include/linux/jbd2.h

   1 /*
   2  * linux/include/linux/jbd2.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_JBD2_H
  17 #define _LINUX_JBD2_H
  18
  19 /* Allow this file to be included directly into e2fsprogs */
  20 #ifndef __KERNEL__
  21 #include "jfs_compat.h"
  22 #define JBD2_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
  34 #include <asm/semaphore.h>
  35 #endif
  36
  37 #define journal_oom_retry 1
  38
  39 /*
  40  * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
  41  * certain classes of error which can occur due to failed IOs.  Under
  42  * normal use we want ext4 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 JBD2_PARANOID_IOFAIL
  47
  48 /*
  49  * The default maximum commit age, in seconds.
  50  */
  51 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
  52
  53 #ifdef CONFIG_JBD2_DEBUG
  54 /*
  55  * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
  56  * consistency checks.  By default we don't do this unless
  57  * CONFIG_JBD2_DEBUG is on.
  58  */
  59 #define JBD2_EXPENSIVE_CHECKING
  60 extern u8 jbd2_journal_enable_debug;
  61
  62 #define jbd_debug(n, f, a...)                                           \
  63         do {                                                            \
  64                 if ((n) <= jbd2_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 *jbd2_alloc(size_t size, gfp_t flags)
  75 {
  76         return (void *)__get_free_pages(flags, get_order(size));
  77 }
  78
  79 static inline void jbd2_free(void *ptr, size_t size)
  80 {
  81         free_pages((unsigned long)ptr, get_order(size));
  82 };
  83
  84 #define JBD2_MIN_JOURNAL_BLOCKS 1024
  85
  86 #ifdef __KERNEL__
  87
  88 /**
  89  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
  90  *
  91  * All filesystem modifications made by the process go
  92  * through this handle.  Recursive operations (such as quota operations)
  93  * are gathered into a single update.
  94  *
  95  * The buffer credits field is used to account for journaled buffers
  96  * being modified by the running process.  To ensure that there is
  97  * enough log space for all outstanding operations, we need to limit the
  98  * number of outstanding buffers possible at any time.  When the
  99  * operation completes, any buffer credits not used are credited back to
 100  * the transaction, so that at all times we know how many buffers the
 101  * outstanding updates on a transaction might possibly touch.
 102  *
 103  * This is an opaque datatype.
 104  **/
 105 typedef struct handle_s         handle_t;       /* Atomic operation type */
 106
 107
 108 /**
 109  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
 110  *
 111  * journal_t is linked to from the fs superblock structure.
 112  *
 113  * We use the journal_t to keep track of all outstanding transaction
 114  * activity on the filesystem, and to manage the state of the log
 115  * writing process.
 116  *
 117  * This is an opaque datatype.
 118  **/
 119 typedef struct journal_s        journal_t;      /* Journal control structure */
 120 #endif
 121
 122 /*
 123  * Internal structures used by the logging mechanism:
 124  */
 125
 126 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
 127
 128 /*
 129  * On-disk structures
 130  */
 131
 132 /*
 133  * Descriptor block types:
 134  */
 135
 136 #define JBD2_DESCRIPTOR_BLOCK   1
 137 #define JBD2_COMMIT_BLOCK       2
 138 #define JBD2_SUPERBLOCK_V1      3
 139 #define JBD2_SUPERBLOCK_V2      4
 140 #define JBD2_REVOKE_BLOCK       5
 141
 142 /*
 143  * Standard header for all descriptor blocks:
 144  */
 145 typedef struct journal_header_s
 146 {
 147         __be32          h_magic;
 148         __be32          h_blocktype;
 149         __be32          h_sequence;
 150 } journal_header_t;
 151
 152
 153 /*
 154  * The block tag: used to describe a single buffer in the journal.
 155  * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
 156  * raw struct shouldn't be used for pointer math or sizeof() - use
 157  * journal_tag_bytes(journal) instead to compute this.
