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 #endif
  37
  38 #define journal_oom_retry 1
  39
  40 /*
  41  * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
  42  * certain classes of error which can occur due to failed IOs.  Under
  43  * normal use we want ext3 to continue after such errors, because
  44  * hardware _can_ fail, but for debugging purposes when running tests on
  45  * known-good hardware we may want to trap these errors.
  46  */
  47 #undef JBD_PARANOID_IOFAIL
  48
  49 /*
  50  * The default maximum commit age, in seconds.
  51  */
  52 #define JBD_DEFAULT_MAX_COMMIT_AGE 5
  53
  54 #ifdef CONFIG_JBD_DEBUG
  55 /*
  56  * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
  57  * consistency checks.  By default we don't do this unless
  58  * CONFIG_JBD_DEBUG is on.
  59  */
  60 #define JBD_EXPENSIVE_CHECKING
  61 extern int journal_enable_debug;
  62
  63 #define jbd_debug(n, f, a...)                                           \
  64         do {                                                            \
  65                 if ((n) <= journal_enable_debug) {                      \
  66                         printk (KERN_DEBUG "(%s, %d): %s: ",            \
  67                                 __FILE__, __LINE__, __FUNCTION__);      \
  68                         printk (f, ## a);                               \
  69                 }                                                       \
  70         } while (0)
  71 #else
  72 #define jbd_debug(f, a...)      /**/
  73 #endif
  74
  75 static inline void *jbd_alloc(size_t size, gfp_t flags)
  76 {
  77         return (void *)__get_free_pages(flags, get_order(size));
  78 }
  79
  80 static inline void jbd_free(void *ptr, size_t size)
  81 {
  82         free_pages((unsigned long)ptr, get_order(size));
  83 };
  84
  85 #define JFS_MIN_JOURNAL_BLOCKS 1024
  86
  87 #ifdef __KERNEL__
  88
  89 /**
  90  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
  91  *
  92  * All filesystem modifications made by the process go
  93  * through this handle.  Recursive operations (such as quota operations)
  94  * are gathered into a single update.
  95  *
  96  * The buffer credits field is used to account for journaled buffers
  97  * being modified by the running process.  To ensure that there is
  98  * enough log space for all outstanding operations, we need to limit the
  99  * number of outstanding buffers possible at any time.  When the
 100  * operation completes, any buffer credits not used are credited back to
 101  * the transaction, so that at all times we know how many buffers the
 102  * outstanding updates on a transaction might possibly touch.
 103  *
 104  * This is an opaque datatype.
 105  **/
 106 typedef struct handle_s         handle_t;       /* Atomic operation type */
 107
 108
 109 /**
 110  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
 111  *
 112  * journal_t is linked to from the fs superblock structure.
 113  *
 114  * We use the journal_t to keep track of all outstanding transaction
 115  * activity on the filesystem, and to manage the state of the log
 116  * writing process.
 117  *
 118  * This is an opaque datatype.
 119  **/
 120 typedef struct journal_s        journal_t;      /* Journal control structure */
 121 #endif
 122
 123 /*
 124  * Internal structures used by the logging mechanism:
 125  */
 126
 127 #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
 128
 129 /*
 130  * On-disk structures
 131  */
 132
 133 /*
 134  * Descriptor block types:
 135  */
 136
 137 #define JFS_DESCRIPTOR_BLOCK    1
 138 #define JFS_COMMIT_BLOCK        2
 139 #define JFS_SUPERBLOCK_V1       3
 140 #define JFS_SUPERBLOCK_V2       4
 141 #define JFS_REVOKE_BLOCK        5
 142
 143 /*
 144  * Standard header for all descriptor blocks:
 145  */
 146 typedef struct journal_header_s
 147 {
 148         __be32          h_magic;
 149         __be32          h_blocktype;
 150         __be32          h_sequence;
 151 } journal_header_t;
 152
 153
 154 /*
 155  * The block tag: used to describe a single buffer in the journal
 156  */
 157 typedef struct journal_block_tag_s
 158 {
 159         __be32          t_blocknr;      /* The on-disk block number */
 160         __be32          t_flags;        /* See below */
 161 } journal_block_tag_t;
 162
 163 /*
 164  * The revoke descriptor: used on disk to describe a series of blocks to
 165  * be revoked from the log
 166  */
 167 typedef struct journal_revoke_header_s
 168 {
 169         journal_header_t r_header;
 170         __be32           r_count;       /* Count of bytes used in the block */
 171 } journal_revoke_header_t;
 172
 173
 174 /* Definitions for the journal tag flags word: */
 175 #define JFS_FLAG_ESCAPE         1       /* on-disk block is escaped */
 176 #define JFS_FLAG_SAME_UUID      2       /* block has same uuid as previous */
 177 #define JFS_FLAG_DELETED        4       /* block deleted by this transaction */
 178 #define JFS_FLAG_LAST_TAG       8       /* last tag in this descriptor block */
 179
 180
 181 /*
 182  * The journal superblock.  All fields are in big-endian byte order.
