jbd2.h

   1/* SPDX-License-Identifier: GPL-2.0-or-later */
   2/*
   3 * linux/include/linux/jbd2.h
   4 *
   5 * Written by Stephen C. Tweedie <sct@redhat.com>
   6 *
   7 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
   8 *
   9 * Definitions for transaction data structures for the buffer cache
  10 * filesystem journaling support.
  11 */
  12
  13#ifndef _LINUX_JBD2_H
  14#define _LINUX_JBD2_H
  15
  16/* Allow this file to be included directly into e2fsprogs */
  17#ifndef __KERNEL__
  18#include "jfs_compat.h"
  19#define JBD2_DEBUG
  20#else
  21
  22#include <linux/types.h>
  23#include <linux/buffer_head.h>
  24#include <linux/journal-head.h>
  25#include <linux/stddef.h>
  26#include <linux/mutex.h>
  27#include <linux/timer.h>
  28#include <linux/slab.h>
  29#include <linux/bit_spinlock.h>
  30#include <linux/blkdev.h>
  31#include <linux/crc32c.h>
  32#endif
  33
  34#define journal_oom_retry 1
  35
  36/*
  37 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
  38 * certain classes of error which can occur due to failed IOs.  Under
  39 * normal use we want ext4 to continue after such errors, because
  40 * hardware _can_ fail, but for debugging purposes when running tests on
  41 * known-good hardware we may want to trap these errors.
  42 */
  43#undef JBD2_PARANOID_IOFAIL
  44
  45/*
  46 * The default maximum commit age, in seconds.
  47 */
  48#define JBD2_DEFAULT_MAX_COMMIT_AGE 5
  49
  50#ifdef CONFIG_JBD2_DEBUG
  51/*
  52 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
  53 * consistency checks.  By default we don't do this unless
  54 * CONFIG_JBD2_DEBUG is on.
  55 */
  56#define JBD2_EXPENSIVE_CHECKING
  57void __jbd2_debug(int level, const char *file, const char *func,
  58		  unsigned int line, const char *fmt, ...);
  59
  60#define jbd2_debug(n, fmt, a...) \
  61	__jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
  62#else
  63#define jbd2_debug(n, fmt, a...)  no_printk(fmt, ##a)
  64#endif
  65
  66extern void *jbd2_alloc(size_t size, gfp_t flags);
  67extern void jbd2_free(void *ptr, size_t size);
  68
  69#define JBD2_MIN_JOURNAL_BLOCKS 1024
  70#define JBD2_DEFAULT_FAST_COMMIT_BLOCKS 256
  71
  72#ifdef __KERNEL__
  73
  74/**
  75 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
  76 *
  77 * All filesystem modifications made by the process go
  78 * through this handle.  Recursive operations (such as quota operations)
  79 * are gathered into a single update.
  80 *
  81 * The buffer credits field is used to account for journaled buffers
  82 * being modified by the running process.  To ensure that there is
  83 * enough log space for all outstanding operations, we need to limit the
  84 * number of outstanding buffers possible at any time.  When the
  85 * operation completes, any buffer credits not used are credited back to
  86 * the transaction, so that at all times we know how many buffers the
  87 * outstanding updates on a transaction might possibly touch.
  88 *
  89 * This is an opaque datatype.
  90 **/
  91typedef struct jbd2_journal_handle handle_t;	/* Atomic operation type */
  92
  93
  94/**
  95 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
  96 *
  97 * journal_t is linked to from the fs superblock structure.
  98 *
  99 * We use the journal_t to keep track of all outstanding transaction
 100 * activity on the filesystem, and to manage the state of the log
 101 * writing process.
 102 *
 103 * This is an opaque datatype.
 104 **/
 105typedef struct journal_s	journal_t;	/* Journal control structure */
 106#endif
 107
 108/*
 109 * Internal structures used by the logging mechanism:
 110 */
 111
 112#define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
 113
 114/*
 115 * On-disk structures
 116 */
 117
 118/*
 119 * Descriptor block types:
 120 */
 121
 122#define JBD2_DESCRIPTOR_BLOCK	1
 123#define JBD2_COMMIT_BLOCK	2
 124#define JBD2_SUPERBLOCK_V1	3
 125#define JBD2_SUPERBLOCK_V2	4
 126#define JBD2_REVOKE_BLOCK	5
 127
 128/*
 129 * Standard header for all descriptor blocks:
 130 */
 131typedef struct journal_header_s
 132{
 133	__be32		h_magic;
 134	__be32		h_blocktype;
 135	__be32		h_sequence;
 136} journal_header_t;
 137
 138/*
 139 * Checksum types.
 140 */
 141#define JBD2_CRC32_CHKSUM   1
 142#define JBD2_MD5_CHKSUM     2
 143#define JBD2_SHA1_CHKSUM    3
 144#define JBD2_CRC32C_CHKSUM  4
 145
 146#define JBD2_CRC32_CHKSUM_SIZE 4
 147
 148#define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
 149/*
 150 * Commit block header for storing transactional checksums:
 151 *
 152 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
 153 * fields are used to store a checksum of the descriptor and data blocks.
 154 *
 155 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
 156 * field is used to store crc32c(uuid+commit_block).  Each journal metadata
 157 * block gets its own checksum, and data block checksums are stored in
 158 * journal_block_tag (in the descriptor).  The other h_chksum* fields are
 159 * not used.
 160 *
 161 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
 162 * journal_block_tag3_t to store a full 32-bit checksum.  Everything else
 163 * is the same as v2.
 164 *
 165 * Checksum v1, v2, and v3 are mutually exclusive features.
 166 */
 167struct commit_header {
 168	__be32		h_magic;
 169	__be32          h_blocktype;
 170	__be32          h_sequence;
 171	unsigned char   h_chksum_type;
 172	unsigned char   h_chksum_size;
 173	unsigned char 	h_padding[2];
 174	__be32 		h_chksum[JBD2_CHECKSUM_BYTES];
 175	__be64		h_commit_sec;
 176	__be32		h_commit_nsec;
 177};
 178
 179/*
 180 * The block tag: used to describe a single buffer in the journal.
 181 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
 182 * raw struct shouldn't be used for pointer math or sizeof() - use
 183 * journal_tag_bytes(journal) instead to compute this.
 184 */
 185typedef struct journal_block_tag3_s
 186{
 187	__be32		t_blocknr;	/* The on-disk block number */
 188	__be32		t_flags;	/* See below */
 189	__be32		t_blocknr_high; /* most-significant high 32bits. */
 190	__be32		t_checksum;	/* crc32c(uuid+seq+block) */
 191} journal_block_tag3_t;
 192
 193typedef struct journal_block_tag_s
 194{
 195	__be32		t_blocknr;	/* The on-disk block number */
 196	__be16		t_checksum;	/* truncated crc32c(uuid+seq+block) */
 197	__be16		t_flags;	/* See below */
 198	__be32		t_blocknr_high; /* most-significant high 32bits. */
 199} journal_block_tag_t;
 200
 201/* Tail of descriptor or revoke block, for checksumming */
 202struct jbd2_journal_block_tail {
 203	__be32		t_checksum;	/* crc32c(uuid+descr_block) */
 204};
 205
 206/*
 207 * The revoke descriptor: used on disk to describe a series of blocks to
 208 * be revoked from the log
 209 */
 210typedef struct jbd2_journal_revoke_header_s
 211{
 212	journal_header_t r_header;
 213	__be32		 r_count;	/* Count of bytes used in the block */
 214} jbd2_journal_revoke_header_t;
 215
 216/* Definitions for the journal tag flags word: */
 217#define JBD2_FLAG_ESCAPE		1	/* on-disk block is escaped */
 218#define JBD2_FLAG_SAME_UUID	2	/* block has same uuid as previous */
 219#define JBD2_FLAG_DELETED	4	/* block deleted by this transaction */
 220#define JBD2_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
 221
 222
 223/*
 224 * The journal superblock.  All fields are in big-endian byte order.
