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/bit_spinlock.h>
  31 #include <linux/mutex.h>
  32 #include <asm/semaphore.h>
  33 #endif
  34
  35 #define journal_oom_retry 1
  36
  37 /*
  38  * Define JBD_PARANIOD_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 int journal_enable_debug;
  59
  60 #define jbd_debug(n, f, a...)                                           \
  61         do {                                                            \
  62                 if ((n) <= journal_enable_debug) {                      \
  63                         printk (KERN_DEBUG "(%s, %d): %s: ",            \
  64                                 __FILE__, __LINE__, __FUNCTION__);      \
  65                         printk (f, ## a);                               \
  66                 }                                                       \
  67         } while (0)
  68 #else
  69 #define jbd_debug(f, a...)      /**/
  70 #endif
  71
  72 extern void * __jbd_kmalloc (const char *where, size_t size, gfp_t flags, int retry);
  73 #define jbd_kmalloc(size, flags) \
  74         __jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
  75 #define jbd_rep_kmalloc(size, flags) \
  76         __jbd_kmalloc(__FUNCTION__, (size), (flags), 1)
  77
  78 #define JFS_MIN_JOURNAL_BLOCKS 1024
  79
  80 #ifdef __KERNEL__
  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 #define JBD_ASSERTIONS
 245 #ifdef JBD_ASSERTIONS
 246 #define J_ASSERT(assert)                                                \
 247 do {                                                                    \
 248         if (!(assert)) {                                                \
 249                 printk (KERN_EMERG                                      \
 250                         "Assertion failure in %s() at %s:%d: \"%s\"\n", \
 251                         __FUNCTION__, __FILE__, __LINE__, # assert);    \
 252                 BUG();                                                  \
 253         }                                                               \
 254 } while (0)
 255
 256 #if defined(CONFIG_BUFFER_DEBUG)
 257 void buffer_assertion_failure(struct buffer_head *bh);
 258 #define J_ASSERT_BH(bh, expr)                                           \
 259         do {                                                            \
 260                 if (!(expr))                                            \
 261                         buffer_assertion_failure(bh);                   \
 262                 J_ASSERT(expr);                                         \
 263         } while (0)
 264 #define J_ASSERT_JH(jh, expr)   J_ASSERT_BH(jh2bh(jh), expr)
 265 #else
 266 #define J_ASSERT_BH(bh, expr)   J_ASSERT(expr)
 267 #define J_ASSERT_JH(jh, expr)   J_ASSERT(expr)
 268 #endif
 269
 270 #else
 271 #define J_ASSERT(assert)        do { } while (0)
 272 #endif          /* JBD_ASSERTIONS */
 273
 274 #if defined(JBD_PARANOID_IOFAIL)
 275 #define J_EXPECT(expr, why...)          J_ASSERT(expr)
 276 #define J_EXPECT_BH(bh, expr, why...)   J_ASSERT_BH(bh, expr)
 277 #define J_EXPECT_JH(jh, expr, why...)   J_ASSERT_JH(jh, expr)
 278 #else
 279 #define __journal_expect(expr, why...)                                       \
 280         ({                                                                   \
 281                 int val = (expr);                                            \
 282                 if (!val) {                                                  \
 283                         printk(KERN_ERR                                      \
 284                                 "EXT3-fs unexpected failure: %s;\n",# expr); \
 285                         printk(KERN_ERR why "\n");                           \
 286                 }                                                            \
 287                 val;                                                         \
 288         })
 289 #define J_EXPECT(expr, why...)          __journal_expect(expr, ## why)
 290 #define J_EXPECT_BH(bh, expr, why...)   __journal_expect(expr, ## why)
 291 #define J_EXPECT_JH(jh, expr, why...)   __journal_expect(expr, ## why)
 292 #endif
 293
 294 enum jbd_state_bits {
 295         BH_JBD                  /* Has an attached ext3 journal_head */
 296           = BH_PrivateStart,
