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 - The jbd_inode type is the structure linking inodes in 422 * ordered mode present in a transaction so that we can sync them during commit. 423 */ 424 struct jbd2_inode { 425 /** 426 * @i_transaction: 427 * 428 * Which transaction does this inode belong to? Either the running 429 * transaction or the committing one. [j_list_lock] 430 */ 431 transaction_t *i_transaction; 432 433 /** 434 * @i_next_transaction: 435 * 436 * Pointer to the running transaction modifying inode's data in case 437 * there is already a committing transaction touching it. [j_list_lock] 438 */ 439 transaction_t *i_next_transaction; 440 441 /** 442 * @i_list: List of inodes in the i_transaction [j_list_lock] 443 */ 444 struct list_head i_list; 445 446 /** 447 * @i_vfs_inode: 448 * 449 * VFS inode this inode belongs to [constant for lifetime of structure] 450 */ 451 struct inode *i_vfs_inode; 452 453 /** 454 * @i_flags: Flags of inode [j_list_lock] 455 */ 456 unsigned long i_flags; 457 }; 458 459 struct jbd2_revoke_table_s; 460 461 /** 462 * struct handle_s - The handle_s type is the concrete type associated with 463 * handle_t. 464 * @h_transaction: Which compound transaction is this update a part of? 465 * @h_journal: Which journal handle belongs to - used iff h_reserved set. 466 * @h_rsv_handle: Handle reserved for finishing the logical operation. 467 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. 468 * @h_ref: Reference count on this handle. 469 * @h_err: Field for caller's use to track errors through large fs operations. 470 * @h_sync: Flag for sync-on-close. 471 * @h_jdata: Flag to force data journaling. 472 * @h_reserved: Flag for handle for reserved credits. 473 * @h_aborted: Flag indicating fatal error on handle. 474 * @h_type: For handle statistics. 475 * @h_line_no: For handle statistics. 476 * @h_start_jiffies: Handle Start time. 477 * @h_requested_credits: Holds @h_buffer_credits after handle is started. 478 * @saved_alloc_context: Saved context while transaction is open. 479 **/ 480 481 /* Docbook can't yet cope with the bit fields, but will leave the documentation 482 * in so it can be fixed later. 483 */ 484 485 struct jbd2_journal_handle 486 { 487 union { 488 transaction_t *h_transaction; 489 /* Which journal handle belongs to - used iff h_reserved set */ 490 journal_t *h_journal; 491 }; 492 493 handle_t *h_rsv_handle; 494 int h_buffer_credits; 495 int h_ref; 496 int h_err; 497 498 /* Flags [no locking] */ 499 unsigned int h_sync: 1; 500 unsigned int h_jdata: 1; 501 unsigned int h_reserved: 1; 502 unsigned int h_aborted: 1; 503 unsigned int h_type: 8; 504 unsigned int h_line_no: 16; 505 506 unsigned long h_start_jiffies; 507 unsigned int h_requested_credits; 508 509 unsigned int saved_alloc_context; 510 }; 511 512 513 /* 514 * Some stats for checkpoint phase 515 */ 516 struct transaction_chp_stats_s { 517 unsigned long cs_chp_time; 518 __u32 cs_forced_to_close; 519 __u32 cs_written; 520 __u32 cs_dropped; 521 }; 522 523 /* The transaction_t type is the guts of the journaling mechanism. It 524 * tracks a compound transaction through its various states: 525 * 526 * RUNNING: accepting new updates 527 * LOCKED: Updates still running but we don't accept new ones 528 * RUNDOWN: Updates are tidying up but have finished requesting 529 * new buffers to modify (state not used for now) 530 * FLUSH: All updates complete, but we are still writing to disk 531 * COMMIT: All data on disk, writing commit record 532 * FINISHED: We still have to keep the transaction for checkpointing. 533 * 534 * The transaction keeps track of all of the buffers modified by a 535 * running transaction, and all of the buffers committed but not yet 536 * flushed to home for finished transactions. 537 */ 538 539 /* 540 * Lock ranking: 541 * 542 * j_list_lock 543 * ->jbd_lock_bh_journal_head() (This is "innermost") 544 * 545 * j_state_lock 546 * ->jbd_lock_bh_state() 547 * 548 * jbd_lock_bh_state() 549 * ->j_list_lock 550 * 551 * j_state_lock 552 * ->t_handle_lock 553 * 554 * j_state_lock 555 * ->j_list_lock (journal_unmap_buffer) 556 * 557 */ 558 559 struct transaction_s 560 { 561 /* Pointer to the journal for this transaction. [no locking] */ 562 journal_t *t_journal; 563 564 /* Sequence number for this transaction [no locking] */ 565 tid_t t_tid; 566 567 /* 568 * Transaction's current state 569 * [no locking - only kjournald2 alters this] 570 * [j_list_lock] guards transition of a transaction into T_FINISHED 571 * state and subsequent