMIT_RUNNING 0 403 #define __JI_WRITE_DATA 1 404 #define __JI_WAIT_DATA 2 405 406 /* 407 * Commit of the inode data in progress. We use this flag to protect us from 408 * concurrent deletion of inode. We cannot use reference to inode for this 409 * since we cannot afford doing last iput() on behalf of kjournald 410 */ 411 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 412 /* Write allocated dirty buffers in this inode before commit */ 413 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA) 414 /* Wait for outstanding data writes for this inode before commit */ 415 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA) 416 417 /** 418 * struct jbd_inode - The jbd_inode type is the structure linking inodes in 419 * ordered mode present in a transaction so that we can sync them during commit. 420 */ 421 struct jbd2_inode { 422 /** 423 * @i_transaction: 424 * 425 * Which transaction does this inode belong to? Either the running 426 * transaction or the committing one. [j_list_lock] 427 */ 428 transaction_t *i_transaction; 429 430 /** 431 * @i_next_transaction: 432 * 433 * Pointer to the running transaction modifying inode's data in case 434 * there is already a committing transaction touching it. [j_list_lock] 435 */ 436 transaction_t *i_next_transaction; 437 438 /** 439 * @i_list: List of inodes in the i_transaction [j_list_lock] 440 */ 441 struct list_head i_list; 442 443 /** 444 * @i_vfs_inode: 445 * 446 * VFS inode this inode belongs to [constant for lifetime of structure] 447 */ 448 struct inode *i_vfs_inode; 449 450 /** 451 * @i_flags: Flags of inode [j_list_lock] 452 */ 453 unsigned long i_flags; 454 }; 455 456 struct jbd2_revoke_table_s; 457 458 /** 459 * struct handle_s - The handle_s type is the concrete type associated with 460 * handle_t. 461 * @h_transaction: Which compound transaction is this update a part of? 462 * @h_journal: Which journal handle belongs to - used iff h_reserved set. 463 * @h_rsv_handle: Handle reserved for finishing the logical operation. 464 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. 465 * @h_ref: Reference count on this handle. 466 * @h_err: Field for caller's use to track errors through large fs operations. 467 * @h_sync: Flag for sync-on-close. 468 * @h_jdata: Flag to force data journaling. 469 * @h_reserved: Flag for handle for reserved credits. 470 * @h_aborted: Flag indicating fatal error on handle. 471 * @h_type: For handle statistics. 472 * @h_line_no: For handle statistics. 473 * @h_start_jiffies: Handle Start time. 474 * @h_requested_credits: Holds @h_buffer_credits after handle is started. 475 * @saved_alloc_context: Saved context while transaction is open. 476 **/ 477 478 /* Docbook can't yet cope with the bit fields, but will leave the documentation 479 * in so it can be fixed later. 480 */ 481 482 struct jbd2_journal_handle 483 { 484 union { 485 transaction_t *h_transaction; 486 /* Which journal handle belongs to - used iff h_reserved set */ 487 journal_t *h_journal; 488 }; 489 490 handle_t *h_rsv_handle; 491 int h_buffer_credits; 492 int h_ref; 493 int h_err; 494 495 /* Flags [no locking] */ 496 unsigned int h_sync: 1; 497 unsigned int h_jdata: 1; 498 unsigned int h_reserved: 1; 499 unsigned int h_aborted: 1; 500 unsigned int h_type: 8; 501 unsigned int h_line_no: 16; 502 503 unsigned long h_start_jiffies; 504 unsigned int h_requested_credits; 505 506 unsigned int saved_alloc_context; 507 }; 508 509 510 /* 511 * Some stats for checkpoint phase 512 */ 513 struct transaction_chp_stats_s { 514 unsigned long cs_chp_time; 515 __u32 cs_forced_to_close; 516 __u32 cs_written; 517 __u32 cs_dropped; 518 }; 519 520 /* The transaction_t type is the guts of the journaling mechanism. It 521 * tracks a compound transaction through its various states: 522 * 523 * RUNNING: accepting new updates 524 * LOCKED: Updates still running but we don't accept new ones 525 * RUNDOWN: Updates are tidying up but have finished requesting 526 * new buffers to modify (state not used for now) 527 * FLUSH: All updates complete, but we are still writing to disk 528 * COMMIT: All data on disk, writing commit record 529 * FINISHED: We still have to keep the transaction for checkpointing. 530 * 531 * The transaction keeps track of all of the buffers modified by a 532 * running transaction, and all of the buffers committed but not yet 533 * flushed to home for finished transactions. 