 183  */
 184 typedef struct journal_superblock_s
 185 {
 186 /* 0x0000 */
 187         journal_header_t s_header;
 188
 189 /* 0x000C */
 190         /* Static information describing the journal */
 191         __be32  s_blocksize;            /* journal device blocksize */
 192         __be32  s_maxlen;               /* total blocks in journal file */
 193         __be32  s_first;                /* first block of log information */
 194
 195 /* 0x0018 */
 196         /* Dynamic information describing the current state of the log */
 197         __be32  s_sequence;             /* first commit ID expected in log */
 198         __be32  s_start;                /* blocknr of start of log */
 199
 200 /* 0x0020 */
 201         /* Error value, as set by journal_abort(). */
 202         __be32  s_errno;
 203
 204 /* 0x0024 */
 205         /* Remaining fields are only valid in a version-2 superblock */
 206         __be32  s_feature_compat;       /* compatible feature set */
 207         __be32  s_feature_incompat;     /* incompatible feature set */
 208         __be32  s_feature_ro_compat;    /* readonly-compatible feature set */
 209 /* 0x0030 */
 210         __u8    s_uuid[16];             /* 128-bit uuid for journal */
 211
 212 /* 0x0040 */
 213         __be32  s_nr_users;             /* Nr of filesystems sharing log */
 214
 215         __be32  s_dynsuper;             /* Blocknr of dynamic superblock copy*/
 216
 217 /* 0x0048 */
 218         __be32  s_max_transaction;      /* Limit of journal blocks per trans.*/
 219         __be32  s_max_trans_data;       /* Limit of data blocks per trans. */
 220
 221 /* 0x0050 */
 222         __u32   s_padding[44];
 223
 224 /* 0x0100 */
 225         __u8    s_users[16*48];         /* ids of all fs'es sharing the log */
 226 /* 0x0400 */
 227 } journal_superblock_t;
 228
 229 #define JFS_HAS_COMPAT_FEATURE(j,mask)                                  \
 230         ((j)->j_format_version >= 2 &&                                  \
 231          ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
 232 #define JFS_HAS_RO_COMPAT_FEATURE(j,mask)                               \
 233         ((j)->j_format_version >= 2 &&                                  \
 234          ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
 235 #define JFS_HAS_INCOMPAT_FEATURE(j,mask)                                \
 236         ((j)->j_format_version >= 2 &&                                  \
 237          ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
 238
 239 #define JFS_FEATURE_INCOMPAT_REVOKE     0x00000001
 240
 241 /* Features known to this kernel version: */
 242 #define JFS_KNOWN_COMPAT_FEATURES       0
 243 #define JFS_KNOWN_ROCOMPAT_FEATURES     0
 244 #define JFS_KNOWN_INCOMPAT_FEATURES     JFS_FEATURE_INCOMPAT_REVOKE
 245
 246 #ifdef __KERNEL__
 247
 248 #include <linux/fs.h>
 249 #include <linux/sched.h>
 250
 251 #define JBD_ASSERTIONS
 252 #ifdef JBD_ASSERTIONS
 253 #define J_ASSERT(assert)                                                \
 254 do {                                                                    \
 255         if (!(assert)) {                                                \
 256                 printk (KERN_EMERG                                      \
 257                         "Assertion failure in %s() at %s:%d: \"%s\"\n", \
 258                         __FUNCTION__, __FILE__, __LINE__, # assert);    \
 259                 BUG();                                                  \
 260         }                                                               \
 261 } while (0)
 262
 263 #if defined(CONFIG_BUFFER_DEBUG)
 264 void buffer_assertion_failure(struct buffer_head *bh);
 265 #define J_ASSERT_BH(bh, expr)                                           \
 266         do {                                                            \
 267                 if (!(expr))                                            \
 268                         buffer_assertion_failure(bh);                   \
 269                 J_ASSERT(expr);                                         \
 270         } while (0)
 271 #define J_ASSERT_JH(jh, expr)   J_ASSERT_BH(jh2bh(jh), expr)
 272 #else
 273 #define J_ASSERT_BH(bh, expr)   J_ASSERT(expr)
 274 #define J_ASSERT_JH(jh, expr)   J_ASSERT(expr)
 275 #endif
 276
 277 #else
 278 #define J_ASSERT(assert)        do { } while (0)
 279 #endif          /* JBD_ASSERTIONS */
 280
 281 #if defined(JBD_PARANOID_IOFAIL)
 282 #define J_EXPECT(expr, why...)          J_ASSERT(expr)
 283 #define J_EXPECT_BH(bh, expr, why...)   J_ASSERT_BH(bh, expr)
 284 #define J_EXPECT_JH(jh, expr, why...)   J_ASSERT_JH(jh, expr)
 285 #else
 286 #define __journal_expect(expr, why...)                                       \
 287         ({                                                                   \
 288                 int val = (expr);                                            \
 289                 if (!val) {                                                  \
 290                         printk(KERN_ERR                                      \
 291                                 "EXT3-fs unexpected failure: %s;\n",# expr); \
