include/linux/jbd.h

   1 /*
   2  * linux/include/linux/jbd.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_JBD_H
  17 #define _LINUX_JBD_H
  18
  19 /* Allow this file to be included directly into e2fsprogs */
  20 #ifndef __KERNEL__
  21 #include "jfs_compat.h"
  22 #define JFS_DEBUG
  23 #define jfs_debug jbd_debug
  24 #else
  25
  26 #include <linux/types.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/journal-head.h>
  29 #include <linux/stddef.h>
  30 #include <linux/mutex.h>
  31 #include <linux/timer.h>
  32 #include <linux/lockdep.h>
  33 #include <linux/slab.h>
  34
  35 #define journal_oom_retry 1
  36
  37 /*
  38  * Define JBD_PARANOID_IOFAIL to cause a kernel BUG() if ext3 finds
  39  * certain classes of error which can occur due to failed IOs.  Under
  40  * normal use we want ext3 to continue after such errors, because
  41  * hardware _can_ fail, but for debugging purposes when running tests on
  42  * known-good hardware we may want to trap these errors.
  43  */
  44 #undef JBD_PARANOID_IOFAIL
  45
  46 /*
  47  * The default maximum commit age, in seconds.
  48  */
  49 #define JBD_DEFAULT_MAX_COMMIT_AGE 5
  50
  51 #ifdef CONFIG_JBD_DEBUG
  52 /*
  53  * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
  54  * consistency checks.  By default we don't do this unless
  55  * CONFIG_JBD_DEBUG is on.
  56  */
  57 #define JBD_EXPENSIVE_CHECKING
  58 extern u8 journal_enable_debug;
  59
  60 void __jbd_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         __jbd_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
  65 #else
  66 #define jbd_debug(n, fmt, a...)    /**/
  67 #endif
  68
  69 static inline void *jbd_alloc(size_t size, gfp_t flags)
  70 {
  71         return (void *)__get_free_pages(flags, get_order(size));
  72 }
  73
  74 static inline void jbd_free(void *ptr, size_t size)
  75 {
  76         free_pages((unsigned long)ptr, get_order(size));
  77 }
  78
  79 #define JFS_MIN_JOURNAL_BLOCKS 1024
  80
  81
  82 /**
  83  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
  84  *
  85  * All filesystem modifications made by the process go
  86  * through this handle.  Recursive operations (such as quota operations)
  87  * are gathered into a single update.
  88  *
  89  * The buffer credits field is used to account for journaled buffers
  90  * being modified by the running process.  To ensure that there is
  91  * enough log space for all outstanding operations, we need to limit the
  92  * number of outstanding buffers possible at any time.  When the
  93  * operation completes, any buffer credits not used are credited back to
  94  * the transaction, so that at all times we know how many buffers the
  95  * outstanding updates on a transaction might possibly touch.
  96  *
  97  * This is an opaque datatype.
  98  **/
  99 typedef struct handle_s         handle_t;       /* Atomic operation type */
 100
 101
 102 /**
 103  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
 104  *
 105  * journal_t is linked to from the fs superblock structure.
 106  *
 107  * We use the journal_t to keep track of all outstanding transaction
 108  * activity on the filesystem, and to manage the state of the log
 109  * writing process.
 110  *
 111  * This is an opaque datatype.
 112  **/
 113 typedef struct journal_s        journal_t;      /* Journal control structure */
 114 #endif
 115
 116 /*
 117  * Internal structures used by the logging mechanism:
 118  */
 119
 120 #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
 121
 122 /*
 123  * On-disk structures
 124  */
 125
 126 /*
 127  * Descriptor block types:
 128  */
 129
 130 #define JFS_DESCRIPTOR_BLOCK    1
 131 #define JFS_COMMIT_BLOCK        2
 132 #define JFS_SUPERBLOCK_V1       3
 133 #define JFS_SUPERBLOCK_V2       4
 134 #define JFS_REVOKE_BLOCK        5
 135
 136 /*
 137  * Standard header for all descriptor blocks:
 138  */
 139 typedef struct journal_header_s
 140 {
 141         __be32          h_magic;
 142         __be32          h_blocktype;
 143         __be32          h_sequence;
 144 } journal_header_t;
 145
 146
 147 /*
 148  * The block tag: used to describe a single buffer in the journal
 149  */
 150 typedef struct journal_block_tag_s
 151 {
 152         __be32          t_blocknr;      /* The on-disk block number */
 153         __be32          t_flags;        /* See below */
 154 } journal_block_tag_t;
 155
 156 /*
 157  * The revoke descriptor: used on disk to describe a series of blocks to
 158  * be revoked from the log
 159  */
 160 typedef struct journal_revoke_header_s
 161 {
 162         journal_header_t r_header;
 163         __be32           r_count;       /* Count of bytes used in the block */
 164 } journal_revoke_header_t;
 165
 166
 167 /* Definitions for the journal tag flags word: */
 168 #define JFS_FLAG_ESCAPE         1       /* on-disk block is escaped */
 169 #define JFS_FLAG_SAME_UUID      2       /* block has same uuid as previous */
 170 #define JFS_FLAG_DELETED        4       /* block deleted by this transaction */
 171 #define JFS_FLAG_LAST_TAG       8       /* last tag in this descriptor block */
 172
 173
 174 /*
 175  * The journal superblock.  All fields are in big-endian byte order.
