1 /* 2 * linux/include/linux/jbd2.h 3 * 4 * Written by Stephen C. Tweedie <sct@redhat.com> 5 * 6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved 7 * 8 * This file is part of the Linux kernel and is made available under 9 * the terms of the GNU General Public License, version 2, or at your 10 * option, any later version, incorporated herein by reference. 11 * 12 * Definitions for transaction data structures for the buffer cache 13 * filesystem journaling support. 14 */ 15 16 #ifndef _LINUX_JBD2_H 17 #define _LINUX_JBD2_H 18 19 /* Allow this file to be included directly into e2fsprogs */ 20 #ifndef __KERNEL__ 21 #include "jfs_compat.h" 22 #define JBD2_DEBUG 23 #else 24 25 #include <linux/types.h> 26 #include <linux/buffer_head.h> 27 #include <linux/journal-head.h> 28 #include <linux/stddef.h> 29 #include <linux/mutex.h> 30 #include <linux/timer.h> 31 #include <linux/slab.h> 32 #include <linux/bit_spinlock.h> 33 #include <crypto/hash.h> 34 #endif 35 36 #define journal_oom_retry 1 37 38 /* 39 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds 40 * certain classes of error which can occur due to failed IOs. Under 41 * normal use we want ext4 to continue after such errors, because 42 * hardware _can_ fail, but for debugging purposes when running tests on 43 * known-good hardware we may want to trap these errors. 44 */ 45 #undef JBD2_PARANOID_IOFAIL 46 47 /* 48 * The default maximum commit age, in seconds. 49 */ 50 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5 51 52 #ifdef CONFIG_JBD2_DEBUG 53 /* 54 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal 55 * consistency checks. By default we don't do this unless 56 * CONFIG_JBD2_DEBUG is on. 57 */ 58 #define JBD2_EXPENSIVE_CHECKING 59 extern ushort jbd2_journal_enable_debug; 60 void __jbd2_debug(int level, const char *file, const char *func, 61 unsigned int line, const char *fmt, ...); 62 63 #define jbd_debug(n, fmt, a...) \ 64 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a) 65 #else 66 #define jbd_debug(n, fmt, a...) /**/ 67 #endif 68 69 extern void *jbd2_alloc(size_t size, gfp_t flags); 70 extern void jbd2_free(void *ptr, size_t size); 71 72 #define JBD2_MIN_JOURNAL_BLOCKS 1024 73 74 #ifdef __KERNEL__ 75 76 /** 77 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 78 * 79 * All filesystem modifications made by the process go 80 * through this handle. Recursive operations (such as quota operations) 81 * are gathered into a single update. 82 * 83 * The buffer credits field is used to account for journaled buffers 84 * being modified by the running process. To ensure that there is 85 * enough log space for all outstanding operations, we need to limit the 86 * number of outstanding buffers possible at any time. When the 87 * operation completes, any buffer credits not used are credited back to 88 * the transaction, so that at all times we know how many buffers the 89 * outstanding updates on a transaction might possibly touch. 90 * 91 * This is an opaque datatype. 92 **/ 93 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ 94 95 96 /** 97 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 98 * 99 * journal_t is linked to from the fs superblock structure. 100 * 101 * We use the journal_t to keep track of all outstanding transaction 102 * activity on the filesystem, and to manage the state of the log 103 * writing process. 104 * 105 * This is an opaque datatype. 106 **/ 107 typedef struct journal_s journal_t; /* Journal control structure */ 108 #endif 109 110 /* 111 * Internal structures used by the logging mechanism: 112 */ 113 114 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 115 116 /* 117 * On-disk structures 118 */ 119 120 /* 121 * Descriptor block types: 122 */ 123 124 #define JBD2_DESCRIPTOR_BLOCK 1 125 #define JBD2_COMMIT_BLOCK 2 126 #define JBD2_SUPERBLOCK_V1 3 127 #define JBD2_SUPERBLOCK_V2 4 128 #define JBD2_REVOKE_BLOCK 5 129 130 /* 131 * Standard header for all descriptor blocks: 132 */ 133 typedef struct journal_header_s 134 { 135 __be32 h_magic; 136 __be32 h_blocktype; 137 __be32 h_sequence; 138 } journal_header_t; 139 140 /* 141 * Checksum types. 142 */ 143 #define JBD2_CRC32_CHKSUM 1 144 #define JBD2_MD5_CHKSUM 2 145 #define JBD2_SHA1_CHKSUM 3 146 #define JBD2_CRC32C_CHKSUM 4 147 148 #define JBD2_CRC32_CHKSUM_SIZE 4 149 150 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 151 /* 152 * Commit block header for storing transactional checksums: 153 * 154 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* 155 * fields are used to store a checksum of the descriptor and data blocks. 156 * 157 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum 158 * field is used to store crc32c(uuid+commit_block). Each journal metadata 159 * block gets its own checksum, and data block checksums are stored in 160 * journal_block_tag (in the descriptor). The other h_chksum* fields are 161 * not used. 162 * 163 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses 164 * journal_block_tag3_t to store a full 32-bit checksum. Everything else 165 * is the same as v2. 166 * 167 * Checksum v1, v2, and v3 are mutually exclusive features. 168 */ 169 struct commit_header { 170 __be32 h_magic; 171 __be32 h_blocktype; 172 __be32 h_sequence; 173 unsigned char h_chksum_type; 174 unsigned char h_chksum_size; 175 unsigned char h_padding[2]; 176 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 177 __be64 h_commit_sec; 178 __be32 h_commit_nsec; 179 }; 180 181 /* 182 * The block tag: used to describe a single buffer in the journal. 183 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 184 * raw struct shouldn't be used for pointer math or sizeof() - use 185 * journal_tag_bytes(journal) instead to compute this. 186 */ 187 typedef struct journal_block_tag3_s 188 { 189 __be32 t_blocknr; /* The on-disk block number */ 190 __be32 t_flags; /* See below */ 191 __be32 t_blocknr_high; /* most-significant high 32bits. */ 192 __be32 t_checksum; /* crc32c(uuid+seq+block) */ 193 } journal_block_tag3_t; 194 195 typedef struct journal_block_tag_s 196 { 197 __be32 t_blocknr; /* The on-disk block number */ 198 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ 199 __be16 t_flags; /* See below */ 200 __be32 t_blocknr_high; /* most-significant high 32bits. */ 201 } journal_block_tag_t; 202 203 /* Tail of descriptor or revoke block, for checksumming */ 204 struct jbd2_journal_block_tail { 205 __be32 t_checksum; /* crc32c(uuid+descr_block) */ 206 }; 207 208 /* 209 * The revoke descriptor: used on disk to describe a series of blocks to 210 * be revoked from the log 211 */ 212 typedef struct jbd2_journal_revoke_header_s 213 { 214 journal_header_t r_header; 215 __be32 r_count; /* Count of bytes used in the block */ 216 } jbd2_journal_revoke_header_t; 217 218 /* Definitions for the journal tag flags word: */ 219 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 220 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 221 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 222 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 223 224 225 /* 226 * The journal superblock. All fields are in big-endian byte order. 