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 <linux/timer.h>
  33 #include <linux/lockdep.h>
  34
  35 #include <asm/semaphore.h>
  36
  37 #define journal_oom_retry 1
  38
  39 /*
  40  * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
  41  * certain classes of error which can occur due to failed IOs.  Under
  42  * normal use we want ext3 to continue after such errors, because
  43  * hardware _can_ fail, but for debugging purposes when running tests on
  44  * known-good hardware we may want to trap these errors.
  45  */
  46 #undef JBD_PARANOID_IOFAIL
  47
  48 /*
  49  * The default maximum commit age, in seconds.
  50  */
  51 #define JBD_DEFAULT_MAX_COMMIT_AGE 5
  52
  53 #ifdef CONFIG_JBD_DEBUG
  54 /*
  55  * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
  56  * consistency checks.  By default we don't do this unless
  57  * CONFIG_JBD_DEBUG is on.
  58  */
  59 #define JBD_EXPENSIVE_CHECKING
  60 extern u8 journal_enable_debug;
  61
  62 #define jbd_debug(n, f, a...)                                           \
  63         do {                                                            \
  64                 if ((n) <= journal_enable_debug) {                      \
  65                         printk (KERN_DEBUG "(%s, %d): %s: ",            \
  66                                 __FILE__, __LINE__, __FUNCTION__);      \
  67                         printk (f, ## a);                               \
  68                 }                                                       \
  69         } while (0)
  70 #else
  71 #define jbd_debug(f, a...)      /**/
  72 #endif
  73
  74 static inline void *jbd_alloc(size_t size, gfp_t flags)
  75 {
  76         return (void *)__get_free_pages(flags, get_order(size));
  77 }
  78
  79 static inline void jbd_free(void *ptr, size_t size)
  80 {
  81         free_pages((unsigned long)ptr, get_order(size));
  82 };
  83
  84 #define JFS_MIN_JOURNAL_BLOCKS 1024
  85
  86
  87 /**
  88  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
  89  *
  90  * All filesystem modifications made by the process go
  91  * through this handle.  Recursive operations (such as quota operations)
  92  * are gathered into a single update.
  93  *
  94  * The buffer credits field is used to account for journaled buffers
  95  * being modified by the running process.  To ensure that there is
  96  * enough log space for all outstanding operations, we need to limit the
  97  * number of outstanding buffers possible at any time.  When the
  98  * operation completes, any buffer credits not used are credited back to
  99  * the transaction, so that at all times we know how many buffers the
 100  * outstanding updates on a transaction might possibly touch.
 101  *
 102  * This is an opaque datatype.
 103  **/
 104 typedef struct handle_s         handle_t;       /* Atomic operation type */
 105
 106
 107 /**
 108  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
 109  *
 110  * journal_t is linked to from the fs superblock structure.
 111  *
 112  * We use the journal_t to keep track of all outstanding transaction
 113  * activity on the filesystem, and to manage the state of the log
 114  * writing process.
 115  *
 116  * This is an opaque datatype.
 117  **/
 118 typedef struct journal_s        journal_t;      /* Journal control structure */
 119 #endif
 120
 121 /*
 122  * Internal structures used by the logging mechanism:
 123  */
 124
 125 #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
 126
 127 /*
 128  * On-disk structures
 129  */
 130
 131 /*
 132  * Descriptor block types:
 133  */
 134
 135 #define JFS_DESCRIPTOR_BLOCK    1
 136 #define JFS_COMMIT_BLOCK        2
 137 #define JFS_SUPERBLOCK_V1       3
 138 #define JFS_SUPERBLOCK_V2       4
 139 #define JFS_REVOKE_BLOCK        5
 140
 141 /*
 142  * Standard header for all descriptor blocks:
 143  */
 144 typedef struct journal_header_s
 145 {
 146         __be32          h_magic;
 147         __be32          h_blocktype;
 148         __be32          h_sequence;
 149 } journal_header_t;
 150
 151
 152 /*
 153  * The block tag: used to describe a single buffer in the journal
 154  */
 155 typedef struct journal_block_tag_s
 156 {
 157         __be32          t_blocknr;      /* The on-disk block number */
 158         __be32          t_flags;        /* See below */
 159 } journal_block_tag_t;
 160
 161 /*
 162  * The revoke descriptor: used on disk to describe a series of blocks to
 163  * be revoked from the log
 164  */
 165 typedef struct journal_revoke_header_s
 166 {
 167         journal_header_t r_header;
 168         __be32           r_count;       /* Count of bytes used in the block */
 169 } journal_revoke_header_t;
 170
 171
 172 /* Definitions for the journal tag flags word: */
 173 #define JFS_FLAG_ESCAPE         1       /* on-disk block is escaped */
 174 #define JFS_FLAG_SAME_UUID      2       /* block has same uuid as previous */
 175 #define JFS_FLAG_DELETED        4       /* block deleted by this transaction */
 176 #define JFS_FLAG_LAST_TAG       8       /* last tag in this descriptor block */
 177
 178
 179 /*
 180  * The journal superblock.  All fields are in big-endian byte order.