 176  */
 177 typedef struct journal_superblock_s
 178 {
 179 /* 0x0000 */
 180         journal_header_t s_header;
 181
 182 /* 0x000C */
 183         /* Static information describing the journal */
 184         __be32  s_blocksize;            /* journal device blocksize */
 185         __be32  s_maxlen;               /* total blocks in journal file */
 186         __be32  s_first;                /* first block of log information */
 187
 188 /* 0x0018 */
 189         /* Dynamic information describing the current state of the log */
 190         __be32  s_sequence;             /* first commit ID expected in log */
 191         __be32  s_start;                /* blocknr of start of log */
 192
 193 /* 0x0020 */
 194         /* Error value, as set by journal_abort(). */
 195         __be32  s_errno;
 196
 197 /* 0x0024 */
 198         /* Remaining fields are only valid in a version-2 superblock */
 199         __be32  s_feature_compat;       /* compatible feature set */
 200         __be32  s_feature_incompat;     /* incompatible feature set */
 201         __be32  s_feature_ro_compat;    /* readonly-compatible feature set */
 202 /* 0x0030 */
 203         __u8    s_uuid[16];             /* 128-bit uuid for journal */
 204
 205 /* 0x0040 */
 206         __be32  s_nr_users;             /* Nr of filesystems sharing log */
 207
 208         __be32  s_dynsuper;             /* Blocknr of dynamic superblock copy*/
 209
 210 /* 0x0048 */
 211         __be32  s_max_transaction;      /* Limit of journal blocks per trans.*/
 212         __be32  s_max_trans_data;       /* Limit of data blocks per trans. */
 213
 214 /* 0x0050 */
 215         __u32   s_padding[44];
 216
 217 /* 0x0100 */
 218         __u8    s_users[16*48];         /* ids of all fs'es sharing the log */
 219 /* 0x0400 */
 220 } journal_superblock_t;
 221
 222 #define JFS_HAS_COMPAT_FEATURE(j,mask)                                  \
 223         ((j)->j_format_version >= 2 &&                                  \
 224          ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
 225 #define JFS_HAS_RO_COMPAT_FEATURE(j,mask)                               \
 226         ((j)->j_format_version >= 2 &&                                  \
 227          ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
 228 #define JFS_HAS_INCOMPAT_FEATURE(j,mask)                                \
 229         ((j)->j_format_version >= 2 &&                                  \
 230          ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
 231
 232 #define JFS_FEATURE_INCOMPAT_REVOKE     0x00000001
 233
 234 /* Features known to this kernel version: */
 235 #define JFS_KNOWN_COMPAT_FEATURES       0
 236 #define JFS_KNOWN_ROCOMPAT_FEATURES     0
 237 #define JFS_KNOWN_INCOMPAT_FEATURES     JFS_FEATURE_INCOMPAT_REVOKE
 238
 239 #ifdef __KERNEL__
 240
 241 #include <linux/fs.h>
 242 #include <linux/sched.h>
 243
 244 enum jbd_state_bits {
 245         BH_JBD                  /* Has an attached ext3 journal_head */
 246           = BH_PrivateStart,
 247         BH_JWrite,              /* Being written to log (@@@ DEBUGGING) */
 248         BH_Freed,               /* Has been freed (truncated) */
 249         BH_Revoked,             /* Has been revoked from the log */
 250         BH_RevokeValid,         /* Revoked flag is valid */
 251         BH_JBDDirty,            /* Is dirty but journaled */
 252         BH_State,               /* Pins most journal_head state */
 253         BH_JournalHead,         /* Pins bh->b_private and jh->b_bh */
 254         BH_Unshadow,            /* Dummy bit, for BJ_Shadow wakeup filtering */
 255         BH_JBDPrivateStart,     /* First bit available for private use by FS */
 256 };
 257
 258 BUFFER_FNS(JBD, jbd)
 259 BUFFER_FNS(JWrite, jwrite)
 260 BUFFER_FNS(JBDDirty, jbddirty)
 261 TAS_BUFFER_FNS(JBDDirty, jbddirty)
 262 BUFFER_FNS(Revoked, revoked)
 263 TAS_BUFFER_FNS(Revoked, revoked)
 264 BUFFER_FNS(RevokeValid, revokevalid)
 265 TAS_BUFFER_FNS(RevokeValid, revokevalid)
 266 BUFFER_FNS(Freed, freed)
 267
 268 #include <linux/jbd_common.h>
 269
 270 #define J_ASSERT(assert)        BUG_ON(!(assert))
 271
 272 #define J_ASSERT_BH(bh, expr)   J_ASSERT(expr)
 273 #define J_ASSERT_JH(jh, expr)   J_ASSERT(expr)
 274
 275 #if defined(JBD_PARANOID_IOFAIL)
 276 #define J_EXPECT(expr, why...)          J_ASSERT(expr)
 277 #define J_EXPECT_BH(bh, expr, why...)   J_ASSERT_BH(bh, expr)
 278 #define J_EXPECT_JH(jh, expr, why...)   J_ASSERT_JH(jh, expr)
 279 #else
 280 #define __journal_expect(expr, why...)                                       \
 281         ({                                                                   \
 282                 int val = (expr);                                            \
 283                 if (!val) {                                                  \
 284                         printk(KERN_ERR                                      \
 285                                 "EXT3-fs unexpected failure: %s;\n",# expr); \
 286                         printk(KERN_ERR why "\n");                           \
 287                 }                                                            \
 288                 val;                                                         \
 289         })
 290 #define J_EXPECT(expr, why...)          __journal_expect(expr, ## why)
 291 #define J_EXPECT_BH(bh, expr, why...)   __journal_expect(expr, ## why)
 292 #define J_EXPECT_JH(jh, expr, why...)   __journal_expect(expr, ## why)
 293 #endif
 294
 295 struct jbd_revoke_table_s;
 296
 297 /**
 298  * struct handle_s - this is the concrete type associated with handle_t.