 158  */
 159 typedef struct journal_block_tag_s
 160 {
 161         __be32          t_blocknr;      /* The on-disk block number */
 162         __be32          t_flags;        /* See below */
 163         __be32          t_blocknr_high; /* most-significant high 32bits. */
 164 } journal_block_tag_t;
 165
 166 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
 167 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
 168
 169 /*
 170  * The revoke descriptor: used on disk to describe a series of blocks to
 171  * be revoked from the log
 172  */
 173 typedef struct jbd2_journal_revoke_header_s
 174 {
 175         journal_header_t r_header;
 176         __be32           r_count;       /* Count of bytes used in the block */
 177 } jbd2_journal_revoke_header_t;
 178
 179
 180 /* Definitions for the journal tag flags word: */
 181 #define JBD2_FLAG_ESCAPE                1       /* on-disk block is escaped */
 182 #define JBD2_FLAG_SAME_UUID     2       /* block has same uuid as previous */
 183 #define JBD2_FLAG_DELETED       4       /* block deleted by this transaction */
 184 #define JBD2_FLAG_LAST_TAG      8       /* last tag in this descriptor block */
 185
 186
 187 /*
 188  * The journal superblock.  All fields are in big-endian byte order.
 189  */
 190 typedef struct journal_superblock_s
 191 {
 192 /* 0x0000 */
 193         journal_header_t s_header;
 194
 195 /* 0x000C */
 196         /* Static information describing the journal */
 197         __be32  s_blocksize;            /* journal device blocksize */
 198         __be32  s_maxlen;               /* total blocks in journal file */
 199         __be32  s_first;                /* first block of log information */
 200
 201 /* 0x0018 */
 202         /* Dynamic information describing the current state of the log */
 203         __be32  s_sequence;             /* first commit ID expected in log */
 204         __be32  s_start;                /* blocknr of start of log */
 205
 206 /* 0x0020 */
 207         /* Error value, as set by jbd2_journal_abort(). */
 208         __be32  s_errno;
 209
 210 /* 0x0024 */
 211         /* Remaining fields are only valid in a version-2 superblock */
 212         __be32  s_feature_compat;       /* compatible feature set */
 213         __be32  s_feature_incompat;     /* incompatible feature set */
 214         __be32  s_feature_ro_compat;    /* readonly-compatible feature set */
 215 /* 0x0030 */
 216         __u8    s_uuid[16];             /* 128-bit uuid for journal */
 217
 218 /* 0x0040 */
 219         __be32  s_nr_users;             /* Nr of filesystems sharing log */
 220
 221         __be32  s_dynsuper;             /* Blocknr of dynamic superblock copy*/
 222
 223 /* 0x0048 */
 224         __be32  s_max_transaction;      /* Limit of journal blocks per trans.*/
 225         __be32  s_max_trans_data;       /* Limit of data blocks per trans. */
 226
 227 /* 0x0050 */
 228         __u32   s_padding[44];
 229
 230 /* 0x0100 */
 231         __u8    s_users[16*48];         /* ids of all fs'es sharing the log */
 232 /* 0x0400 */
 233 } journal_superblock_t;
 234
 235 #define JBD2_HAS_COMPAT_FEATURE(j,mask)                                 \
 236         ((j)->j_format_version >= 2 &&                                  \
 237          ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
 238 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask)                              \
 239         ((j)->j_format_version >= 2 &&                                  \
 240          ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
 241 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask)                               \
 242         ((j)->j_format_version >= 2 &&                                  \
 243          ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
 244
 245 #define JBD2_FEATURE_INCOMPAT_REVOKE    0x00000001