 181  */
 182 typedef struct journal_superblock_s
 183 {
 184 /* 0x0000 */
 185         journal_header_t s_header;
 186
 187 /* 0x000C */
 188         /* Static information describing the journal */
 189         __be32  s_blocksize;            /* journal device blocksize */
 190         __be32  s_maxlen;               /* total blocks in journal file */
 191         __be32  s_first;                /* first block of log information */
 192
 193 /* 0x0018 */
 194         /* Dynamic information describing the current state of the log */
 195         __be32  s_sequence;             /* first commit ID expected in log */
 196         __be32  s_start;                /* blocknr of start of log */
 197
 198 /* 0x0020 */
 199         /* Error value, as set by journal_abort(). */
 200         __be32  s_errno;
 201
 202 /* 0x0024 */
 203         /* Remaining fields are only valid in a version-2 superblock */
 204         __be32  s_feature_compat;       /* compatible feature set */
 205         __be32  s_feature_incompat;     /* incompatible feature set */
 206         __be32  s_feature_ro_compat;    /* readonly-compatible feature set */
 207 /* 0x0030 */
 208         __u8    s_uuid[16];             /* 128-bit uuid for journal */
 209
 210 /* 0x0040 */
 211         __be32  s_nr_users;             /* Nr of filesystems sharing log */
 212
 213         __be32  s_dynsuper;             /* Blocknr of dynamic superblock copy*/
 214
 215 /* 0x0048 */
 216         __be32  s_max_transaction;      /* Limit of journal blocks per trans.*/
 217         __be32  s_max_trans_data;       /* Limit of data blocks per trans. */
 218
 219 /* 0x0050 */
 220         __u32   s_padding[44];
 221
 222 /* 0x0100 */
 223         __u8    s_users[16*48];         /* ids of all fs'es sharing the log */
 224 /* 0x0400 */
 225 } journal_superblock_t;
 226
 227 #define JFS_HAS_COMPAT_FEATURE(j,mask)                                  \
 228         ((j)->j_format_version >= 2 &&                                  \
 229          ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
 230 #define JFS_HAS_RO_COMPAT_FEATURE(j,mask)                               \
 231         ((j)->j_format_version >= 2 &&                                  \
 232          ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
 233 #define JFS_HAS_INCOMPAT_FEATURE(j,mask)                                \
 234         ((j)->j_format_version >= 2 &&                                  \
 235          ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
 236
 237 #define JFS_FEATURE_INCOMPAT_REVOKE     0x00000001
 238
 239 /* Features known to this kernel version: */
 240 #define JFS_KNOWN_COMPAT_FEATURES       0
 241 #define JFS_KNOWN_ROCOMPAT_FEATURES     0
 242 #define JFS_KNOWN_INCOMPAT_FEATURES     JFS_FEATURE_INCOMPAT_REVOKE
 243
 244 #ifdef __KERNEL__
 245
 246 #include <linux/fs.h>
 247 #include <linux/sched.h>
 248
 249 #define J_ASSERT(assert)        BUG_ON(!(assert))
 250
 251 #if defined(CONFIG_BUFFER_DEBUG)
 252 void buffer_assertion_failure(struct buffer_head *bh);
 253 #define J_ASSERT_BH(bh, expr)                                           \
 254         do {                                                            \
 255                 if (!(expr))                                            \
 256                         buffer_assertion_failure(bh);                   \
 257                 J_ASSERT(expr);                                         \
 258         } while (0)
 259 #define J_ASSERT_JH(jh, expr)   J_ASSERT_BH(jh2bh(jh), expr)
 260 #else
 261 #define J_ASSERT_BH(bh, expr)   J_ASSERT(expr)
 262 #define J_ASSERT_JH(jh, expr)   J_ASSERT(expr)
 263 #endif
 264
 265 #if defined(JBD_PARANOID_IOFAIL)
 266 #define J_EXPECT(expr, why...)          J_ASSERT(expr)
 267 #define J_EXPECT_BH(bh, expr, why...)   J_ASSERT_BH(bh, expr)
 268 #define J_EXPECT_JH(jh, expr, why...)   J_ASSERT_JH(jh, expr)
 269 #else
 270 #define __journal_expect(expr, why...)                                       \
 271         ({                                                                   \
 272                 int val = (expr);                                            \
 273                 if (!val) {                                                  \
 274                         printk(KERN_ERR                                      \
 275                                 "EXT3-fs unexpected failure: %s;\n",# expr); \
 276                         printk(KERN_ERR why "\n");                           \
 277                 }                                                            \
 278                 val;                                                         \
 279         })
 280 #define J_EXPECT(expr, why...)          __journal_expect(expr, ## why)