call of __jbd2_journal_drop_transaction() 572 * FIXME: needs barriers 573 * KLUDGE: [use j_state_lock] 574 */ 575 enum { 576 T_RUNNING, 577 T_LOCKED, 578 T_FLUSH, 579 T_COMMIT, 580 T_COMMIT_DFLUSH, 581 T_COMMIT_JFLUSH, 582 T_COMMIT_CALLBACK, 583 T_FINISHED 584 } t_state; 585 586 /* 587 * Where in the log does this transaction's commit start? [no locking] 588 */ 589 unsigned long t_log_start; 590 591 /* Number of buffers on the t_buffers list [j_list_lock] */ 592 int t_nr_buffers; 593 594 /* 595 * Doubly-linked circular list of all buffers reserved but not yet 596 * modified by this transaction [j_list_lock] 597 */ 598 struct journal_head *t_reserved_list; 599 600 /* 601 * Doubly-linked circular list of all metadata buffers owned by this 602 * transaction [j_list_lock] 603 */ 604 struct journal_head *t_buffers; 605 606 /* 607 * Doubly-linked circular list of all forget buffers (superseded 608 * buffers which we can un-checkpoint once this transaction commits) 609 * [j_list_lock] 610 */ 611 struct journal_head *t_forget; 612 613 /* 614 * Doubly-linked circular list of all buffers still to be flushed before 615 * this transaction can be checkpointed. [j_list_lock] 616 */ 617 struct journal_head *t_checkpoint_list; 618 619 /* 620 * Doubly-linked circular list of all buffers submitted for IO while 621 * checkpointing. [j_list_lock] 622 */ 623 struct journal_head *t_checkpoint_io_list; 624 625 /* 626 * Doubly-linked circular list of metadata buffers being shadowed by log 627 * IO. The IO buffers on the iobuf list and the shadow buffers on this 628 * list match each other one for one at all times. [j_list_lock] 629 */ 630 struct journal_head *t_shadow_list; 631 632 /* 633 * List of inodes whose data we've modified in data=ordered mode. 634 * [j_list_lock] 635 */ 636 struct list_head t_inode_list; 637 638 /* 639 * Protects info related to handles 640 */ 641 spinlock_t t_handle_lock; 642 643 /* 644 * Longest time some handle had to wait for running transaction 645 */ 646 unsigned long t_max_wait; 647 648 /* 649 * When transaction started 650 */ 651 unsigned long t_start; 652 653 /* 654 * When commit was requested 655 */ 656 unsigned long t_requested; 657 658 /* 659 * Checkpointing stats [j_checkpoint_sem] 660 */ 661 struct transaction_chp_stats_s t_chp_stats; 662 663 /* 664 * Number of outstanding updates running on this transaction 665 * [t_handle_lock] 666 */ 667 atomic_t t_updates; 668 669 /* 670 * Number of buffers reserved for use by all handles in this transaction 671 * handle but not yet modified. [t_handle_lock] 672 */ 673 atomic_t t_outstanding_credits; 674 675 /* 676 * Forward and backward links for the circular list of all transactions 677 * awaiting checkpoint. [j_list_lock] 678 */ 679 transaction_t *t_cpnext, *t_cpprev; 680 681 /* 682 * When will the transaction expire (become due for commit), in jiffies? 683 * [no locking] 684 */ 685 unsigned long t_expires; 686 687 /* 688 * When this transaction started, in nanoseconds [no locking] 689 */ 690 ktime_t t_start_time; 691 692 /* 693 * How many handles used this transaction? [t_handle_lock] 694 */ 695 atomic_t t_handle_count; 696 697 /* 698 * This transaction is being forced and some process is 699 * waiting for it to finish. 700 */ 701 unsigned int t_synchronous_commit:1; 702 703 /* Disk flush needs to be sent to fs partition [no locking] */ 704 int t_need_data_flush; 705 706 /* 707 * For use by the filesystem to store fs-specific data 708 * structures associated with the transaction 709 */ 710 struct list_head t_private_list; 711 }; 712 713 struct transaction_run_stats_s { 714 unsigned long rs_wait; 715 unsigned long rs_request_delay; 716 unsigned long rs_running; 717 unsigned long rs_locked; 718 unsigned long rs_flushing; 719 unsigned long rs_logging; 720 721 __u32 rs_handle_count; 722 __u32 rs_blocks; 723 __u32 rs_blocks_logged; 724 }; 725 726 struct transaction_stats_s { 727 unsigned long ts_tid; 728 unsigned long ts_requested; 729 struct transaction_run_stats_s run; 730 }; 731 732 static inline unsigned long 733 jbd2_time_diff(unsigned long start, unsigned long end) 734 { 735 if (end >= start) 736 return end - start; 737 738 return end + (MAX_JIFFY_OFFSET - start); 739 } 740 741 #define JBD2_NR_BATCH 64 742 743 /** 744 * struct journal_s - The journal_s type is the concrete type associated with 745 * journal_t. 746 */ 747 struct journal_s 748 { 749 /** 750 * @j_flags: General journaling state flags [j_state_lock] 751 */ 752 unsigned long j_flags; 753 754 /** 755 * @j_errno: 756 * 757 * Is there an outstanding uncleared error on the journal (from a prior 758 * abort)? [j_state_lock] 759 */ 760 int j_errno; 761 762 /** 763 * @j_sb_buffer: The first part of the superblock buffer. 