227 */ 228 typedef struct journal_superblock_s 229 { 230 /* 0x0000 */ 231 journal_header_t s_header; 232 233 /* 0x000C */ 234 /* Static information describing the journal */ 235 __be32 s_blocksize; /* journal device blocksize */ 236 __be32 s_maxlen; /* total blocks in journal file */ 237 __be32 s_first; /* first block of log information */ 238 239 /* 0x0018 */ 240 /* Dynamic information describing the current state of the log */ 241 __be32 s_sequence; /* first commit ID expected in log */ 242 __be32 s_start; /* blocknr of start of log */ 243 244 /* 0x0020 */ 245 /* Error value, as set by jbd2_journal_abort(). */ 246 __be32 s_errno; 247 248 /* 0x0024 */ 249 /* Remaining fields are only valid in a version-2 superblock */ 250 __be32 s_feature_compat; /* compatible feature set */ 251 __be32 s_feature_incompat; /* incompatible feature set */ 252 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 253 /* 0x0030 */ 254 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 255 256 /* 0x0040 */ 257 __be32 s_nr_users; /* Nr of filesystems sharing log */ 258 259 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 260 261 /* 0x0048 */ 262 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 263 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 264 265 /* 0x0050 */ 266 __u8 s_checksum_type; /* checksum type */ 267 __u8 s_padding2[3]; 268 __u32 s_padding[42]; 269 __be32 s_checksum; /* crc32c(superblock) */ 270 271 /* 0x0100 */ 272 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 273 /* 0x0400 */ 274 } journal_superblock_t; 275 276 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */ 277 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 278 ((j)->j_format_version >= 2 && \ 279 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 280 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 281 ((j)->j_format_version >= 2 && \ 282 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 283 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 284 ((j)->j_format_version >= 2 && \ 285 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 286 287 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 288 289 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 290 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 291 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 292 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 293 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 294 295 /* See "journal feature predicate functions" below */ 296 297 /* Features known to this kernel version: */ 298 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 299 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 300 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 301 JBD2_FEATURE_INCOMPAT_64BIT | \ 302 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 303 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 304 JBD2_FEATURE_INCOMPAT_CSUM_V3) 305 306 #ifdef __KERNEL__ 307 308 #include <linux/fs.h> 309 #include <linux/sched.h> 310 311 enum jbd_state_bits { 312 BH_JBD /* Has an attached ext3 journal_head */ 313 = BH_PrivateStart, 314 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 315 BH_Freed, /* Has been freed (truncated) */ 316 BH_Revoked, /* Has been revoked from the log */ 317 BH_RevokeValid, /* Revoked flag is valid */ 318 BH_JBDDirty, /* Is dirty but journaled */ 319 BH_State, /* Pins most journal_head state */ 320 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 321 BH_Shadow, /* IO on shadow buffer is running */ 322 BH_Verified, /* Metadata block has been verified ok */ 323 BH_JBDPrivateStart, /* First bit available for private use by FS */ 324 }; 325 326 BUFFER_FNS(JBD, jbd) 327 BUFFER_FNS(JWrite, jwrite) 328 BUFFER_FNS(JBDDirty, jbddirty) 329 TAS_BUFFER_FNS(JBDDirty, jbddirty) 330 BUFFER_FNS(Revoked, revoked) 331 TAS_BUFFER_FNS(Revoked, revoked) 332 BUFFER_FNS(RevokeValid, revokevalid) 333 TAS_BUFFER_FNS(RevokeValid, revokevalid) 334 BUFFER_FNS(Freed, freed) 335 BUFFER_FNS(Shadow, shadow) 336 BUFFER_FNS(Verified, verified) 337 338 static inline struct buffer_head *jh2bh(struct journal_head *jh) 339 { 340 return jh->b_bh; 341 } 342 343 static inline struct journal_head *bh2jh(struct buffer_head *bh) 344 { 345 return bh->b_private; 346 } 347 348 static inline void jbd_lock_bh_state(struct buffer_head *bh) 349 { 350 bit_spin_lock(BH_State, &bh->b_state); 351 } 352 353 static inline int jbd_trylock_bh_state(struct buffer_head *bh) 354 { 355 return bit_spin_trylock(BH_State, &bh->b_state); 356 } 357 358 static inline int jbd_is_locked_bh_state(struct buffer_head *bh) 359 { 360 return bit_spin_is_locked(BH_State, &bh->b_state); 361 } 362 363 static inline void jbd_unlock_bh_state(struct buffer_head *bh) 364 { 365 bit_spin_unlock(BH_State, &bh->b_state); 366 } 367 368 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 369 { 370 bit_spin_lock(BH_JournalHead, &bh->b_state); 371 } 372 373 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 374 { 375 bit_spin_unlock(BH_JournalHead, &bh->b_state); 376 } 377 378 #define J_ASSERT(assert) BUG_ON(!