 281 #define J_EXPECT_BH(bh, expr, why...)   __journal_expect(expr, ## why)
 282 #define J_EXPECT_JH(jh, expr, why...)   __journal_expect(expr, ## why)
 283 #endif
 284
 285 enum jbd_state_bits {
 286         BH_JBD                  /* Has an attached ext3 journal_head */
 287           = BH_PrivateStart,
 288         BH_JWrite,              /* Being written to log (@@@ DEBUGGING) */
 289         BH_Freed,               /* Has been freed (truncated) */
 290         BH_Revoked,             /* Has been revoked from the log */
 291         BH_RevokeValid,         /* Revoked flag is valid */
 292         BH_JBDDirty,            /* Is dirty but journaled */
 293         BH_State,               /* Pins most journal_head state */
 294         BH_JournalHead,         /* Pins bh->b_private and jh->b_bh */
 295         BH_Unshadow,            /* Dummy bit, for BJ_Shadow wakeup filtering */
 296 };
 297
 298 BUFFER_FNS(JBD, jbd)
 299 BUFFER_FNS(JWrite, jwrite)
 300 BUFFER_FNS(JBDDirty, jbddirty)
 301 TAS_BUFFER_FNS(JBDDirty, jbddirty)
 302 BUFFER_FNS(Revoked, revoked)
 303 TAS_BUFFER_FNS(Revoked, revoked)
 304 BUFFER_FNS(RevokeValid, revokevalid)
 305 TAS_BUFFER_FNS(RevokeValid, revokevalid)
 306 BUFFER_FNS(Freed, freed)
 307
 308 static inline struct buffer_head *jh2bh(struct journal_head *jh)
 309 {
 310         return jh->b_bh;
 311 }
 312
 313 static inline struct journal_head *bh2jh(struct buffer_head *bh)
 314 {
 315         return bh->b_private;
 316 }
 317
 318 static inline void jbd_lock_bh_state(struct buffer_head *bh)
 319 {
 320         bit_spin_lock(BH_State, &bh->b_state);
 321 }
 322
 323 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
 324 {
 325         return bit_spin_trylock(BH_State, &bh->b_state);
 326 }
 327
 328 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
 329 {
 330         return bit_spin_is_locked(BH_State, &bh->b_state);
 331 }
 332
 333 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
 334 {
 335         bit_spin_unlock(BH_State, &bh->b_state);
 336 }
 337
 338 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
 339 {
 340         bit_spin_lock(BH_JournalHead, &bh->b_state);
 341 }
 342
 343 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
 344 {
 345         bit_spin_unlock(BH_JournalHead, &bh->b_state);
 346 }
 347
 348 struct jbd_revoke_table_s;
 349
 350 /**
 351  * struct handle_s - this is the concrete type associated with handle_t.
 352  * @h_transaction: Which compound transaction is this update a part of?
 353  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
 354  * @h_ref: Reference count on this handle
 355  * @h_err: Field for caller's use to track errors through large fs operations
 356  * @h_sync: flag for sync-on-close
 357  * @h_jdata: flag to force data journaling
 358  * @h_aborted: flag indicating fatal error on handle
 359  * @h_lockdep_map: lockdep info for debugging lock problems
 360  */
 361 struct handle_s
 362 {
 363         /* Which compound transaction is this update a part of? */
 364         transaction_t           *h_transaction;
 365
 366         /* Number of remaining buffers we are allowed to dirty: */
 367         int                     h_buffer_credits;
 368
 369         /* Reference count on this handle */
 370         int                     h_ref;
 371
 372         /* Field for caller's use to track errors through large fs */
 373         /* operations */
 374         int                     h_err;
 375
 376         /* Flags [no locking] */
 377         unsigned int    h_sync:         1;      /* sync-on-close */
 378         unsigned int    h_jdata:        1;      /* force data journaling */
 379         unsigned int    h_aborted:      1;      /* fatal error on handle */
 380
 381 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 382         struct lockdep_map      h_lockdep_map;
 383 #endif
 384 };
 385
 386
 387 /* The transaction_t type is the guts of the journaling mechanism.  It
 388  * tracks a compound transaction through its various states:
 389  *
 390  * RUNNING:     accepting new updates
 391  * LOCKED:      Updates still running but we don't accept new ones
 392  * RUNDOWN:     Updates are tidying up but have finished requesting
 393  *              new buffers to modify (state not used for now)
 394  * FLUSH:       All updates complete, but we are still writing to disk
 395  * COMMIT:      All data on disk, writing commit record
 396  * FINISHED:    We still have to keep the transaction for checkpointing.
 397  *
 398  * The transaction keeps track of all of the buffers modified by a
 399  * running transaction, and all of the buffers committed but not yet
 400  * flushed to home for finished transactions.