764 */ 765 struct buffer_head *j_sb_buffer; 766 767 /** 768 * @j_superblock: The second part of the superblock buffer. 769 */ 770 journal_superblock_t *j_superblock; 771 772 /** 773 * @j_format_version: Version of the superblock format. 774 */ 775 int j_format_version; 776 777 /** 778 * @j_state_lock: Protect the various scalars in the journal. 779 */ 780 rwlock_t j_state_lock; 781 782 /** 783 * @j_barrier_count: 784 * 785 * Number of processes waiting to create a barrier lock [j_state_lock] 786 */ 787 int j_barrier_count; 788 789 /** 790 * @j_barrier: The barrier lock itself. 791 */ 792 struct mutex j_barrier; 793 794 /** 795 * @j_running_transaction: 796 * 797 * Transactions: The current running transaction... 798 * [j_state_lock] [caller holding open handle] 799 */ 800 transaction_t *j_running_transaction; 801 802 /** 803 * @j_committing_transaction: 804 * 805 * the transaction we are pushing to disk 806 * [j_state_lock] [caller holding open handle] 807 */ 808 transaction_t *j_committing_transaction; 809 810 /** 811 * @j_checkpoint_transactions: 812 * 813 * ... and a linked circular list of all transactions waiting for 814 * checkpointing. [j_list_lock] 815 */ 816 transaction_t *j_checkpoint_transactions; 817 818 /** 819 * @j_wait_transaction_locked: 820 * 821 * Wait queue for waiting for a locked transaction to start committing, 822 * or for a barrier lock to be released. 823 */ 824 wait_queue_head_t j_wait_transaction_locked; 825 826 /** 827 * @j_wait_done_commit: Wait queue for waiting for commit to complete. 828 */ 829 wait_queue_head_t j_wait_done_commit; 830 831 /** 832 * @j_wait_commit: Wait queue to trigger commit. 833 */ 834 wait_queue_head_t j_wait_commit; 835 836 /** 837 * @j_wait_updates: Wait queue to wait for updates to complete. 838 */ 839 wait_queue_head_t j_wait_updates; 840 841 /** 842 * @j_wait_reserved: 843 * 844 * Wait queue to wait for reserved buffer credits to drop. 845 */ 846 wait_queue_head_t j_wait_reserved; 847 848 /** 849 * @j_checkpoint_mutex: 850 * 851 * Semaphore for locking against concurrent checkpoints. 852 */ 853 struct mutex j_checkpoint_mutex; 854 855 /** 856 * @j_chkpt_bhs: 857 * 858 * List of buffer heads used by the checkpoint routine. This 859 * was moved from jbd2_log_do_checkpoint() to reduce stack 860 * usage. Access to this array is controlled by the 861 * @j_checkpoint_mutex. [j_checkpoint_mutex] 862 */ 863 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 864 865 /** 866 * @j_head: 867 * 868 * Journal head: identifies the first unused block in the journal. 869 * [j_state_lock] 870 */ 871 unsigned long j_head; 872 873 /** 874 * @j_tail: 875 * 876 * Journal tail: identifies the oldest still-used block in the journal. 877 * [j_state_lock] 878 */ 879 unsigned long j_tail; 880 881 /** 882 * @j_free: 883 * 884 * Journal free: how many free blocks are there in the journal? 885 * [j_state_lock] 886 */ 887 unsigned long j_free; 888 889 /** 890 * @j_first: 891 * 892 * The block number of the first usable block in the journal 893 * [j_state_lock]. 894 */ 895 unsigned long j_first; 896 897 /** 898 * @j_last: 899 * 900 * The block number one beyond the last usable block in the journal 901 * [j_state_lock]. 902 */ 903 unsigned long j_last; 904 905 /** 906 * @j_dev: Device where we store the journal. 907 */ 908 struct block_device *j_dev; 909 910 /** 911 * @j_blocksize: Block size for the location where we store the journal. 912 */ 913 int j_blocksize; 914 915 /** 916 * @j_blk_offset: 917 * 918 * Starting block offset into the device where we store the journal. 919 */ 920 unsigned long long j_blk_offset; 921 922 /** 923 * @j_devname: Journal device name. 924 */ 925 char j_devname[BDEVNAME_SIZE+24]; 926 927 /** 928 * @j_fs_dev: 929 * 930 * Device which holds the client fs. For internal journal this will be 931 * equal to j_dev. 932 */ 933 struct block_device *j_fs_dev; 934 935 /** 936 * @j_maxlen: Total maximum capacity of the journal region on disk. 937 */ 938 unsigned int j_maxlen; 939 940 /** 941 * @j_reserved_credits: 942 * 943 * Number of buffers reserved from the running transaction. 944 */ 945 atomic_t j_reserved_credits; 946 947 /** 948 * @j_list_lock: Protects the buffer lists and internal buffer state. 