 246 #define JBD2_FEATURE_INCOMPAT_64BIT     0x00000002
 247
 248 /* Features known to this kernel version: */
 249 #define JBD2_KNOWN_COMPAT_FEATURES      0
 250 #define JBD2_KNOWN_ROCOMPAT_FEATURES    0
 251 #define JBD2_KNOWN_INCOMPAT_FEATURES    (JBD2_FEATURE_INCOMPAT_REVOKE | \
 252                                          JBD2_FEATURE_INCOMPAT_64BIT)
 253
 254 #ifdef __KERNEL__
 255
 256 #include <linux/fs.h>
 257 #include <linux/sched.h>
 258
 259 #define JBD2_ASSERTIONS
 260 #ifdef JBD2_ASSERTIONS
 261 #define J_ASSERT(assert)                                                \
 262 do {                                                                    \
 263         if (!(assert)) {                                                \
 264                 printk (KERN_EMERG                                      \
 265                         "Assertion failure in %s() at %s:%d: \"%s\"\n", \
 266                         __FUNCTION__, __FILE__, __LINE__, # assert);    \
 267                 BUG();                                                  \
 268         }                                                               \
 269 } while (0)
 270
 271 #if defined(CONFIG_BUFFER_DEBUG)
 272 void buffer_assertion_failure(struct buffer_head *bh);
 273 #define J_ASSERT_BH(bh, expr)                                           \
 274         do {                                                            \
 275                 if (!(expr))                                            \
 276                         buffer_assertion_failure(bh);                   \
 277                 J_ASSERT(expr);                                         \
 278         } while (0)
 279 #define J_ASSERT_JH(jh, expr)   J_ASSERT_BH(jh2bh(jh), expr)
 280 #else
 281 #define J_ASSERT_BH(bh, expr)   J_ASSERT(expr)
 282 #define J_ASSERT_JH(jh, expr)   J_ASSERT(expr)
 283 #endif
 284
 285 #else
 286 #define J_ASSERT(assert)        do { } while (0)
 287 #endif          /* JBD2_ASSERTIONS */
 288
 289 #if defined(JBD2_PARANOID_IOFAIL)
 290 #define J_EXPECT(expr, why...)          J_ASSERT(expr)
 291 #define J_EXPECT_BH(bh, expr, why...)   J_ASSERT_BH(bh, expr)
 292 #define J_EXPECT_JH(jh, expr, why...)   J_ASSERT_JH(jh, expr)
 293 #else
 294 #define __journal_expect(expr, why...)                                       \
 295         ({                                                                   \
 296                 int val = (expr);                                            \
 297                 if (!val) {                                                  \
 298                         printk(KERN_ERR                                      \
 299                                 "EXT3-fs unexpected failure: %s;\n",# expr); \
 300                         printk(KERN_ERR why "\n");                           \
 301                 }                                                            \
 302                 val;                                                         \
 303         })
 304 #define J_EXPECT(expr, why...)          __journal_expect(expr, ## why)
 305 #define J_EXPECT_BH(bh, expr, why...)   __journal_expect(expr, ## why)
 306 #define J_EXPECT_JH(jh, expr, why...)   __journal_expect(expr, ## why)
 307 #endif
 308
 309 enum jbd_state_bits {
 310         BH_JBD                  /* Has an attached ext3 journal_head */
 311           = BH_PrivateStart,
 312         BH_JWrite,              /* Being written to log (@@@ DEBUGGING) */
 313         BH_Freed,               /* Has been freed (truncated) */
 314         BH_Revoked,             /* Has been revoked from the log */
 315         BH_RevokeValid,         /* Revoked flag is valid */
 316         BH_JBDDirty,            /* Is dirty but journaled */
 317         BH_State,               /* Pins most journal_head state */
 318         BH_JournalHead,         /* Pins bh->b_private and jh->b_bh */