534 */ 535 536 /* 537 * Lock ranking: 538 * 539 * j_list_lock 540 * ->jbd_lock_bh_journal_head() (This is "innermost") 541 * 542 * j_state_lock 543 * ->jbd_lock_bh_state() 544 * 545 * jbd_lock_bh_state() 546 * ->j_list_lock 547 * 548 * j_state_lock 549 * ->t_handle_lock 550 * 551 * j_state_lock 552 * ->j_list_lock (journal_unmap_buffer) 553 * 554 */ 555 556 struct transaction_s 557 { 558 /* Pointer to the journal for this transaction. [no locking] */ 559 journal_t *t_journal; 560 561 /* Sequence number for this transaction [no locking] */ 562 tid_t t_tid; 563 564 /* 565 * Transaction's current state 566 * [no locking - only kjournald2 alters this] 567 * [j_list_lock] guards transition of a transaction into T_FINISHED 568 * state and subsequent call of __jbd2_journal_drop_transaction() 569 * FIXME: needs barriers 570 * KLUDGE: [use j_state_lock] 571 */ 572 enum { 573 T_RUNNING, 574 T_LOCKED, 575 T_SWITCH, 576 T_FLUSH, 577 T_COMMIT, 578 T_COMMIT_DFLUSH, 579 T_COMMIT_JFLUSH, 580 T_COMMIT_CALLBACK, 581 T_FINISHED 582 } t_state; 583 584 /* 585 * Where in the log does this transaction's commit start? [no locking] 586 */ 587 unsigned long t_log_start; 588 589 /* Number of buffers on the t_buffers list [j_list_lock] */ 590 int t_nr_buffers; 591 592 /* 593 * Doubly-linked circular list of all buffers reserved but not yet 594 * modified by this transaction [j_list_lock] 595 */ 596 struct journal_head *t_reserved_list; 597 598 /* 599 * Doubly-linked circular list of all metadata buffers owned by this 600 * transaction [j_list_lock] 601 */ 602 struct journal_head *t_buffers; 603 604 /* 605 * Doubly-linked circular list of all forget buffers (superseded 606 * buffers which we can un-checkpoint once this transaction commits) 607 * [j_list_lock] 608 */ 609 struct journal_head *t_forget; 610 611 /* 612 * Doubly-linked circular list of all buffers still to be flushed before 613 * this transaction can be checkpointed. [j_list_lock] 614 */ 615 struct journal_head *t_checkpoint_list; 616 617 /* 618 * Doubly-linked circular list of all buffers submitted for IO while 619 * checkpointing. [j_list_lock] 620 */ 621 struct journal_head *t_checkpoint_io_list; 622 623 /* 624 * Doubly-linked circular list of metadata buffers being shadowed by log 625 * IO. The IO buffers on the iobuf list and the shadow buffers on this 626 * list match each other one for one at all times. [j_list_lock] 627 */ 628 struct journal_head *t_shadow_list; 629 630 /* 631 * List of inodes whose data we've modified in data=ordered mode. 632 * [j_list_lock] 633 */ 634 struct list_head t_inode_list; 635 636 /* 637 * Protects info related to handles 638 */ 639 spinlock_t t_handle_lock; 640 641 /* 642 * Longest time some handle had to wait for running transaction 643 */ 644 unsigned long t_max_wait; 645 646 /* 647 * When transaction started 648 */ 649 unsigned long t_start; 650 651 /* 652 * When commit was requested 653 */ 654 unsigned long t_requested; 655 656 /* 657 * Checkpointing stats [j_checkpoint_sem] 658 */ 659 struct transaction_chp_stats_s t_chp_stats; 660 661 /* 662 * Number of outstanding updates running on this transaction 663 * [none] 664 */ 665 atomic_t t_updates; 666 667 /* 668 * Number of buffers reserved for use by all handles in this transaction 669 * handle but not yet modified. [none] 670 */ 671 atomic_t t_outstanding_credits; 672 673 /* 674 * Forward and backward links for the circular list of all transactions 675 * awaiting checkpoint. [j_list_lock] 676 */ 677 transaction_t *t_cpnext, *t_cpprev; 678 679 /* 680 * When will the transaction expire (become due for commit), in jiffies? 681 * [no locking] 682 */ 683 unsigned long t_expires; 684 685 /* 686 * When this transaction started, in nanoseconds [no locking] 687 */ 688 ktime_t t_start_time; 689 690 /* 691 * How many handles used this transaction? [none] 692 */ 693 atomic_t t_handle_count; 694 695 /* 696 * This transaction is being forced and some process is 697 * waiting for it to finish. 698 */ 699 unsigned int t_synchronous_commit:1; 700 701 /* Disk flush needs to be sent to fs partition [no locking] */ 702 int t_need_data_flush; 703 704 /* 705 * For use by the filesystem to store fs-specific data 706 * structures associated with the transaction 707 */ 708 struct list_head t_private_list; 709 }; 710 711 struct transaction_run_stats_s { 712 unsigned long rs_wait; 713 unsigned long rs_request_delay; 714 unsigned long rs_running; 715 unsigned long rs_locked; 716 unsigned long rs_flushing; 717 unsigned long rs_logging; 718 719 __u32 rs_handle_count; 720 __u32 rs_blocks; 721 __u32 rs_blocks_logged; 722 }; 723 724 struct transaction_stats_s { 725 unsigned long ts_tid; 726 unsigned long ts_requested; 727 struct transaction_run_stats_s run; 728 }; 729 730 static inline unsigned long 731 jbd2_time_diff(unsigned long start, unsigned long end) 732 { 733 if (end >= start) 734 return end - start; 735 736 return end + (MAX_JIFFY_OFFSET - start); 737 } 738 739 #define JBD2_NR_BATCH 64 740 741 /** 742 * struct journal_s - The journal_s type is the concrete type associated with 743 * journal_t. 744 */ 745 struct journal_s 746 { 747 /** 748 * @j_flags: General journaling state flags [j_state_lock] 749 */ 750 unsigned long j_flags; 751 752 /** 753 * @j_errno: 754 * 755 * Is there an outstanding uncleared error on the journal (from a prior 756 * abort)? [j_state_lock] 757 */ 758 int j_errno; 759 760 /** 761 * @j_sb_buffer: The first part of the superblock buffer. 