 297         BH_JWrite,              /* Being written to log (@@@ DEBUGGING) */
 298         BH_Freed,               /* Has been freed (truncated) */
 299         BH_Revoked,             /* Has been revoked from the log */
 300         BH_RevokeValid,         /* Revoked flag is valid */
 301         BH_JBDDirty,            /* Is dirty but journaled */
 302         BH_State,               /* Pins most journal_head state */
 303         BH_JournalHead,         /* Pins bh->b_private and jh->b_bh */
 304         BH_Unshadow,            /* Dummy bit, for BJ_Shadow wakeup filtering */
 305 };
 306
 307 BUFFER_FNS(JBD, jbd)
 308 BUFFER_FNS(JWrite, jwrite)
 309 BUFFER_FNS(JBDDirty, jbddirty)
 310 TAS_BUFFER_FNS(JBDDirty, jbddirty)
 311 BUFFER_FNS(Revoked, revoked)
 312 TAS_BUFFER_FNS(Revoked, revoked)
 313 BUFFER_FNS(RevokeValid, revokevalid)
 314 TAS_BUFFER_FNS(RevokeValid, revokevalid)
 315 BUFFER_FNS(Freed, freed)
 316
 317 static inline struct buffer_head *jh2bh(struct journal_head *jh)
 318 {
 319         return jh->b_bh;
 320 }
 321
 322 static inline struct journal_head *bh2jh(struct buffer_head *bh)
 323 {
 324         return bh->b_private;
 325 }
 326
 327 static inline void jbd_lock_bh_state(struct buffer_head *bh)
 328 {
 329         bit_spin_lock(BH_State, &bh->b_state);
 330 }
 331
 332 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
 333 {
 334         return bit_spin_trylock(BH_State, &bh->b_state);
 335 }
 336
 337 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
 338 {
 339         return bit_spin_is_locked(BH_State, &bh->b_state);
 340 }
 341
 342 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
 343 {
 344         bit_spin_unlock(BH_State, &bh->b_state);
 345 }
 346
 347 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
 348 {
 349         bit_spin_lock(BH_JournalHead, &bh->b_state);
 350 }
 351
 352 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
 353 {
 354         bit_spin_unlock(BH_JournalHead, &bh->b_state);
 355 }
 356
 357 struct jbd_revoke_table_s;
 358
 359 /**
 360  * struct handle_s - The handle_s type is the concrete type associated with
 361  *     handle_t.
 362  * @h_transaction: Which compound transaction is this update a part of?
 363  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
 364  * @h_ref: Reference count on this handle
 365  * @h_err: Field for caller's use to track errors through large fs operations
 366  * @h_sync: flag for sync-on-close
 367  * @h_jdata: flag to force data journaling
 368  * @h_aborted: flag indicating fatal error on handle
 369  **/
 370
 371 /* Docbook can't yet cope with the bit fields, but will leave the documentation
 372  * in so it can be fixed later.
 373  */
 374
 375 struct handle_s
 376 {
 377         /* Which compound transaction is this update a part of? */
 378         transaction_t           *h_transaction;
 379
 380         /* Number of remaining buffers we are allowed to dirty: */
 381         int                     h_buffer_credits;
 382
 383         /* Reference count on this handle */
 384         int                     h_ref;
 385
 386         /* Field for caller's use to track errors through large fs */
 387         /* operations */
 388         int                     h_err;
 389
 390         /* Flags [no locking] */
 391         unsigned int    h_sync:         1;      /* sync-on-close */
 392         unsigned int    h_jdata:        1;      /* force data journaling */
 393         unsigned int    h_aborted:      1;      /* fatal error on handle */
 394 };
 395
 396
 397 /* The transaction_t type is the guts of the journaling mechanism.  It
 398  * tracks a compound transaction through its various states:
 399  *
 400  * RUNNING:     accepting new updates
 401  * LOCKED:      Updates still running but we don't accept new ones
 402  * RUNDOWN:     Updates are tidying up but have finished requesting
 403  *              new buffers to modify (state not used for now)
 404  * FLUSH:       All updates complete, but we are still writing to disk
 405  * COMMIT:      All data on disk, writing commit record
 406  * FINISHED:    We still have to keep the transaction for checkpointing.
 407  *
 408  * The transaction keeps track of all of the buffers modified by a
 409  * running transaction, and all of the buffers committed but not yet
 410  * flushed to home for finished transactions.