 292                         printk(KERN_ERR why "\n");                           \
 293                 }                                                            \
 294                 val;                                                         \
 295         })
 296 #define J_EXPECT(expr, why...)          __journal_expect(expr, ## why)
 297 #define J_EXPECT_BH(bh, expr, why...)   __journal_expect(expr, ## why)
 298 #define J_EXPECT_JH(jh, expr, why...)   __journal_expect(expr, ## why)
 299 #endif
 300
 301 enum jbd_state_bits {
 302         BH_JBD                  /* Has an attached ext3 journal_head */
 303           = BH_PrivateStart,
 304         BH_JWrite,              /* Being written to log (@@@ DEBUGGING) */
 305         BH_Freed,               /* Has been freed (truncated) */
 306         BH_Revoked,             /* Has been revoked from the log */
 307         BH_RevokeValid,         /* Revoked flag is valid */
 308         BH_JBDDirty,            /* Is dirty but journaled */
 309         BH_State,               /* Pins most journal_head state */
 310         BH_JournalHead,         /* Pins bh->b_private and jh->b_bh */
 311         BH_Unshadow,            /* Dummy bit, for BJ_Shadow wakeup filtering */
 312 };
 313
 314 BUFFER_FNS(JBD, jbd)
 315 BUFFER_FNS(JWrite, jwrite)
 316 BUFFER_FNS(JBDDirty, jbddirty)
 317 TAS_BUFFER_FNS(JBDDirty, jbddirty)
 318 BUFFER_FNS(Revoked, revoked)
 319 TAS_BUFFER_FNS(Revoked, revoked)
 320 BUFFER_FNS(RevokeValid, revokevalid)
 321 TAS_BUFFER_FNS(RevokeValid, revokevalid)
 322 BUFFER_FNS(Freed, freed)
 323
 324 static inline struct buffer_head *jh2bh(struct journal_head *jh)
 325 {
 326         return jh->b_bh;
 327 }
 328
 329 static inline struct journal_head *bh2jh(struct buffer_head *bh)
 330 {
 331         return bh->b_private;
 332 }
 333
 334 static inline void jbd_lock_bh_state(struct buffer_head *bh)
 335 {
 336         bit_spin_lock(BH_State, &bh->b_state);
 337 }
 338
 339 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
 340 {
 341         return bit_spin_trylock(BH_State, &bh->b_state);
 342 }
 343
 344 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
 345 {
 346         return bit_spin_is_locked(BH_State, &bh->b_state);
 347 }
 348
 349 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
 350 {
 351         bit_spin_unlock(BH_State, &bh->b_state);
 352 }
 353
 354 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
 355 {
 356         bit_spin_lock(BH_JournalHead, &bh->b_state);
 357 }
 358
 359 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
 360 {
 361         bit_spin_unlock(BH_JournalHead, &bh->b_state);
 362 }
 363
 364 struct jbd_revoke_table_s;
 365
 366 /**
 367  * struct handle_s - The handle_s type is the concrete type associated with
 368  *     handle_t.
 369  * @h_transaction: Which compound transaction is this update a part of?
 370  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
 371  * @h_ref: Reference count on this handle
 372  * @h_err: Field for caller's use to track errors through large fs operations
 373  * @h_sync: flag for sync-on-close
 374  * @h_jdata: flag to force data journaling
 375  * @h_aborted: flag indicating fatal error on handle
 376  * @h_lockdep_map: lockdep info for debugging lock problems
 377  **/
 378
 379 /* Docbook can't yet cope with the bit fields, but will leave the documentation
 380  * in so it can be fixed later.
 381  */
 382
 383 struct handle_s
 384 {
 385         /* Which compound transaction is this update a part of? */
 386         transaction_t           *h_transaction;
 387
 388         /* Number of remaining buffers we are allowed to dirty: */
 389         int                     h_buffer_credits;
 390
 391         /* Reference count on this handle */
 392         int                     h_ref;
 393
 394         /* Field for caller's use to track errors through large fs */
 395         /* operations */
 396         int                     h_err;
 397
 398         /* Flags [no locking] */
 399         unsigned int    h_sync:         1;      /* sync-on-close */
 400         unsigned int    h_jdata:        1;      /* force data journaling */
 401         unsigned int    h_aborted:      1;      /* fatal error on handle */
 402
 403 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 404         struct lockdep_map      h_lockdep_map;
 405 #endif
 406 };
 407
 408
 409 /* The transaction_t type is the guts of the journaling mechanism.  It
 410  * tracks a compound transaction through its various states:
 411  *
 412  * RUNNING:     accepting new updates
 413  * LOCKED:      Updates still running but we don't accept new ones
 414  * RUNDOWN:     Updates are tidying up but have finished requesting
 415  *              new buffers to modify (state not used for now)
 416  * FLUSH:       All updates complete, but we are still writing to disk
 417  * COMMIT:      All data on disk, writing commit record
 418  * FINISHED:    We still have to keep the transaction for checkpointing.
 419  *
 420  * The transaction keeps track of all of the buffers modified by a
 421  * running transaction, and all of the buffers committed but not yet
 422  * flushed to home for finished transactions.