762 */ 763 struct buffer_head *j_sb_buffer; 764 765 /** 766 * @j_superblock: The second part of the superblock buffer. 767 */ 768 journal_superblock_t *j_superblock; 769 770 /** 771 * @j_format_version: Version of the superblock format. 772 */ 773 int j_format_version; 774 775 /** 776 * @j_state_lock: Protect the various scalars in the journal. 777 */ 778 rwlock_t j_state_lock; 779 780 /** 781 * @j_barrier_count: 782 * 783 * Number of processes waiting to create a barrier lock [j_state_lock] 784 */ 785 int j_barrier_count; 786 787 /** 788 * @j_barrier: The barrier lock itself. 789 */ 790 struct mutex j_barrier; 791 792 /** 793 * @j_running_transaction: 794 * 795 * Transactions: The current running transaction... 796 * [j_state_lock] [caller holding open handle] 797 */ 798 transaction_t *j_running_transaction; 799 800 /** 801 * @j_committing_transaction: 802 * 803 * the transaction we are pushing to disk 804 * [j_state_lock] [caller holding open handle] 805 */ 806 transaction_t *j_committing_transaction; 807 808 /** 809 * @j_checkpoint_transactions: 810 * 811 * ... and a linked circular list of all transactions waiting for 812 * checkpointing. [j_list_lock] 813 */ 814 transaction_t *j_checkpoint_transactions; 815 816 /** 817 * @j_wait_transaction_locked: 818 * 819 * Wait queue for waiting for a locked transaction to start committing, 820 * or for a barrier lock to be released. 821 */ 822 wait_queue_head_t j_wait_transaction_locked; 823 824 /** 825 * @j_wait_done_commit: Wait queue for waiting for commit to complete. 826 */ 827 wait_queue_head_t j_wait_done_commit; 828 829 /** 830 * @j_wait_commit: Wait queue to trigger commit. 831 */ 832 wait_queue_head_t j_wait_commit; 833 834 /** 835 * @j_wait_updates: Wait queue to wait for updates to complete. 836 */ 837 wait_queue_head_t j_wait_updates; 838 839 /** 840 * @j_wait_reserved: 841 * 842 * Wait queue to wait for reserved buffer credits to drop. 843 */ 844 wait_queue_head_t j_wait_reserved; 845 846 /** 847 * @j_checkpoint_mutex: 848 * 849 * Semaphore for locking against concurrent checkpoints. 850 */ 851 struct mutex j_checkpoint_mutex; 852 853 /** 854 * @j_chkpt_bhs: 855 * 856 * List of buffer heads used by the checkpoint routine. This 857 * was moved from jbd2_log_do_checkpoint() to reduce stack 858 * usage. Access to this array is controlled by the 859 * @j_checkpoint_mutex. [j_checkpoint_mutex] 860 */ 861 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 862 863 /** 864 * @j_head: 865 * 866 * Journal head: identifies the first unused block in the journal. 867 * [j_state_lock] 868 */ 869 unsigned long j_head; 870 871 /** 872 * @j_tail: 873 * 874 * Journal tail: identifies the oldest still-used block in the journal. 875 * [j_state_lock] 876 */ 877 unsigned long j_tail; 878 879 /** 880 * @j_free: 881 * 882 * Journal free: how many free blocks are there in the journal? 883 * [j_state_lock] 884 */ 885 unsigned long j_free; 886 887 /** 888 * @j_first: 889 * 890 * The block number of the first usable block in the journal 891 * [j_state_lock]. 892 */ 893 unsigned long j_first; 894 895 /** 896 * @j_last: 897 * 898 * The block number one beyond the last usable block in the journal 899 * [j_state_lock]. 900 */ 901 unsigned long j_last; 902 903 /** 904 * @j_dev: Device where we store the journal. 905 */ 906 struct block_device *j_dev; 907 908 /** 909 * @j_blocksize: Block size for the location where we store the journal. 910 */ 911 int j_blocksize; 912 913 /** 914 * @j_blk_offset: 915 * 916 * Starting block offset into the device where we store the journal. 917 */ 918 unsigned long long j_blk_offset; 919 920 /** 921 * @j_devname: Journal device name. 922 */ 923 char j_devname[BDEVNAME_SIZE+24]; 924 925 /** 926 * @j_fs_dev: 927 * 928 * Device which holds the client fs. For internal journal this will be 929 * equal to j_dev. 930 */ 931 struct block_device *j_fs_dev; 932 933 /** 934 * @j_maxlen: Total maximum capacity of the journal region on disk. 935 */ 936 unsigned int j_maxlen; 937 938 /** 939 * @j_reserved_credits: 940 * 941 * Number of buffers reserved from the running transaction. 