(assert)) 379 380 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 381 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 382 383 #if defined(JBD2_PARANOID_IOFAIL) 384 #define J_EXPECT(expr, why...) J_ASSERT(expr) 385 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 386 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 387 #else 388 #define __journal_expect(expr, why...) \ 389 ({ \ 390 int val = (expr); \ 391 if (!val) { \ 392 printk(KERN_ERR \ 393 "JBD2 unexpected failure: %s: %s;\n", \ 394 __func__, #expr); \ 395 printk(KERN_ERR why "\n"); \ 396 } \ 397 val; \ 398 }) 399 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 400 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 401 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 402 #endif 403 404 /* Flags in jbd_inode->i_flags */ 405 #define __JI_COMMIT_RUNNING 0 406 #define __JI_WRITE_DATA 1 407 #define __JI_WAIT_DATA 2 408 409 /* 410 * Commit of the inode data in progress. We use this flag to protect us from 411 * concurrent deletion of inode. We cannot use reference to inode for this 412 * since we cannot afford doing last iput() on behalf of kjournald 413 */ 414 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 415 /* Write allocated dirty buffers in this inode before commit */ 416 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA) 417 /* Wait for outstanding data writes for this inode before commit */ 418 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA) 419 420 /** 421 * struct jbd_inode is the structure linking inodes in ordered mode 422 * present in a transaction so that we can sync them during commit. 423 */ 424 struct jbd2_inode { 425 /* Which transaction does this inode belong to? Either the running 426 * transaction or the committing one. [j_list_lock] */ 427 transaction_t *i_transaction; 428 429 /* Pointer to the running transaction modifying inode's data in case 430 * there is already a committing transaction touching it. [j_list_lock] */ 431 transaction_t *i_next_transaction; 432 433 /* List of inodes in the i_transaction [j_list_lock] */ 434 struct list_head i_list; 435 436 /* VFS inode this inode belongs to [constant during the lifetime 437 * of the structure] */ 438 struct inode *i_vfs_inode; 439 440 /* Flags of inode [j_list_lock] */ 441 unsigned long i_flags; 442 }; 443 444 struct jbd2_revoke_table_s; 445 446 /** 447 * struct handle_s - The handle_s type is the concrete type associated with 448 * handle_t. 449 * @h_transaction: Which compound transaction is this update a part of? 450 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. 451 * @h_ref: Reference count on this handle 452 * @h_err: Field for caller's use to track errors through large fs operations 453 * @h_sync: flag for sync-on-close 454 * @h_jdata: flag to force data journaling 455 * @h_aborted: flag indicating fatal error on handle 456 **/ 457 458 /* Docbook can't yet cope with the bit fields, but will leave the documentation 459 * in so it can be fixed later. 460 */ 461 462 struct jbd2_journal_handle 463 { 464 union { 465 /* Which compound transaction is this update a part of? */ 466 transaction_t *h_transaction; 467 /* Which journal handle belongs to - used iff h_reserved set */ 468 journal_t *h_journal; 469 }; 470 471 /* Handle reserved for finishing the logical operation */ 472 handle_t *h_rsv_handle; 473 474 /* Number of remaining buffers we are allowed to dirty: */ 475 int h_buffer_credits; 476 477 /* Reference count on this handle */ 478 int h_ref; 479 480 /* Field for caller's use to track errors through large fs */ 481 /* operations */ 482 int h_err; 483 484 /* Flags [no locking] */ 485 unsigned int h_sync: 1; /* sync-on-close */ 486 unsigned int h_jdata: 1; /* force data journaling */ 487 unsigned int h_reserved: 1; /* handle with reserved credits */ 488 unsigned int h_aborted: 1; /* fatal error on handle */ 489 unsigned int h_type: 8; /* for handle statistics */ 490 unsigned int h_line_no: 16; /* for handle statistics */ 491 492 unsigned long h_start_jiffies; 493 unsigned int h_requested_credits; 494 }; 495 496 497 /* 498 * Some stats for checkpoint phase 499 */ 500 struct transaction_chp_stats_s { 501 unsigned long cs_chp_time; 502 __u32 cs_forced_to_close; 503 __u32 cs_written; 504 __u32 cs_dropped; 505 }; 506 507 /* The transaction_t type is the guts of the journaling mechanism. It 508 * tracks a compound transaction through its various states: 509 * 510 * RUNNING: accepting new updates 511 * LOCKED: Updates still running but we don't accept new ones 512 * RUNDOWN: Updates are tidying up but have finished requesting 513 * new buffers to modify (state not used for now) 514 * FLUSH: All updates complete, but we are still writing to disk 515 * COMMIT: All data on disk, writing commit record 516 * FINISHED: We still have to keep the transaction for checkpointing. 517 * 518 * The transaction keeps track of all of the buffers modified by a 519 * running transaction, and all of the buffers committed but not yet 520 * flushed to home for finished transactions. 521 */ 522 523 /* 524 * Lock ranking: 525 * 526 * j_list_lock 527 * ->jbd_lock_bh_journal_head() (This is "innermost") 528 * 529 * j_state_lock 530 * ->jbd_lock_bh_state() 531 * 532 * jbd_lock_bh_state() 533 * ->j_list_lock 534 * 535 * j_state_lock 536 * ->t_handle_lock 537 * 538 * j_state_lock 539 * ->j_list_lock (journal_unmap_buffer) 540 * 541 */ 542 543 struct transaction_s 544 { 545 /* Pointer to the journal for this transaction. [no locking] */ 546 journal_t *t_journal; 547 548 /* Sequence number for this transaction [no locking] */ 549 tid_t t_tid; 550 551 /* 552 * Transaction's current state 553 * [no locking - only kjournald2 alters this] 554 * [j_list_lock] guards transition of a transaction into T_FINISHED 555 * state and subsequent call of __jbd2_journal_drop_transaction() 556 * FIXME: needs barriers 557 * KLUDGE: [use j_state_lock] 558 */ 559 enum { 560 T_RUNNING, 561 T_LOCKED, 562 T_FLUSH, 563 T_COMMIT, 564 T_COMMIT_DFLUSH, 565 T_COMMIT_JFLUSH, 566 T_COMMIT_CALLBACK, 567 T_FINISHED 568 } t_state; 569 570 /* 571 * Where in the log does this transaction's commit start? [no locking] 572 */ 573 unsigned long t_log_start; 574 575 /* Number of buffers on the t_buffers list [j_list_lock] */ 576 int t_nr_buffers; 577 578 /* 579 * Doubly-linked circular list of all buffers reserved but not yet 580 * modified by this transaction [j_list_lock] 581 */ 582 struct journal_head *t_reserved_list; 583 584 /* 585 * Doubly-linked circular list of all metadata buffers owned by this 586 * transaction [j_list_lock] 587 */ 588 struct journal_head *t_buffers; 589 590 /* 591 * Doubly-linked circular list of all forget buffers (superseded 592 * buffers which we can un-checkpoint once this transaction commits) 593 * [j_list_lock] 594 */ 595 struct journal_head *t_forget; 596 597 /* 598 * Doubly-linked circular list of all buffers still to be flushed before 599 * this transaction can be checkpointed. [j_list_lock] 