 225 */
 226typedef struct journal_superblock_s
 227{
 228/* 0x0000 */
 229	journal_header_t s_header;
 230
 231/* 0x000C */
 232	/* Static information describing the journal */
 233	__be32	s_blocksize;		/* journal device blocksize */
 234	__be32	s_maxlen;		/* total blocks in journal file */
 235	__be32	s_first;		/* first block of log information */
 236
 237/* 0x0018 */
 238	/* Dynamic information describing the current state of the log */
 239	__be32	s_sequence;		/* first commit ID expected in log */
 240	__be32	s_start;		/* blocknr of start of log */
 241
 242/* 0x0020 */
 243	/* Error value, as set by jbd2_journal_abort(). */
 244	__be32	s_errno;
 245
 246/* 0x0024 */
 247	/* Remaining fields are only valid in a version-2 superblock */
 248	__be32	s_feature_compat;	/* compatible feature set */
 249	__be32	s_feature_incompat;	/* incompatible feature set */
 250	__be32	s_feature_ro_compat;	/* readonly-compatible feature set */
 251/* 0x0030 */
 252	__u8	s_uuid[16];		/* 128-bit uuid for journal */
 253
 254/* 0x0040 */
 255	__be32	s_nr_users;		/* Nr of filesystems sharing log */
 256
 257	__be32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
 258
 259/* 0x0048 */
 260	__be32	s_max_transaction;	/* Limit of journal blocks per trans.*/
 261	__be32	s_max_trans_data;	/* Limit of data blocks per trans. */
 262
 263/* 0x0050 */
 264	__u8	s_checksum_type;	/* checksum type */
 265	__u8	s_padding2[3];
 266/* 0x0054 */
 267	__be32	s_num_fc_blks;		/* Number of fast commit blocks */
 268	__be32	s_head;			/* blocknr of head of log, only uptodate
 269					 * while the filesystem is clean */
 270/* 0x005C */
 271	__u32	s_padding[40];
 272	__be32	s_checksum;		/* crc32c(superblock) */
 273
 274/* 0x0100 */
 275	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
 276/* 0x0400 */
 277} journal_superblock_t;
 278
 279#define JBD2_FEATURE_COMPAT_CHECKSUM		0x00000001
 280
 281#define JBD2_FEATURE_INCOMPAT_REVOKE		0x00000001
 282#define JBD2_FEATURE_INCOMPAT_64BIT		0x00000002
 283#define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
 284#define JBD2_FEATURE_INCOMPAT_CSUM_V2		0x00000008
 285#define JBD2_FEATURE_INCOMPAT_CSUM_V3		0x00000010
 286#define JBD2_FEATURE_INCOMPAT_FAST_COMMIT	0x00000020
 287
 288/* See "journal feature predicate functions" below */
 289
 290/* Features known to this kernel version: */
 291#define JBD2_KNOWN_COMPAT_FEATURES	JBD2_FEATURE_COMPAT_CHECKSUM
 292#define JBD2_KNOWN_ROCOMPAT_FEATURES	0
 293#define JBD2_KNOWN_INCOMPAT_FEATURES	(JBD2_FEATURE_INCOMPAT_REVOKE | \
 294					JBD2_FEATURE_INCOMPAT_64BIT | \
 295					JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
 296					JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
 297					JBD2_FEATURE_INCOMPAT_CSUM_V3 | \
 298					JBD2_FEATURE_INCOMPAT_FAST_COMMIT)
 299
 300#ifdef __KERNEL__
 301
 302#include <linux/fs.h>
 303#include <linux/sched.h>
 304
 305enum jbd_state_bits {
 306	BH_JBD			/* Has an attached ext3 journal_head */
 307	  = BH_PrivateStart,
 308	BH_JWrite,		/* Being written to log (@@@ DEBUGGING) */
 309	BH_Freed,		/* Has been freed (truncated) */
 310	BH_Revoked,		/* Has been revoked from the log */
 311	BH_RevokeValid,		/* Revoked flag is valid */
 312	BH_JBDDirty,		/* Is dirty but journaled */
 313	BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
 314	BH_Shadow,		/* IO on shadow buffer is running */
 315	BH_Verified,		/* Metadata block has been verified ok */
 316	BH_JBDPrivateStart,	/* First bit available for private use by FS */
 317};
 318
 319BUFFER_FNS(JBD, jbd)
 320BUFFER_FNS(JWrite, jwrite)
 321BUFFER_FNS(JBDDirty, jbddirty)
 322TAS_BUFFER_FNS(JBDDirty, jbddirty)
 323BUFFER_FNS(Revoked, revoked)
 324TAS_BUFFER_FNS(Revoked, revoked)
 325BUFFER_FNS(RevokeValid, revokevalid)
 326TAS_BUFFER_FNS(RevokeValid, revokevalid)
 327BUFFER_FNS(Freed, freed)
 328BUFFER_FNS(Shadow, shadow)
 329BUFFER_FNS(Verified, verified)
 330
 331static inline struct buffer_head *jh2bh(struct journal_head *jh)
 332{
 333	return jh->b_bh;
 334}
 335
 336static inline struct journal_head *bh2jh(struct buffer_head *bh)
 337{
 338	return bh->b_private;
 339}
 340
 341static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
 342{
 343	bit_spin_lock(BH_JournalHead, &bh->b_state);
 344}
 345
 346static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
 347{
 348	bit_spin_unlock(BH_JournalHead, &bh->b_state);
 349}
 350
 351#define J_ASSERT(assert)	BUG_ON(!(assert))
 352
 353#define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
 354#define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
 355
 356#if defined(JBD2_PARANOID_IOFAIL)
 357#define J_EXPECT(expr, why...)		J_ASSERT(expr)
 358#define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
 359#define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
 360#else
 361#define __journal_expect(expr, why...)					     \
 362	({								     \
 363		int val = (expr);					     \
 364		if (!val) {						     \
 365			printk(KERN_ERR					     \
 366			       "JBD2 unexpected failure: %s: %s;\n",	     \
 367			       __func__, #expr);			     \
 368			printk(KERN_ERR why "\n");			     \
 369		}							     \
 370		val;							     \
 371	})
 372#define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
 373#define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
 374#define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
 375#endif
 376
 377/* Flags in jbd_inode->i_flags */
 378#define __JI_COMMIT_RUNNING 0
 379#define __JI_WRITE_DATA 1
 380#define __JI_WAIT_DATA 2
 381
 382/*
 383 * Commit of the inode data in progress. We use this flag to protect us from
 384 * concurrent deletion of inode. We cannot use reference to inode for this
 385 * since we cannot afford doing last iput() on behalf of kjournald
 386 */
 387#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
 388/* Write allocated dirty buffers in this inode before commit */
 389#define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
 390/* Wait for outstanding data writes for this inode before commit */
 391#define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
 392
 393/**
 394 * struct jbd2_inode - The jbd_inode type is the structure linking inodes in
 395 * ordered mode present in a transaction so that we can sync them during commit.
 396 */
 397struct jbd2_inode {
 398	/**
 399	 * @i_transaction:
 400	 *
 401	 * Which transaction does this inode belong to? Either the running
 402	 * transaction or the committing one. [j_list_lock]
 403	 */
 404	transaction_t *i_transaction;
 405
 406	/**
 407	 * @i_next_transaction:
 408	 *
 409	 * Pointer to the running transaction modifying inode's data in case
 410	 * there is already a committing transaction touching it. [j_list_lock]
 411	 */
 412	transaction_t *i_next_transaction;
 413
 414	/**
 415	 * @i_list: List of inodes in the i_transaction [j_list_lock]
 416	 */
 417	struct list_head i_list;
 418
 419	/**
 420	 * @i_vfs_inode:
 421	 *
 422	 * VFS inode this inode belongs to [constant for lifetime of structure]
 423	 */
 424	struct inode *i_vfs_inode;
 425
 426	/**
 427	 * @i_flags: Flags of inode [j_list_lock]
 428	 */
 429	unsigned long i_flags;
 430
 431	/**
 432	 * @i_dirty_start:
 433	 *
 434	 * Offset in bytes where the dirty range for this inode starts.
