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