 299  * @h_transaction: Which compound transaction is this update a part of?
 300  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
 301  * @h_ref: Reference count on this handle
 302  * @h_err: Field for caller's use to track errors through large fs operations
 303  * @h_sync: flag for sync-on-close
 304  * @h_jdata: flag to force data journaling
 305  * @h_aborted: flag indicating fatal error on handle
 306  * @h_lockdep_map: lockdep info for debugging lock problems
 307  */
 308 struct handle_s
 309 {
 310         /* Which compound transaction is this update a part of? */
 311         transaction_t           *h_transaction;
 312
 313         /* Number of remaining buffers we are allowed to dirty: */
 314         int                     h_buffer_credits;
 315
 316         /* Reference count on this handle */
 317         int                     h_ref;
 318
 319         /* Field for caller's use to track errors through large fs */
 320         /* operations */
 321         int                     h_err;
 322
 323         /* Flags [no locking] */
 324         unsigned int    h_sync:         1;      /* sync-on-close */
 325         unsigned int    h_jdata:        1;      /* force data journaling */
 326         unsigned int    h_aborted:      1;      /* fatal error on handle */
 327
 328 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 329         struct lockdep_map      h_lockdep_map;
 330 #endif
 331 };
 332
 333
 334 /* The transaction_t type is the guts of the journaling mechanism.  It
 335  * tracks a compound transaction through its various states:
 336  *
 337  * RUNNING:     accepting new updates
 338  * LOCKED:      Updates still running but we don't accept new ones
 339  * RUNDOWN:     Updates are tidying up but have finished requesting
 340  *              new buffers to modify (state not used for now)
 341  * FLUSH:       All updates complete, but we are still writing to disk
 342  * COMMIT:      All data on disk, writing commit record
 343  * FINISHED:    We still have to keep the transaction for checkpointing.
 344  *
 345  * The transaction keeps track of all of the buffers modified by a
 346  * running transaction, and all of the buffers committed but not yet
 347  * flushed to home for finished transactions.
 348  */
 349
 350 /*
 351  * Lock ranking:
 352  *
 353  *    j_list_lock
 354  *      ->jbd_lock_bh_journal_head()    (This is "innermost")
 355  *
 356  *    j_state_lock
 357  *    ->jbd_lock_bh_state()
 358  *
 359  *    jbd_lock_bh_state()
 360  *    ->j_list_lock
 361  *
 362  *    j_state_lock
 363  *    ->t_handle_lock
 364  *
 365  *    j_state_lock
 366  *    ->j_list_lock                     (journal_unmap_buffer)
 367  *
 368  */
 369
 370 struct transaction_s
 371 {
 372         /* Pointer to the journal for this transaction. [no locking] */
 373         journal_t               *t_journal;
 374
 375         /* Sequence number for this transaction [no locking] */
 376         tid_t                   t_tid;
 377
 378         /*
 379          * Transaction's current state
 380          * [no locking - only kjournald alters this]
 381          * [j_list_lock] guards transition of a transaction into T_FINISHED
 382          * state and subsequent call of __journal_drop_transaction()
 383          * FIXME: needs barriers
 384          * KLUDGE: [use j_state_lock]
 385          */
 386         enum {
 387                 T_RUNNING,
 388                 T_LOCKED,
 389                 T_FLUSH,
 390                 T_COMMIT,
 391                 T_COMMIT_RECORD,
 392                 T_FINISHED
 393         }                       t_state;
 394
 395         /*
 396          * Where in the log does this transaction's commit start? [no locking]
 397          */
 398         unsigned int            t_log_start;
 399
 400         /* Number of buffers on the t_buffers list [j_list_lock] */
 401         int                     t_nr_buffers;
 402
 403         /*
 404          * Doubly-linked circular list of all buffers reserved but not yet
 405          * modified by this transaction [j_list_lock]
 406          */
 407         struct journal_head     *t_reserved_list;
 408
 409         /*
 410          * Doubly-linked circular list of all buffers under writeout during
 411          * commit [j_list_lock]
 412          */
 413         struct journal_head     *t_locked_list;
 414
 415         /*
 416          * Doubly-linked circular list of all metadata buffers owned by this
 417          * transaction [j_list_lock]
 418          */
 419         struct journal_head     *t_buffers;
 420
 421         /*
 422          * Doubly-linked circular list of all data buffers still to be
 423          * flushed before this transaction can be committed [j_list_lock]
 424          */
 425         struct journal_head     *t_sync_datalist;
 426
 427         /*