 435	 * [j_list_lock]
 436	 */
 437	loff_t i_dirty_start;
 438
 439	/**
 440	 * @i_dirty_end:
 441	 *
 442	 * Inclusive offset in bytes where the dirty range for this inode
 443	 * ends. [j_list_lock]
 444	 */
 445	loff_t i_dirty_end;
 446};
 447
 448struct jbd2_revoke_table_s;
 449
 450/**
 451 * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete
 452 *     type associated with handle_t.
 453 * @h_transaction: Which compound transaction is this update a part of?
 454 * @h_journal: Which journal handle belongs to - used iff h_reserved set.
 455 * @h_rsv_handle: Handle reserved for finishing the logical operation.
 456 * @h_total_credits: Number of remaining buffers we are allowed to add to
 457 *	journal. These are dirty buffers and revoke descriptor blocks.
 458 * @h_revoke_credits: Number of remaining revoke records available for handle
 459 * @h_ref: Reference count on this handle.
 460 * @h_err: Field for caller's use to track errors through large fs operations.
 461 * @h_sync: Flag for sync-on-close.
 462 * @h_reserved: Flag for handle for reserved credits.
 463 * @h_aborted: Flag indicating fatal error on handle.
 464 * @h_type: For handle statistics.
 465 * @h_line_no: For handle statistics.
 466 * @h_start_jiffies: Handle Start time.
 467 * @h_requested_credits: Holds @h_total_credits after handle is started.
 468 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
 469 * @saved_alloc_context: Saved context while transaction is open.
 470 **/
 471
 472/* Docbook can't yet cope with the bit fields, but will leave the documentation
 473 * in so it can be fixed later.
 474 */
 475
 476struct jbd2_journal_handle
 477{
 478	union {
 479		transaction_t	*h_transaction;
 480		/* Which journal handle belongs to - used iff h_reserved set */
 481		journal_t	*h_journal;
 482	};
 483
 484	handle_t		*h_rsv_handle;
 485	int			h_total_credits;
 486	int			h_revoke_credits;
 487	int			h_revoke_credits_requested;
 488	int			h_ref;
 489	int			h_err;
 490
 491	/* Flags [no locking] */
 492	unsigned int	h_sync:		1;
 493	unsigned int	h_reserved:	1;
 494	unsigned int	h_aborted:	1;
 495	unsigned int	h_type:		8;
 496	unsigned int	h_line_no:	16;
 497
 498	unsigned long		h_start_jiffies;
 499	unsigned int		h_requested_credits;
 500
 501	unsigned int		saved_alloc_context;
 502};
 503
 504
 505/*
 506 * Some stats for checkpoint phase
 507 */
 508struct transaction_chp_stats_s {
 509	unsigned long		cs_chp_time;
 510	__u32			cs_forced_to_close;
 511	__u32			cs_written;
 512	__u32			cs_dropped;
 513};
 514
 515/* The transaction_t type is the guts of the journaling mechanism.  It
 516 * tracks a compound transaction through its various states:
 517 *
 518 * RUNNING:	accepting new updates
 519 * LOCKED:	Updates still running but we don't accept new ones
 520 * RUNDOWN:	Updates are tidying up but have finished requesting
 521 *		new buffers to modify (state not used for now)
 522 * FLUSH:       All updates complete, but we are still writing to disk
 523 * COMMIT:      All data on disk, writing commit record
 524 * FINISHED:	We still have to keep the transaction for checkpointing.
 525 *
 526 * The transaction keeps track of all of the buffers modified by a
 527 * running transaction, and all of the buffers committed but not yet
 528 * flushed to home for finished transactions.
 529 * (Locking Documentation improved by LockDoc)
 530 */
 531
 532/*
 533 * Lock ranking:
 534 *
 535 *    j_list_lock
 536 *      ->jbd_lock_bh_journal_head()	(This is "innermost")
 537 *
 538 *    j_state_lock
 539 *    ->b_state_lock
 540 *
 541 *    b_state_lock
 542 *    ->j_list_lock
 543 *
 544 *    j_state_lock
 545 *    ->j_list_lock			(journal_unmap_buffer)
 546 *
 547 */
 548
 549struct transaction_s
 550{
 551	/* Pointer to the journal for this transaction. [no locking] */
 552	journal_t		*t_journal;
 553
 554	/* Sequence number for this transaction [no locking] */
 555	tid_t			t_tid;
 556
 557	/*
 558	 * Transaction's current state
 559	 * [no locking - only kjournald2 alters this]
 560	 * [j_list_lock] guards transition of a transaction into T_FINISHED
 561	 * state and subsequent call of __jbd2_journal_drop_transaction()
 562	 * FIXME: needs barriers
 563	 * KLUDGE: [use j_state_lock]
 564	 */
 565	enum {
 566		T_RUNNING,
 567		T_LOCKED,
 568		T_SWITCH,
 569		T_FLUSH,
 570		T_COMMIT,
 571		T_COMMIT_DFLUSH,
 572		T_COMMIT_JFLUSH,
 573		T_COMMIT_CALLBACK,
 574		T_FINISHED
 575	}			t_state;
 576
 577	/*
 578	 * Where in the log does this transaction's commit start? [no locking]
 579	 */
 580	unsigned long		t_log_start;
 581
 582	/*
 583	 * Number of buffers on the t_buffers list [j_list_lock, no locks
 584	 * needed for jbd2 thread]
 585	 */
 586	int			t_nr_buffers;
 587
 588	/*
 589	 * Doubly-linked circular list of all buffers reserved but not yet
 590	 * modified by this transaction [j_list_lock, no locks needed fo
 591	 * jbd2 thread]
 592	 */
 593	struct journal_head	*t_reserved_list;
 594
 595	/*
 596	 * Doubly-linked circular list of all metadata buffers owned by this
 597	 * transaction [j_list_lock, no locks needed for jbd2 thread]
 598	 */
 599	struct journal_head	*t_buffers;
 600
 601	/*
 602	 * Doubly-linked circular list of all forget buffers (superseded
 603	 * buffers which we can un-checkpoint once this transaction commits)
 604	 * [j_list_lock]
 605	 */
 606	struct journal_head	*t_forget;
 607
 608	/*
 609	 * Doubly-linked circular list of all buffers still to be flushed before
 610	 * this transaction can be checkpointed. [j_list_lock]
 611	 */
 612	struct journal_head	*t_checkpoint_list;
 613
 614	/*
 615	 * Doubly-linked circular list of metadata buffers being
 616	 * shadowed by log IO.  The IO buffers on the iobuf list and
 617	 * the shadow buffers on this list match each other one for
 618	 * one at all times. [j_list_lock, no locks needed for jbd2
 619	 * thread]
 620	 */
 621	struct journal_head	*t_shadow_list;
 622
 623	/*
 624	 * List of inodes associated with the transaction; e.g., ext4 uses
 625	 * this to track inodes in data=ordered and data=journal mode that
 626	 * need special handling on transaction commit; also used by ocfs2.
 627	 * [j_list_lock]
 628	 */
 629	struct list_head	t_inode_list;
 630
 631	/*
 632	 * Longest time some handle had to wait for running transaction
 633	 */
 634	unsigned long		t_max_wait;
 635
 636	/*
 637	 * When transaction started
 638	 */
 639	unsigned long		t_start;
 640
 641	/*
 642	 * When commit was requested [j_state_lock]
 643	 */
 644	unsigned long		t_requested;
 645
 646	/*
 647	 * Checkpointing stats [j_list_lock]
 648	 */
 649	struct transaction_chp_stats_s t_chp_stats;
 650
 651	/*
 652	 * Number of outstanding updates running on this transaction
 653	 * [none]
 654	 */
 655	atomic_t		t_updates;
 656
 657	/*
 658	 * Number of blocks reserved for this transaction in the journal.