 411  */
 412
 413 /*
 414  * Lock ranking:
 415  *
 416  *    j_list_lock
 417  *      ->jbd_lock_bh_journal_head()    (This is "innermost")
 418  *
 419  *    j_state_lock
 420  *    ->jbd_lock_bh_state()
 421  *
 422  *    jbd_lock_bh_state()
 423  *    ->j_list_lock
 424  *
 425  *    j_state_lock
 426  *    ->t_handle_lock
 427  *
 428  *    j_state_lock
 429  *    ->j_list_lock                     (journal_unmap_buffer)
 430  *
 431  */
 432
 433 struct transaction_s
 434 {
 435         /* Pointer to the journal for this transaction. [no locking] */
 436         journal_t               *t_journal;
 437
 438         /* Sequence number for this transaction [no locking] */
 439         tid_t                   t_tid;
 440
 441         /*
 442          * Transaction's current state
 443          * [no locking - only kjournald alters this]
 444          * FIXME: needs barriers
 445          * KLUDGE: [use j_state_lock]
 446          */
 447         enum {
 448                 T_RUNNING,
 449                 T_LOCKED,
 450                 T_RUNDOWN,
 451                 T_FLUSH,
 452                 T_COMMIT,
 453                 T_FINISHED
 454         }                       t_state;
 455
 456         /*
 457          * Where in the log does this transaction's commit start? [no locking]
 458          */
 459         unsigned long           t_log_start;
 460
 461         /* Number of buffers on the t_buffers list [j_list_lock] */
 462         int                     t_nr_buffers;
 463
 464         /*
 465          * Doubly-linked circular list of all buffers reserved but not yet
 466          * modified by this transaction [j_list_lock]
 467          */
 468         struct journal_head     *t_reserved_list;
 469
 470         /*
 471          * Doubly-linked circular list of all buffers under writeout during
 472          * commit [j_list_lock]
 473          */
 474         struct journal_head     *t_locked_list;
 475
 476         /*
 477          * Doubly-linked circular list of all metadata buffers owned by this
 478          * transaction [j_list_lock]
 479          */
 480         struct journal_head     *t_buffers;
 481
 482         /*
 483          * Doubly-linked circular list of all data buffers still to be
 484          * flushed before this transaction can be committed [j_list_lock]
 485          */
 486         struct journal_head     *t_sync_datalist;
 487
 488         /*
 489          * Doubly-linked circular list of all forget buffers (superseded
 490          * buffers which we can un-checkpoint once this transaction commits)
 491          * [j_list_lock]
 492          */
 493         struct journal_head     *t_forget;
 494
 495         /*
 496          * Doubly-linked circular list of all buffers still to be flushed before
 497          * this transaction can be checkpointed. [j_list_lock]
 498          */
 499         struct journal_head     *t_checkpoint_list;
 500
 501         /*
 502          * Doubly-linked circular list of temporary buffers currently undergoing
 503          * IO in the log [j_list_lock]
 504          */
 505         struct journal_head     *t_iobuf_list;
 506
 507         /*
 508          * Doubly-linked circular list of metadata buffers being shadowed by log
 509          * IO.  The IO buffers on the iobuf list and the shadow buffers on this
 510          * list match each other one for one at all times. [j_list_lock]
 511          */
 512         struct journal_head     *t_shadow_list;
 513
 514         /*
 515          * Doubly-linked circular list of control buffers being written to the
 516          * log. [j_list_lock]
 517          */
 518         struct journal_head     *t_log_list;
 519
 520         /*
 521          * Protects info related to handles
 522          */
 523         spinlock_t              t_handle_lock;
 524
 525         /*
 526          * Number of outstanding updates running on this transaction
 527          * [t_handle_lock]
 528          */
 529         int                     t_updates;
 530
 531         /*
 532          * Number of buffers reserved for use by all handles in this transaction
 533          * handle but not yet modified. [t_handle_lock]
 534          */
 535         int                     t_outstanding_credits;
 536
 537         /*
 538          * Forward and backward links for the circular list of all transactions
 539          * awaiting checkpoint. [j_list_lock]
 540          */
 541         transaction_t           *t_cpnext, *t_cpprev;
 542
 543         /*
 544          * When will the transaction expire (become due for commit), in jiffies?