600 */ 601 struct journal_head *t_checkpoint_list; 602 603 /* 604 * Doubly-linked circular list of all buffers submitted for IO while 605 * checkpointing. [j_list_lock] 606 */ 607 struct journal_head *t_checkpoint_io_list; 608 609 /* 610 * Doubly-linked circular list of metadata buffers being shadowed by log 611 * IO. The IO buffers on the iobuf list and the shadow buffers on this 612 * list match each other one for one at all times. [j_list_lock] 613 */ 614 struct journal_head *t_shadow_list; 615 616 /* 617 * List of inodes whose data we've modified in data=ordered mode. 618 * [j_list_lock] 619 */ 620 struct list_head t_inode_list; 621 622 /* 623 * Protects info related to handles 624 */ 625 spinlock_t t_handle_lock; 626 627 /* 628 * Longest time some handle had to wait for running transaction 629 */ 630 unsigned long t_max_wait; 631 632 /* 633 * When transaction started 634 */ 635 unsigned long t_start; 636 637 /* 638 * When commit was requested 639 */ 640 unsigned long t_requested; 641 642 /* 643 * Checkpointing stats [j_checkpoint_sem] 644 */ 645 struct transaction_chp_stats_s t_chp_stats; 646 647 /* 648 * Number of outstanding updates running on this transaction 649 * [t_handle_lock] 650 */ 651 atomic_t t_updates; 652 653 /* 654 * Number of buffers reserved for use by all handles in this transaction 655 * handle but not yet modified. [t_handle_lock] 656 */ 657 atomic_t t_outstanding_credits; 658 659 /* 660 * Forward and backward links for the circular list of all transactions 661 * awaiting checkpoint. [j_list_lock] 662 */ 663 transaction_t *t_cpnext, *t_cpprev; 664 665 /* 666 * When will the transaction expire (become due for commit), in jiffies? 667 * [no locking] 668 */ 669 unsigned long t_expires; 670 671 /* 672 * When this transaction started, in nanoseconds [no locking] 673 */ 674 ktime_t t_start_time; 675 676 /* 677 * How many handles used this transaction? [t_handle_lock] 678 */ 679 atomic_t t_handle_count; 680 681 /* 682 * This transaction is being forced and some process is 683 * waiting for it to finish. 684 */ 685 unsigned int t_synchronous_commit:1; 686 687 /* Disk flush needs to be sent to fs partition [no locking] */ 688 int t_need_data_flush; 689 690 /* 691 * For use by the filesystem to store fs-specific data 692 * structures associated with the transaction 693 */ 694 struct list_head t_private_list; 695 }; 696 697 struct transaction_run_stats_s { 698 unsigned long rs_wait; 699 unsigned long rs_request_delay; 700 unsigned long rs_running; 701 unsigned long rs_locked; 702 unsigned long rs_flushing; 703 unsigned long rs_logging; 704 705 __u32 rs_handle_count; 706 __u32 rs_blocks; 707 __u32 rs_blocks_logged; 708 }; 709 710 struct transaction_stats_s { 711 unsigned long ts_tid; 712 unsigned long ts_requested; 713 struct transaction_run_stats_s run; 714 }; 715 716 static inline unsigned long 717 jbd2_time_diff(unsigned long start, unsigned long end) 718 { 719 if (end >= start) 720 return end - start; 721 722 return end + (MAX_JIFFY_OFFSET - start); 723 } 724 725 #define JBD2_NR_BATCH 64 726 727 /** 728 * struct journal_s - The journal_s type is the concrete type associated with 729 * journal_t. 730 * @j_flags: General journaling state flags 731 * @j_errno: Is there an outstanding uncleared error on the journal (from a 732 * prior abort)? 733 * @j_sb_buffer: First part of superblock buffer 734 * @j_superblock: Second part of superblock buffer 735 * @j_format_version: Version of the superblock format 736 * @j_state_lock: Protect the various scalars in the journal 737 * @j_barrier_count: Number of processes waiting to create a barrier lock 738 * @j_barrier: The barrier lock itself 739 * @j_running_transaction: The current running transaction.. 740 * @j_committing_transaction: the transaction we are pushing to disk 741 * @j_checkpoint_transactions: a linked circular list of all transactions 742 * waiting for checkpointing 743 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction 744 * to start committing, or for a barrier lock to be released 745 * @j_wait_done_commit: Wait queue for waiting for commit to complete 746 * @j_wait_commit: Wait queue to trigger commit 747 * @j_wait_updates: Wait queue to wait for updates to complete 748 * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop 749 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints 750 * @j_head: Journal head - identifies the first unused block in the journal 751 * @j_tail: Journal tail - identifies the oldest still-used block in the 752 * journal. 753 * @j_free: Journal free - how many free blocks are there in the journal? 754 * @j_first: The block number of the first usable block 755 * @j_last: The block number one beyond the last usable block 756 * @j_dev: Device where we store the journal 757 * @j_blocksize: blocksize for the location where we store the journal. 758 * @j_blk_offset: starting block offset for into the device where we store the 759 * journal 760 * @j_fs_dev: Device which holds the client fs. For internal journal this will 761 * be equal to j_dev 762 * @j_reserved_credits: Number of buffers reserved from the running transaction 763 * @j_maxlen: Total maximum capacity of the journal region on disk. 764 * @j_list_lock: Protects the buffer lists and internal buffer state. 765 * @j_inode: Optional inode where we store the journal. If present, all journal 766 * block numbers are mapped into this inode via bmap(). 767 * @j_tail_sequence: Sequence number of the oldest transaction in the log 768 * @j_transaction_sequence: Sequence number of the next transaction to grant 769 * @j_commit_sequence: Sequence number of the most recently committed 770 * transaction 771 * @j_commit_request: Sequence number of the most recent transaction wanting 772 * commit 773 * @j_uuid: Uuid of client object. 774 * @j_task: Pointer to the current commit thread for this journal 775 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a 776 * single compound commit transaction 777 * @j_commit_interval: What is the maximum transaction lifetime before we begin 778 * a commit? 779 * @j_commit_timer: The timer used to wakeup the commit thread 780 * @j_revoke_lock: Protect the revoke table 781 * @j_revoke: The revoke table - maintains the list of revoked blocks in the 782 * current transaction. 