 659	 * This is including all credits reserved when starting transaction
 660	 * handles as well as all journal descriptor blocks needed for this
 661	 * transaction. [none]
 662	 */
 663	atomic_t		t_outstanding_credits;
 664
 665	/*
 666	 * Number of revoke records for this transaction added by already
 667	 * stopped handles. [none]
 668	 */
 669	atomic_t		t_outstanding_revokes;
 670
 671	/*
 672	 * How many handles used this transaction? [none]
 673	 */
 674	atomic_t		t_handle_count;
 675
 676	/*
 677	 * Forward and backward links for the circular list of all transactions
 678	 * awaiting checkpoint. [j_list_lock]
 679	 */
 680	transaction_t		*t_cpnext, *t_cpprev;
 681
 682	/*
 683	 * When will the transaction expire (become due for commit), in jiffies?
 684	 * [no locking]
 685	 */
 686	unsigned long		t_expires;
 687
 688	/*
 689	 * When this transaction started, in nanoseconds [no locking]
 690	 */
 691	ktime_t			t_start_time;
 692
 693	/*
 694	 * This transaction is being forced and some process is
 695	 * waiting for it to finish.
 696	 */
 697	unsigned int t_synchronous_commit:1;
 698
 699	/* Disk flush needs to be sent to fs partition [no locking] */
 700	int			t_need_data_flush;
 701};
 702
 703struct transaction_run_stats_s {
 704	unsigned long		rs_wait;
 705	unsigned long		rs_request_delay;
 706	unsigned long		rs_running;
 707	unsigned long		rs_locked;
 708	unsigned long		rs_flushing;
 709	unsigned long		rs_logging;
 710
 711	__u32			rs_handle_count;
 712	__u32			rs_blocks;
 713	__u32			rs_blocks_logged;
 714};
 715
 716struct transaction_stats_s {
 717	unsigned long		ts_tid;
 718	unsigned long		ts_requested;
 719	struct transaction_run_stats_s run;
 720};
 721
 722static inline unsigned long
 723jbd2_time_diff(unsigned long start, unsigned long end)
 724{
 725	if (end >= start)
 726		return end - start;
 727
 728	return end + (MAX_JIFFY_OFFSET - start);
 729}
 730
 731#define JBD2_NR_BATCH	64
 732
 733enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
 734
 735#define JBD2_FC_REPLAY_STOP	0
 736#define JBD2_FC_REPLAY_CONTINUE	1
 737
 738/**
 739 * struct journal_s - The journal_s type is the concrete type associated with
 740 *     journal_t.
 741 */
 742struct journal_s
 743{
 744	/**
 745	 * @j_flags: General journaling state flags [j_state_lock,
 746	 * no lock for quick racy checks]
 747	 */
 748	unsigned long		j_flags;
 749
 750	/**
 751	 * @j_errno:
 752	 *
 753	 * Is there an outstanding uncleared error on the journal (from a prior
 754	 * abort)? [j_state_lock]
 755	 */
 756	int			j_errno;
 757
 758	/**
 759	 * @j_abort_mutex: Lock the whole aborting procedure.
 760	 */
 761	struct mutex		j_abort_mutex;
 762
 763	/**
 764	 * @j_sb_buffer: The first part of the superblock buffer.
 765	 */
 766	struct buffer_head	*j_sb_buffer;
 767
 768	/**
 769	 * @j_superblock: The second part of the superblock buffer.
 770	 */
 771	journal_superblock_t	*j_superblock;
 772
 773	/**
 774	 * @j_state_lock: Protect the various scalars in the journal.
 775	 */
 776	rwlock_t		j_state_lock;
 777
 778	/**
 779	 * @j_barrier_count:
 780	 *
 781	 * Number of processes waiting to create a barrier lock [j_state_lock,
 782	 * no lock for quick racy checks]
 783	 */
 784	int			j_barrier_count;
 785
 786	/**
 787	 * @j_barrier: The barrier lock itself.
 788	 */
 789	struct mutex		j_barrier;
 790
 791	/**
 792	 * @j_running_transaction:
 793	 *
 794	 * Transactions: The current running transaction...
 795	 * [j_state_lock, no lock for quick racy checks] [caller holding
 796	 * open handle]
 797	 */
 798	transaction_t		*j_running_transaction;
 799
 800	/**
 801	 * @j_committing_transaction:
 802	 *
 803	 * the transaction we are pushing to disk
 804	 * [j_state_lock] [caller holding open handle]
 805	 */
 806	transaction_t		*j_committing_transaction;
 807
 808	/**
 809	 * @j_checkpoint_transactions:
 810	 *
 811	 * ... and a linked circular list of all transactions waiting for
 812	 * checkpointing. [j_list_lock]
 813	 */
 814	transaction_t		*j_checkpoint_transactions;
 815
 816	/**
 817	 * @j_wait_transaction_locked:
 818	 *
 819	 * Wait queue for waiting for a locked transaction to start committing,
 820	 * or for a barrier lock to be released.
 821	 */
 822	wait_queue_head_t	j_wait_transaction_locked;
 823
 824	/**
 825	 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
 826	 */
 827	wait_queue_head_t	j_wait_done_commit;
 828
 829	/**
 830	 * @j_wait_commit: Wait queue to trigger commit.
 831	 */
 832	wait_queue_head_t	j_wait_commit;
 833
 834	/**
 835	 * @j_wait_updates: Wait queue to wait for updates to complete.
 836	 */
 837	wait_queue_head_t	j_wait_updates;
 838
 839	/**
 840	 * @j_wait_reserved:
 841	 *
 842	 * Wait queue to wait for reserved buffer credits to drop.
 843	 */
 844	wait_queue_head_t	j_wait_reserved;
 845
 846	/**
 847	 * @j_fc_wait:
 848	 *
 849	 * Wait queue to wait for completion of async fast commits.
 850	 */
 851	wait_queue_head_t	j_fc_wait;
 852
 853	/**
 854	 * @j_checkpoint_mutex:
 855	 *
 856	 * Semaphore for locking against concurrent checkpoints.
 857	 */
 858	struct mutex		j_checkpoint_mutex;
 859
 860	/**
 861	 * @j_chkpt_bhs:
 862	 *
 863	 * List of buffer heads used by the checkpoint routine.  This
 864	 * was moved from jbd2_log_do_checkpoint() to reduce stack
 865	 * usage.  Access to this array is controlled by the
 866	 * @j_checkpoint_mutex.  [j_checkpoint_mutex]
 867	 */
 868	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
 869
 870	/**
 871	 * @j_shrinker:
 872	 *
 873	 * Journal head shrinker, reclaim buffer's journal head which
 874	 * has been written back.
 875	 */
 876	struct shrinker		*j_shrinker;
 877
 878	/**
 879	 * @j_checkpoint_jh_count:
 880	 *
 881	 * Number of journal buffers on the checkpoint list. [j_list_lock]
 882	 */
 883	struct percpu_counter	j_checkpoint_jh_count;
 884
 885	/**
 886	 * @j_shrink_transaction:
 887	 *
 888	 * Record next transaction will shrink on the checkpoint list.
 889	 * [j_list_lock]
 890	 */
 891	transaction_t		*j_shrink_transaction;
 892
 893	/**
 894	 * @j_head:
 895	 *
 896	 * Journal head: identifies the first unused block in the journal.