 545          * [no locking]
 546          */
 547         unsigned long           t_expires;
 548
 549         /*
 550          * How many handles used this transaction? [t_handle_lock]
 551          */
 552         int t_handle_count;
 553
 554 };
 555
 556 /**
 557  * struct journal_s - The journal_s type is the concrete type associated with
 558  *     journal_t.
 559  * @j_flags:  General journaling state flags
 560  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
 561  *     prior abort)?
 562  * @j_sb_buffer: First part of superblock buffer
 563  * @j_superblock: Second part of superblock buffer
 564  * @j_format_version: Version of the superblock format
 565  * @j_state_lock: Protect the various scalars in the journal
 566  * @j_barrier_count:  Number of processes waiting to create a barrier lock
 567  * @j_barrier: The barrier lock itself
 568  * @j_running_transaction: The current running transaction..
 569  * @j_committing_transaction: the transaction we are pushing to disk
 570  * @j_checkpoint_transactions: a linked circular list of all transactions
 571  *  waiting for checkpointing
 572  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
 573  *  to start committing, or for a barrier lock to be released
 574  * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
 575  * @j_wait_done_commit: Wait queue for waiting for commit to complete
 576  * @j_wait_checkpoint:  Wait queue to trigger checkpointing
 577  * @j_wait_commit: Wait queue to trigger commit
 578  * @j_wait_updates: Wait queue to wait for updates to complete
 579  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
 580  * @j_head: Journal head - identifies the first unused block in the journal
 581  * @j_tail: Journal tail - identifies the oldest still-used block in the
 582  *  journal.
 583  * @j_free: Journal free - how many free blocks are there in the journal?
 584  * @j_first: The block number of the first usable block
 585  * @j_last: The block number one beyond the last usable block
 586  * @j_dev: Device where we store the journal
 587  * @j_blocksize: blocksize for the location where we store the journal.
 588  * @j_blk_offset: starting block offset for into the device where we store the
 589  *     journal
 590  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
 591  *     be equal to j_dev
 592  * @j_maxlen: Total maximum capacity of the journal region on disk.
 593  * @j_list_lock: Protects the buffer lists and internal buffer state.
 594  * @j_inode: Optional inode where we store the journal.  If present, all journal
 595  *     block numbers are mapped into this inode via bmap().
 596  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
 597  * @j_transaction_sequence: Sequence number of the next transaction to grant
 598  * @j_commit_sequence: Sequence number of the most recently committed
 599  *  transaction
 600  * @j_commit_request: Sequence number of the most recent transaction wanting
 601  *     commit
 602  * @j_uuid: Uuid of client object.
 603  * @j_task: Pointer to the current commit thread for this journal
 604  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
 605  *     single compound commit transaction
 606  * @j_commit_interval: What is the maximum transaction lifetime before we begin
 607  *  a commit?
 608  * @j_commit_timer:  The timer used to wakeup the commit thread
 609  * @j_revoke_lock: Protect the revoke table
 610  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
 611  *     current transaction.
 612  * @j_revoke_table: alternate revoke tables for j_revoke
 613  * @j_wbuf: array of buffer_heads for journal_commit_transaction
 614  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
 615  *      number that will fit in j_blocksize
 616  * @j_last_sync_writer: most recent pid which did a synchronous write
 617  * @j_private: An opaque pointer to fs-private information.
 618  */
 619
 620 struct journal_s
 621 {
 622         /* General journaling state flags [j_state_lock] */
 623         unsigned long           j_flags;
 624
 625         /*
 626          * Is there an outstanding uncleared error on the journal (from a prior
 627          * abort)? [j_state_lock]
 628          */
 629         int                     j_errno;
 630
 631         /* The superblock buffer */
 632         struct buffer_head      *j_sb_buffer;
 633         journal_superblock_t    *j_superblock;
 634
 635         /* Version of the superblock format */
 636         int                     j_format_version;
 637
 638         /*
 639          * Protect the various scalars in the journal
 640          */
 641         spinlock_t              j_state_lock;
 642
 643         /*
 644          * Number of processes waiting to create a barrier lock [j_state_lock]
 645          */
 646         int                     j_barrier_count;
 647
 648         /* The barrier lock itself */
 649         struct mutex            j_barrier;
 650
 651         /*
 652          * Transactions: The current running transaction...