 428          * Doubly-linked circular list of all forget buffers (superseded
 429          * buffers which we can un-checkpoint once this transaction commits)
 430          * [j_list_lock]
 431          */
 432         struct journal_head     *t_forget;
 433
 434         /*
 435          * Doubly-linked circular list of all buffers still to be flushed before
 436          * this transaction can be checkpointed. [j_list_lock]
 437          */
 438         struct journal_head     *t_checkpoint_list;
 439
 440         /*
 441          * Doubly-linked circular list of all buffers submitted for IO while
 442          * checkpointing. [j_list_lock]
 443          */
 444         struct journal_head     *t_checkpoint_io_list;
 445
 446         /*
 447          * Doubly-linked circular list of temporary buffers currently undergoing
 448          * IO in the log [j_list_lock]
 449          */
 450         struct journal_head     *t_iobuf_list;
 451
 452         /*
 453          * Doubly-linked circular list of metadata buffers being shadowed by log
 454          * IO.  The IO buffers on the iobuf list and the shadow buffers on this
 455          * list match each other one for one at all times. [j_list_lock]
 456          */
 457         struct journal_head     *t_shadow_list;
 458
 459         /*
 460          * Doubly-linked circular list of control buffers being written to the
 461          * log. [j_list_lock]
 462          */
 463         struct journal_head     *t_log_list;
 464
 465         /*
 466          * Protects info related to handles
 467          */
 468         spinlock_t              t_handle_lock;
 469
 470         /*
 471          * Number of outstanding updates running on this transaction
 472          * [t_handle_lock]
 473          */
 474         int                     t_updates;
 475
 476         /*
 477          * Number of buffers reserved for use by all handles in this transaction
 478          * handle but not yet modified. [t_handle_lock]
 479          */
 480         int                     t_outstanding_credits;
 481
 482         /*
 483          * Forward and backward links for the circular list of all transactions
 484          * awaiting checkpoint. [j_list_lock]
 485          */
 486         transaction_t           *t_cpnext, *t_cpprev;
 487
 488         /*
 489          * When will the transaction expire (become due for commit), in jiffies?
 490          * [no locking]
 491          */
 492         unsigned long           t_expires;
 493
 494         /*
 495          * When this transaction started, in nanoseconds [no locking]
 496          */
 497         ktime_t                 t_start_time;
 498
 499         /*
 500          * How many handles used this transaction? [t_handle_lock]
 501          */
 502         int t_handle_count;
 503 };
 504
 505 /**
 506  * struct journal_s - this is the concrete type associated with journal_t.
 507  * @j_flags:  General journaling state flags
 508  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
 509  *     prior abort)?
 510  * @j_sb_buffer: First part of superblock buffer
 511  * @j_superblock: Second part of superblock buffer
 512  * @j_format_version: Version of the superblock format
 513  * @j_state_lock: Protect the various scalars in the journal
 514  * @j_barrier_count:  Number of processes waiting to create a barrier lock
 515  * @j_running_transaction: The current running transaction..
 516  * @j_committing_transaction: the transaction we are pushing to disk
 517  * @j_checkpoint_transactions: a linked circular list of all transactions
 518  *  waiting for checkpointing
 519  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
 520  *  to start committing, or for a barrier lock to be released
 521  * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
 522  * @j_wait_done_commit: Wait queue for waiting for commit to complete
 523  * @j_wait_checkpoint:  Wait queue to trigger checkpointing
 524  * @j_wait_commit: Wait queue to trigger commit
 525  * @j_wait_updates: Wait queue to wait for updates to complete
 526  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
 527  * @j_head: Journal head - identifies the first unused block in the journal
 528  * @j_tail: Journal tail - identifies the oldest still-used block in the
 529  *  journal.
 530  * @j_free: Journal free - how many free blocks are there in the journal?
 531  * @j_first: The block number of the first usable block
 532  * @j_last: The block number one beyond the last usable block
 533  * @j_dev: Device where we store the journal
 534  * @j_blocksize: blocksize for the location where we store the journal.
 535  * @j_blk_offset: starting block offset for into the device where we store the
 536  *     journal
 537  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
 538  *     be equal to j_dev
 539  * @j_maxlen: Total maximum capacity of the journal region on disk.