942 */ 943 atomic_t j_reserved_credits; 944 945 /** 946 * @j_list_lock: Protects the buffer lists and internal buffer state. 947 */ 948 spinlock_t j_list_lock; 949 950 /** 951 * @j_inode: 952 * 953 * Optional inode where we store the journal. If present, all 954 * journal block numbers are mapped into this inode via bmap(). 955 */ 956 struct inode *j_inode; 957 958 /** 959 * @j_tail_sequence: 960 * 961 * Sequence number of the oldest transaction in the log [j_state_lock] 962 */ 963 tid_t j_tail_sequence; 964 965 /** 966 * @j_transaction_sequence: 967 * 968 * Sequence number of the next transaction to grant [j_state_lock] 969 */ 970 tid_t j_transaction_sequence; 971 972 /** 973 * @j_commit_sequence: 974 * 975 * Sequence number of the most recently committed transaction 976 * [j_state_lock]. 977 */ 978 tid_t j_commit_sequence; 979 980 /** 981 * @j_commit_request: 982 * 983 * Sequence number of the most recent transaction wanting commit 984 * [j_state_lock] 985 */ 986 tid_t j_commit_request; 987 988 /** 989 * @j_uuid: 990 * 991 * Journal uuid: identifies the object (filesystem, LVM volume etc) 992 * backed by this journal. This will eventually be replaced by an array 993 * of uuids, allowing us to index multiple devices within a single 994 * journal and to perform atomic updates across them. 995 */ 996 __u8 j_uuid[16]; 997 998 /** 999 * @j_task: Pointer to the current commit thread for this journal. 1000 */ 1001 struct task_struct *j_task; 1002 1003 /** 1004 * @j_max_transaction_buffers: 1005 * 1006 * Maximum number of metadata buffers to allow in a single compound 1007 * commit transaction. 1008 */ 1009 int j_max_transaction_buffers; 1010 1011 /** 1012 * @j_commit_interval: 1013 * 1014 * What is the maximum transaction lifetime before we begin a commit? 1015 */ 1016 unsigned long j_commit_interval; 1017 1018 /** 1019 * @j_commit_timer: The timer used to wakeup the commit thread. 1020 */ 1021 struct timer_list j_commit_timer; 1022 1023 /** 1024 * @j_revoke_lock: Protect the revoke table. 1025 */ 1026 spinlock_t j_revoke_lock; 1027 1028 /** 1029 * @j_revoke: 1030 * 1031 * The revoke table - maintains the list of revoked blocks in the 1032 * current transaction. 1033 */ 1034 struct jbd2_revoke_table_s *j_revoke; 1035 1036 /** 1037 * @j_revoke_table: Alternate revoke tables for j_revoke. 1038 */ 1039 struct jbd2_revoke_table_s *j_revoke_table[2]; 1040 1041 /** 1042 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction. 1043 */ 1044 struct buffer_head **j_wbuf; 1045 1046 /** 1047 * @j_wbufsize: 1048 * 1049 * Size of @j_wbuf array. 1050 */ 1051 int j_wbufsize; 1052 1053 /** 1054 * @j_last_sync_writer: 1055 * 1056 * The pid of the last person to run a synchronous operation 1057 * through the journal. 1058 */ 1059 pid_t j_last_sync_writer; 1060 1061 /** 1062 * @j_average_commit_time: 1063 * 1064 * The average amount of time in nanoseconds it takes to commit a 1065 * transaction to disk. [j_state_lock] 1066 */ 1067 u64 j_average_commit_time; 1068 1069 /** 1070 * @j_min_batch_time: 1071 * 1072 * Minimum time that we should wait for additional filesystem operations 1073 * to get batched into a synchronous handle in microseconds. 1074 */ 1075 u32 j_min_batch_time; 1076 1077 /** 1078 * @j_max_batch_time: 1079 * 1080 * Maximum time that we should wait for additional filesystem operations 1081 * to get batched into a synchronous handle in microseconds. 1082 */ 1083 u32 j_max_batch_time; 1084 1085 /** 1086 * @j_commit_callback: 1087 * 1088 * This function is called when a transaction is closed. 1089 */ 1090 void (*j_commit_callback)(journal_t *, 1091 transaction_t *); 1092 1093 /* 1094 * Journal statistics 1095 */ 1096 1097 /** 1098 * @j_history_lock: Protect the transactions statistics history. 1099 */ 1100 spinlock_t j_history_lock; 1101 1102 /** 1103 * @j_proc_entry: procfs entry for the jbd statistics directory. 1104 */ 1105 struct proc_dir_entry *j_proc_entry; 1106 1107 /** 1108 * @j_stats: Overall statistics. 1109 */ 1110 struct transaction_stats_s j_stats; 1111 1112 /** 1113 * @j_failed_commit: Failed journal commit ID. 1114 */ 1115 unsigned int j_failed_commit; 1116 1117 /** 1118 * @j_private: 1119 * 1120 * An opaque pointer to fs-private information. ext3 puts its 1121 * superblock pointer here. 1122 */ 1123 void *j_private; 1124 1125 /** 1126 * @j_chksum_driver: 1127 * 1128 * Reference to checksum algorithm driver via cryptoapi. 1129 */ 1130 struct crypto_shash *j_chksum_driver; 1131 1132 /** 1133 * @j_csum_seed: 1134 * 1135 * Precomputed journal UUID checksum for seeding other checksums. 