 897	 * [j_state_lock]
 898	 */
 899	unsigned long		j_head;
 900
 901	/**
 902	 * @j_tail:
 903	 *
 904	 * Journal tail: identifies the oldest still-used block in the journal.
 905	 * [j_state_lock]
 906	 */
 907	unsigned long		j_tail;
 908
 909	/**
 910	 * @j_free:
 911	 *
 912	 * Journal free: how many free blocks are there in the journal?
 913	 * [j_state_lock]
 914	 */
 915	unsigned long		j_free;
 916
 917	/**
 918	 * @j_first:
 919	 *
 920	 * The block number of the first usable block in the journal
 921	 * [j_state_lock].
 922	 */
 923	unsigned long		j_first;
 924
 925	/**
 926	 * @j_last:
 927	 *
 928	 * The block number one beyond the last usable block in the journal
 929	 * [j_state_lock].
 930	 */
 931	unsigned long		j_last;
 932
 933	/**
 934	 * @j_fc_first:
 935	 *
 936	 * The block number of the first fast commit block in the journal
 937	 * [j_state_lock].
 938	 */
 939	unsigned long		j_fc_first;
 940
 941	/**
 942	 * @j_fc_off:
 943	 *
 944	 * Number of fast commit blocks currently allocated. Accessed only
 945	 * during fast commit. Currently only process can do fast commit, so
 946	 * this field is not protected by any lock.
 947	 */
 948	unsigned long		j_fc_off;
 949
 950	/**
 951	 * @j_fc_last:
 952	 *
 953	 * The block number one beyond the last fast commit block in the journal
 954	 * [j_state_lock].
 955	 */
 956	unsigned long		j_fc_last;
 957
 958	/**
 959	 * @j_dev: Device where we store the journal.
 960	 */
 961	struct block_device	*j_dev;
 962
 963	/**
 964	 * @j_blocksize: Block size for the location where we store the journal.
 965	 */
 966	int			j_blocksize;
 967
 968	/**
 969	 * @j_blk_offset:
 970	 *
 971	 * Starting block offset into the device where we store the journal.
 972	 */
 973	unsigned long long	j_blk_offset;
 974
 975	/**
 976	 * @j_devname: Journal device name.
 977	 */
 978	char			j_devname[BDEVNAME_SIZE+24];
 979
 980	/**
 981	 * @j_fs_dev:
 982	 *
 983	 * Device which holds the client fs.  For internal journal this will be
 984	 * equal to j_dev.
 985	 */
 986	struct block_device	*j_fs_dev;
 987
 988	/**
 989	 * @j_fs_dev_wb_err:
 990	 *
 991	 * Records the errseq of the client fs's backing block device.
 992	 */
 993	errseq_t		j_fs_dev_wb_err;
 994
 995	/**
 996	 * @j_total_len: Total maximum capacity of the journal region on disk.
 997	 */
 998	unsigned int		j_total_len;
 999
1000	/**
1001	 * @j_reserved_credits:
1002	 *
1003	 * Number of buffers reserved from the running transaction.
1004	 */
1005	atomic_t		j_reserved_credits;
1006
1007	/**
1008	 * @j_list_lock: Protects the buffer lists and internal buffer state.
1009	 */
1010	spinlock_t		j_list_lock;
1011
1012	/**
1013	 * @j_inode:
1014	 *
1015	 * Optional inode where we store the journal.  If present, all
1016	 * journal block numbers are mapped into this inode via bmap().
1017	 */
1018	struct inode		*j_inode;
1019
1020	/**
1021	 * @j_tail_sequence:
1022	 *
1023	 * Sequence number of the oldest transaction in the log [j_state_lock]
1024	 */
1025	tid_t			j_tail_sequence;
1026
1027	/**
1028	 * @j_transaction_sequence:
1029	 *
1030	 * Sequence number of the next transaction to grant [j_state_lock]
1031	 */
1032	tid_t			j_transaction_sequence;
1033
1034	/**
1035	 * @j_commit_sequence:
1036	 *
1037	 * Sequence number of the most recently committed transaction
1038	 * [j_state_lock, no lock for quick racy checks]
1039	 */
1040	tid_t			j_commit_sequence;
1041
1042	/**
1043	 * @j_commit_request:
1044	 *
1045	 * Sequence number of the most recent transaction wanting commit
1046	 * [j_state_lock, no lock for quick racy checks]
1047	 */
1048	tid_t			j_commit_request;
1049
1050	/**
1051	 * @j_uuid:
1052	 *
1053	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1054	 * backed by this journal.  This will eventually be replaced by an array
1055	 * of uuids, allowing us to index multiple devices within a single
1056	 * journal and to perform atomic updates across them.
1057	 */
1058	__u8			j_uuid[16];
1059
1060	/**
1061	 * @j_task: Pointer to the current commit thread for this journal.
1062	 */
1063	struct task_struct	*j_task;
1064
1065	/**
1066	 * @j_max_transaction_buffers:
1067	 *
1068	 * Maximum number of metadata buffers to allow in a single compound
1069	 * commit transaction.
1070	 */
1071	int			j_max_transaction_buffers;
1072
1073	/**
1074	 * @j_revoke_records_per_block:
1075	 *
1076	 * Number of revoke records that fit in one descriptor block.
1077	 */
1078	int			j_revoke_records_per_block;
1079
1080	/**
1081	 * @j_transaction_overhead_buffers:
1082	 *
1083	 * Number of blocks each transaction needs for its own bookkeeping
1084	 */
1085	int			j_transaction_overhead_buffers;
1086
1087	/**
1088	 * @j_commit_interval:
1089	 *
1090	 * What is the maximum transaction lifetime before we begin a commit?
1091	 */
1092	unsigned long		j_commit_interval;
1093
1094	/**
1095	 * @j_commit_timer: The timer used to wakeup the commit thread.
1096	 */
1097	struct timer_list	j_commit_timer;
1098
1099	/**
1100	 * @j_revoke_lock: Protect the revoke table.
1101	 */
1102	spinlock_t		j_revoke_lock;
1103
1104	/**
1105	 * @j_revoke:
1106	 *
1107	 * The revoke table - maintains the list of revoked blocks in the
1108	 * current transaction.
1109	 */
1110	struct jbd2_revoke_table_s *j_revoke;
1111
1112	/**
1113	 * @j_revoke_table: Alternate revoke tables for j_revoke.
1114	 */
1115	struct jbd2_revoke_table_s *j_revoke_table[2];
1116
1117	/**
1118	 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1119	 */
1120	struct buffer_head	**j_wbuf;
1121
1122	/**
1123	 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only
1124	 * during a fast commit. Currently only process can do fast commit, so
1125	 * this field is not protected by any lock.
1126	 */
1127	struct buffer_head	**j_fc_wbuf;
1128
1129	/**
1130	 * @j_wbufsize:
1131	 *
1132	 * Size of @j_wbuf array.
1133	 */
1134	int			j_wbufsize;
1135
1136	/**
1137	 * @j_fc_wbufsize:
1138	 *
1139	 * Size of @j_fc_wbuf array.
1140	 */
1141	int			j_fc_wbufsize;
1142
1143	/**
1144	 * @j_last_sync_writer:
1145	 *
1146	 * The pid of the last person to run a synchronous operation
1147	 * through the journal.
1148	 */
1149	pid_t			j_last_sync_writer;
1150
1151	/**
1152	 * @j_average_commit_time:
1153	 *
1154	 * The average amount of time in nanoseconds it takes to commit a
1155	 * transaction to disk. [j_state_lock]
1156	 */
1157	u64			j_average_commit_time;
1158
1159	/**
1160	 * @j_min_batch_time:
1161	 *
1162	 * Minimum time that we should wait for additional filesystem operations
1163	 * to get batched into a synchronous handle in microseconds.
1164	 */
1165	u32			j_min_batch_time;
1166
1167	/**
1168	 * @j_max_batch_time:
1169	 *
1170	 * Maximum time that we should wait for additional filesystem operations
1171	 * to get batched into a synchronous handle in microseconds.