 540  * @j_list_lock: Protects the buffer lists and internal buffer state.
 541  * @j_inode: Optional inode where we store the journal.  If present, all journal
 542  *     block numbers are mapped into this inode via bmap().
 543  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
 544  * @j_transaction_sequence: Sequence number of the next transaction to grant
 545  * @j_commit_sequence: Sequence number of the most recently committed
 546  *  transaction
 547  * @j_commit_request: Sequence number of the most recent transaction wanting
 548  *     commit
 549  * @j_commit_waited: Sequence number of the most recent transaction someone
 550  *     is waiting for to commit.
 551  * @j_uuid: Uuid of client object.
 552  * @j_task: Pointer to the current commit thread for this journal
 553  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
 554  *     single compound commit transaction
 555  * @j_commit_interval: What is the maximum transaction lifetime before we begin
 556  *  a commit?
 557  * @j_commit_timer:  The timer used to wakeup the commit thread
 558  * @j_revoke_lock: Protect the revoke table
 559  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
 560  *     current transaction.
 561  * @j_revoke_table: alternate revoke tables for j_revoke
 562  * @j_wbuf: array of buffer_heads for journal_commit_transaction
 563  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
 564  *      number that will fit in j_blocksize
 565  * @j_last_sync_writer: most recent pid which did a synchronous write
 566  * @j_average_commit_time: the average amount of time in nanoseconds it
 567  *      takes to commit a transaction to the disk.
 568  * @j_private: An opaque pointer to fs-private information.
 569  */
 570
 571 struct journal_s
 572 {
 573         /* General journaling state flags [j_state_lock] */
 574         unsigned long           j_flags;
 575
 576         /*
 577          * Is there an outstanding uncleared error on the journal (from a prior
 578          * abort)? [j_state_lock]
 579          */
 580         int                     j_errno;
 581
 582         /* The superblock buffer */
 583         struct buffer_head      *j_sb_buffer;
 584         journal_superblock_t    *j_superblock;
 585
 586         /* Version of the superblock format */
 587         int                     j_format_version;
 588
 589         /*
 590          * Protect the various scalars in the journal
 591          */
 592         spinlock_t              j_state_lock;
 593
 594         /*
 595          * Number of processes waiting to create a barrier lock [j_state_lock]
 596          */
 597         int                     j_barrier_count;
 598
 599         /*
 600          * Transactions: The current running transaction...
 601          * [j_state_lock] [caller holding open handle]
 602          */
 603         transaction_t           *j_running_transaction;
 604
 605         /*
 606          * the transaction we are pushing to disk
 607          * [j_state_lock] [caller holding open handle]
 608          */
 609         transaction_t           *j_committing_transaction;
 610
 611         /*
 612          * ... and a linked circular list of all transactions waiting for
 613          * checkpointing. [j_list_lock]
 614          */
 615         transaction_t           *j_checkpoint_transactions;
 616
 617         /*
 618          * Wait queue for waiting for a locked transaction to start committing,
 619          * or for a barrier lock to be released
 620          */
 621         wait_queue_head_t       j_wait_transaction_locked;
 622
 623         /* Wait queue for waiting for checkpointing to complete */
 624         wait_queue_head_t       j_wait_logspace;
 625
 626         /* Wait queue for waiting for commit to complete */
 627         wait_queue_head_t       j_wait_done_commit;
 628
 629         /* Wait queue to trigger checkpointing */
 630         wait_queue_head_t       j_wait_checkpoint;
 631
 632         /* Wait queue to trigger commit */
 633         wait_queue_head_t       j_wait_commit;
 634
 635         /* Wait queue to wait for updates to complete */
 636         wait_queue_head_t       j_wait_updates;
 637
 638         /* Semaphore for locking against concurrent checkpoints */
 639         struct mutex            j_checkpoint_mutex;
 640
 641         /*
 642          * Journal head: identifies the first unused block in the journal.
 643          * [j_state_lock]
 644          */
 645         unsigned int            j_head;
 646
 647         /*
 648          * Journal tail: identifies the oldest still-used block in the journal.
 649          * [j_state_lock]
 650          */
 651         unsigned int            j_tail;
 652
 653         /*
 654          * Journal free: how many free blocks are there in the journal?
 655          * [j_state_lock]
 656          */
 657         unsigned int            j_free;
 658
 659         /*
 660          * Journal start and end: the block numbers of the first usable block
 661          * and one beyond the last usable block in the journal. [j_state_lock]
 662          */
 663         unsigned int            j_first;
 664         unsigned int            j_last;
 665
 666         /*
 667          * Device, blocksize and starting block offset for the location where we
 668          * store the journal.
 669          */
 670         struct block_device     *j_dev;
 671         int                     j_blocksize;
 672         unsigned int            j_blk_offset;
 673
 674         /*
 675          * Device which holds the client fs.  For internal journal this will be
 676          * equal to j_dev.