 653          * [j_state_lock] [caller holding open handle]
 654          */
 655         transaction_t           *j_running_transaction;
 656
 657         /*
 658          * the transaction we are pushing to disk
 659          * [j_state_lock] [caller holding open handle]
 660          */
 661         transaction_t           *j_committing_transaction;
 662
 663         /*
 664          * ... and a linked circular list of all transactions waiting for
 665          * checkpointing. [j_list_lock]
 666          */
 667         transaction_t           *j_checkpoint_transactions;
 668
 669         /*
 670          * Wait queue for waiting for a locked transaction to start committing,
 671          * or for a barrier lock to be released
 672          */
 673         wait_queue_head_t       j_wait_transaction_locked;
 674
 675         /* Wait queue for waiting for checkpointing to complete */
 676         wait_queue_head_t       j_wait_logspace;
 677
 678         /* Wait queue for waiting for commit to complete */
 679         wait_queue_head_t       j_wait_done_commit;
 680
 681         /* Wait queue to trigger checkpointing */
 682         wait_queue_head_t       j_wait_checkpoint;
 683
 684         /* Wait queue to trigger commit */
 685         wait_queue_head_t       j_wait_commit;
 686
 687         /* Wait queue to wait for updates to complete */
 688         wait_queue_head_t       j_wait_updates;
 689
 690         /* Semaphore for locking against concurrent checkpoints */
 691         struct mutex            j_checkpoint_mutex;
 692
 693         /*
 694          * Journal head: identifies the first unused block in the journal.
 695          * [j_state_lock]
 696          */
 697         unsigned long           j_head;
 698
 699         /*
 700          * Journal tail: identifies the oldest still-used block in the journal.
 701          * [j_state_lock]
 702          */
 703         unsigned long           j_tail;
 704
 705         /*
 706          * Journal free: how many free blocks are there in the journal?
 707          * [j_state_lock]
 708          */
 709         unsigned long           j_free;
 710
 711         /*
 712          * Journal start and end: the block numbers of the first usable block
 713          * and one beyond the last usable block in the journal. [j_state_lock]
 714          */
 715         unsigned long           j_first;
 716         unsigned long           j_last;
 717
 718         /*
 719          * Device, blocksize and starting block offset for the location where we
 720          * store the journal.
 721          */
 722         struct block_device     *j_dev;
 723         int                     j_blocksize;
 724         unsigned int            j_blk_offset;
 725
 726         /*
 727          * Device which holds the client fs.  For internal journal this will be
 728          * equal to j_dev.
 729          */
 730         struct block_device     *j_fs_dev;
 731
 732         /* Total maximum capacity of the journal region on disk. */
 733         unsigned int            j_maxlen;
 734
 735         /*
 736          * Protects the buffer lists and internal buffer state.
 737          */
 738         spinlock_t              j_list_lock;
 739
 740         /* Optional inode where we store the journal.  If present, all */
 741         /* journal block numbers are mapped into this inode via */
 742         /* bmap(). */
 743         struct inode            *j_inode;
 744
 745         /*
 746          * Sequence number of the oldest transaction in the log [j_state_lock]
 747          */
 748         tid_t                   j_tail_sequence;
 749
 750         /*
 751          * Sequence number of the next transaction to grant [j_state_lock]
 752          */
 753         tid_t                   j_transaction_sequence;
 754
 755         /*
 756          * Sequence number of the most recently committed transaction
 757          * [j_state_lock].
 758          */
 759         tid_t                   j_commit_sequence;
 760
 761         /*
 762          * Sequence number of the most recent transaction wanting commit
 763          * [j_state_lock]
 764          */
 765         tid_t                   j_commit_request;
 766
 767         /*
 768          * Journal uuid: identifies the object (filesystem, LVM volume etc)
 769          * backed by this journal.  This will eventually be replaced by an array
 770          * of uuids, allowing us to index multiple devices within a single
 771          * journal and to perform atomic updates across them.
 772          */
 773         __u8                    j_uuid[16];
 774
 775         /* Pointer to the current commit thread for this journal */
 776         struct task_struct      *j_task;
 777
 778         /*
 779          * Maximum number of metadata buffers to allow in a single compound
 780          * commit transaction
 781          */
 782         int                     j_max_transaction_buffers;
 783
 784         /*
 785          * What is the maximum transaction lifetime before we begin a commit?