1136 */ 1137 __u32 j_csum_seed; 1138 1139 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1140 /** 1141 * @j_trans_commit_map: 1142 * 1143 * Lockdep entity to track transaction commit dependencies. Handles 1144 * hold this "lock" for read, when we wait for commit, we acquire the 1145 * "lock" for writing. This matches the properties of jbd2 journalling 1146 * where the running transaction has to wait for all handles to be 1147 * dropped to commit that transaction and also acquiring a handle may 1148 * require transaction commit to finish. 1149 */ 1150 struct lockdep_map j_trans_commit_map; 1151 #endif 1152 }; 1153 1154 #define jbd2_might_wait_for_commit(j) \ 1155 do { \ 1156 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \ 1157 rwsem_release(&j->j_trans_commit_map, 1, _THIS_IP_); \ 1158 } while (0) 1159 1160 /* journal feature predicate functions */ 1161 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1162 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1163 { \ 1164 return ((j)->j_format_version >= 2 && \ 1165 ((j)->j_superblock->s_feature_compat & \ 1166 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1167 } \ 1168 static inline void jbd2_set_feature_##name(journal_t *j) \ 1169 { \ 1170 (j)->j_superblock->s_feature_compat |= \ 1171 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1172 } \ 1173 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1174 { \ 1175 (j)->j_superblock->s_feature_compat &= \ 1176 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1177 } 1178 1179 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1180 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1181 { \ 1182 return ((j)->j_format_version >= 2 && \ 1183 ((j)->j_superblock->s_feature_ro_compat & \ 1184 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1185 } \ 1186 static inline void jbd2_set_feature_##name(journal_t *j) \ 1187 { \ 1188 (j)->j_superblock->s_feature_ro_compat |= \ 1189 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1190 } \ 1191 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1192 { \ 1193 (j)->j_superblock->s_feature_ro_compat &= \ 1194 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1195 } 1196 1197 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1198 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1199 { \ 1200 return ((j)->j_format_version >= 2 && \ 1201 ((j)->j_superblock->s_feature_incompat & \ 1202 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1203 } \ 1204 static inline void jbd2_set_feature_##name(journal_t *j) \ 1205 { \ 1206 (j)->j_superblock->s_feature_incompat |= \ 1207 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1208 } \ 1209 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1210 { \ 1211 (j)->j_superblock->s_feature_incompat &= \ 1212 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1213 } 1214 1215 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1216 1217 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1218 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1219 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1220 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1221 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1222 1223 /* 1224 * Journal flag definitions 1225 */ 1226 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1227 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1228 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1229 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1230 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1231 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1232 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1233 * data write error in ordered 1234 * mode */ 1235 #define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */ 1236 1237 /* 1238 * Function declarations for the journaling transaction and buffer 1239 * management 1240 */ 1241 1242 /* Filing buffers */ 1243 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1244 extern void __jbd2_journal_refile_buffer(struct journal_head *); 1245 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1246 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1247 extern void __journal_free_buffer(struct journal_head *bh); 1248 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1249 extern void __journal_clean_data_list(transaction_t *transaction); 1250 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1251 { 1252 list_add_tail(&bh->b_assoc_buffers, head); 1253 } 1254 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1255 { 1256 list_del_init(&bh->b_assoc_buffers); 1257 } 1258 1259 /* Log buffer allocation */ 1260 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1261 