 401  */
 402
 403 /*
 404  * Lock ranking:
 405  *
 406  *    j_list_lock
 407  *      ->jbd_lock_bh_journal_head()    (This is "innermost")
 408  *
 409  *    j_state_lock
 410  *    ->jbd_lock_bh_state()
 411  *
 412  *    jbd_lock_bh_state()
 413  *    ->j_list_lock
 414  *
 415  *    j_state_lock
 416  *    ->t_handle_lock
 417  *
 418  *    j_state_lock
 419  *    ->j_list_lock                     (journal_unmap_buffer)
 420  *
 421  */
 422
 423 struct transaction_s
 424 {
 425         /* Pointer to the journal for this transaction. [no locking] */
 426         journal_t               *t_journal;
 427
 428         /* Sequence number for this transaction [no locking] */
 429         tid_t                   t_tid;
 430
 431         /*
 432          * Transaction's current state
 433          * [no locking - only kjournald alters this]
 434          * [j_list_lock] guards transition of a transaction into T_FINISHED
 435          * state and subsequent call of __journal_drop_transaction()
 436          * FIXME: needs barriers
 437          * KLUDGE: [use j_state_lock]
 438          */
 439         enum {
 440                 T_RUNNING,
 441                 T_LOCKED,
 442                 T_RUNDOWN,
 443                 T_FLUSH,
 444                 T_COMMIT,
 445                 T_FINISHED
 446         }                       t_state;
 447
 448         /*
 449          * Where in the log does this transaction's commit start? [no locking]
 450          */
 451         unsigned long           t_log_start;
 452
 453         /* Number of buffers on the t_buffers list [j_list_lock] */
 454         int                     t_nr_buffers;
 455
 456         /*
 457          * Doubly-linked circular list of all buffers reserved but not yet
 458          * modified by this transaction [j_list_lock]
 459          */
 460         struct journal_head     *t_reserved_list;
 461
 462         /*
 463          * Doubly-linked circular list of all buffers under writeout during
 464          * commit [j_list_lock]
 465          */
 466         struct journal_head     *t_locked_list;
 467
 468         /*
 469          * Doubly-linked circular list of all metadata buffers owned by this
 470          * transaction [j_list_lock]
 471          */
 472         struct journal_head     *t_buffers;
 473
 474         /*
 475          * Doubly-linked circular list of all data buffers still to be
 476          * flushed before this transaction can be committed [j_list_lock]
 477          */
 478         struct journal_head     *t_sync_datalist;
 479
 480         /*
 481          * Doubly-linked circular list of all forget buffers (superseded
 482          * buffers which we can un-checkpoint once this transaction commits)
 483          * [j_list_lock]
 484          */
 485         struct journal_head     *t_forget;
 486
 487         /*
 488          * Doubly-linked circular list of all buffers still to be flushed before
 489          * this transaction can be checkpointed. [j_list_lock]
 490          */
 491         struct journal_head     *t_checkpoint_list;
 492
 493         /*
 494          * Doubly-linked circular list of all buffers submitted for IO while
 495          * checkpointing. [j_list_lock]
 496          */
 497         struct journal_head     *t_checkpoint_io_list;
 498
 499         /*
 500          * Doubly-linked circular list of temporary buffers currently undergoing
 501          * IO in the log [j_list_lock]
 502          */
 503         struct journal_head     *t_iobuf_list;
 504
 505         /*
 506          * Doubly-linked circular list of metadata buffers being shadowed by log
 507          * IO.  The IO buffers on the iobuf list and the shadow buffers on this
 508          * list match each other one for one at all times. [j_list_lock]
 509          */
 510         struct journal_head     *t_shadow_list;
 511
 512         /*
 513          * Doubly-linked circular list of control buffers being written to the
 514          * log. [j_list_lock]
 515          */
 516         struct journal_head     *t_log_list;
 517
 518         /*
 519          * Protects info related to handles
 520          */
 521         spinlock_t              t_handle_lock;
 522
 523         /*
 524          * Number of outstanding updates running on this transaction
 525          * [t_handle_lock]
 526          */
 527         int                     t_updates;
 528
 529         /*
 530          * Number of buffers reserved for use by all handles in this transaction
 531          * handle but not yet modified. [t_handle_lock]
 532          */
 533         int                     t_outstanding_credits;
 534
 535         /*
 536          * Forward and backward links for the circular list of all transactions
 537          * awaiting checkpoint. [j_list_lock]
 538          */
 539         transaction_t           *t_cpnext, *t_cpprev;
 540
 541         /*
 542          * When will the transaction expire (become due for commit), in jiffies?
 543          * [no locking]
 544          */
 545         unsigned long           t_expires;
 546
 547         /*
 548          * How many handles used this transaction? [t_handle_lock]
 549          */
 550         int t_handle_count;
 551
 552 };
 553
 554 /**
 555  * struct journal_s - this is the concrete type associated with journal_t.
 556  * @j_flags:  General journaling state flags
 557  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
 558  *     prior abort)?
 559  * @j_sb_buffer: First part of superblock buffer
 560  * @j_superblock: Second part of superblock buffer
 561  * @j_format_version: Version of the superblock format
 562  * @j_state_lock: Protect the various scalars in the journal
 563  * @j_barrier_count:  Number of processes waiting to create a barrier lock
 564  * @j_barrier: The barrier lock itself
 565  * @j_running_transaction: The current running transaction..