 786          */
 787         unsigned long           j_commit_interval;
 788
 789         /* The timer used to wakeup the commit thread: */
 790         struct timer_list       *j_commit_timer;
 791
 792         /*
 793          * The revoke table: maintains the list of revoked blocks in the
 794          * current transaction.  [j_revoke_lock]
 795          */
 796         spinlock_t              j_revoke_lock;
 797         struct jbd_revoke_table_s *j_revoke;
 798         struct jbd_revoke_table_s *j_revoke_table[2];
 799
 800         /*
 801          * array of bhs for journal_commit_transaction
 802          */
 803         struct buffer_head      **j_wbuf;
 804         int                     j_wbufsize;
 805
 806         pid_t                   j_last_sync_writer;
 807
 808         /*
 809          * An opaque pointer to fs-private information.  ext3 puts its
 810          * superblock pointer here
 811          */
 812         void *j_private;
 813 };
 814
 815 /*
 816  * Journal flag definitions
 817  */
 818 #define JFS_UNMOUNT     0x001   /* Journal thread is being destroyed */
 819 #define JFS_ABORT       0x002   /* Journaling has been aborted for errors. */
 820 #define JFS_ACK_ERR     0x004   /* The errno in the sb has been acked */
 821 #define JFS_FLUSHED     0x008   /* The journal superblock has been flushed */
 822 #define JFS_LOADED      0x010   /* The journal superblock has been loaded */
 823 #define JFS_BARRIER     0x020   /* Use IDE barriers */
 824
 825 /*
 826  * Function declarations for the journaling transaction and buffer
 827  * management
 828  */
 829
 830 /* Filing buffers */
 831 extern void __journal_temp_unlink_buffer(struct journal_head *jh);
 832 extern void journal_unfile_buffer(journal_t *, struct journal_head *);
 833 extern void __journal_unfile_buffer(struct journal_head *);
 834 extern void __journal_refile_buffer(struct journal_head *);
 835 extern void journal_refile_buffer(journal_t *, struct journal_head *);
 836 extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
 837 extern void __journal_free_buffer(struct journal_head *bh);
 838 extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
 839 extern void __journal_clean_data_list(transaction_t *transaction);
 840
 841 /* Log buffer allocation */
 842 extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
 843 int journal_next_log_block(journal_t *, unsigned long *);
 844
 845 /* Commit management */
 846 extern void journal_commit_transaction(journal_t *);
 847
 848 /* Checkpoint list management */
 849 int __journal_clean_checkpoint_list(journal_t *journal);
 850 void __journal_remove_checkpoint(struct journal_head *);
 851 void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
 852
 853 /* Buffer IO */
 854 extern int
 855 journal_write_metadata_buffer(transaction_t       *transaction,
 856                               struct journal_head  *jh_in,
 857                               struct journal_head **jh_out,
 858                               int                  blocknr);
 859
 860 /* Transaction locking */
 861 extern void             __wait_on_journal (journal_t *);
 862
 863 /*
 864  * Journal locking.
 865  *
 866  * We need to lock the journal during transaction state changes so that nobody
 867  * ever tries to take a handle on the running transaction while we are in the
 868  * middle of moving it to the commit phase.  j_state_lock does this.
 869  *
 870  * Note that the locking is completely interrupt unsafe.  We never touch
 871  * journal structures from interrupts.
 872  */
 873
 874 static inline handle_t *journal_current_handle(void)
 875 {
 876         return current->journal_info;
 877 }
 878
 879 /* The journaling code user interface:
 880  *
 881  * Create and destroy handles
 882  * Register buffer modifications against the current transaction.