 319         BH_Unshadow,            /* Dummy bit, for BJ_Shadow wakeup filtering */
 320 };
 321
 322 BUFFER_FNS(JBD, jbd)
 323 BUFFER_FNS(JWrite, jwrite)
 324 BUFFER_FNS(JBDDirty, jbddirty)
 325 TAS_BUFFER_FNS(JBDDirty, jbddirty)
 326 BUFFER_FNS(Revoked, revoked)
 327 TAS_BUFFER_FNS(Revoked, revoked)
 328 BUFFER_FNS(RevokeValid, revokevalid)
 329 TAS_BUFFER_FNS(RevokeValid, revokevalid)
 330 BUFFER_FNS(Freed, freed)
 331
 332 static inline struct buffer_head *jh2bh(struct journal_head *jh)
 333 {
 334         return jh->b_bh;
 335 }
 336
 337 static inline struct journal_head *bh2jh(struct buffer_head *bh)
 338 {
 339         return bh->b_private;
 340 }
 341
 342 static inline void jbd_lock_bh_state(struct buffer_head *bh)
 343 {
 344         bit_spin_lock(BH_State, &bh->b_state);
 345 }
 346
 347 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
 348 {
 349         return bit_spin_trylock(BH_State, &bh->b_state);
 350 }
 351
 352 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
 353 {
 354         return bit_spin_is_locked(BH_State, &bh->b_state);
 355 }
 356
 357 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
 358 {
 359         bit_spin_unlock(BH_State, &bh->b_state);
 360 }
 361
 362 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
 363 {
 364         bit_spin_lock(BH_JournalHead, &bh->b_state);
 365 }
 366
 367 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
 368 {
 369         bit_spin_unlock(BH_JournalHead, &bh->b_state);
 370 }
 371
 372 struct jbd2_revoke_table_s;
 373
 374 /**
 375  * struct handle_s - The handle_s type is the concrete type associated with
 376  *     handle_t.
 377  * @h_transaction: Which compound transaction is this update a part of?
 378  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
 379  * @h_ref: Reference count on this handle
 380  * @h_err: Field for caller's use to track errors through large fs operations
 381  * @h_sync: flag for sync-on-close
 382  * @h_jdata: flag to force data journaling
 383  * @h_aborted: flag indicating fatal error on handle
 384  **/
 385
 386 /* Docbook can't yet cope with the bit fields, but will leave the documentation
 387  * in so it can be fixed later.
 388  */
 389
 390 struct handle_s
 391 {
 392         /* Which compound transaction is this update a part of? */
 393         transaction_t           *h_transaction;
 394
 395         /* Number of remaining buffers we are allowed to dirty: */
 396         int                     h_buffer_credits;
 397
 398         /* Reference count on this handle */
 399         int                     h_ref;
 400
 401         /* Field for caller's use to track errors through large fs */
 402         /* operations */
 403         int                     h_err;
 404
 405         /* Flags [no locking] */
 406         unsigned int    h_sync:         1;      /* sync-on-close */
 407         unsigned int    h_jdata:        1;      /* force data journaling */
 408         unsigned int    h_aborted:      1;      /* fatal error on handle */
 409 };
 410
 411
 412 /* The transaction_t type is the guts of the journaling mechanism.  It
 413  * tracks a compound transaction through its various states:
 414  *
 415  * RUNNING:     accepting new updates
 416  * LOCKED:      Updates still running but we don't accept new ones
 417  * RUNDOWN:     Updates are tidying up but have finished requesting
 418  *              new buffers to modify (state not used for now)
 419  * FLUSH:       All updates complete, but we are still writing to disk
 420  * COMMIT:      All data on disk, writing commit record
 421  * FINISHED:    We still have to keep the transaction for checkpointing.
 422  *
 423  * The transaction keeps track of all of the buffers modified by a
 424  * running transaction, and all of the buffers committed but not yet
 425  * flushed to home for finished transactions.