 423  */
 424
 425 /*
 426  * Lock ranking:
 427  *
 428  *    j_list_lock
 429  *      ->jbd_lock_bh_journal_head()    (This is "innermost")
 430  *
 431  *    j_state_lock
 432  *    ->jbd_lock_bh_state()
 433  *
 434  *    jbd_lock_bh_state()
 435  *    ->j_list_lock
 436  *
 437  *    j_state_lock
 438  *    ->t_handle_lock
 439  *
 440  *    j_state_lock
 441  *    ->j_list_lock                     (journal_unmap_buffer)
 442  *
 443  */
 444
 445 struct transaction_s
 446 {
 447         /* Pointer to the journal for this transaction. [no locking] */
 448         journal_t               *t_journal;
 449
 450         /* Sequence number for this transaction [no locking] */
 451         tid_t                   t_tid;
 452
 453         /*
 454          * Transaction's current state
 455          * [no locking - only kjournald alters this]
 456          * FIXME: needs barriers
 457          * KLUDGE: [use j_state_lock]
 458          */
 459         enum {
 460                 T_RUNNING,
 461                 T_LOCKED,
 462                 T_RUNDOWN,
 463                 T_FLUSH,
 464                 T_COMMIT,
 465                 T_FINISHED
 466         }                       t_state;
 467
 468         /*
 469          * Where in the log does this transaction's commit start? [no locking]
 470          */
 471         unsigned long           t_log_start;
 472
 473         /* Number of buffers on the t_buffers list [j_list_lock] */
 474         int                     t_nr_buffers;
 475
 476         /*
 477          * Doubly-linked circular list of all buffers reserved but not yet
 478          * modified by this transaction [j_list_lock]
 479          */
 480         struct journal_head     *t_reserved_list;
 481
 482         /*
 483          * Doubly-linked circular list of all buffers under writeout during
 484          * commit [j_list_lock]
 485          */
 486         struct journal_head     *t_locked_list;
 487
 488         /*
 489          * Doubly-linked circular list of all metadata buffers owned by this
 490          * transaction [j_list_lock]
 491          */
 492         struct journal_head     *t_buffers;
 493
 494         /*
 495          * Doubly-linked circular list of all data buffers still to be
 496          * flushed before this transaction can be committed [j_list_lock]
 497          */
 498         struct journal_head     *t_sync_datalist;
 499
 500         /*
 501          * Doubly-linked circular list of all forget buffers (superseded
 502          * buffers which we can un-checkpoint once this transaction commits)
 503          * [j_list_lock]
 504          */
 505         struct journal_head     *t_forget;
 506
 507         /*
 508          * Doubly-linked circular list of all buffers still to be flushed before
 509          * this transaction can be checkpointed. [j_list_lock]
 510          */
 511         struct journal_head     *t_checkpoint_list;
 512
 513         /*
 514          * Doubly-linked circular list of all buffers submitted for IO while
 515          * checkpointing. [j_list_lock]
 516          */
 517         struct journal_head     *t_checkpoint_io_list;
 518
 519         /*
 520          * Doubly-linked circular list of temporary buffers currently undergoing
 521          * IO in the log [j_list_lock]
 522          */
 523         struct journal_head     *t_iobuf_list;
 524
 525         /*
 526          * Doubly-linked circular list of metadata buffers being shadowed by log
 527          * IO.  The IO buffers on the iobuf list and the shadow buffers on this
 528          * list match each other one for one at all times. [j_list_lock]
 529          */
 530         struct journal_head     *t_shadow_list;
 531
 532         /*
 533          * Doubly-linked circular list of control buffers being written to the
 534          * log. [j_list_lock]
 535          */
 536         struct journal_head     *t_log_list;
 537
 538         /*
 539          * Protects info related to handles
 540          */
 541         spinlock_t              t_handle_lock;
 542
 543         /*
 544          * Number of outstanding updates running on this transaction
 545          * [t_handle_lock]
 546          */
 547         int                     t_updates;
 548
 549         /*
 550          * Number of buffers reserved for use by all handles in this transaction
 551          * handle but not yet modified. [t_handle_lock]
 552          */
 553         int                     t_outstanding_credits;
 554
 555         /*
 556          * Forward and backward links for the circular list of all transactions
 557          * awaiting checkpoint. [j_list_lock]
 558          */
 559         transaction_t           *t_cpnext, *t_cpprev;
 560
 561         /*
 562          * When will the transaction expire (become due for commit), in jiffies?
 563          * [no locking]
 564          */
 565         unsigned long           t_expires;
 566
 567         /*
 568          * How many handles used this transaction? [t_handle_lock]
 569          */
 570         int t_handle_count;
 571
 572 };
 573
 574 /**
 575  * struct journal_s - The journal_s type is the concrete type associated with
 576  *     journal_t.
 577  * @j_flags:  General journaling state flags
 578  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
 579  *     prior abort)?
 580  * @j_sb_buffer: First part of superblock buffer
 581  * @j_superblock: Second part of superblock buffer
 582  * @j_format_version: Version of the superblock format
 583  * @j_state_lock: Protect the various scalars in the journal
 584  * @j_barrier_count:  Number of processes waiting to create a barrier lock
 585  * @j_barrier: The barrier lock itself
 586  * @j_running_transaction: The current running transaction..