 566  * @j_committing_transaction: the transaction we are pushing to disk
 567  * @j_checkpoint_transactions: a linked circular list of all transactions
 568  *  waiting for checkpointing
 569  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
 570  *  to start committing, or for a barrier lock to be released
 571  * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
 572  * @j_wait_done_commit: Wait queue for waiting for commit to complete
 573  * @j_wait_checkpoint:  Wait queue to trigger checkpointing
 574  * @j_wait_commit: Wait queue to trigger commit
 575  * @j_wait_updates: Wait queue to wait for updates to complete
 576  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
 577  * @j_head: Journal head - identifies the first unused block in the journal
 578  * @j_tail: Journal tail - identifies the oldest still-used block in the
 579  *  journal.
 580  * @j_free: Journal free - how many free blocks are there in the journal?
 581  * @j_first: The block number of the first usable block
 582  * @j_last: The block number one beyond the last usable block
 583  * @j_dev: Device where we store the journal
 584  * @j_blocksize: blocksize for the location where we store the journal.
 585  * @j_blk_offset: starting block offset for into the device where we store the
 586  *     journal
 587  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
 588  *     be equal to j_dev
 589  * @j_maxlen: Total maximum capacity of the journal region on disk.
 590  * @j_list_lock: Protects the buffer lists and internal buffer state.
 591  * @j_inode: Optional inode where we store the journal.  If present, all journal
 592  *     block numbers are mapped into this inode via bmap().
 593  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
 594  * @j_transaction_sequence: Sequence number of the next transaction to grant
 595  * @j_commit_sequence: Sequence number of the most recently committed
 596  *  transaction
 597  * @j_commit_request: Sequence number of the most recent transaction wanting
 598  *     commit
 599  * @j_uuid: Uuid of client object.
 600  * @j_task: Pointer to the current commit thread for this journal
 601  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
 602  *     single compound commit transaction
 603  * @j_commit_interval: What is the maximum transaction lifetime before we begin
 604  *  a commit?
 605  * @j_commit_timer:  The timer used to wakeup the commit thread
 606  * @j_revoke_lock: Protect the revoke table
 607  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
 608  *     current transaction.
 609  * @j_revoke_table: alternate revoke tables for j_revoke
 610  * @j_wbuf: array of buffer_heads for journal_commit_transaction
 611  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
 612  *      number that will fit in j_blocksize
 613  * @j_last_sync_writer: most recent pid which did a synchronous write
 614  * @j_private: An opaque pointer to fs-private information.
 615  */
 616
 617 struct journal_s
 618 {
 619         /* General journaling state flags [j_state_lock] */
 620         unsigned long           j_flags;
 621
 622         /*
 623          * Is there an outstanding uncleared error on the journal (from a prior
 624          * abort)? [j_state_lock]
 625          */
 626         int                     j_errno;
 627
 628         /* The superblock buffer */
 629         struct buffer_head      *j_sb_buffer;
 630         journal_superblock_t    *j_superblock;
 631
 632         /* Version of the superblock format */
 633         int                     j_format_version;
 634
 635         /*
 636          * Protect the various scalars in the journal
 637          */
 638         spinlock_t              j_state_lock;
 639
 640         /*
 641          * Number of processes waiting to create a barrier lock [j_state_lock]
 642          */
 643         int                     j_barrier_count;
 644
 645         /* The barrier lock itself */
 646         struct mutex            j_barrier;
 647
 648         /*
 649          * Transactions: The current running transaction...
 650          * [j_state_lock] [caller holding open handle]
 651          */
 652         transaction_t           *j_running_transaction;
 653
 654         /*
 655          * the transaction we are pushing to disk
 656          * [j_state_lock] [caller holding open handle]
 657          */
 658         transaction_t           *j_committing_transaction;
 659
 660         /*
 661          * ... and a linked circular list of all transactions waiting for
 662          * checkpointing. [j_list_lock]
 663          */
 664         transaction_t           *j_checkpoint_transactions;
 665
 666         /*
 667          * Wait queue for waiting for a locked transaction to start committing,
 668          * or for a barrier lock to be released
 669          */
 670         wait_queue_head_t       j_wait_transaction_locked;
 671
 672         /* Wait queue for waiting for checkpointing to complete */
 673         wait_queue_head_t       j_wait_logspace;
 674
 675         /* Wait queue for waiting for commit to complete */
 676         wait_queue_head_t       j_wait_done_commit;
 677
 678         /* Wait queue to trigger checkpointing */
 679         wait_queue_head_t       j_wait_checkpoint;
 680
 681         /* Wait queue to trigger commit */
 682         wait_queue_head_t       j_wait_commit;
 683
 684         /* Wait queue to wait for updates to complete */
 685         wait_queue_head_t       j_wait_updates;
 686
 687         /* Semaphore for locking against concurrent checkpoints */
 688         struct mutex            j_checkpoint_mutex;
 689
 690         /*
 691          * Journal head: identifies the first unused block in the journal.
 692          * [j_state_lock]
 693          */
 694         unsigned long           j_head;
 695
 696         /*
 697          * Journal tail: identifies the oldest still-used block in the journal.