 426  */
 427
 428 /*
 429  * Lock ranking:
 430  *
 431  *    j_list_lock
 432  *      ->jbd_lock_bh_journal_head()    (This is "innermost")
 433  *
 434  *    j_state_lock
 435  *    ->jbd_lock_bh_state()
 436  *
 437  *    jbd_lock_bh_state()
 438  *    ->j_list_lock
 439  *
 440  *    j_state_lock
 441  *    ->t_handle_lock
 442  *
 443  *    j_state_lock
 444  *    ->j_list_lock                     (journal_unmap_buffer)
 445  *
 446  */
 447
 448 struct transaction_s
 449 {
 450         /* Pointer to the journal for this transaction. [no locking] */
 451         journal_t               *t_journal;
 452
 453         /* Sequence number for this transaction [no locking] */
 454         tid_t                   t_tid;
 455
 456         /*
 457          * Transaction's current state
 458          * [no locking - only kjournald2 alters this]
 459          * FIXME: needs barriers
 460          * KLUDGE: [use j_state_lock]
 461          */
 462         enum {
 463                 T_RUNNING,
 464                 T_LOCKED,
 465                 T_RUNDOWN,
 466                 T_FLUSH,
 467                 T_COMMIT,
 468                 T_FINISHED
 469         }                       t_state;
 470
 471         /*
 472          * Where in the log does this transaction's commit start? [no locking]
 473          */
 474         unsigned long           t_log_start;
 475
 476         /* Number of buffers on the t_buffers list [j_list_lock] */
 477         int                     t_nr_buffers;
 478
 479         /*
 480          * Doubly-linked circular list of all buffers reserved but not yet
 481          * modified by this transaction [j_list_lock]
 482          */
 483         struct journal_head     *t_reserved_list;
 484
 485         /*
 486          * Doubly-linked circular list of all buffers under writeout during
 487          * commit [j_list_lock]
 488          */
 489         struct journal_head     *t_locked_list;
 490
 491         /*
 492          * Doubly-linked circular list of all metadata buffers owned by this
 493          * transaction [j_list_lock]
 494          */
 495         struct journal_head     *t_buffers;
 496
 497         /*
 498          * Doubly-linked circular list of all data buffers still to be
 499          * flushed before this transaction can be committed [j_list_lock]
 500          */
 501         struct journal_head     *t_sync_datalist;
 502
 503         /*
 504          * Doubly-linked circular list of all forget buffers (superseded
 505          * buffers which we can un-checkpoint once this transaction commits)
 506          * [j_list_lock]
 507          */
 508         struct journal_head     *t_forget;
 509
 510         /*
 511          * Doubly-linked circular list of all buffers still to be flushed before
 512          * this transaction can be checkpointed. [j_list_lock]
 513          */
 514         struct journal_head     *t_checkpoint_list;
 515
 516         /*
 517          * Doubly-linked circular list of all buffers submitted for IO while
 518          * checkpointing. [j_list_lock]
 519          */
 520         struct journal_head     *t_checkpoint_io_list;
 521
 522         /*
 523          * Doubly-linked circular list of temporary buffers currently undergoing
 524          * IO in the log [j_list_lock]
 525          */
 526         struct journal_head     *t_iobuf_list;
 527
 528         /*
 529          * Doubly-linked circular list of metadata buffers being shadowed by log
 530          * IO.  The IO buffers on the iobuf list and the shadow buffers on this
 531          * list match each other one for one at all times. [j_list_lock]
 532          */
 533         struct journal_head     *t_shadow_list;
 534
 535         /*
 536          * Doubly-linked circular list of control buffers being written to the
 537          * log. [j_list_lock]
 538          */
 539         struct journal_head     *t_log_list;
 540
 541         /*
 542          * Protects info related to handles
 543          */
 544         spinlock_t              t_handle_lock;
 545
 546         /*
 547          * Number of outstanding updates running on this transaction
 548          * [t_handle_lock]
 549          */
 550         int                     t_updates;
 551
 552         /*
 553          * Number of buffers reserved for use by all handles in this transaction
 554          * handle but not yet modified. [t_handle_lock]
 555          */
 556         int                     t_outstanding_credits;
 557
 558         /*
 559          * Forward and backward links for the circular list of all transactions
 560          * awaiting checkpoint. [j_list_lock]
 561          */
 562         transaction_t           *t_cpnext, *t_cpprev;
 563
 564         /*
 565          * When will the transaction expire (become due for commit), in jiffies?
 566          * [no locking]
 567          */
 568         unsigned long           t_expires;
 569
 570         /*
 571          * How many handles used this transaction? [t_handle_lock]
 572          */
 573         int t_handle_count;
 574
 575 };
 576
 577 /**
 578  * struct journal_s - The journal_s type is the concrete type associated with
 579  *     journal_t.