949 */ 950 spinlock_t j_list_lock; 951 952 /** 953 * @j_inode: 954 * 955 * Optional inode where we store the journal. If present, all 956 * journal block numbers are mapped into this inode via bmap(). 957 */ 958 struct inode *j_inode; 959 960 /** 961 * @j_tail_sequence: 962 * 963 * Sequence number of the oldest transaction in the log [j_state_lock] 964 */ 965 tid_t j_tail_sequence; 966 967 /** 968 * @j_transaction_sequence: 969 * 970 * Sequence number of the next transaction to grant [j_state_lock] 971 */ 972 tid_t j_transaction_sequence; 973 974 /** 975 * @j_commit_sequence: 976 * 977 * Sequence number of the most recently committed transaction 978 * [j_state_lock]. 979 */ 980 tid_t j_commit_sequence; 981 982 /** 983 * @j_commit_request: 984 * 985 * Sequence number of the most recent transaction wanting commit 986 * [j_state_lock] 987 */ 988 tid_t j_commit_request; 989 990 /** 991 * @j_uuid: 992 * 993 * Journal uuid: identifies the object (filesystem, LVM volume etc) 994 * backed by this journal. This will eventually be replaced by an array 995 * of uuids, allowing us to index multiple devices within a single 996 * journal and to perform atomic updates across them. 997 */ 998 __u8 j_uuid[16]; 999 1000 /** 1001 * @j_task: Pointer to the current commit thread for this journal. 1002 */ 1003 struct task_struct *j_task; 1004 1005 /** 1006 * @j_max_transaction_buffers: 1007 * 1008 * Maximum number of metadata buffers to allow in a single compound 1009 * commit transaction. 1010 */ 1011 int j_max_transaction_buffers; 1012 1013 /** 1014 * @j_commit_interval: 1015 * 1016 * What is the maximum transaction lifetime before we begin a commit? 1017 */ 1018 unsigned long j_commit_interval; 1019 1020 /** 1021 * @j_commit_timer: The timer used to wakeup the commit thread. 1022 */ 1023 struct timer_list j_commit_timer; 1024 1025 /** 1026 * @j_revoke_lock: Protect the revoke table. 1027 */ 1028 spinlock_t j_revoke_lock; 1029 1030 /** 1031 * @j_revoke: 1032 * 1033 * The revoke table - maintains the list of revoked blocks in the 1034 * current transaction. 1035 */ 1036 struct jbd2_revoke_table_s *j_revoke; 1037 1038 /** 1039 * @j_revoke_table: Alternate revoke tables for j_revoke. 1040 */ 1041 struct jbd2_revoke_table_s *j_revoke_table[2]; 1042 1043 /** 1044 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction. 1045 */ 1046 struct buffer_head **j_wbuf; 1047 1048 /** 1049 * @j_wbufsize: 1050 * 1051 * Size of @j_wbuf array. 1052 */ 1053 int j_wbufsize; 1054 1055 /** 1056 * @j_last_sync_writer: 1057 * 1058 * The pid of the last person to run a synchronous operation 1059 * through the journal. 1060 */ 1061 pid_t j_last_sync_writer; 1062 1063 /** 1064 * @j_average_commit_time: 1065 * 1066 * The average amount of time in nanoseconds it takes to commit a 1067 * transaction to disk. [j_state_lock] 1068 */ 1069 u64 j_average_commit_time; 1070 1071 /** 1072 * @j_min_batch_time: 1073 * 1074 * Minimum time that we should wait for additional filesystem operations 1075 * to get batched into a synchronous handle in microseconds. 1076 */ 1077 u32 j_min_batch_time; 1078 1079 /** 1080 * @j_max_batch_time: 1081 * 1082 * Maximum time that we should wait for additional filesystem operations 1083 * to get batched into a synchronous handle in microseconds. 1084 */ 1085 u32 j_max_batch_time; 1086 1087 /** 1088 * @j_commit_callback: 1089 * 1090 * This function is called when a transaction is closed. 1091 */ 1092 void (*j_commit_callback)(journal_t *, 1093 transaction_t *); 1094 1095 /* 1096 * Journal statistics 1097 */ 1098 1099 /** 1100 * @j_history_lock: Protect the transactions statistics history. 1101 */ 1102 spinlock_t j_history_lock; 1103 1104 /** 1105 * @j_proc_entry: procfs entry for the jbd statistics directory. 1106 */ 1107 struct proc_dir_entry *j_proc_entry; 1108 1109 /** 1110 * @j_stats: Overall statistics. 1111 */ 1112 struct transaction_stats_s j_stats; 1113 1114 /** 1115 * @j_failed_commit: Failed journal commit ID. 1116 */ 1117 unsigned int j_failed_commit; 1118 1119 /** 1120 * @j_private: 1121 * 1122 * An opaque pointer to fs-private information. ext3 puts its 1123 * superblock pointer here. 1124 */ 1125 void *j_private; 1126 1127 /** 1128 * @j_chksum_driver: 1129 * 1130 * Reference to checksum algorithm driver via cryptoapi. 1131 */ 1132 struct crypto_shash *j_chksum_driver; 1133 1134 /** 1135 * @j_csum_seed: 1136 * 1137 * Precomputed journal UUID checksum for seeding other checksums. 