783 * @j_revoke_table: alternate revoke tables for j_revoke 784 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction 785 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the 786 * number that will fit in j_blocksize 787 * @j_last_sync_writer: most recent pid which did a synchronous write 788 * @j_history_lock: Protect the transactions statistics history 789 * @j_proc_entry: procfs entry for the jbd statistics directory 790 * @j_stats: Overall statistics 791 * @j_private: An opaque pointer to fs-private information. 792 * @j_trans_commit_map: Lockdep entity to track transaction commit dependencies 793 */ 794 795 struct journal_s 796 { 797 /* General journaling state flags [j_state_lock] */ 798 unsigned long j_flags; 799 800 /* 801 * Is there an outstanding uncleared error on the journal (from a prior 802 * abort)? [j_state_lock] 803 */ 804 int j_errno; 805 806 /* The superblock buffer */ 807 struct buffer_head *j_sb_buffer; 808 journal_superblock_t *j_superblock; 809 810 /* Version of the superblock format */ 811 int j_format_version; 812 813 /* 814 * Protect the various scalars in the journal 815 */ 816 rwlock_t j_state_lock; 817 818 /* 819 * Number of processes waiting to create a barrier lock [j_state_lock] 820 */ 821 int j_barrier_count; 822 823 /* The barrier lock itself */ 824 struct mutex j_barrier; 825 826 /* 827 * Transactions: The current running transaction... 828 * [j_state_lock] [caller holding open handle] 829 */ 830 transaction_t *j_running_transaction; 831 832 /* 833 * the transaction we are pushing to disk 834 * [j_state_lock] [caller holding open handle] 835 */ 836 transaction_t *j_committing_transaction; 837 838 /* 839 * ... and a linked circular list of all transactions waiting for 840 * checkpointing. [j_list_lock] 841 */ 842 transaction_t *j_checkpoint_transactions; 843 844 /* 845 * Wait queue for waiting for a locked transaction to start committing, 846 * or for a barrier lock to be released 847 */ 848 wait_queue_head_t j_wait_transaction_locked; 849 850 /* Wait queue for waiting for commit to complete */ 851 wait_queue_head_t j_wait_done_commit; 852 853 /* Wait queue to trigger commit */ 854 wait_queue_head_t j_wait_commit; 855 856 /* Wait queue to wait for updates to complete */ 857 wait_queue_head_t j_wait_updates; 858 859 /* Wait queue to wait for reserved buffer credits to drop */ 860 wait_queue_head_t j_wait_reserved; 861 862 /* Semaphore for locking against concurrent checkpoints */ 863 struct mutex j_checkpoint_mutex; 864 865 /* 866 * List of buffer heads used by the checkpoint routine. This 867 * was moved from jbd2_log_do_checkpoint() to reduce stack 868 * usage. Access to this array is controlled by the 869 * j_checkpoint_mutex. [j_checkpoint_mutex] 870 */ 871 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 872 873 /* 874 * Journal head: identifies the first unused block in the journal. 875 * [j_state_lock] 876 */ 877 unsigned long j_head; 878 879 /* 880 * Journal tail: identifies the oldest still-used block in the journal. 881 * [j_state_lock] 882 */ 883 unsigned long j_tail; 884 885 /* 886 * Journal free: how many free blocks are there in the journal? 887 * [j_state_lock] 888 */ 889 unsigned long j_free; 890 891 /* 892 * Journal start and end: the block numbers of the first usable block 893 * and one beyond the last usable block in the journal. [j_state_lock] 894 */ 895 unsigned long j_first; 896 unsigned long j_last; 897 898 /* 899 * Device, blocksize and starting block offset for the location where we 900 * store the journal. 901 */ 902 struct block_device *j_dev; 903 int j_blocksize; 904 unsigned long long j_blk_offset; 905 char j_devname[BDEVNAME_SIZE+24]; 906 907 /* 908 * Device which holds the client fs. For internal journal this will be 909 * equal to j_dev. 910 */ 911 struct block_device *j_fs_dev; 912 913 /* Total maximum capacity of the journal region on disk. */ 914 unsigned int j_maxlen; 915 916 /* Number of buffers reserved from the running transaction */ 917 atomic_t j_reserved_credits; 918 919 /* 920 * Protects the buffer lists and internal buffer state. 921 */ 922 spinlock_t j_list_lock; 923 924 /* Optional inode where we store the journal. If present, all */ 925 /* journal block numbers are mapped into this inode via */ 926 /* bmap(). */ 927 struct inode *j_inode; 928 929 /* 930 * Sequence number of the oldest transaction in the log [j_state_lock] 931 */ 932 tid_t j_tail_sequence; 933 934 /* 935 * Sequence number of the next transaction to grant [j_state_lock] 936 */ 937 tid_t j_transaction_sequence; 938 939 /* 940 * Sequence number of the most recently committed transaction 941 * [j_state_lock]. 942 */ 943 tid_t j_commit_sequence; 944 945 /* 946 * Sequence number of the most recent transaction wanting commit 947 * [j_state_lock] 948 */ 949 tid_t j_commit_request; 950 951 /* 952 * Journal uuid: identifies the object (filesystem, LVM volume etc) 953 * backed by this journal. This will eventually be replaced by an array 954 * of uuids, allowing us to index multiple devices within a single 955 * journal and to perform atomic updates across them. 956 */ 957 __u8 j_uuid[16]; 958 959 /* Pointer to the current commit thread for this journal */ 960 struct task_struct *j_task; 961 962 /* 963 * Maximum number of metadata buffers to allow in a single compound 964 * commit transaction 965 */ 966 int j_max_transaction_buffers; 967 968 /* 969 * What is the maximum transaction lifetime before we begin a commit? 970 */ 971 unsigned long j_commit_interval; 972 973 /* The timer used to wakeup the commit thread: */ 974 struct timer_list j_commit_timer; 975 976 /* 977 * The revoke table: maintains the list of revoked blocks in the 978 * current transaction. [j_revoke_lock] 979 */ 980 spinlock_t j_revoke_lock; 981 struct jbd2_revoke_table_s *j_revoke; 982 struct jbd2_revoke_table_s *j_revoke_table[2]; 983 984 /* 985 * array of bhs for jbd2_journal_commit_transaction 986 */ 987 struct buffer_head **j_wbuf; 988 int j_wbufsize; 989 990 /* 991 * this is the pid of hte last person to run a synchronous operation 992 * through the journal 993 */ 994 pid_t j_last_sync_writer; 995 996 /* 997 * the average amount of time in nanoseconds it takes to commit a 998 * transaction to disk. [j_state_lock] 999 */ 1000 u64 j_average_commit_time; 1001 1002 /* 1003 * minimum and maximum times that we should wait for 1004 * additional filesystem operations to get batched into a 1005 * synchronous handle in microseconds 1006 */ 1007 u32 