 580  * @j_flags:  General journaling state flags
 581  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
 582  *     prior abort)?
 583  * @j_sb_buffer: First part of superblock buffer
 584  * @j_superblock: Second part of superblock buffer
 585  * @j_format_version: Version of the superblock format
 586  * @j_state_lock: Protect the various scalars in the journal
 587  * @j_barrier_count:  Number of processes waiting to create a barrier lock
 588  * @j_barrier: The barrier lock itself
 589  * @j_running_transaction: The current running transaction..
 590  * @j_committing_transaction: the transaction we are pushing to disk
 591  * @j_checkpoint_transactions: a linked circular list of all transactions
 592  *  waiting for checkpointing
 593  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
 594  *  to start committing, or for a barrier lock to be released
 595  * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
 596  * @j_wait_done_commit: Wait queue for waiting for commit to complete
 597  * @j_wait_checkpoint:  Wait queue to trigger checkpointing
 598  * @j_wait_commit: Wait queue to trigger commit
 599  * @j_wait_updates: Wait queue to wait for updates to complete
 600  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
 601  * @j_head: Journal head - identifies the first unused block in the journal
 602  * @j_tail: Journal tail - identifies the oldest still-used block in the
 603  *  journal.
 604  * @j_free: Journal free - how many free blocks are there in the journal?
 605  * @j_first: The block number of the first usable block
 606  * @j_last: The block number one beyond the last usable block
 607  * @j_dev: Device where we store the journal
 608  * @j_blocksize: blocksize for the location where we store the journal.
 609  * @j_blk_offset: starting block offset for into the device where we store the
 610  *     journal
 611  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
 612  *     be equal to j_dev
 613  * @j_maxlen: Total maximum capacity of the journal region on disk.
 614  * @j_list_lock: Protects the buffer lists and internal buffer state.
 615  * @j_inode: Optional inode where we store the journal.  If present, all journal
 616  *     block numbers are mapped into this inode via bmap().
 617  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
 618  * @j_transaction_sequence: Sequence number of the next transaction to grant
 619  * @j_commit_sequence: Sequence number of the most recently committed
 620  *  transaction
 621  * @j_commit_request: Sequence number of the most recent transaction wanting
 622  *     commit
 623  * @j_uuid: Uuid of client object.
 624  * @j_task: Pointer to the current commit thread for this journal
 625  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
 626  *     single compound commit transaction
 627  * @j_commit_interval: What is the maximum transaction lifetime before we begin
 628  *  a commit?
 629  * @j_commit_timer:  The timer used to wakeup the commit thread
 630  * @j_revoke_lock: Protect the revoke table
 631  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
 632  *     current transaction.
 633  * @j_revoke_table: alternate revoke tables for j_revoke
 634  * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
 635  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
 636  *      number that will fit in j_blocksize
 637  * @j_last_sync_writer: most recent pid which did a synchronous write
 638  * @j_private: An opaque pointer to fs-private information.
 639  */
 640
 641 struct journal_s
 642 {
 643         /* General journaling state flags [j_state_lock] */
 644         unsigned long           j_flags;
 645
 646         /*
 647          * Is there an outstanding uncleared error on the journal (from a prior
 648          * abort)? [j_state_lock]
 649          */
 650         int                     j_errno;
 651
 652         /* The superblock buffer */
 653         struct buffer_head      *j_sb_buffer;
 654         journal_superblock_t    *j_superblock;
 655
 656         /* Version of the superblock format */
 657         int                     j_format_version;
 658
 659         /*
 660          * Protect the various scalars in the journal
 661          */
 662         spinlock_t              j_state_lock;
 663
 664         /*
 665          * Number of processes waiting to create a barrier lock [j_state_lock]
 666          */
 667         int                     j_barrier_count;
 668
 669         /* The barrier lock itself */
 670         struct mutex            j_barrier;
 671
 672         /*
 673          * Transactions: The current running transaction...