 587  * @j_committing_transaction: the transaction we are pushing to disk
 588  * @j_checkpoint_transactions: a linked circular list of all transactions
 589  *  waiting for checkpointing
 590  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
 591  *  to start committing, or for a barrier lock to be released
 592  * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
 593  * @j_wait_done_commit: Wait queue for waiting for commit to complete
 594  * @j_wait_checkpoint:  Wait queue to trigger checkpointing
 595  * @j_wait_commit: Wait queue to trigger commit
 596  * @j_wait_updates: Wait queue to wait for updates to complete
 597  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
 598  * @j_head: Journal head - identifies the first unused block in the journal
 599  * @j_tail: Journal tail - identifies the oldest still-used block in the
 600  *  journal.
 601  * @j_free: Journal free - how many free blocks are there in the journal?
 602  * @j_first: The block number of the first usable block
 603  * @j_last: The block number one beyond the last usable block
 604  * @j_dev: Device where we store the journal
 605  * @j_blocksize: blocksize for the location where we store the journal.
 606  * @j_blk_offset: starting block offset for into the device where we store the
 607  *     journal
 608  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
 609  *     be equal to j_dev
 610  * @j_maxlen: Total maximum capacity of the journal region on disk.
 611  * @j_list_lock: Protects the buffer lists and internal buffer state.
 612  * @j_inode: Optional inode where we store the journal.  If present, all journal
 613  *     block numbers are mapped into this inode via bmap().
 614  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
 615  * @j_transaction_sequence: Sequence number of the next transaction to grant
 616  * @j_commit_sequence: Sequence number of the most recently committed
 617  *  transaction
 618  * @j_commit_request: Sequence number of the most recent transaction wanting
 619  *     commit
 620  * @j_uuid: Uuid of client object.
 621  * @j_task: Pointer to the current commit thread for this journal
 622  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
 623  *     single compound commit transaction
 624  * @j_commit_interval: What is the maximum transaction lifetime before we begin
 625  *  a commit?
 626  * @j_commit_timer:  The timer used to wakeup the commit thread
 627  * @j_revoke_lock: Protect the revoke table
 628  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
 629  *     current transaction.
 630  * @j_revoke_table: alternate revoke tables for j_revoke
 631  * @j_wbuf: array of buffer_heads for journal_commit_transaction
 632  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
 633  *      number that will fit in j_blocksize
 634  * @j_last_sync_writer: most recent pid which did a synchronous write
 635  * @j_private: An opaque pointer to fs-private information.
 636  */
 637
 638 struct journal_s
 639 {
 640         /* General journaling state flags [j_state_lock] */
 641         unsigned long           j_flags;
 642
 643         /*
 644          * Is there an outstanding uncleared error on the journal (from a prior
 645          * abort)? [j_state_lock]
 646          */
 647         int                     j_errno;
 648
 649         /* The superblock buffer */
 650         struct buffer_head      *j_sb_buffer;
 651         journal_superblock_t    *j_superblock;
 652
 653         /* Version of the superblock format */
 654         int                     j_format_version;
 655
 656         /*
 657          * Protect the various scalars in the journal
 658          */
 659         spinlock_t              j_state_lock;
 660
 661         /*
 662          * Number of processes waiting to create a barrier lock [j_state_lock]
 663          */
 664         int                     j_barrier_count;
 665
 666         /* The barrier lock itself */
 667         struct mutex            j_barrier;
 668
 669         /*
 670          * Transactions: The current running transaction...
 671          * [j_state_lock] [caller holding open handle]
 672          */
 673         transaction_t           *j_running_transaction;
 674
 675         /*
 676          * the transaction we are pushing to disk
 677          * [j_state_lock] [caller holding open handle]
 678          */
 679         transaction_t           *j_committing_transaction;
 680
 681         /*
 682          * ... and a linked circular list of all transactions waiting for
 683          * checkpointing. [j_list_lock]
 684          */
 685         transaction_t           *j_checkpoint_transactions;
 686
 687         /*
 688          * Wait queue for waiting for a locked transaction to start committing,
 689          * or for a barrier lock to be released
 690          */
 691         wait_queue_head_t       j_wait_transaction_locked;
 692
 693         /* Wait queue for waiting for checkpointing to complete */
 694         wait_queue_head_t       j_wait_logspace;
 695
 696         /* Wait queue for waiting for commit to complete */
 697         wait_queue_head_t       j_wait_done_commit;
 698
 699         /* Wait queue to trigger checkpointing */
 700         wait_queue_head_t       j_wait_checkpoint;
 701
 702         /* Wait queue to trigger commit */
 703         wait_queue_head_t       j_wait_commit;
 704
 705         /* Wait queue to wait for updates to complete */
 706         wait_queue_head_t       j_wait_updates;
 707
 708         /* Semaphore for locking against concurrent checkpoints */
 709         struct mutex            j_checkpoint_mutex;
 710
 711         /*
 712          * Journal head: identifies the first unused block in the journal.