 677          */
 678         struct block_device     *j_fs_dev;
 679
 680         /* Total maximum capacity of the journal region on disk. */
 681         unsigned int            j_maxlen;
 682
 683         /*
 684          * Protects the buffer lists and internal buffer state.
 685          */
 686         spinlock_t              j_list_lock;
 687
 688         /* Optional inode where we store the journal.  If present, all */
 689         /* journal block numbers are mapped into this inode via */
 690         /* bmap(). */
 691         struct inode            *j_inode;
 692
 693         /*
 694          * Sequence number of the oldest transaction in the log [j_state_lock]
 695          */
 696         tid_t                   j_tail_sequence;
 697
 698         /*
 699          * Sequence number of the next transaction to grant [j_state_lock]
 700          */
 701         tid_t                   j_transaction_sequence;
 702
 703         /*
 704          * Sequence number of the most recently committed transaction
 705          * [j_state_lock].
 706          */
 707         tid_t                   j_commit_sequence;
 708
 709         /*
 710          * Sequence number of the most recent transaction wanting commit
 711          * [j_state_lock]
 712          */
 713         tid_t                   j_commit_request;
 714
 715         /*
 716          * Sequence number of the most recent transaction someone is waiting
 717          * for to commit.
 718          * [j_state_lock]
 719          */
 720         tid_t                   j_commit_waited;
 721
 722         /*
 723          * Journal uuid: identifies the object (filesystem, LVM volume etc)
 724          * backed by this journal.  This will eventually be replaced by an array
 725          * of uuids, allowing us to index multiple devices within a single
 726          * journal and to perform atomic updates across them.
 727          */
 728         __u8                    j_uuid[16];
 729
 730         /* Pointer to the current commit thread for this journal */
 731         struct task_struct      *j_task;
 732
 733         /*
 734          * Maximum number of metadata buffers to allow in a single compound
 735          * commit transaction
 736          */
 737         int                     j_max_transaction_buffers;
 738
 739         /*
 740          * What is the maximum transaction lifetime before we begin a commit?
 741          */
 742         unsigned long           j_commit_interval;
 743
 744         /* The timer used to wakeup the commit thread: */
 745         struct timer_list       j_commit_timer;
 746
 747         /*
 748          * The revoke table: maintains the list of revoked blocks in the
 749          * current transaction.  [j_revoke_lock]
 750          */
 751         spinlock_t              j_revoke_lock;
 752         struct jbd_revoke_table_s *j_revoke;
 753         struct jbd_revoke_table_s *j_revoke_table[2];
 754
 755         /*
 756          * array of bhs for journal_commit_transaction
 757          */
 758         struct buffer_head      **j_wbuf;
 759         int                     j_wbufsize;
 760
 761         /*
 762          * this is the pid of the last person to run a synchronous operation
 763          * through the journal.
 764          */
 765         pid_t                   j_last_sync_writer;
 766
 767         /*
 768          * the average amount of time in nanoseconds it takes to commit a
 769          * transaction to the disk.  [j_state_lock]
 770          */
 771         u64                     j_average_commit_time;
 772
 773         /*
 774          * An opaque pointer to fs-private information.  ext3 puts its
 775          * superblock pointer here
 776          */
 777         void *j_private;
 778 };
 779
 780 /*
 781  * Journal flag definitions
 782  */
 783 #define JFS_UNMOUNT     0x001   /* Journal thread is being destroyed */
 784 #define JFS_ABORT       0x002   /* Journaling has been aborted for errors. */
 785 #define JFS_ACK_ERR     0x004   /* The errno in the sb has been acked */
 786 #define JFS_FLUSHED     0x008   /* The journal superblock has been flushed */
 787 #define JFS_LOADED      0x010   /* The journal superblock has been loaded */
 788 #define JFS_BARRIER     0x020   /* Use IDE barriers */
 789 #define JFS_ABORT_ON_SYNCDATA_ERR       0x040  /* Abort the journal on file
 790                                                 * data write error in ordered
 791                                                 * mode */
 792
 793 /*
 794  * Function declarations for the journaling transaction and buffer
 795  * management
 796  */
 797
 798 /* Filing buffers */
 799 extern void journal_unfile_buffer(journal_t *, struct journal_head *);
 800 extern void __journal_unfile_buffer(struct journal_head *);
 801 extern void __journal_refile_buffer(struct journal_head *);
 802 extern void journal_refile_buffer(journal_t *, struct journal_head *);
 803 extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
 804 extern void __journal_free_buffer(struct journal_head *bh);
 805 extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
 806 extern void __journal_clean_data_list(transaction_t *transaction);
 807
 808 /* Log buffer allocation */
 809 extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
 810 int journal_next_log_block(journal_t *, unsigned int *);
 811
 812 /* Commit management */
 813 extern void journal_commit_transaction(journal_t *);
 814
 815 /* Checkpoint list management */
 816 int __journal_clean_checkpoint_list(journal_t *journal);
 817 int __journal_remove_checkpoint(struct journal_head *);
 818 void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
 819
 820 /* Buffer IO */
 821 extern int
 822 journal_write_metadata_buffer(transaction_t       *transaction,
 823                               struct journal_head  *jh_in,
 824                               struct journal_head **jh_out,
 825                               unsigned int blocknr);
 826
 827 /* Transaction locking */
 828 extern void             __wait_on_journal (journal_t *);
 829
 830 /*
 831  * Journal locking.