1138 */ 1139 __u32 j_csum_seed; 1140 1141 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1142 /** 1143 * @j_trans_commit_map: 1144 * 1145 * Lockdep entity to track transaction commit dependencies. Handles 1146 * hold this "lock" for read, when we wait for commit, we acquire the 1147 * "lock" for writing. This matches the properties of jbd2 journalling 1148 * where the running transaction has to wait for all handles to be 1149 * dropped to commit that transaction and also acquiring a handle may 1150 * require transaction commit to finish. 1151 */ 1152 struct lockdep_map j_trans_commit_map; 1153 #endif 1154 }; 1155 1156 #define jbd2_might_wait_for_commit(j) \ 1157 do { \ 1158 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \ 1159 rwsem_release(&j->j_trans_commit_map, 1, _THIS_IP_); \ 1160 } while (0) 1161 1162 /* journal feature predicate functions */ 1163 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1164 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1165 { \ 1166 return ((j)->j_format_version >= 2 && \ 1167 ((j)->j_superblock->s_feature_compat & \ 1168 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1169 } \ 1170 static inline void jbd2_set_feature_##name(journal_t *j) \ 1171 { \ 1172 (j)->j_superblock->s_feature_compat |= \ 1173 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1174 } \ 1175 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1176 { \ 1177 (j)->j_superblock->s_feature_compat &= \ 1178 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1179 } 1180 1181 #define JBD2_FEATURE_RO_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_ro_compat & \ 1186 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1187 } \ 1188 static inline void jbd2_set_feature_##name(journal_t *j) \ 1189 { \ 1190 (j)->j_superblock->s_feature_ro_compat |= \ 1191 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1192 } \ 1193 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1194 { \ 1195 (j)->j_superblock->s_feature_ro_compat &= \ 1196 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1197 } 1198 1199 #define JBD2_FEATURE_INCOMPAT_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_incompat & \ 1204 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1205 } \ 1206 static inline void jbd2_set_feature_##name(journal_t *j) \ 1207 { \ 1208 (j)->j_superblock->s_feature_incompat |= \ 1209 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1210 } \ 1211 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1212 { \ 1213 (j)->j_superblock->s_feature_incompat &= \ 1214 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1215 } 1216 1217 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1218 1219 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1220 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1221 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1222 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1223 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1224 1225 /* 1226 * Journal flag definitions 1227 */ 1228 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1229 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1230 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1231 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1232 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1233 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1234 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1235 * data write error in ordered 1236 * mode */ 1237 #define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */ 1238 1239 /* 1240 * Function declarations for the journaling transaction and buffer 1241 * management 1242 */ 1243 1244 /* Filing buffers */ 1245 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1246 extern void __jbd2_journal_refile_buffer(struct journal_head *); 1247 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1248 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1249 extern void __journal_free_buffer(struct journal_head *bh); 1250 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1251 extern void __journal_clean_data_list(transaction_t *transaction); 1252 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1253 { 1254 list_add_tail(&bh->b_assoc_buffers, head); 1255 } 1256 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1257 { 1258 list_del_init(&bh->b_assoc_buffers); 1259 } 1260 1261 /* Log buffer allocation */ 1262 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1263 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1264 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1265 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1266 unsigned long *block); 1267 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1268 