 713          * [j_state_lock]
 714          */
 715         unsigned long           j_head;
 716
 717         /*
 718          * Journal tail: identifies the oldest still-used block in the journal.
 719          * [j_state_lock]
 720          */
 721         unsigned long           j_tail;
 722
 723         /*
 724          * Journal free: how many free blocks are there in the journal?
 725          * [j_state_lock]
 726          */
 727         unsigned long           j_free;
 728
 729         /*
 730          * Journal start and end: the block numbers of the first usable block
 731          * and one beyond the last usable block in the journal. [j_state_lock]
 732          */
 733         unsigned long           j_first;
 734         unsigned long           j_last;
 735
 736         /*
 737          * Device, blocksize and starting block offset for the location where we
 738          * store the journal.
 739          */
 740         struct block_device     *j_dev;
 741         int                     j_blocksize;
 742         unsigned long           j_blk_offset;
 743
 744         /*
 745          * Device which holds the client fs.  For internal journal this will be
 746          * equal to j_dev.
 747          */
 748         struct block_device     *j_fs_dev;
 749
 750         /* Total maximum capacity of the journal region on disk. */
 751         unsigned int            j_maxlen;
 752
 753         /*
 754          * Protects the buffer lists and internal buffer state.
 755          */
 756         spinlock_t              j_list_lock;
 757
 758         /* Optional inode where we store the journal.  If present, all */
 759         /* journal block numbers are mapped into this inode via */
 760         /* bmap(). */
 761         struct inode            *j_inode;
 762
 763         /*
 764          * Sequence number of the oldest transaction in the log [j_state_lock]
 765          */
 766         tid_t                   j_tail_sequence;
 767
 768         /*
 769          * Sequence number of the next transaction to grant [j_state_lock]
 770          */
 771         tid_t                   j_transaction_sequence;
 772
 773         /*
 774          * Sequence number of the most recently committed transaction
 775          * [j_state_lock].
 776          */
 777         tid_t                   j_commit_sequence;
 778
 779         /*
 780          * Sequence number of the most recent transaction wanting commit
 781          * [j_state_lock]
 782          */
 783         tid_t                   j_commit_request;
 784
 785         /*
 786          * Journal uuid: identifies the object (filesystem, LVM volume etc)
 787          * backed by this journal.  This will eventually be replaced by an array
 788          * of uuids, allowing us to index multiple devices within a single
 789          * journal and to perform atomic updates across them.
 790          */
 791         __u8                    j_uuid[16];
 792
 793         /* Pointer to the current commit thread for this journal */
 794         struct task_struct      *j_task;
 795
 796         /*
 797          * Maximum number of metadata buffers to allow in a single compound
 798          * commit transaction
 799          */
 800         int                     j_max_transaction_buffers;
 801
 802         /*
 803          * What is the maximum transaction lifetime before we begin a commit?
 804          */
 805         unsigned long           j_commit_interval;
 806
 807         /* The timer used to wakeup the commit thread: */
 808         struct timer_list       j_commit_timer;
 809
 810         /*
 811          * The revoke table: maintains the list of revoked blocks in the
 812          * current transaction.  [j_revoke_lock]
 813          */
 814         spinlock_t              j_revoke_lock;
 815         struct jbd_revoke_table_s *j_revoke;
 816         struct jbd_revoke_table_s *j_revoke_table[2];
 817
 818         /*
 819          * array of bhs for journal_commit_transaction
 820          */
 821         struct buffer_head      **j_wbuf;
 822         int                     j_wbufsize;
 823
 824         pid_t                   j_last_sync_writer;
 825
 826         /*
 827          * An opaque pointer to fs-private information.  ext3 puts its
 828          * superblock pointer here
 829          */
 830         void *j_private;
 831 };
 832
 833 /*
 834  * Journal flag definitions
 835  */
 836 #define JFS_UNMOUNT     0x001   /* Journal thread is being destroyed */
 837 #define JFS_ABORT       0x002   /* Journaling has been aborted for errors. */
 838 #define JFS_ACK_ERR     0x004   /* The errno in the sb has been acked */
 839 #define JFS_FLUSHED     0x008   /* The journal superblock has been flushed */
 840 #define JFS_LOADED      0x010   /* The journal superblock has been loaded */
 841 #define JFS_BARRIER     0x020   /* Use IDE barriers */
 842
 843 /*
 844  * Function declarations for the journaling transaction and buffer
 845  * management
 846  */
 847
 848 /* Filing buffers */
 849 extern void journal_unfile_buffer(journal_t *, struct journal_head *);
 850 extern void __journal_unfile_buffer(struct journal_head *);
 851 extern void __journal_refile_buffer(struct journal_head *);
 852 extern void journal_refile_buffer(journal_t *, struct journal_head *);
 853 extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
 854 extern void __journal_free_buffer(struct journal_head *bh);
 855 extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
 856 extern void __journal_clean_data_list(transaction_t *transaction);
 857
 858 /* Log buffer allocation */
 859 extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
 860 int journal_next_log_block(journal_t *, unsigned long *);
 861
 862 /* Commit management */
 863 extern void journal_commit_transaction(journal_t *);
 864
 865 /* Checkpoint list management */
 866 int __journal_clean_checkpoint_list(journal_t *journal);
 867 int __journal_remove_checkpoint(struct journal_head *);
 868 void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
 869
 870 /* Buffer IO */
 871 extern int
 872 journal_write_metadata_buffer(transaction_t       *transaction,
 873                               struct journal_head  *jh_in,
 874                               struct journal_head **jh_out,
 875                               unsigned long        blocknr);
 876
 877 /* Transaction locking */
 878 extern void             __wait_on_journal (journal_t *);
 879
 880 /*
 881  * Journal locking.