 674          * [j_state_lock] [caller holding open handle]
 675          */
 676         transaction_t           *j_running_transaction;
 677
 678         /*
 679          * the transaction we are pushing to disk
 680          * [j_state_lock] [caller holding open handle]
 681          */
 682         transaction_t           *j_committing_transaction;
 683
 684         /*
 685          * ... and a linked circular list of all transactions waiting for
 686          * checkpointing. [j_list_lock]
 687          */
 688         transaction_t           *j_checkpoint_transactions;
 689
 690         /*
 691          * Wait queue for waiting for a locked transaction to start committing,
 692          * or for a barrier lock to be released
 693          */
 694         wait_queue_head_t       j_wait_transaction_locked;
 695
 696         /* Wait queue for waiting for checkpointing to complete */
 697         wait_queue_head_t       j_wait_logspace;
 698
 699         /* Wait queue for waiting for commit to complete */
 700         wait_queue_head_t       j_wait_done_commit;
 701
 702         /* Wait queue to trigger checkpointing */
 703         wait_queue_head_t       j_wait_checkpoint;
 704
 705         /* Wait queue to trigger commit */
 706         wait_queue_head_t       j_wait_commit;
 707
 708         /* Wait queue to wait for updates to complete */
 709         wait_queue_head_t       j_wait_updates;
 710
 711         /* Semaphore for locking against concurrent checkpoints */
 712         struct mutex            j_checkpoint_mutex;
 713
 714         /*
 715          * Journal head: identifies the first unused block in the journal.
 716          * [j_state_lock]
 717          */
 718         unsigned long           j_head;
 719
 720         /*
 721          * Journal tail: identifies the oldest still-used block in the journal.
 722          * [j_state_lock]
 723          */
 724         unsigned long           j_tail;
 725
 726         /*
 727          * Journal free: how many free blocks are there in the journal?
 728          * [j_state_lock]
 729          */
 730         unsigned long           j_free;
 731
 732         /*
 733          * Journal start and end: the block numbers of the first usable block
 734          * and one beyond the last usable block in the journal. [j_state_lock]
 735          */
 736         unsigned long           j_first;
 737         unsigned long           j_last;
 738
 739         /*
 740          * Device, blocksize and starting block offset for the location where we
 741          * store the journal.
 742          */
 743         struct block_device     *j_dev;
 744         int                     j_blocksize;
 745         unsigned long long              j_blk_offset;
 746
 747         /*
 748          * Device which holds the client fs.  For internal journal this will be
 749          * equal to j_dev.
 750          */
 751         struct block_device     *j_fs_dev;
 752
 753         /* Total maximum capacity of the journal region on disk. */
 754         unsigned int            j_maxlen;
 755
 756         /*
 757          * Protects the buffer lists and internal buffer state.
 758          */
 759         spinlock_t              j_list_lock;
 760
 761         /* Optional inode where we store the journal.  If present, all */
 762         /* journal block numbers are mapped into this inode via */
 763         /* bmap(). */
 764         struct inode            *j_inode;
 765
 766         /*
 767          * Sequence number of the oldest transaction in the log [j_state_lock]
 768          */
 769         tid_t                   j_tail_sequence;
 770
 771         /*
 772          * Sequence number of the next transaction to grant [j_state_lock]
 773          */
 774         tid_t                   j_transaction_sequence;
 775
 776         /*
 777          * Sequence number of the most recently committed transaction
 778          * [j_state_lock].
 779          */
 780         tid_t                   j_commit_sequence;
 781
 782         /*
 783          * Sequence number of the most recent transaction wanting commit
 784          * [j_state_lock]
 785          */
 786         tid_t                   j_commit_request;
 787
 788         /*
 789          * Journal uuid: identifies the object (filesystem, LVM volume etc)
 790          * backed by this journal.  This will eventually be replaced by an array
 791          * of uuids, allowing us to index multiple devices within a single
 792          * journal and to perform atomic updates across them.