1172	 */
1173	u32			j_max_batch_time;
1174
1175	/**
1176	 * @j_commit_callback:
1177	 *
1178	 * This function is called when a transaction is closed.
1179	 */
1180	void			(*j_commit_callback)(journal_t *,
1181						     transaction_t *);
1182
1183	/**
1184	 * @j_submit_inode_data_buffers:
1185	 *
1186	 * This function is called for all inodes associated with the
1187	 * committing transaction marked with JI_WRITE_DATA flag
1188	 * before we start to write out the transaction to the journal.
1189	 */
1190	int			(*j_submit_inode_data_buffers)
1191					(struct jbd2_inode *);
1192
1193	/**
1194	 * @j_finish_inode_data_buffers:
1195	 *
1196	 * This function is called for all inodes associated with the
1197	 * committing transaction marked with JI_WAIT_DATA flag
1198	 * after we have written the transaction to the journal
1199	 * but before we write out the commit block.
1200	 */
1201	int			(*j_finish_inode_data_buffers)
1202					(struct jbd2_inode *);
1203
1204	/*
1205	 * Journal statistics
1206	 */
1207
1208	/**
1209	 * @j_history_lock: Protect the transactions statistics history.
1210	 */
1211	spinlock_t		j_history_lock;
1212
1213	/**
1214	 * @j_proc_entry: procfs entry for the jbd statistics directory.
1215	 */
1216	struct proc_dir_entry	*j_proc_entry;
1217
1218	/**
1219	 * @j_stats: Overall statistics.
1220	 */
1221	struct transaction_stats_s j_stats;
1222
1223	/**
1224	 * @j_failed_commit: Failed journal commit ID.
1225	 */
1226	unsigned int		j_failed_commit;
1227
1228	/**
1229	 * @j_private:
1230	 *
1231	 * An opaque pointer to fs-private information.  ext3 puts its
1232	 * superblock pointer here.
1233	 */
1234	void *j_private;
1235
1236	/**
1237	 * @j_csum_seed:
1238	 *
1239	 * Precomputed journal UUID checksum for seeding other checksums.
1240	 */
1241	__u32 j_csum_seed;
1242
1243#ifdef CONFIG_DEBUG_LOCK_ALLOC
1244	/**
1245	 * @j_trans_commit_map:
1246	 *
1247	 * Lockdep entity to track transaction commit dependencies. Handles
1248	 * hold this "lock" for read, when we wait for commit, we acquire the
1249	 * "lock" for writing. This matches the properties of jbd2 journalling
1250	 * where the running transaction has to wait for all handles to be
1251	 * dropped to commit that transaction and also acquiring a handle may
1252	 * require transaction commit to finish.
1253	 */
1254	struct lockdep_map	j_trans_commit_map;
1255#endif
1256	/**
1257	 * @jbd2_trans_commit_key:
1258	 *
1259	 * "struct lock_class_key" for @j_trans_commit_map
1260	 */
1261	struct lock_class_key	jbd2_trans_commit_key;
1262
1263	/**
1264	 * @j_fc_cleanup_callback:
1265	 *
1266	 * Clean-up after fast commit or full commit. JBD2 calls this function
1267	 * after every commit operation.
1268	 */
1269	void (*j_fc_cleanup_callback)(struct journal_s *journal, int full, tid_t tid);
1270
1271	/**
1272	 * @j_fc_replay_callback:
1273	 *
1274	 * File-system specific function that performs replay of a fast
1275	 * commit. JBD2 calls this function for each fast commit block found in
1276	 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1277	 * to indicate that the block was processed correctly and more fast
1278	 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1279	 * indicates the end of replay (no more blocks remaining). A negative
1280	 * return value indicates error.
1281	 */
1282	int (*j_fc_replay_callback)(struct journal_s *journal,
1283				    struct buffer_head *bh,
1284				    enum passtype pass, int off,
1285				    tid_t expected_commit_id);
1286
1287	/**
1288	 * @j_bmap:
1289	 *
1290	 * Bmap function that should be used instead of the generic
1291	 * VFS bmap function.
1292	 */
1293	int (*j_bmap)(struct journal_s *journal, sector_t *block);
1294};
1295
1296#define jbd2_might_wait_for_commit(j) \
1297	do { \
1298		rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1299		rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1300	} while (0)
1301
1302/*
1303 * We can support any known requested features iff the
1304 * superblock is not in version 1.  Otherwise we fail to support any
1305 * extended sb features.
1306 */
1307static inline bool jbd2_format_support_feature(journal_t *j)
1308{
1309	return j->j_superblock->s_header.h_blocktype !=
1310					cpu_to_be32(JBD2_SUPERBLOCK_V1);
1311}
1312
1313/* journal feature predicate functions */
1314#define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1315static inline bool jbd2_has_feature_##name(journal_t *j) \
1316{ \
1317	return (jbd2_format_support_feature(j) && \
1318		((j)->j_superblock->s_feature_compat & \
1319		 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1320} \
1321static inline void jbd2_set_feature_##name(journal_t *j) \
1322{ \
1323	(j)->j_superblock->s_feature_compat |= \
1324		cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1325} \
1326static inline void jbd2_clear_feature_##name(journal_t *j) \
1327{ \
1328	(j)->j_superblock->s_feature_compat &= \
1329		~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1330}
1331
1332#define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1333static inline bool jbd2_has_feature_##name(journal_t *j) \
1334{ \
1335	return (jbd2_format_support_feature(j) && \
1336		((j)->j_superblock->s_feature_ro_compat & \
1337		 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1338} \
1339static inline void jbd2_set_feature_##name(journal_t *j) \
1340{ \
1341	(j)->j_superblock->s_feature_ro_compat |= \
1342		cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1343} \
1344static inline void jbd2_clear_feature_##name(journal_t *j) \
1345{ \
1346	(j)->j_superblock->s_feature_ro_compat &= \
1347		~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1348}
1349
1350#define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1351static inline bool jbd2_has_feature_##name(journal_t *j) \
1352{ \
1353	return (jbd2_format_support_feature(j) && \
1354		((j)->j_superblock->s_feature_incompat & \
1355		 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1356} \
1357static inline void jbd2_set_feature_##name(journal_t *j) \
1358{ \
1359	(j)->j_superblock->s_feature_incompat |= \
1360		cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1361} \
1362static inline void jbd2_clear_feature_##name(journal_t *j) \
1363{ \
1364	(j)->j_superblock->s_feature_incompat &= \
1365		~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1366}
1367
1368JBD2_FEATURE_COMPAT_FUNCS(checksum,		CHECKSUM)
1369
1370JBD2_FEATURE_INCOMPAT_FUNCS(revoke,		REVOKE)
1371JBD2_FEATURE_INCOMPAT_FUNCS(64bit,		64BIT)
1372JBD2_FEATURE_INCOMPAT_FUNCS(async_commit,	ASYNC_COMMIT)
1373JBD2_FEATURE_INCOMPAT_FUNCS(csum2,		CSUM_V2)
1374JBD2_FEATURE_INCOMPAT_FUNCS(csum3,		CSUM_V3)
1375JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit,	FAST_COMMIT)
1376
1377/* Journal high priority write IO operation flags */
1378#define JBD2_JOURNAL_REQ_FLAGS		(REQ_META | REQ_SYNC | REQ_IDLE)
1379
1380/*
1381 * Journal flag definitions
1382 */
1383#define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1384#define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1385#define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1386#define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1387#define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1388#define JBD2_BARRIER	0x020	/* Use IDE barriers */
1389#define JBD2_CYCLE_RECORD		0x080	/* Journal cycled record log on
1390						 * clean and empty filesystem
1391						 * logging area */
1392#define JBD2_FAST_COMMIT_ONGOING	0x100	/* Fast commit is ongoing */
1393#define JBD2_FULL_COMMIT_ONGOING	0x200	/* Full commit is ongoing */
1394#define JBD2_JOURNAL_FLUSH_DISCARD	0x0001
1395#define JBD2_JOURNAL_FLUSH_ZEROOUT	0x0002
1396#define JBD2_JOURNAL_FLUSH_VALID	(JBD2_JOURNAL_FLUSH_DISCARD | \
1397					JBD2_JOURNAL_FLUSH_ZEROOUT)
1398
1399/*
1400 * Function declarations for the journaling transaction and buffer
1401 * management
1402 */
1403
1404/* Filing buffers */
1405extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1406extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1407extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1408extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1409static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1410{
1411	list_add_tail(&bh->b_assoc_buffers, head);
1412}
1413static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1414{
1415	list_del_init(&bh->b_assoc_buffers);
1416}
1417
1418/* Log buffer allocation */
1419struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1420void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1421int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1422int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1423			      unsigned long *block);
1424int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1425void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1426
1427/* Commit management */
1428extern void jbd2_journal_commit_transaction(journal_t *);
1429
1430/* Checkpoint list management */
1431enum jbd2_shrink_type {JBD2_SHRINK_DESTROY, JBD2_SHRINK_BUSY_STOP, JBD2_SHRINK_BUSY_SKIP};
1432
1433void __jbd2_journal_clean_checkpoint_list(journal_t *journal, enum jbd2_shrink_type type);
1434unsigned long jbd2_journal_shrink_checkpoint_list(journal_t *journal, unsigned long *nr_to_scan);
1435int __jbd2_journal_remove_checkpoint(struct journal_head *);
1436int jbd2_journal_try_remove_checkpoint(struct journal_head *jh);
1437void jbd2_journal_destroy_checkpoint(journal_t *journal);
1438void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1439
1440
1441/*
1442 * Triggers
1443 */
1444
1445struct jbd2_buffer_trigger_type {
1446	/*
1447	 * Fired a the moment data to write to the journal are known to be
1448	 * stable - so either at the moment b_frozen_data is created or just
1449	 * before a buffer is written to the journal.  mapped_data is a mapped
1450	 * buffer that is the frozen data for commit.