 882  *
 883  * We need to lock the journal during transaction state changes so that nobody
 884  * ever tries to take a handle on the running transaction while we are in the
 885  * middle of moving it to the commit phase.  j_state_lock does this.
 886  *
 887  * Note that the locking is completely interrupt unsafe.  We never touch
 888  * journal structures from interrupts.
 889  */
 890
 891 static inline handle_t *journal_current_handle(void)
 892 {
 893         return current->journal_info;
 894 }
 895
 896 /* The journaling code user interface:
 897  *
 898  * Create and destroy handles
 899  * Register buffer modifications against the current transaction.
 900  */
 901
 902 extern handle_t *journal_start(journal_t *, int nblocks);
 903 extern int       journal_restart (handle_t *, int nblocks);
 904 extern int       journal_extend (handle_t *, int nblocks);
 905 extern int       journal_get_write_access(handle_t *, struct buffer_head *);
 906 extern int       journal_get_create_access (handle_t *, struct buffer_head *);
 907 extern int       journal_get_undo_access(handle_t *, struct buffer_head *);
 908 extern int       journal_dirty_data (handle_t *, struct buffer_head *);
 909 extern int       journal_dirty_metadata (handle_t *, struct buffer_head *);
 910 extern void      journal_release_buffer (handle_t *, struct buffer_head *);
 911 extern int       journal_forget (handle_t *, struct buffer_head *);
 912 extern void      journal_sync_buffer (struct buffer_head *);
 913 extern void      journal_invalidatepage(journal_t *,
 914                                 struct page *, unsigned long);
 915 extern int       journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
 916 extern int       journal_stop(handle_t *);
 917 extern int       journal_flush (journal_t *);
 918 extern void      journal_lock_updates (journal_t *);
 919 extern void      journal_unlock_updates (journal_t *);
 920
 921 extern journal_t * journal_init_dev(struct block_device *bdev,
 922                                 struct block_device *fs_dev,
 923                                 int start, int len, int bsize);
 924 extern journal_t * journal_init_inode (struct inode *);
 925 extern int         journal_update_format (journal_t *);
 926 extern int         journal_check_used_features
 927                    (journal_t *, unsigned long, unsigned long, unsigned long);
 928 extern int         journal_check_available_features
 929                    (journal_t *, unsigned long, unsigned long, unsigned long);
 930 extern int         journal_set_features
 931                    (journal_t *, unsigned long, unsigned long, unsigned long);
 932 extern int         journal_create     (journal_t *);
 933 extern int         journal_load       (journal_t *journal);
 934 extern void        journal_destroy    (journal_t *);
 935 extern int         journal_recover    (journal_t *journal);
 936 extern int         journal_wipe       (journal_t *, int);
 937 extern int         journal_skip_recovery        (journal_t *);
 938 extern void        journal_update_superblock    (journal_t *, int);
 939 extern void        __journal_abort_hard (journal_t *);
 940 extern void        journal_abort      (journal_t *, int);
 941 extern int         journal_errno      (journal_t *);
 942 extern void        journal_ack_err    (journal_t *);
 943 extern int         journal_clear_err  (journal_t *);
 944 extern int         journal_bmap(journal_t *, unsigned long, unsigned long *);
 945 extern int         journal_force_commit(journal_t *);
 946
 947 /*
 948  * journal_head management
 949  */
 950 struct journal_head *journal_add_journal_head(struct buffer_head *bh);
 951 struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
 952 void journal_remove_journal_head(struct buffer_head *bh);
 953 void journal_put_journal_head(struct journal_head *jh);
 954
 955 /*
 956  * handle management
 957  */
 958 extern struct kmem_cache *jbd_handle_cache;
 959
 960 static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
 961 {
 962         return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
 963 }
 964
 965 static inline void jbd_free_handle(handle_t *handle)
 966 {
 967         kmem_cache_free(jbd_handle_cache, handle);
 968 }
 969
 970 /* Primary revoke support */
 971 #define JOURNAL_REVOKE_DEFAULT_HASH 256
 972 extern int         journal_init_revoke(journal_t *, int);
 973 extern void        journal_destroy_revoke_caches(void);
 974 extern int         journal_init_revoke_caches(void);
 975
 976 extern void        journal_destroy_revoke(journal_t *);
 977 extern int         journal_revoke (handle_t *,
 978                                 unsigned long, struct buffer_head *);
 979 extern int         journal_cancel_revoke(handle_t *, struct journal_head *);
 980 extern void        journal_write_revoke_records(journal_t *, transaction_t *);
 981
 982 /* Recovery revoke support */
 983 extern int      journal_set_revoke(journal_t *, unsigned long, tid_t);
 984 extern int      journal_test_revoke(journal_t *, unsigned long, tid_t);
 985 extern void     journal_clear_revoke(journal_t *);
 986 extern void     journal_switch_revoke_table(journal_t *journal);
 987
 988 /*
 989  * The log thread user interface:
 990  *
 991  * Request space in the current transaction, and force transaction commit
 992  * transitions on demand.