 832  *
 833  * We need to lock the journal during transaction state changes so that nobody
 834  * ever tries to take a handle on the running transaction while we are in the
 835  * middle of moving it to the commit phase.  j_state_lock does this.
 836  *
 837  * Note that the locking is completely interrupt unsafe.  We never touch
 838  * journal structures from interrupts.
 839  */
 840
 841 static inline handle_t *journal_current_handle(void)
 842 {
 843         return current->journal_info;
 844 }
 845
 846 /* The journaling code user interface:
 847  *
 848  * Create and destroy handles
 849  * Register buffer modifications against the current transaction.
 850  */
 851
 852 extern handle_t *journal_start(journal_t *, int nblocks);
 853 extern int       journal_restart (handle_t *, int nblocks);
 854 extern int       journal_extend (handle_t *, int nblocks);
 855 extern int       journal_get_write_access(handle_t *, struct buffer_head *);
 856 extern int       journal_get_create_access (handle_t *, struct buffer_head *);
 857 extern int       journal_get_undo_access(handle_t *, struct buffer_head *);
 858 extern int       journal_dirty_data (handle_t *, struct buffer_head *);
 859 extern int       journal_dirty_metadata (handle_t *, struct buffer_head *);
 860 extern void      journal_release_buffer (handle_t *, struct buffer_head *);
 861 extern int       journal_forget (handle_t *, struct buffer_head *);
 862 extern void      journal_sync_buffer (struct buffer_head *);
 863 extern void      journal_invalidatepage(journal_t *,
 864                                 struct page *, unsigned int, unsigned int);
 865 extern int       journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
 866 extern int       journal_stop(handle_t *);
 867 extern int       journal_flush (journal_t *);
 868 extern void      journal_lock_updates (journal_t *);
 869 extern void      journal_unlock_updates (journal_t *);
 870
 871 extern journal_t * journal_init_dev(struct block_device *bdev,
 872                                 struct block_device *fs_dev,
 873                                 int start, int len, int bsize);
 874 extern journal_t * journal_init_inode (struct inode *);
 875 extern int         journal_update_format (journal_t *);
 876 extern int         journal_check_used_features
 877                    (journal_t *, unsigned long, unsigned long, unsigned long);
 878 extern int         journal_check_available_features
 879                    (journal_t *, unsigned long, unsigned long, unsigned long);
 880 extern int         journal_set_features
 881                    (journal_t *, unsigned long, unsigned long, unsigned long);
 882 extern int         journal_create     (journal_t *);
 883 extern int         journal_load       (journal_t *journal);
 884 extern int         journal_destroy    (journal_t *);
 885 extern int         journal_recover    (journal_t *journal);
 886 extern int         journal_wipe       (journal_t *, int);
 887 extern int         journal_skip_recovery        (journal_t *);
 888 extern void        journal_update_sb_log_tail   (journal_t *, tid_t, unsigned int,
 889                                                  int);
 890 extern void        journal_abort      (journal_t *, int);
 891 extern int         journal_errno      (journal_t *);
 892 extern void        journal_ack_err    (journal_t *);
 893 extern int         journal_clear_err  (journal_t *);
 894 extern int         journal_bmap(journal_t *, unsigned int, unsigned int *);
 895 extern int         journal_force_commit(journal_t *);
 896
 897 /*
 898  * journal_head management
 899  */
 900 struct journal_head *journal_add_journal_head(struct buffer_head *bh);
 901 struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
 902 void journal_put_journal_head(struct journal_head *jh);
 903
 904 /*
 905  * handle management
 906  */
 907 extern struct kmem_cache *jbd_handle_cache;
 908
 909 static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
 910 {
 911         return kmem_cache_zalloc(jbd_handle_cache, gfp_flags);
 912 }
 913
 914 static inline void jbd_free_handle(handle_t *handle)
 915 {
 916         kmem_cache_free(jbd_handle_cache, handle);
 917 }
 918
 919 /* Primary revoke support */
 920 #define JOURNAL_REVOKE_DEFAULT_HASH 256
 921 extern int         journal_init_revoke(journal_t *, int);
 922 extern void        journal_destroy_revoke_caches(void);
 923 extern int         journal_init_revoke_caches(void);
 924
 925 extern void        journal_destroy_revoke(journal_t *);
 926 extern int         journal_revoke (handle_t *,
 927                                 unsigned int, struct buffer_head *);
 928 extern int         journal_cancel_revoke(handle_t *, struct journal_head *);
 929 extern void        journal_write_revoke_records(journal_t *,
 930                                                 transaction_t *, int);
 931
 932 /* Recovery revoke support */
 933 extern int      journal_set_revoke(journal_t *, unsigned int, tid_t);
 934 extern int      journal_test_revoke(journal_t *, unsigned int, tid_t);
 935 extern void     journal_clear_revoke(journal_t *);
 936 extern void     journal_switch_revoke_table(journal_t *journal);
 937 extern void     journal_clear_buffer_revoked_flags(journal_t *journal);
 938
 939 /*
 940  * The log thread user interface:
 941  *
 942  * Request space in the current transaction, and force transaction commit
 943  * transitions on demand.