j_min_batch_time; 1008 u32 j_max_batch_time; 1009 1010 /* This function is called when a transaction is closed */ 1011 void (*j_commit_callback)(journal_t *, 1012 transaction_t *); 1013 1014 /* 1015 * Journal statistics 1016 */ 1017 spinlock_t j_history_lock; 1018 struct proc_dir_entry *j_proc_entry; 1019 struct transaction_stats_s j_stats; 1020 1021 /* Failed journal commit ID */ 1022 unsigned int j_failed_commit; 1023 1024 /* 1025 * An opaque pointer to fs-private information. ext3 puts its 1026 * superblock pointer here 1027 */ 1028 void *j_private; 1029 1030 /* Reference to checksum algorithm driver via cryptoapi */ 1031 struct crypto_shash *j_chksum_driver; 1032 1033 /* Precomputed journal UUID checksum for seeding other checksums */ 1034 __u32 j_csum_seed; 1035 1036 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1037 /* 1038 * Lockdep entity to track transaction commit dependencies. Handles 1039 * hold this "lock" for read, when we wait for commit, we acquire the 1040 * "lock" for writing. This matches the properties of jbd2 journalling 1041 * where the running transaction has to wait for all handles to be 1042 * dropped to commit that transaction and also acquiring a handle may 1043 * require transaction commit to finish. 1044 */ 1045 struct lockdep_map j_trans_commit_map; 1046 #endif 1047 }; 1048 1049 #define jbd2_might_wait_for_commit(j) \ 1050 do { \ 1051 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \ 1052 rwsem_release(&j->j_trans_commit_map, 1, _THIS_IP_); \ 1053 } while (0) 1054 1055 /* journal feature predicate functions */ 1056 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1057 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1058 { \ 1059 return ((j)->j_format_version >= 2 && \ 1060 ((j)->j_superblock->s_feature_compat & \ 1061 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1062 } \ 1063 static inline void jbd2_set_feature_##name(journal_t *j) \ 1064 { \ 1065 (j)->j_superblock->s_feature_compat |= \ 1066 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1067 } \ 1068 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1069 { \ 1070 (j)->j_superblock->s_feature_compat &= \ 1071 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1072 } 1073 1074 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1075 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1076 { \ 1077 return ((j)->j_format_version >= 2 && \ 1078 ((j)->j_superblock->s_feature_ro_compat & \ 1079 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1080 } \ 1081 static inline void jbd2_set_feature_##name(journal_t *j) \ 1082 { \ 1083 (j)->j_superblock->s_feature_ro_compat |= \ 1084 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1085 } \ 1086 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1087 { \ 1088 (j)->j_superblock->s_feature_ro_compat &= \ 1089 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1090 } 1091 1092 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1093 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1094 { \ 1095 return ((j)->j_format_version >= 2 && \ 1096 ((j)->j_superblock->s_feature_incompat & \ 1097 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1098 } \ 1099 static inline void jbd2_set_feature_##name(journal_t *j) \ 1100 { \ 1101 (j)->j_superblock->s_feature_incompat |= \ 1102 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1103 } \ 1104 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1105 { \ 1106 (j)->j_superblock->s_feature_incompat &= \ 1107 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1108 } 1109 1110 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1111 1112 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1113 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1114 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1115 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1116 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1117 1118 /* 1119 * Journal flag definitions 1120 */ 1121 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1122 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1123 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1124 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1125 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1126 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1127 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1128 * data write error in ordered 1129 * mode */ 1130 #define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */ 1131 1132 /* 1133 * Function declarations for the journaling transaction and buffer 1134 * management 1135 */ 1136 1137 /* Filing buffers */ 1138 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1139 extern void __jbd2_journal_refile_buffer(struct journal_head *); 1140 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1141 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1142 extern void __journal_free_buffer(struct journal_head *bh); 1143 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1144 extern void __journal_clean_data_list(transaction_t *transaction); 1145 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1146 { 1147 list_add_tail(&bh->b_assoc_buffers, head); 1148 } 1149 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1150 { 1151 list_del_init(&bh->b_assoc_buffers); 1152 } 1153 1154 /* Log buffer allocation */ 1155 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1156 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1157 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1158 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1159 unsigned long *block); 1160 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1161 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1162 1163 /* Commit management */ 1164 extern void jbd2_journal_commit_transaction(journal_t *); 1165 1166 /* Checkpoint list management */ 1167 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1168 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1169 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1170 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1171 1172 1173 /* 1174 * Triggers 1175 */ 1176 1177 struct jbd2_buffer_trigger_type { 1178 /* 1179 * Fired a the moment data to write to the journal are known to be 1180 * stable - so either at the moment b_frozen_data is created or just 1181 * before a buffer is written to the journal. mapped_data is a mapped 1182 * buffer that is the frozen data for commit. 