 993  */
 994
 995 int __log_space_left(journal_t *); /* Called with journal locked */
 996 int log_start_commit(journal_t *journal, tid_t tid);
 997 int __log_start_commit(journal_t *journal, tid_t tid);
 998 int journal_start_commit(journal_t *journal, tid_t *tid);
 999 int journal_force_commit_nested(journal_t *journal);
1000 int log_wait_commit(journal_t *journal, tid_t tid);
1001 int log_do_checkpoint(journal_t *journal);
1002
1003 void __log_wait_for_space(journal_t *journal);
1004 extern void     __journal_drop_transaction(journal_t *, transaction_t *);
1005 extern int      cleanup_journal_tail(journal_t *);
1006
1007 /* Debugging code only: */
1008
1009 #define jbd_ENOSYS() \
1010 do {                                                                       \
1011         printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1012         current->state = TASK_UNINTERRUPTIBLE;                             \
1013         schedule();                                                        \
1014 } while (1)
1015
1016 /*
1017  * is_journal_abort
1018  *
1019  * Simple test wrapper function to test the JFS_ABORT state flag.  This
1020  * bit, when set, indicates that we have had a fatal error somewhere,
1021  * either inside the journaling layer or indicated to us by the client
1022  * (eg. ext3), and that we and should not commit any further
1023  * transactions.
1024  */
1025
1026 static inline int is_journal_aborted(journal_t *journal)
1027 {
1028         return journal->j_flags & JFS_ABORT;
1029 }
1030
1031 static inline int is_handle_aborted(handle_t *handle)
1032 {
1033         if (handle->h_aborted)
1034                 return 1;
1035         return is_journal_aborted(handle->h_transaction->t_journal);
1036 }
1037
1038 static inline void journal_abort_handle(handle_t *handle)
1039 {
1040         handle->h_aborted = 1;
1041 }
1042
1043 #endif /* __KERNEL__   */
1044
1045 /* Comparison functions for transaction IDs: perform comparisons using
1046  * modulo arithmetic so that they work over sequence number wraps. */
1047
1048 static inline int tid_gt(tid_t x, tid_t y)
1049 {
1050         int difference = (x - y);
1051         return (difference > 0);
1052 }
1053
1054 static inline int tid_geq(tid_t x, tid_t y)
1055 {
1056         int difference = (x - y);
1057         return (difference >= 0);
1058 }
1059
1060 extern int journal_blocks_per_page(struct inode *inode);
1061
1062 /*
1063  * Return the minimum number of blocks which must be free in the journal
1064  * before a new transaction may be started.  Must be called under j_state_lock.
1065  */
1066 static inline int jbd_space_needed(journal_t *journal)
1067 {
1068         int nblocks = journal->j_max_transaction_buffers;
1069         if (journal->j_committing_transaction)
1070                 nblocks += journal->j_committing_transaction->
1071                                         t_outstanding_credits;
1072         return nblocks;
1073 }
1074
1075 /*
1076  * Definitions which augment the buffer_head layer
1077  */
1078
1079 /* journaling buffer types */
1080 #define BJ_None         0       /* Not journaled */
1081 #define BJ_SyncData     1       /* Normal data: flush before commit */
1082 #define BJ_Metadata     2       /* Normal journaled metadata */
1083 #define BJ_Forget       3       /* Buffer superseded by this transaction */
1084 #define BJ_IO           4       /* Buffer is for temporary IO use */
1085 #define BJ_Shadow       5       /* Buffer contents being shadowed to the log */
1086 #define BJ_LogCtl       6       /* Buffer contains log descriptors */
1087 #define BJ_Reserved     7       /* Buffer is reserved for access by journal */
1088 #define BJ_Locked       8       /* Locked for I/O during commit */
1089 #define BJ_Types        9
1090
1091 extern int jbd_blocks_per_page(struct inode *inode);
1092
1093 #ifdef __KERNEL__
1094
1095 #define buffer_trace_init(bh)   do {} while (0)
1096 #define print_buffer_fields(bh) do {} while (0)
1097 #define print_buffer_trace(bh)  do {} while (0)
1098 #define BUFFER_TRACE(bh, info)  do {} while (0)
1099 #define BUFFER_TRACE2(bh, bh2, info)    do {} while (0)
1100 #define JBUFFER_TRACE(jh, info) do {} while (0)
1101
1102 #endif  /* __KERNEL__ */
1103
1104 #endif  /* _LINUX_JBD_H */