 698          * [j_state_lock]
 699          */
 700         unsigned long           j_tail;
 701
 702         /*
 703          * Journal free: how many free blocks are there in the journal?
 704          * [j_state_lock]
 705          */
 706         unsigned long           j_free;
 707
 708         /*
 709          * Journal start and end: the block numbers of the first usable block
 710          * and one beyond the last usable block in the journal. [j_state_lock]
 711          */
 712         unsigned long           j_first;
 713         unsigned long           j_last;
 714
 715         /*
 716          * Device, blocksize and starting block offset for the location where we
 717          * store the journal.
 718          */
 719         struct block_device     *j_dev;
 720         int                     j_blocksize;
 721         unsigned long           j_blk_offset;
 722
 723         /*
 724          * Device which holds the client fs.  For internal journal this will be
 725          * equal to j_dev.
 726          */
 727         struct block_device     *j_fs_dev;
 728
 729         /* Total maximum capacity of the journal region on disk. */
 730         unsigned int            j_maxlen;
 731
 732         /*
 733          * Protects the buffer lists and internal buffer state.
 734          */
 735         spinlock_t              j_list_lock;
 736
 737         /* Optional inode where we store the journal.  If present, all */
 738         /* journal block numbers are mapped into this inode via */
 739         /* bmap(). */
 740         struct inode            *j_inode;
 741
 742         /*
 743          * Sequence number of the oldest transaction in the log [j_state_lock]
 744          */
 745         tid_t                   j_tail_sequence;
 746
 747         /*
 748          * Sequence number of the next transaction to grant [j_state_lock]
 749          */
 750         tid_t                   j_transaction_sequence;
 751
 752         /*
 753          * Sequence number of the most recently committed transaction
 754          * [j_state_lock].
 755          */
 756         tid_t                   j_commit_sequence;
 757
 758         /*
 759          * Sequence number of the most recent transaction wanting commit
 760          * [j_state_lock]
 761          */
 762         tid_t                   j_commit_request;
 763
 764         /*
 765          * Journal uuid: identifies the object (filesystem, LVM volume etc)
 766          * backed by this journal.  This will eventually be replaced by an array
 767          * of uuids, allowing us to index multiple devices within a single
 768          * journal and to perform atomic updates across them.
 769          */
 770         __u8                    j_uuid[16];
 771
 772         /* Pointer to the current commit thread for this journal */
 773         struct task_struct      *j_task;
 774
 775         /*
 776          * Maximum number of metadata buffers to allow in a single compound
 777          * commit transaction
 778          */
 779         int                     j_max_transaction_buffers;
 780
 781         /*
 782          * What is the maximum transaction lifetime before we begin a commit?
 783          */
 784         unsigned long           j_commit_interval;
 785
 786         /* The timer used to wakeup the commit thread: */
 787         struct timer_list       j_commit_timer;
 788
 789         /*
 790          * The revoke table: maintains the list of revoked blocks in the
 791          * current transaction.  [j_revoke_lock]
 792          */
 793         spinlock_t              j_revoke_lock;
 794         struct jbd_revoke_table_s *j_revoke;
 795         struct jbd_revoke_table_s *j_revoke_table[2];
 796
 797         /*
 798          * array of bhs for journal_commit_transaction
 799          */
 800         struct buffer_head      **j_wbuf;
 801         int                     j_wbufsize;
 802
 803         pid_t                   j_last_sync_writer;
 804
 805         /*
 806          * An opaque pointer to fs-private information.  ext3 puts its
 807          * superblock pointer here
 808          */
 809         void *j_private;
 810 };
 811
 812 /*
 813  * Journal flag definitions
 814  */
 815 #define JFS_UNMOUNT     0x001   /* Journal thread is being destroyed */
 816 #define JFS_ABORT       0x002   /* Journaling has been aborted for errors. */
 817 #define JFS_ACK_ERR     0x004   /* The errno in the sb has been acked */
 818 #define JFS_FLUSHED     0x008   /* The journal superblock has been flushed */
 819 #define JFS_LOADED      0x010   /* The journal superblock has been loaded */
 820 #define JFS_BARRIER     0x020   /* Use IDE barriers */
 821
 822 /*
 823  * Function declarations for the journaling transaction and buffer
 824  * management
 825  */
 826
 827 /* Filing buffers */
 828 extern void journal_unfile_buffer(journal_t *, struct journal_head *);
 829 extern void __journal_unfile_buffer(struct journal_head *);
 830 extern void __journal_refile_buffer(struct journal_head *);
 831 extern void journal_refile_buffer(journal_t *, struct journal_head *);
 832 extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
 833 extern void __journal_free_buffer(struct journal_head *bh);
 834 extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
 835 extern void __journal_clean_data_list(transaction_t *transaction);
 836
 837 /* Log buffer allocation */
 838 extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
 839 int journal_next_log_block(journal_t *, unsigned long *);
 840
 841 /* Commit management */
 842 extern void journal_commit_transaction(journal_t *);
 843
 844 /* Checkpoint list management */
 845 int __journal_clean_checkpoint_list(journal_t *journal);
 846 int __journal_remove_checkpoint(struct journal_head *);
 847 void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
 848
 849 /* Buffer IO */
 850 extern int
 851 journal_write_metadata_buffer(transaction_t       *transaction,
 852                               struct journal_head  *jh_in,
 853                               struct journal_head **jh_out,
 854                               unsigned long        blocknr);
 855
 856 /* Transaction locking */
 857 extern void             __wait_on_journal (journal_t *);
 858
 859 /*
 860  * Journal locking.