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1269 1270 /* Commit management */ 1271 extern void jbd2_journal_commit_transaction(journal_t *); 1272 1273 /* Checkpoint list management */ 1274 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1275 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1276 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1277 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1278 1279 1280 /* 1281 * Triggers 1282 */ 1283 1284 struct jbd2_buffer_trigger_type { 1285 /* 1286 * Fired a the moment data to write to the journal are known to be 1287 * stable - so either at the moment b_frozen_data is created or just 1288 * before a buffer is written to the journal. mapped_data is a mapped 1289 * buffer that is the frozen data for commit. 1290 */ 1291 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1292 struct buffer_head *bh, void *mapped_data, 1293 size_t size); 1294 1295 /* 1296 * Fired during journal abort for dirty buffers that will not be 1297 * committed. 1298 */ 1299 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1300 struct buffer_head *bh); 1301 }; 1302 1303 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1304 void *mapped_data, 1305 struct jbd2_buffer_trigger_type *triggers); 1306 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1307 struct jbd2_buffer_trigger_type *triggers); 1308 1309 /* Buffer IO */ 1310 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1311 struct journal_head *jh_in, 1312 struct buffer_head **bh_out, 1313 sector_t blocknr); 1314 1315 /* Transaction locking */ 1316 extern void __wait_on_journal (journal_t *); 1317 1318 /* Transaction cache support */ 1319 extern void jbd2_journal_destroy_transaction_cache(void); 1320 extern int jbd2_journal_init_transaction_cache(void); 1321 extern void jbd2_journal_free_transaction(transaction_t *); 1322 1323 /* 1324 * Journal locking. 1325 * 1326 * We need to lock the journal during transaction state changes so that nobody 1327 * ever tries to take a handle on the running transaction while we are in the 1328 * middle of moving it to the commit phase. j_state_lock does this. 1329 * 1330 * Note that the locking is completely interrupt unsafe. We never touch 1331 * journal structures from interrupts. 1332 */ 1333 1334 static inline handle_t *journal_current_handle(void) 1335 { 1336 return current->journal_info; 1337 } 1338 1339 /* The journaling code user interface: 1340 * 1341 * Create and destroy handles 1342 * Register buffer modifications against the current transaction. 1343 */ 1344 1345 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1346 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1347 gfp_t gfp_mask, unsigned int type, 1348 unsigned int line_no); 1349 extern int jbd2_journal_restart(handle_t *, int nblocks); 1350 extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask); 1351 extern int jbd2_journal_start_reserved(handle_t *handle, 1352 unsigned int type, unsigned int line_no); 1353 extern void jbd2_journal_free_reserved(handle_t *handle); 1354 extern int jbd2_journal_extend (handle_t *, int nblocks); 1355 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1356 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1357 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1358 void jbd2_journal_set_triggers(struct buffer_head *, 1359 struct jbd2_buffer_trigger_type *type); 1360 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1361 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1362 extern void journal_sync_buffer (struct buffer_head *); 1363 extern int jbd2_journal_invalidatepage(journal_t *, 1364 struct page *, unsigned int, unsigned int); 1365 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); 1366 extern int jbd2_journal_stop(handle_t *); 1367 extern int jbd2_journal_flush (journal_t *); 1368 extern void jbd2_journal_lock_updates (journal_t *); 1369 extern void jbd2_journal_unlock_updates (journal_t *); 1370 1371 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1372 struct block_device *fs_dev, 1373 unsigned long long start, int len, int bsize); 1374 extern journal_t * jbd2_journal_init_inode (struct inode *); 1375 extern int jbd2_journal_update_format (journal_t *); 1376 extern int jbd2_journal_check_used_features 1377 (journal_t *, unsigned long, unsigned long, unsigned long); 1378 extern int jbd2_journal_check_available_features 1379 (journal_t *, unsigned long, unsigned long, unsigned long); 1380 extern int jbd2_journal_set_features 1381 (journal_t *, unsigned long, unsigned long, unsigned long); 1382 extern void jbd2_journal_clear_features 1383 (journal_t *, unsigned long, unsigned long, unsigned long); 1384 extern int jbd2_journal_load (journal_t *journal); 1385 extern int jbd2_journal_destroy (journal_t *); 1386 extern int jbd2_journal_recover (journal_t *journal); 1387 extern int jbd2_journal_wipe (journal_t *, int); 1388 extern int jbd2_journal_skip_recovery (journal_t *); 1389 extern void jbd2_journal_update_sb_errno(journal_t *); 1390 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1391 unsigned long, int); 1392 extern void __jbd2_journal_abort_hard (journal_t *); 1393 extern void jbd2_journal_abort (journal_t *, int); 1394 extern int jbd2_journal_errno (journal_t *); 1395 extern void jbd2_journal_ack_err (journal_t *); 1396 extern int jbd2_journal_clear_err (journal_t *); 1397 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1398 extern int jbd2_journal_force_commit(journal_t *); 1399 extern int jbd2_journal_force_commit_nested(journal_t *); 1400 extern int jbd2_journal_inode_add_write(handle_t *handle, struct jbd2_inode *inode); 1401 extern int jbd2_journal_inode_add_wait(handle_t *handle, struct jbd2_inode *inode); 1402 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1403 struct jbd2_inode *inode, loff_t new_size); 1404 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1405 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1406 1407 /* 1408 * journal_head management 1409 */ 1410 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1411 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1412 void jbd2_journal_put_journal_head(struct journal_head *jh); 1413 1414 /* 1415 * handle management 1416 */ 1417 extern struct kmem_cache *jbd2_handle_cache; 1418 1419 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1420 { 1421 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1422 } 1423 1424 static inline void jbd2_free_handle(handle_t *handle) 1425 { 1426 kmem_cache_free(jbd2_handle_cache, handle); 1427 } 1428 1429 /* 1430 * jbd2_inode management (optional, for those file systems that want to use 1431 * dynamically allocated jbd2_inode structures) 1432 */ 1433 extern struct kmem_cache *jbd2_inode_cache; 1434 1435 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1436 { 1437 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1438 } 1439 1440 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1441 { 1442 kmem_cache_free(jbd2_inode_cache, jinode); 1443 } 1444 1445 /* Primary revoke support */ 1446 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1447 extern int jbd2_journal_init_revoke(journal_t *, int); 1448 extern void jbd2_journal_destroy_revoke_caches(void); 1449 extern int jbd2_journal_init_revoke_caches(void); 1450 1451 extern void jbd2_journal_destroy_revoke(journal_t *); 1452 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1453 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1454 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1455 struct list_head *log_bufs); 1456 1457 /* Recovery revoke support */ 1458 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1459 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1460 extern void jbd2_journal_clear_revoke(journal_t *); 1461 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1462 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1463 1464 /* 1465 * The log thread user interface: 1466 * 1467 * Request space in the current transaction, and force transaction commit 1468 * transitions on demand. 1469 */ 1470 1471 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1472 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1473 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1474 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1475 int jbd2_transaction_committed(journal_t *journal, tid_t tid); 1476 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1477 int jbd2_log_do_checkpoint(journal_t *journal); 1478 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1479 1480 void __jbd2_log_wait_for_space(journal_t *journal); 1481 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1482 extern int jbd2_cleanup_journal_tail(journal_t *); 1483 1484 /* 1485 * is_journal_abort 1486 * 1487 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1488 * bit, when set, indicates that we have had a fatal error somewhere, 1489 * either inside the journaling layer or indicated to us by the client 1490 * (eg. ext3), and that we and should not commit any further 1491 * transactions. 