 883  */
 884
 885 extern handle_t *journal_start(journal_t *, int nblocks);
 886 extern int       journal_restart (handle_t *, int nblocks);
 887 extern int       journal_extend (handle_t *, int nblocks);
 888 extern int       journal_get_write_access(handle_t *, struct buffer_head *);
 889 extern int       journal_get_create_access (handle_t *, struct buffer_head *);
 890 extern int       journal_get_undo_access(handle_t *, struct buffer_head *);
 891 extern int       journal_dirty_data (handle_t *, struct buffer_head *);
 892 extern int       journal_dirty_metadata (handle_t *, struct buffer_head *);
 893 extern void      journal_release_buffer (handle_t *, struct buffer_head *);
 894 extern int       journal_forget (handle_t *, struct buffer_head *);
 895 extern void      journal_sync_buffer (struct buffer_head *);
 896 extern int       journal_invalidatepage(journal_t *,
 897                                 struct page *, unsigned long);
 898 extern int       journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
 899 extern int       journal_stop(handle_t *);
 900 extern int       journal_flush (journal_t *);
 901 extern void      journal_lock_updates (journal_t *);
 902 extern void      journal_unlock_updates (journal_t *);
 903
 904 extern journal_t * journal_init_dev(struct block_device *bdev,
 905                                 struct block_device *fs_dev,
 906                                 int start, int len, int bsize);
 907 extern journal_t * journal_init_inode (struct inode *);
 908 extern int         journal_update_format (journal_t *);
 909 extern int         journal_check_used_features
 910                    (journal_t *, unsigned long, unsigned long, unsigned long);
 911 extern int         journal_check_available_features
 912                    (journal_t *, unsigned long, unsigned long, unsigned long);
 913 extern int         journal_set_features
 914                    (journal_t *, unsigned long, unsigned long, unsigned long);
 915 extern int         journal_create     (journal_t *);
 916 extern int         journal_load       (journal_t *journal);
 917 extern void        journal_destroy    (journal_t *);
 918 extern int         journal_recover    (journal_t *journal);
 919 extern int         journal_wipe       (journal_t *, int);
 920 extern int         journal_skip_recovery        (journal_t *);
 921 extern void        journal_update_superblock    (journal_t *, int);
 922 extern void        __journal_abort_hard (journal_t *);
 923 extern void        journal_abort      (journal_t *, int);
 924 extern int         journal_errno      (journal_t *);
 925 extern void        journal_ack_err    (journal_t *);
 926 extern int         journal_clear_err  (journal_t *);
 927 extern int         journal_bmap(journal_t *, unsigned long, unsigned long *);
 928 extern int         journal_force_commit(journal_t *);
 929
 930 /*
 931  * journal_head management
 932  */
 933 struct journal_head *journal_add_journal_head(struct buffer_head *bh);
 934 struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
 935 void journal_remove_journal_head(struct buffer_head *bh);
 936 void journal_put_journal_head(struct journal_head *jh);
 937
 938 /*
 939  * handle management
 940  */
 941 extern kmem_cache_t *jbd_handle_cache;
 942
 943 static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
 944 {
 945         return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
 946 }
 947
 948 static inline void jbd_free_handle(handle_t *handle)
 949 {
 950         kmem_cache_free(jbd_handle_cache, handle);
 951 }
 952
 953 /* Primary revoke support */
 954 #define JOURNAL_REVOKE_DEFAULT_HASH 256
 955 extern int         journal_init_revoke(journal_t *, int);
 956 extern void        journal_destroy_revoke_caches(void);
 957 extern int         journal_init_revoke_caches(void);
 958
 959 extern void        journal_destroy_revoke(journal_t *);
 960 extern int         journal_revoke (handle_t *,
 961                                 unsigned long, struct buffer_head *);
 962 extern int         journal_cancel_revoke(handle_t *, struct journal_head *);
 963 extern void        journal_write_revoke_records(journal_t *, transaction_t *);
 964
 965 /* Recovery revoke support */
 966 extern int      journal_set_revoke(journal_t *, unsigned long, tid_t);
 967 extern int      journal_test_revoke(journal_t *, unsigned long, tid_t);
 968 extern void     journal_clear_revoke(journal_t *);
 969 extern void     journal_brelse_array(struct buffer_head *b[], int n);
 970 extern void     journal_switch_revoke_table(journal_t *journal);
 971
 972 /*
 973  * The log thread user interface:
 974  *
 975  * Request space in the current transaction, and force transaction commit
 976  * transitions on demand.