 861  *
 862  * We need to lock the journal during transaction state changes so that nobody
 863  * ever tries to take a handle on the running transaction while we are in the
 864  * middle of moving it to the commit phase.  j_state_lock does this.
 865  *
 866  * Note that the locking is completely interrupt unsafe.  We never touch
 867  * journal structures from interrupts.
 868  */
 869
 870 static inline handle_t *journal_current_handle(void)
 871 {
 872         return current->journal_info;
 873 }
 874
 875 /* The journaling code user interface:
 876  *
 877  * Create and destroy handles
 878  * Register buffer modifications against the current transaction.
 879  */
 880
 881 extern handle_t *journal_start(journal_t *, int nblocks);
 882 extern int       journal_restart (handle_t *, int nblocks);
 883 extern int       journal_extend (handle_t *, int nblocks);
 884 extern int       journal_get_write_access(handle_t *, struct buffer_head *);
 885 extern int       journal_get_create_access (handle_t *, struct buffer_head *);
 886 extern int       journal_get_undo_access(handle_t *, struct buffer_head *);
 887 extern int       journal_dirty_data (handle_t *, struct buffer_head *);
 888 extern int       journal_dirty_metadata (handle_t *, struct buffer_head *);
 889 extern void      journal_release_buffer (handle_t *, struct buffer_head *);
 890 extern int       journal_forget (handle_t *, struct buffer_head *);
 891 extern void      journal_sync_buffer (struct buffer_head *);
 892 extern void      journal_invalidatepage(journal_t *,
 893                                 struct page *, unsigned long);
 894 extern int       journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
 895 extern int       journal_stop(handle_t *);
 896 extern int       journal_flush (journal_t *);
 897 extern void      journal_lock_updates (journal_t *);
 898 extern void      journal_unlock_updates (journal_t *);
 899
 900 extern journal_t * journal_init_dev(struct block_device *bdev,
 901                                 struct block_device *fs_dev,
 902                                 int start, int len, int bsize);
 903 extern journal_t * journal_init_inode (struct inode *);
 904 extern int         journal_update_format (journal_t *);
 905 extern int         journal_check_used_features
 906                    (journal_t *, unsigned long, unsigned long, unsigned long);
 907 extern int         journal_check_available_features
 908                    (journal_t *, unsigned long, unsigned long, unsigned long);
 909 extern int         journal_set_features
 910                    (journal_t *, unsigned long, unsigned long, unsigned long);
 911 extern int         journal_create     (journal_t *);
 912 extern int         journal_load       (journal_t *journal);
 913 extern void        journal_destroy    (journal_t *);
 914 extern int         journal_recover    (journal_t *journal);
 915 extern int         journal_wipe       (journal_t *, int);
 916 extern int         journal_skip_recovery        (journal_t *);
 917 extern void        journal_update_superblock    (journal_t *, int);
 918 extern void        journal_abort      (journal_t *, int);
 919 extern int         journal_errno      (journal_t *);
 920 extern void        journal_ack_err    (journal_t *);
 921 extern int         journal_clear_err  (journal_t *);
 922 extern int         journal_bmap(journal_t *, unsigned long, unsigned long *);
 923 extern int         journal_force_commit(journal_t *);
 924
 925 /*
 926  * journal_head management
 927  */
 928 struct journal_head *journal_add_journal_head(struct buffer_head *bh);
 929 struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
 930 void journal_remove_journal_head(struct buffer_head *bh);
 931 void journal_put_journal_head(struct journal_head *jh);
 932
 933 /*
 934  * handle management
 935  */
 936 extern struct kmem_cache *jbd_handle_cache;
 937
 938 static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
 939 {
 940         return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
 941 }
 942
 943 static inline void jbd_free_handle(handle_t *handle)
 944 {
 945         kmem_cache_free(jbd_handle_cache, handle);
 946 }
 947
 948 /* Primary revoke support */
 949 #define JOURNAL_REVOKE_DEFAULT_HASH 256
 950 extern int         journal_init_revoke(journal_t *, int);
 951 extern void        journal_destroy_revoke_caches(void);
 952 extern int         journal_init_revoke_caches(void);
 953
 954 extern void        journal_destroy_revoke(journal_t *);
 955 extern int         journal_revoke (handle_t *,
 956                                 unsigned long, struct buffer_head *);
 957 extern int         journal_cancel_revoke(handle_t *, struct journal_head *);
 958 extern void        journal_write_revoke_records(journal_t *, transaction_t *);
 959
 960 /* Recovery revoke support */
 961 extern int      journal_set_revoke(journal_t *, unsigned long, tid_t);
 962 extern int      journal_test_revoke(journal_t *, unsigned long, tid_t);
 963 extern void     journal_clear_revoke(journal_t *);
 964 extern void     journal_switch_revoke_table(journal_t *journal);
 965
 966 /*
 967  * The log thread user interface:
 968  *
 969  * Request space in the current transaction, and force transaction commit
 970  * transitions on demand.