1183 */ 1184 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1185 struct buffer_head *bh, void *mapped_data, 1186 size_t size); 1187 1188 /* 1189 * Fired during journal abort for dirty buffers that will not be 1190 * committed. 1191 */ 1192 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1193 struct buffer_head *bh); 1194 }; 1195 1196 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1197 void *mapped_data, 1198 struct jbd2_buffer_trigger_type *triggers); 1199 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1200 struct jbd2_buffer_trigger_type *triggers); 1201 1202 /* Buffer IO */ 1203 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1204 struct journal_head *jh_in, 1205 struct buffer_head **bh_out, 1206 sector_t blocknr); 1207 1208 /* Transaction locking */ 1209 extern void __wait_on_journal (journal_t *); 1210 1211 /* Transaction cache support */ 1212 extern void jbd2_journal_destroy_transaction_cache(void); 1213 extern int jbd2_journal_init_transaction_cache(void); 1214 extern void jbd2_journal_free_transaction(transaction_t *); 1215 1216 /* 1217 * Journal locking. 1218 * 1219 * We need to lock the journal during transaction state changes so that nobody 1220 * ever tries to take a handle on the running transaction while we are in the 1221 * middle of moving it to the commit phase. j_state_lock does this. 1222 * 1223 * Note that the locking is completely interrupt unsafe. We never touch 1224 * journal structures from interrupts. 1225 */ 1226 1227 static inline handle_t *journal_current_handle(void) 1228 { 1229 return current->journal_info; 1230 } 1231 1232 /* The journaling code user interface: 1233 * 1234 * Create and destroy handles 1235 * Register buffer modifications against the current transaction. 1236 */ 1237 1238 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1239 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1240 gfp_t gfp_mask, unsigned int type, 1241 unsigned int line_no); 1242 extern int jbd2_journal_restart(handle_t *, int nblocks); 1243 extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask); 1244 extern int jbd2_journal_start_reserved(handle_t *handle, 1245 unsigned int type, unsigned int line_no); 1246 extern void jbd2_journal_free_reserved(handle_t *handle); 1247 extern int jbd2_journal_extend (handle_t *, int nblocks); 1248 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1249 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1250 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1251 void jbd2_journal_set_triggers(struct buffer_head *, 1252 struct jbd2_buffer_trigger_type *type); 1253 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1254 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1255 extern void journal_sync_buffer (struct buffer_head *); 1256 extern int jbd2_journal_invalidatepage(journal_t *, 1257 struct page *, unsigned int, unsigned int); 1258 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); 1259 extern int jbd2_journal_stop(handle_t *); 1260 extern int jbd2_journal_flush (journal_t *); 1261 extern void jbd2_journal_lock_updates (journal_t *); 1262 extern void jbd2_journal_unlock_updates (journal_t *); 1263 1264 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1265 struct block_device *fs_dev, 1266 unsigned long long start, int len, int bsize); 1267 extern journal_t * jbd2_journal_init_inode (struct inode *); 1268 extern int jbd2_journal_update_format (journal_t *); 1269 extern int jbd2_journal_check_used_features 1270 (journal_t *, unsigned long, unsigned long, unsigned long); 1271 extern int jbd2_journal_check_available_features 1272 (journal_t *, unsigned long, unsigned long, unsigned long); 1273 extern int jbd2_journal_set_features 1274 (journal_t *, unsigned long, unsigned long, unsigned long); 1275 extern void jbd2_journal_clear_features 1276 (journal_t *, unsigned long, unsigned long, unsigned long); 1277 extern int jbd2_journal_load (journal_t *journal); 1278 extern int jbd2_journal_destroy (journal_t *); 1279 extern int jbd2_journal_recover (journal_t *journal); 1280 extern int jbd2_journal_wipe (journal_t *, int); 1281 extern int jbd2_journal_skip_recovery (journal_t *); 1282 extern void jbd2_journal_update_sb_errno(journal_t *); 1283 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1284 unsigned long, int); 1285 extern void __jbd2_journal_abort_hard (journal_t *); 1286 extern void jbd2_journal_abort (journal_t *, int); 1287 extern int jbd2_journal_errno (journal_t *); 1288 extern void jbd2_journal_ack_err (journal_t *); 1289 extern int jbd2_journal_clear_err (journal_t *); 1290 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1291 extern int jbd2_journal_force_commit(journal_t *); 1292 extern int jbd2_journal_force_commit_nested(journal_t *); 1293 extern int jbd2_journal_inode_add_write(handle_t *handle, struct jbd2_inode *inode); 1294 extern int jbd2_journal_inode_add_wait(handle_t *handle, struct jbd2_inode *inode); 1295 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1296 struct jbd2_inode *inode, loff_t new_size); 1297 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1298 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1299 1300 /* 1301 * journal_head management 1302 */ 1303 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1304 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1305 void jbd2_journal_put_journal_head(struct journal_head *jh); 1306 1307 /* 1308 * handle management 1309 */ 1310 extern struct kmem_cache *jbd2_handle_cache; 1311 1312 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1313 { 1314 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1315 } 1316 1317 static inline void jbd2_free_handle(handle_t *handle) 1318 { 1319 kmem_cache_free(jbd2_handle_cache, handle); 1320 } 1321 1322 /* 1323 * jbd2_inode management (optional, for those file systems that want to use 1324 * dynamically allocated jbd2_inode structures) 1325 */ 1326 extern struct kmem_cache *jbd2_inode_cache; 1327 1328 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1329 { 1330 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1331 } 1332 1333 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1334 { 1335 kmem_cache_free(jbd2_inode_cache, jinode); 1336 } 1337 1338 /* Primary revoke support */ 1339 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1340 extern int jbd2_journal_init_revoke(journal_t *, int); 1341 extern void jbd2_journal_destroy_revoke_caches(void); 1342 extern int jbd2_journal_init_revoke_caches(void); 1343 1344 extern void jbd2_journal_destroy_revoke(journal_t *); 1345 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1346 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1347 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1348 struct list_head *log_bufs); 1349 1350 /* Recovery revoke support */ 1351 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1352 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1353 extern void jbd2_journal_clear_revoke(journal_t *); 1354 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1355 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1356 1357 /* 1358 * The log thread user interface: 1359 * 1360 * Request space in the current transaction, and force transaction commit 1361 * transitions on demand. 