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1262 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1263 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1264 unsigned long *block); 1265 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1266 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1267 1268 /* Commit management */ 1269 extern void jbd2_journal_commit_transaction(journal_t *); 1270 1271 /* Checkpoint list management */ 1272 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1273 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1274 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1275 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1276 1277 1278 /* 1279 * Triggers 1280 */ 1281 1282 struct jbd2_buffer_trigger_type { 1283 /* 1284 * Fired a the moment data to write to the journal are known to be 1285 * stable - so either at the moment b_frozen_data is created or just 1286 * before a buffer is written to the journal. mapped_data is a mapped 1287 * buffer that is the frozen data for commit. 1288 */ 1289 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1290 struct buffer_head *bh, void *mapped_data, 1291 size_t size); 1292 1293 /* 1294 * Fired during journal abort for dirty buffers that will not be 1295 * committed. 1296 */ 1297 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1298 struct buffer_head *bh); 1299 }; 1300 1301 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1302 void *mapped_data, 1303 struct jbd2_buffer_trigger_type *triggers); 1304 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1305 struct jbd2_buffer_trigger_type *triggers); 1306 1307 /* Buffer IO */ 1308 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1309 struct journal_head *jh_in, 1310 struct buffer_head **bh_out, 1311 sector_t blocknr); 1312 1313 /* Transaction locking */ 1314 extern void __wait_on_journal (journal_t *); 1315 1316 /* Transaction cache support */ 1317 extern void jbd2_journal_destroy_transaction_cache(void); 1318 extern int __init jbd2_journal_init_transaction_cache(void); 1319 extern void jbd2_journal_free_transaction(transaction_t *); 1320 1321 /* 1322 * Journal locking. 1323 * 1324 * We need to lock the journal during transaction state changes so that nobody 1325 * ever tries to take a handle on the running transaction while we are in the 1326 * middle of moving it to the commit phase. j_state_lock does this. 1327 * 1328 * Note that the locking is completely interrupt unsafe. We never touch 1329 * journal structures from interrupts. 1330 */ 1331 1332 static inline handle_t *journal_current_handle(void) 1333 { 1334 return current->journal_info; 1335 } 1336 1337 /* The journaling code user interface: 1338 * 1339 * Create and destroy handles 1340 * Register buffer modifications against the current transaction. 1341 */ 1342 1343 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1344 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1345 gfp_t gfp_mask, unsigned int type, 1346 unsigned int line_no); 1347 extern int jbd2_journal_restart(handle_t *, int nblocks); 1348 extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask); 1349 extern int jbd2_journal_start_reserved(handle_t *handle, 1350 unsigned int type, unsigned int line_no); 1351 extern void jbd2_journal_free_reserved(handle_t *handle); 1352 extern int jbd2_journal_extend (handle_t *, int nblocks); 1353 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1354 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1355 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1356 void jbd2_journal_set_triggers(struct buffer_head *, 1357 struct jbd2_buffer_trigger_type *type); 1358 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1359 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1360 extern void journal_sync_buffer (struct buffer_head *); 1361 extern int jbd2_journal_invalidatepage(journal_t *, 1362 struct page *, unsigned int, unsigned int); 1363 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); 1364 extern int jbd2_journal_stop(handle_t *); 1365 extern int jbd2_journal_flush (journal_t *); 1366 extern void jbd2_journal_lock_updates (journal_t *); 1367 extern void jbd2_journal_unlock_updates (journal_t *); 1368 1369 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1370 struct block_device *fs_dev, 1371 unsigned long long start, int len, int bsize); 1372 extern journal_t * jbd2_journal_init_inode (struct inode *); 1373 extern int jbd2_journal_update_format (journal_t *); 1374 extern int jbd2_journal_check_used_features 1375 (journal_t *, unsigned long, unsigned long, unsigned long); 1376 extern int jbd2_journal_check_available_features 1377 (journal_t *, unsigned long, unsigned long, unsigned long); 1378 extern int jbd2_journal_set_features 1379 (journal_t *, unsigned long, unsigned long, unsigned long); 1380 extern void jbd2_journal_clear_features 1381 (journal_t *, unsigned long, unsigned long, unsigned long); 1382 extern int jbd2_journal_load (journal_t *journal); 1383 extern int jbd2_journal_destroy (journal_t *); 1384 extern int jbd2_journal_recover (journal_t *journal); 1385 extern int jbd2_journal_wipe (journal_t *, int); 1386 extern int jbd2_journal_skip_recovery (journal_t *); 1387 extern void jbd2_journal_update_sb_errno(journal_t *); 1388 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1389 unsigned long, int); 1390 extern void __jbd2_journal_abort_hard (journal_t *); 1391 extern void jbd2_journal_abort (journal_t *, int); 1392 extern int jbd2_journal_errno (journal_t *); 1393 extern void jbd2_journal_ack_err (journal_t *); 1394 extern int jbd2_journal_clear_err (journal_t *); 1395 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1396 extern int jbd2_journal_force_commit(journal_t *); 1397 extern int jbd2_journal_force_commit_nested(journal_t *); 1398 extern int jbd2_journal_inode_add_write(handle_t *handle, struct jbd2_inode *inode); 1399 extern int jbd2_journal_inode_add_wait(handle_t *handle, struct jbd2_inode *inode); 1400 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1401 struct jbd2_inode *inode, loff_t new_size); 1402 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1403 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1404 1405 /* 1406 * journal_head management 1407 */ 1408 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1409 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1410 void jbd2_journal_put_journal_head(struct journal_head *jh); 1411 1412 /* 1413 * handle management 1414 */ 1415 extern struct kmem_cache *jbd2_handle_cache; 1416 1417 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1418 { 1419 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1420 } 1421 1422 static inline void jbd2_free_handle(handle_t *handle) 1423 { 1424 kmem_cache_free(jbd2_handle_cache, handle); 1425 } 1426 1427 /* 1428 * jbd2_inode management (optional, for those file systems that want to use 1429 * dynamically allocated jbd2_inode structures) 1430 */ 1431 extern struct kmem_cache *jbd2_inode_cache; 1432 1433 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1434 { 1435 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1436 } 1437 1438 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1439 { 1440 kmem_cache_free(jbd2_inode_cache, jinode); 1441 } 1442 1443 /* Primary revoke support */ 1444 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1445 extern int jbd2_journal_init_revoke(journal_t *, int); 1446 extern void jbd2_journal_destroy_revoke_record_cache(void); 1447 extern void jbd2_journal_destroy_revoke_table_cache(void); 1448 extern int __init jbd2_journal_init_revoke_record_cache(void); 1449 extern int __init jbd2_journal_init_revoke_table_cache(void); 1450 1451 extern void jbd2_journal_destroy_revoke(journal_t *); 1452 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1453 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1454 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1455 struct list_head *log_bufs); 1456 1457 /* Recovery revoke support */ 1458 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1459 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1460 extern void jbd2_journal_clear_revoke(journal_t *); 1461 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1462 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1463 1464 /* 1465 * The log thread user interface: 1466 * 1467 * Request space in the current transaction, and force transaction commit 1468 * transitions on demand. 1469 */ 1470 1471 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1472 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1473 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1474 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1475 int jbd2_transaction_committed(journal_t *journal, tid_t tid); 1476 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1477 int jbd2_log_do_checkpoint(journal_t *journal); 1478 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1479 1480 void __jbd2_log_wait_for_space(journal_t *journal); 1481 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1482 extern int jbd2_cleanup_journal_tail(journal_t *); 1483 1484 /* 1485 * is_journal_abort 1486 * 1487 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1488 * bit, when set, indicates that we have had a fatal error somewhere, 1489 * either inside the journaling layer or indicated to us by the client 1490 * (eg. ext3), and that we and should not commit any further 1491 * transactions. 