1451	 */
1452	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1453			 struct buffer_head *bh, void *mapped_data,
1454			 size_t size);
1455
1456	/*
1457	 * Fired during journal abort for dirty buffers that will not be
1458	 * committed.
1459	 */
1460	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1461			struct buffer_head *bh);
1462};
1463
1464extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1465				       void *mapped_data,
1466				       struct jbd2_buffer_trigger_type *triggers);
1467extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1468				      struct jbd2_buffer_trigger_type *triggers);
1469
1470/* Buffer IO */
1471extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1472					      struct journal_head *jh_in,
1473					      struct buffer_head **bh_out,
1474					      sector_t blocknr);
1475
1476/* Transaction cache support */
1477extern void jbd2_journal_destroy_transaction_cache(void);
1478extern int __init jbd2_journal_init_transaction_cache(void);
1479extern void jbd2_journal_free_transaction(transaction_t *);
1480
1481/*
1482 * Journal locking.
1483 *
1484 * We need to lock the journal during transaction state changes so that nobody
1485 * ever tries to take a handle on the running transaction while we are in the
1486 * middle of moving it to the commit phase.  j_state_lock does this.
1487 *
1488 * Note that the locking is completely interrupt unsafe.  We never touch
1489 * journal structures from interrupts.
1490 */
1491
1492static inline handle_t *journal_current_handle(void)
1493{
1494	return current->journal_info;
1495}
1496
1497/* The journaling code user interface:
1498 *
1499 * Create and destroy handles
1500 * Register buffer modifications against the current transaction.
1501 */
1502
1503extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1504extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1505				     int revoke_records, gfp_t gfp_mask,
1506				     unsigned int type, unsigned int line_no);
1507extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1508extern int	 jbd2__journal_restart(handle_t *, int nblocks,
1509				       int revoke_records, gfp_t gfp_mask);
1510extern int	 jbd2_journal_start_reserved(handle_t *handle,
1511				unsigned int type, unsigned int line_no);
1512extern void	 jbd2_journal_free_reserved(handle_t *handle);
1513extern int	 jbd2_journal_extend(handle_t *handle, int nblocks,
1514				     int revoke_records);
1515extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1516extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1517extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1518void		 jbd2_journal_set_triggers(struct buffer_head *,
1519					   struct jbd2_buffer_trigger_type *type);
1520extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1521extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1522int jbd2_journal_invalidate_folio(journal_t *, struct folio *,
1523					size_t offset, size_t length);
1524bool jbd2_journal_try_to_free_buffers(journal_t *journal, struct folio *folio);
1525extern int	 jbd2_journal_stop(handle_t *);
1526extern int	 jbd2_journal_flush(journal_t *journal, unsigned int flags);
1527extern void	 jbd2_journal_lock_updates (journal_t *);
1528extern void	 jbd2_journal_unlock_updates (journal_t *);
1529
1530void jbd2_journal_wait_updates(journal_t *);
1531
1532extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1533				struct block_device *fs_dev,
1534				unsigned long long start, int len, int bsize);
1535extern journal_t * jbd2_journal_init_inode (struct inode *);
1536extern int	   jbd2_journal_update_format (journal_t *);
1537extern int	   jbd2_journal_check_used_features
1538		   (journal_t *, unsigned long, unsigned long, unsigned long);
1539extern int	   jbd2_journal_check_available_features
1540		   (journal_t *, unsigned long, unsigned long, unsigned long);
1541extern int	   jbd2_journal_set_features
1542		   (journal_t *, unsigned long, unsigned long, unsigned long);
1543extern void	   jbd2_journal_clear_features
1544		   (journal_t *, unsigned long, unsigned long, unsigned long);
1545extern int	   jbd2_journal_load       (journal_t *journal);
1546extern int	   jbd2_journal_destroy    (journal_t *);
1547extern int	   jbd2_journal_recover    (journal_t *journal);
1548extern int	   jbd2_journal_wipe       (journal_t *, int);
1549extern int	   jbd2_journal_skip_recovery	(journal_t *);
1550extern void	   jbd2_journal_update_sb_errno(journal_t *);
1551extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1552				unsigned long, blk_opf_t);
1553extern void	   jbd2_journal_abort      (journal_t *, int);
1554extern int	   jbd2_journal_errno      (journal_t *);
1555extern void	   jbd2_journal_ack_err    (journal_t *);
1556extern int	   jbd2_journal_clear_err  (journal_t *);
1557extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1558extern int	   jbd2_journal_force_commit(journal_t *);
1559extern int	   jbd2_journal_force_commit_nested(journal_t *);
1560extern int	   jbd2_journal_inode_ranged_write(handle_t *handle,
1561			struct jbd2_inode *inode, loff_t start_byte,
1562			loff_t length);
1563extern int	   jbd2_journal_inode_ranged_wait(handle_t *handle,
1564			struct jbd2_inode *inode, loff_t start_byte,
1565			loff_t length);
1566extern int	   jbd2_journal_finish_inode_data_buffers(
1567			struct jbd2_inode *jinode);
1568extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1569				struct jbd2_inode *inode, loff_t new_size);
1570extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1571extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1572
1573/*
1574 * journal_head management
1575 */
1576struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1577struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1578void jbd2_journal_put_journal_head(struct journal_head *jh);
1579
1580/*
1581 * handle management
1582 */
1583extern struct kmem_cache *jbd2_handle_cache;
1584
1585/*
1586 * This specialized allocator has to be a macro for its allocations to be
1587 * accounted separately (to have a separate alloc_tag). The typecast is
1588 * intentional to enforce typesafety.