1362 */ 1363 1364 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1365 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1366 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1367 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1368 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1369 int jbd2_log_do_checkpoint(journal_t *journal); 1370 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1371 1372 void __jbd2_log_wait_for_space(journal_t *journal); 1373 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1374 extern int jbd2_cleanup_journal_tail(journal_t *); 1375 1376 /* 1377 * is_journal_abort 1378 * 1379 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1380 * bit, when set, indicates that we have had a fatal error somewhere, 1381 * either inside the journaling layer or indicated to us by the client 1382 * (eg. ext3), and that we and should not commit any further 1383 * transactions. 1384 */ 1385 1386 static inline int is_journal_aborted(journal_t *journal) 1387 { 1388 return journal->j_flags & JBD2_ABORT; 1389 } 1390 1391 static inline int is_handle_aborted(handle_t *handle) 1392 { 1393 if (handle->h_aborted || !handle->h_transaction) 1394 return 1; 1395 return is_journal_aborted(handle->h_transaction->t_journal); 1396 } 1397 1398 static inline void jbd2_journal_abort_handle(handle_t *handle) 1399 { 1400 handle->h_aborted = 1; 1401 } 1402 1403 #endif /* __KERNEL__ */ 1404 1405 /* Comparison functions for transaction IDs: perform comparisons using 1406 * modulo arithmetic so that they work over sequence number wraps. */ 1407 1408 static inline int tid_gt(tid_t x, tid_t y) 1409 { 1410 int difference = (x - y); 1411 return (difference > 0); 1412 } 1413 1414 static inline int tid_geq(tid_t x, tid_t y) 1415 { 1416 int difference = (x - y); 1417 return (difference >= 0); 1418 } 1419 1420 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1421 extern size_t journal_tag_bytes(journal_t *journal); 1422 1423 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1424 { 1425 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1426 } 1427 1428 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1429 { 1430 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1431 journal->j_chksum_driver == NULL); 1432 1433 return journal->j_chksum_driver != NULL; 1434 } 1435 1436 /* 1437 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for 1438 * transaction control blocks. 1439 */ 1440 #define JBD2_CONTROL_BLOCKS_SHIFT 5 1441 1442 /* 1443 * Return the minimum number of blocks which must be free in the journal 1444 * before a new transaction may be started. Must be called under j_state_lock. 1445 */ 1446 static inline int jbd2_space_needed(journal_t *journal) 1447 { 1448 int nblocks = journal->j_max_transaction_buffers; 1449 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT); 1450 } 1451 1452 /* 1453 * Return number of free blocks in the log. Must be called under j_state_lock. 1454 */ 1455 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1456 { 1457 /* Allow for rounding errors */ 1458 unsigned long free = journal->j_free - 32; 1459 1460 if (journal->j_committing_transaction) { 1461 unsigned long committing = atomic_read(&journal-> 1462 j_committing_transaction->t_outstanding_credits); 1463 1464 /* Transaction + control blocks */ 1465 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT); 1466 } 1467 return free; 1468 } 1469 1470 /* 1471 * Definitions which augment the buffer_head layer 1472 */ 1473 1474 /* journaling buffer types */ 1475 #define BJ_None 0 /* Not journaled */ 1476 #define BJ_Metadata 1 /* Normal journaled metadata */ 1477 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1478 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1479 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1480 #define BJ_Types 5 1481 1482 extern int jbd_blocks_per_page(struct inode *inode); 1483 1484 /* JBD uses a CRC32 checksum */ 1485 #define JBD_MAX_CHECKSUM_SIZE 4 1486 1487 static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1488 const void *address, unsigned int length) 1489 { 1490 struct { 1491 struct shash_desc shash; 1492 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1493 } desc; 1494 int err; 1495 1496 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1497 JBD_MAX_CHECKSUM_SIZE); 1498 1499 desc.shash.tfm = journal->j_chksum_driver; 1500 desc.shash.flags = 0; 1501 *(u32 *)desc.ctx = crc; 1502 1503 err = crypto_shash_update(&desc.shash, address, length); 1504 BUG_ON(err); 1505 1506 return *(u32 *)desc.ctx; 1507 } 1508 1509 /* Return most recent uncommitted transaction */ 1510 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1511 { 1512 tid_t tid; 1513 1514 read_lock(&journal->j_state_lock); 1515 tid = journal->j_commit_request; 1516 if (journal->j_running_transaction) 1517 tid = journal->j_running_transaction->t_tid; 1518 read_unlock(&journal->j_state_lock); 1519 return tid; 1520 } 1521 1522 #ifdef __KERNEL__ 1523 1524 #define buffer_trace_init(bh) do {} while (0) 1525 #define print_buffer_fields(bh) do {} while (0) 1526 #define print_buffer_trace(bh) do {} while (0) 1527 #define BUFFER_TRACE(bh, info) do {} while (0) 1528 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1529 #define JBUFFER_TRACE(jh, info) do {} while (0) 1530 1531 #endif /* __KERNEL__ */ 1532 1533 #define EFSBADCRC EBADMSG /* Bad CRC detected */ 1534 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1535 1536 #endif /* _LINUX_JBD2_H */ 1537