 944  */
 945
 946 int __log_space_left(journal_t *); /* Called with journal locked */
 947 int log_start_commit(journal_t *journal, tid_t tid);
 948 int __log_start_commit(journal_t *journal, tid_t tid);
 949 int journal_start_commit(journal_t *journal, tid_t *tid);
 950 int journal_force_commit_nested(journal_t *journal);
 951 int log_wait_commit(journal_t *journal, tid_t tid);
 952 int log_do_checkpoint(journal_t *journal);
 953 int journal_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
 954
 955 void __log_wait_for_space(journal_t *journal);
 956 extern void     __journal_drop_transaction(journal_t *, transaction_t *);
 957 extern int      cleanup_journal_tail(journal_t *);
 958
 959 /* Debugging code only: */
 960
 961 #define jbd_ENOSYS() \
 962 do {                                                                       \
 963         printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
 964         current->state = TASK_UNINTERRUPTIBLE;                             \
 965         schedule();                                                        \
 966 } while (1)
 967
 968 /*
 969  * is_journal_abort
 970  *
 971  * Simple test wrapper function to test the JFS_ABORT state flag.  This
 972  * bit, when set, indicates that we have had a fatal error somewhere,
 973  * either inside the journaling layer or indicated to us by the client
 974  * (eg. ext3), and that we and should not commit any further
 975  * transactions.
 976  */
 977
 978 static inline int is_journal_aborted(journal_t *journal)
 979 {
 980         return journal->j_flags & JFS_ABORT;
 981 }
 982
 983 static inline int is_handle_aborted(handle_t *handle)
 984 {
 985         if (handle->h_aborted)
 986                 return 1;
 987         return is_journal_aborted(handle->h_transaction->t_journal);
 988 }
 989
 990 static inline void journal_abort_handle(handle_t *handle)
 991 {
 992         handle->h_aborted = 1;
 993 }
 994
 995 #endif /* __KERNEL__   */
 996
 997 /* Comparison functions for transaction IDs: perform comparisons using
 998  * modulo arithmetic so that they work over sequence number wraps. */
 999
1000 static inline int tid_gt(tid_t x, tid_t y)
1001 {
1002         int difference = (x - y);
1003         return (difference > 0);
1004 }
1005
1006 static inline int tid_geq(tid_t x, tid_t y)
1007 {
1008         int difference = (x - y);
1009         return (difference >= 0);
1010 }
1011
1012 extern int journal_blocks_per_page(struct inode *inode);
1013
1014 /*
1015  * Return the minimum number of blocks which must be free in the journal
1016  * before a new transaction may be started.  Must be called under j_state_lock.
1017  */
1018 static inline int jbd_space_needed(journal_t *journal)
1019 {
1020         int nblocks = journal->j_max_transaction_buffers;
1021         if (journal->j_committing_transaction)
1022                 nblocks += journal->j_committing_transaction->
1023                                         t_outstanding_credits;
1024         return nblocks;
1025 }
1026
1027 /*
1028  * Definitions which augment the buffer_head layer
1029  */
1030
1031 /* journaling buffer types */
1032 #define BJ_None         0       /* Not journaled */
1033 #define BJ_SyncData     1       /* Normal data: flush before commit */
1034 #define BJ_Metadata     2       /* Normal journaled metadata */
1035 #define BJ_Forget       3       /* Buffer superseded by this transaction */
1036 #define BJ_IO           4       /* Buffer is for temporary IO use */
1037 #define BJ_Shadow       5       /* Buffer contents being shadowed to the log */
1038 #define BJ_LogCtl       6       /* Buffer contains log descriptors */
1039 #define BJ_Reserved     7       /* Buffer is reserved for access by journal */
1040 #define BJ_Locked       8       /* Locked for I/O during commit */
1041 #define BJ_Types        9
1042
1043 extern int jbd_blocks_per_page(struct inode *inode);
1044
1045 #ifdef __KERNEL__
1046
1047 #define buffer_trace_init(bh)   do {} while (0)
1048 #define print_buffer_fields(bh) do {} while (0)
1049 #define print_buffer_trace(bh)  do {} while (0)
1050 #define BUFFER_TRACE(bh, info)  do {} while (0)
1051 #define BUFFER_TRACE2(bh, bh2, info)    do {} while (0)
1052 #define JBUFFER_TRACE(jh, info) do {} while (0)
1053
1054 #endif  /* __KERNEL__ */
1055
1056 #endif  /* _LINUX_JBD_H */