 977  */
 978
 979 int __log_space_left(journal_t *); /* Called with journal locked */
 980 int log_start_commit(journal_t *journal, tid_t tid);
 981 int __log_start_commit(journal_t *journal, tid_t tid);
 982 int journal_start_commit(journal_t *journal, tid_t *tid);
 983 int journal_force_commit_nested(journal_t *journal);
 984 int log_wait_commit(journal_t *journal, tid_t tid);
 985 int log_do_checkpoint(journal_t *journal);
 986
 987 void __log_wait_for_space(journal_t *journal);
 988 extern void     __journal_drop_transaction(journal_t *, transaction_t *);
 989 extern int      cleanup_journal_tail(journal_t *);
 990
 991 /* Debugging code only: */
 992
 993 #define jbd_ENOSYS() \
 994 do {                                                                       \
 995         printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
 996         current->state = TASK_UNINTERRUPTIBLE;                             \
 997         schedule();                                                        \
 998 } while (1)
 999
1000 /*
1001  * is_journal_abort
1002  *
1003  * Simple test wrapper function to test the JFS_ABORT state flag.  This
1004  * bit, when set, indicates that we have had a fatal error somewhere,
1005  * either inside the journaling layer or indicated to us by the client
1006  * (eg. ext3), and that we and should not commit any further
1007  * transactions.
1008  */
1009
1010 static inline int is_journal_aborted(journal_t *journal)
1011 {
1012         return journal->j_flags & JFS_ABORT;
1013 }
1014
1015 static inline int is_handle_aborted(handle_t *handle)
1016 {
1017         if (handle->h_aborted)
1018                 return 1;
1019         return is_journal_aborted(handle->h_transaction->t_journal);
1020 }
1021
1022 static inline void journal_abort_handle(handle_t *handle)
1023 {
1024         handle->h_aborted = 1;
1025 }
1026
1027 #endif /* __KERNEL__   */
1028
1029 /* Comparison functions for transaction IDs: perform comparisons using
1030  * modulo arithmetic so that they work over sequence number wraps. */
1031
1032 static inline int tid_gt(tid_t x, tid_t y)
1033 {
1034         int difference = (x - y);
1035         return (difference > 0);
1036 }
1037
1038 static inline int tid_geq(tid_t x, tid_t y)
1039 {
1040         int difference = (x - y);
1041         return (difference >= 0);
1042 }
1043
1044 extern int journal_blocks_per_page(struct inode *inode);
1045
1046 /*
1047  * Return the minimum number of blocks which must be free in the journal
1048  * before a new transaction may be started.  Must be called under j_state_lock.
1049  */
1050 static inline int jbd_space_needed(journal_t *journal)
1051 {
1052         int nblocks = journal->j_max_transaction_buffers;
1053         if (journal->j_committing_transaction)
1054                 nblocks += journal->j_committing_transaction->
1055                                         t_outstanding_credits;
1056         return nblocks;
1057 }
1058
1059 /*
1060  * Definitions which augment the buffer_head layer
1061  */
1062
1063 /* journaling buffer types */
1064 #define BJ_None         0       /* Not journaled */
1065 #define BJ_SyncData     1       /* Normal data: flush before commit */
1066 #define BJ_Metadata     2       /* Normal journaled metadata */
1067 #define BJ_Forget       3       /* Buffer superseded by this transaction */
1068 #define BJ_IO           4       /* Buffer is for temporary IO use */
1069 #define BJ_Shadow       5       /* Buffer contents being shadowed to the log */
1070 #define BJ_LogCtl       6       /* Buffer contains log descriptors */
1071 #define BJ_Reserved     7       /* Buffer is reserved for access by journal */
1072 #define BJ_Locked       8       /* Locked for I/O during commit */
1073 #define BJ_Types        9
1074
1075 extern int jbd_blocks_per_page(struct inode *inode);
1076
1077 #ifdef __KERNEL__
1078
1079 #define buffer_trace_init(bh)   do {} while (0)
1080 #define print_buffer_fields(bh) do {} while (0)
1081 #define print_buffer_trace(bh)  do {} while (0)
1082 #define BUFFER_TRACE(bh, info)  do {} while (0)
1083 #define BUFFER_TRACE2(bh, bh2, info)    do {} while (0)
1084 #define JBUFFER_TRACE(jh, info) do {} while (0)
1085
1086 #endif  /* __KERNEL__ */
1087
1088 #endif  /* _LINUX_JBD_H */