1492 */ 1493 1494 static inline int is_journal_aborted(journal_t *journal) 1495 { 1496 return journal->j_flags & JBD2_ABORT; 1497 } 1498 1499 static inline int is_handle_aborted(handle_t *handle) 1500 { 1501 if (handle->h_aborted || !handle->h_transaction) 1502 return 1; 1503 return is_journal_aborted(handle->h_transaction->t_journal); 1504 } 1505 1506 static inline void jbd2_journal_abort_handle(handle_t *handle) 1507 { 1508 handle->h_aborted = 1; 1509 } 1510 1511 #endif /* __KERNEL__ */ 1512 1513 /* Comparison functions for transaction IDs: perform comparisons using 1514 * modulo arithmetic so that they work over sequence number wraps. */ 1515 1516 static inline int tid_gt(tid_t x, tid_t y) 1517 { 1518 int difference = (x - y); 1519 return (difference > 0); 1520 } 1521 1522 static inline int tid_geq(tid_t x, tid_t y) 1523 { 1524 int difference = (x - y); 1525 return (difference >= 0); 1526 } 1527 1528 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1529 extern size_t journal_tag_bytes(journal_t *journal); 1530 1531 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1532 { 1533 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1534 } 1535 1536 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1537 { 1538 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1539 journal->j_chksum_driver == NULL); 1540 1541 return journal->j_chksum_driver != NULL; 1542 } 1543 1544 /* 1545 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for 1546 * transaction control blocks. 1547 */ 1548 #define JBD2_CONTROL_BLOCKS_SHIFT 5 1549 1550 /* 1551 * Return the minimum number of blocks which must be free in the journal 1552 * before a new transaction may be started. Must be called under j_state_lock. 1553 */ 1554 static inline int jbd2_space_needed(journal_t *journal) 1555 { 1556 int nblocks = journal->j_max_transaction_buffers; 1557 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT); 1558 } 1559 1560 /* 1561 * Return number of free blocks in the log. Must be called under j_state_lock. 1562 */ 1563 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1564 { 1565 /* Allow for rounding errors */ 1566 unsigned long free = journal->j_free - 32; 1567 1568 if (journal->j_committing_transaction) { 1569 unsigned long committing = atomic_read(&journal-> 1570 j_committing_transaction->t_outstanding_credits); 1571 1572 /* Transaction + control blocks */ 1573 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT); 1574 } 1575 return free; 1576 } 1577 1578 /* 1579 * Definitions which augment the buffer_head layer 1580 */ 1581 1582 /* journaling buffer types */ 1583 #define BJ_None 0 /* Not journaled */ 1584 #define BJ_Metadata 1 /* Normal journaled metadata */ 1585 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1586 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1587 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1588 #define BJ_Types 5 1589 1590 extern int jbd_blocks_per_page(struct inode *inode); 1591 1592 /* JBD uses a CRC32 checksum */ 1593 #define JBD_MAX_CHECKSUM_SIZE 4 1594 1595 static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1596 const void *address, unsigned int length) 1597 { 1598 struct { 1599 struct shash_desc shash; 1600 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1601 } desc; 1602 int err; 1603 1604 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1605 JBD_MAX_CHECKSUM_SIZE); 1606 1607 desc.shash.tfm = journal->j_chksum_driver; 1608 desc.shash.flags = 0; 1609 *(u32 *)desc.ctx = crc; 1610 1611 err = crypto_shash_update(&desc.shash, address, length); 1612 BUG_ON(err); 1613 1614 return *(u32 *)desc.ctx; 1615 } 1616 1617 /* Return most recent uncommitted transaction */ 1618 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1619 { 1620 tid_t tid; 1621 1622 read_lock(&journal->j_state_lock); 1623 tid = journal->j_commit_request; 1624 if (journal->j_running_transaction) 1625 tid = journal->j_running_transaction->t_tid; 1626 read_unlock(&journal->j_state_lock); 1627 return tid; 1628 } 1629 1630 #ifdef __KERNEL__ 1631 1632 #define buffer_trace_init(bh) do {} while (0) 1633 #define print_buffer_fields(bh) do {} while (0) 1634 #define print_buffer_trace(bh) do {} while (0) 1635 #define BUFFER_TRACE(bh, info) do {} while (0) 1636 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1637 #define JBUFFER_TRACE(jh, info) do {} while (0) 1638 1639 #endif /* __KERNEL__ */ 1640 1641 #define EFSBADCRC EBADMSG /* Bad CRC detected */ 1642 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1643 1644 #endif /* _LINUX_JBD2_H */ 1645