 971  */
 972
 973 int __log_space_left(journal_t *); /* Called with journal locked */
 974 int log_start_commit(journal_t *journal, tid_t tid);
 975 int __log_start_commit(journal_t *journal, tid_t tid);
 976 int journal_start_commit(journal_t *journal, tid_t *tid);
 977 int journal_force_commit_nested(journal_t *journal);
 978 int log_wait_commit(journal_t *journal, tid_t tid);
 979 int log_do_checkpoint(journal_t *journal);
 980
 981 void __log_wait_for_space(journal_t *journal);
 982 extern void     __journal_drop_transaction(journal_t *, transaction_t *);
 983 extern int      cleanup_journal_tail(journal_t *);
 984
 985 /* Debugging code only: */
 986
 987 #define jbd_ENOSYS() \
 988 do {                                                                       \
 989         printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
 990         current->state = TASK_UNINTERRUPTIBLE;                             \
 991         schedule();                                                        \
 992 } while (1)
 993
 994 /*
 995  * is_journal_abort
 996  *
 997  * Simple test wrapper function to test the JFS_ABORT state flag.  This
 998  * bit, when set, indicates that we have had a fatal error somewhere,
 999  * either inside the journaling layer or indicated to us by the client
1000  * (eg. ext3), and that we and should not commit any further
1001  * transactions.
1002  */
1003
1004 static inline int is_journal_aborted(journal_t *journal)
1005 {
1006         return journal->j_flags & JFS_ABORT;
1007 }
1008
1009 static inline int is_handle_aborted(handle_t *handle)
1010 {
1011         if (handle->h_aborted)
1012                 return 1;
1013         return is_journal_aborted(handle->h_transaction->t_journal);
1014 }
1015
1016 static inline void journal_abort_handle(handle_t *handle)
1017 {
1018         handle->h_aborted = 1;
1019 }
1020
1021 #endif /* __KERNEL__   */
1022
1023 /* Comparison functions for transaction IDs: perform comparisons using
1024  * modulo arithmetic so that they work over sequence number wraps. */
1025
1026 static inline int tid_gt(tid_t x, tid_t y)
1027 {
1028         int difference = (x - y);
1029         return (difference > 0);
1030 }
1031
1032 static inline int tid_geq(tid_t x, tid_t y)
1033 {
1034         int difference = (x - y);
1035         return (difference >= 0);
1036 }
1037
1038 extern int journal_blocks_per_page(struct inode *inode);
1039
1040 /*
1041  * Return the minimum number of blocks which must be free in the journal
1042  * before a new transaction may be started.  Must be called under j_state_lock.
1043  */
1044 static inline int jbd_space_needed(journal_t *journal)
1045 {
1046         int nblocks = journal->j_max_transaction_buffers;
1047         if (journal->j_committing_transaction)
1048                 nblocks += journal->j_committing_transaction->
1049                                         t_outstanding_credits;
1050         return nblocks;
1051 }
1052
1053 /*
1054  * Definitions which augment the buffer_head layer
1055  */
1056
1057 /* journaling buffer types */
1058 #define BJ_None         0       /* Not journaled */
1059 #define BJ_SyncData     1       /* Normal data: flush before commit */
1060 #define BJ_Metadata     2       /* Normal journaled metadata */
1061 #define BJ_Forget       3       /* Buffer superseded by this transaction */
1062 #define BJ_IO           4       /* Buffer is for temporary IO use */
1063 #define BJ_Shadow       5       /* Buffer contents being shadowed to the log */
1064 #define BJ_LogCtl       6       /* Buffer contains log descriptors */
1065 #define BJ_Reserved     7       /* Buffer is reserved for access by journal */
1066 #define BJ_Locked       8       /* Locked for I/O during commit */
1067 #define BJ_Types        9
1068
1069 extern int jbd_blocks_per_page(struct inode *inode);
1070
1071 #ifdef __KERNEL__
1072
1073 #define buffer_trace_init(bh)   do {} while (0)
1074 #define print_buffer_fields(bh) do {} while (0)
1075 #define print_buffer_trace(bh)  do {} while (0)
1076 #define BUFFER_TRACE(bh, info)  do {} while (0)
1077 #define BUFFER_TRACE2(bh, bh2, info)    do {} while (0)
1078 #define JBUFFER_TRACE(jh, info) do {} while (0)
1079
1080 #endif  /* __KERNEL__ */
1081
1082 #endif  /* _LINUX_JBD_H */