1492 */ 1493 1494 static inline int is_journal_aborted(journal_t *journal) 1495 { 1496 return journal->j_flags & JBD2_ABORT; 1497 } 1498 1499 static inline int is_handle_aborted(handle_t *handle) 1500 { 1501 if (handle->h_aborted || !handle->h_transaction) 1502 return 1; 1503 return is_journal_aborted(handle->h_transaction->t_journal); 1504 } 1505 1506 static inline void jbd2_journal_abort_handle(handle_t *handle) 1507 { 1508 handle->h_aborted = 1; 1509 } 1510 1511 #endif /* __KERNEL__ */ 1512 1513 /* Comparison functions for transaction IDs: perform comparisons using 1514 * modulo arithmetic so that they work over sequence number wraps. */ 1515 1516 static inline int tid_gt(tid_t x, tid_t y) 1517 { 1518 int difference = (x - y); 1519 return (difference > 0); 1520 } 1521 1522 static inline int tid_geq(tid_t x, tid_t y) 1523 { 1524 int difference = (x - y); 1525 return (difference >= 0); 1526 } 1527 1528 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1529 extern size_t journal_tag_bytes(journal_t *journal); 1530 1531 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1532 { 1533 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1534 } 1535 1536 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1537 { 1538 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1539 journal->j_chksum_driver == NULL); 1540 1541 return journal->j_chksum_driver != NULL; 1542 } 1543 1544 /* 1545 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for 1546 * transaction control blocks. 1547 */ 1548 #define JBD2_CONTROL_BLOCKS_SHIFT 5 1549 1550 /* 1551 * Return the minimum number of blocks which must be free in the journal 1552 * before a new transaction may be started. Must be called under j_state_lock. 1553 */ 1554 static inline int jbd2_space_needed(journal_t *journal) 1555 { 1556 int nblocks = journal->j_max_transaction_buffers; 1557 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT); 1558 } 1559 1560 /* 1561 * Return number of free blocks in the log. Must be called under j_state_lock. 1562 */ 1563 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1564 { 1565 /* Allow for rounding errors */ 1566 unsigned long free = journal->j_free - 32; 1567 1568 if (journal->j_committing_transaction) { 1569 unsigned long committing = atomic_read(&journal-> 1570 j_committing_transaction->t_outstanding_credits); 1571 1572 /* Transaction + control blocks */ 1573 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT); 1574 } 1575 return free; 1576 } 1577 1578 /* 1579 * Definitions which augment the buffer_head layer 1580 */ 1581 1582 /* journaling buffer types */ 1583 #define BJ_None 0 /* Not journaled */ 1584 #define BJ_Metadata 1 /* Normal journaled metadata */ 1585 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1586 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1587 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1588 #define BJ_Types 5 1589 1590 extern int jbd_blocks_per_page(struct inode *inode); 1591 1592 /* JBD uses a CRC32 checksum */ 1593 #define JBD_MAX_CHECKSUM_SIZE 4 1594 1595 static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1596 const void *address, unsigned int length) 1597 { 1598 struct { 1599 struct shash_desc shash; 1600 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1601 } desc; 1602 int err; 1603 1604 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1605 JBD_MAX_CHECKSUM_SIZE); 1606 1607 desc.shash.tfm = journal->j_chksum_driver; 1608 *(u32 *)desc.ctx = crc; 1609 1610 err = crypto_shash_update(&desc.shash, address, length); 1611 BUG_ON(err); 1612 1613 return *(u32 *)desc.ctx; 1614 } 1615 1616 /* Return most recent uncommitted transaction */ 1617 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1618 { 1619 tid_t tid; 1620 1621 read_lock(&journal->j_state_lock); 1622 tid = journal->j_commit_request; 1623 if (journal->j_running_transaction) 1624 tid = journal->j_running_transaction->t_tid; 1625 read_unlock(&journal->j_state_lock); 1626 return tid; 1627 } 1628 1629 #ifdef __KERNEL__ 1630 1631 #define buffer_trace_init(bh) do {} while (0) 1632 #define print_buffer_fields(bh) do {} while (0) 1633 #define print_buffer_trace(bh) do {} while (0) 1634 #define BUFFER_TRACE(bh, info) do {} while (0) 1635 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1636 #define JBUFFER_TRACE(jh, info) do {} while (0) 1637 1638 #endif /* __KERNEL__ */ 1639 1640 #define EFSBADCRC EBADMSG /* Bad CRC detected */ 1641 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1642 1643 #endif /* _LINUX_JBD2_H */ 1644