1589 */
1590#define jbd2_alloc_handle(_gfp_flags)	\
1591		((handle_t *)kmem_cache_zalloc(jbd2_handle_cache, _gfp_flags))
1592
1593static inline void jbd2_free_handle(handle_t *handle)
1594{
1595	kmem_cache_free(jbd2_handle_cache, handle);
1596}
1597
1598/*
1599 * jbd2_inode management (optional, for those file systems that want to use
1600 * dynamically allocated jbd2_inode structures)
1601 */
1602extern struct kmem_cache *jbd2_inode_cache;
1603
1604/*
1605 * This specialized allocator has to be a macro for its allocations to be
1606 * accounted separately (to have a separate alloc_tag). The typecast is
1607 * intentional to enforce typesafety.
1608 */
1609#define jbd2_alloc_inode(_gfp_flags)	\
1610		((struct jbd2_inode *)kmem_cache_alloc(jbd2_inode_cache, _gfp_flags))
1611
1612static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1613{
1614	kmem_cache_free(jbd2_inode_cache, jinode);
1615}
1616
1617/* Primary revoke support */
1618#define JOURNAL_REVOKE_DEFAULT_HASH 256
1619extern int	   jbd2_journal_init_revoke(journal_t *, int);
1620extern void	   jbd2_journal_destroy_revoke_record_cache(void);
1621extern void	   jbd2_journal_destroy_revoke_table_cache(void);
1622extern int __init jbd2_journal_init_revoke_record_cache(void);
1623extern int __init jbd2_journal_init_revoke_table_cache(void);
1624struct jbd2_revoke_table_s *jbd2_journal_init_revoke_table(int hash_size);
1625void jbd2_journal_destroy_revoke_table(struct jbd2_revoke_table_s *table);
1626
1627extern void	   jbd2_journal_destroy_revoke(journal_t *);
1628extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1629extern void	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1630extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
1631						     struct list_head *log_bufs);
1632
1633/* Recovery revoke support */
1634extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1635extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1636extern void	jbd2_journal_clear_revoke(journal_t *);
1637extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1638extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1639
1640/*
1641 * The log thread user interface:
1642 *
1643 * Request space in the current transaction, and force transaction commit
1644 * transitions on demand.
1645 */
1646
1647int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1648int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1649int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1650int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1651int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1652int jbd2_log_do_checkpoint(journal_t *journal);
1653int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1654
1655void __jbd2_log_wait_for_space(journal_t *journal);
1656extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1657extern int jbd2_cleanup_journal_tail(journal_t *);
1658
1659/* Fast commit related APIs */
1660int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1661int jbd2_fc_end_commit(journal_t *journal);
1662int jbd2_fc_end_commit_fallback(journal_t *journal);
1663int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1664int jbd2_submit_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1665int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1666int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1667void jbd2_fc_release_bufs(journal_t *journal);
1668
1669/*
1670 * is_journal_abort
1671 *
1672 * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1673 * bit, when set, indicates that we have had a fatal error somewhere,
1674 * either inside the journaling layer or indicated to us by the client
1675 * (eg. ext3), and that we and should not commit any further
1676 * transactions.
1677 */
1678
1679static inline int is_journal_aborted(journal_t *journal)
1680{
1681	return journal->j_flags & JBD2_ABORT;
1682}
1683
1684static inline int is_handle_aborted(handle_t *handle)
1685{
1686	if (handle->h_aborted || !handle->h_transaction)
1687		return 1;
1688	return is_journal_aborted(handle->h_transaction->t_journal);
1689}
1690
1691static inline void jbd2_journal_abort_handle(handle_t *handle)
1692{
1693	handle->h_aborted = 1;
1694}
1695
1696static inline void jbd2_init_fs_dev_write_error(journal_t *journal)
1697{
1698	struct address_space *mapping = journal->j_fs_dev->bd_mapping;
1699
1700	/*
1701	 * Save the original wb_err value of client fs's bdev mapping which
1702	 * could be used to detect the client fs's metadata async write error.
1703	 */
1704	errseq_check_and_advance(&mapping->wb_err, &journal->j_fs_dev_wb_err);
1705}
1706
1707static inline int jbd2_check_fs_dev_write_error(journal_t *journal)
1708{
1709	struct address_space *mapping = journal->j_fs_dev->bd_mapping;
1710
1711	return errseq_check(&mapping->wb_err,
1712			    READ_ONCE(journal->j_fs_dev_wb_err));
1713}
1714
1715#endif /* __KERNEL__   */
1716
1717/* Comparison functions for transaction IDs: perform comparisons using
1718 * modulo arithmetic so that they work over sequence number wraps. */
1719
1720static inline int tid_gt(tid_t x, tid_t y)
1721{
1722	int difference = (x - y);
1723	return (difference > 0);
1724}
1725
1726static inline int tid_geq(tid_t x, tid_t y)
1727{
1728	int difference = (x - y);
1729	return (difference >= 0);
1730}
1731
1732extern int jbd2_journal_blocks_per_folio(struct inode *inode);
1733extern size_t journal_tag_bytes(journal_t *journal);
1734
1735static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1736{
1737	return jbd2_has_feature_csum2(journal) ||
1738	       jbd2_has_feature_csum3(journal);
1739}
1740
1741static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb)
1742{
1743	int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks);
1744
1745	return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS;
1746}
1747
1748/*
1749 * Return number of free blocks in the log. Must be called under j_state_lock.
1750 */
1751static inline unsigned long jbd2_log_space_left(journal_t *journal)
1752{
1753	/* Allow for rounding errors */
1754	long free = journal->j_free - 32;
1755
1756	if (journal->j_committing_transaction) {
1757		free -= atomic_read(&journal->
1758                        j_committing_transaction->t_outstanding_credits);
1759	}
1760	return max_t(long, free, 0);
1761}
1762
1763/*
1764 * Definitions which augment the buffer_head layer
1765 */
1766
1767/* journaling buffer types */
1768#define BJ_None		0	/* Not journaled */
1769#define BJ_Metadata	1	/* Normal journaled metadata */
1770#define BJ_Forget	2	/* Buffer superseded by this transaction */
1771#define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1772#define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1773#define BJ_Types	5
1774
1775static inline u32 jbd2_chksum(u32 crc, const void *address, unsigned int length)
1776{
1777	return crc32c(crc, address, length);
1778}
1779
1780/* Return most recent uncommitted transaction */
1781static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1782{
1783	tid_t tid;
1784
1785	read_lock(&journal->j_state_lock);
1786	tid = journal->j_commit_request;
1787	if (journal->j_running_transaction)
1788		tid = journal->j_running_transaction->t_tid;
1789	read_unlock(&journal->j_state_lock);
1790	return tid;
1791}
1792
1793static inline int jbd2_handle_buffer_credits(handle_t *handle)
1794{
1795	journal_t *journal;
1796
1797	if (!handle->h_reserved)
1798		journal = handle->h_transaction->t_journal;
1799	else
1800		journal = handle->h_journal;
1801
1802	return handle->h_total_credits -
1803		DIV_ROUND_UP(handle->h_revoke_credits_requested,
1804			     journal->j_revoke_records_per_block);
1805}
1806
1807#ifdef __KERNEL__
1808
1809#define buffer_trace_init(bh)	do {} while (0)
1810#define print_buffer_fields(bh)	do {} while (0)
1811#define print_buffer_trace(bh)	do {} while (0)
1812#define BUFFER_TRACE(bh, info)	do {} while (0)
1813#define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1814#define JBUFFER_TRACE(jh, info)	do {} while (0)
1815
1816#endif	/* __KERNEL__ */
1817
1818#define EFSBADCRC	EBADMSG		/* Bad CRC detected */
1819#define EFSCORRUPTED	EUCLEAN		/* Filesystem is corrupted */
1820
1821#endif	/* _LINUX_JBD2_H */