 793          */
 794         __u8                    j_uuid[16];
 795
 796         /* Pointer to the current commit thread for this journal */
 797         struct task_struct      *j_task;
 798
 799         /*
 800          * Maximum number of metadata buffers to allow in a single compound
 801          * commit transaction
 802          */
 803         int                     j_max_transaction_buffers;
 804
 805         /*
 806          * What is the maximum transaction lifetime before we begin a commit?
 807          */
 808         unsigned long           j_commit_interval;
 809
 810         /* The timer used to wakeup the commit thread: */
 811         struct timer_list       j_commit_timer;
 812
 813         /*
 814          * The revoke table: maintains the list of revoked blocks in the
 815          * current transaction.  [j_revoke_lock]
 816          */
 817         spinlock_t              j_revoke_lock;
 818         struct jbd2_revoke_table_s *j_revoke;
 819         struct jbd2_revoke_table_s *j_revoke_table[2];
 820
 821         /*
 822          * array of bhs for jbd2_journal_commit_transaction
 823          */
 824         struct buffer_head      **j_wbuf;
 825         int                     j_wbufsize;
 826
 827         pid_t                   j_last_sync_writer;
 828
 829         /*
 830          * An opaque pointer to fs-private information.  ext3 puts its
 831          * superblock pointer here
 832          */
 833         void *j_private;
 834 };
 835
 836 /*
 837  * Journal flag definitions
 838  */
 839 #define JBD2_UNMOUNT    0x001   /* Journal thread is being destroyed */
 840 #define JBD2_ABORT      0x002   /* Journaling has been aborted for errors. */
 841 #define JBD2_ACK_ERR    0x004   /* The errno in the sb has been acked */
 842 #define JBD2_FLUSHED    0x008   /* The journal superblock has been flushed */
 843 #define JBD2_LOADED     0x010   /* The journal superblock has been loaded */
 844 #define JBD2_BARRIER    0x020   /* Use IDE barriers */
 845
 846 /*
 847  * Function declarations for the journaling transaction and buffer
 848  * management
 849  */
 850
 851 /* Filing buffers */
 852 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
 853 extern void __jbd2_journal_unfile_buffer(struct journal_head *);
 854 extern void __jbd2_journal_refile_buffer(struct journal_head *);
 855 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
 856 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
 857 extern void __journal_free_buffer(struct journal_head *bh);
 858 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
 859 extern void __journal_clean_data_list(transaction_t *transaction);
 860
 861 /* Log buffer allocation */
 862 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
 863 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
 864
 865 /* Commit management */
 866 extern void jbd2_journal_commit_transaction(journal_t *);
 867
 868 /* Checkpoint list management */
 869 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
 870 int __jbd2_journal_remove_checkpoint(struct journal_head *);
 871 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
 872
 873 /* Buffer IO */
 874 extern int
 875 jbd2_journal_write_metadata_buffer(transaction_t          *transaction,
 876                               struct journal_head  *jh_in,
 877                               struct journal_head **jh_out,
 878                               unsigned long long   blocknr);
 879
 880 /* Transaction locking */
 881 extern void             __wait_on_journal (journal_t *);
 882
 883 /*
 884  * Journal locking.
 885  *
 886  * We need to lock the journal during transaction state changes so that nobody
 887  * ever tries to take a handle on the running transaction while we are in the
 888  * middle of moving it to the commit phase.  j_state_lock does this.
 889  *
 890  * Note that the locking is completely interrupt unsafe.  We never touch
 891  * journal structures from interrupts.
 892  */
 893
 894 static inline handle_t *journal_current_handle(void)
 895 {
 896         return current->journal_info;
 897 }
 898
 899 /* The journaling code user interface:
 900  *
 901  * Create and destroy handles
 902  * Register buffer modifications against the current transaction.
 903  */
 904
 905 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
 906 extern int       jbd2_journal_restart (handle_t *, int nblocks);
 907 extern int       jbd2_journal_extend (handle_t *, int nblocks);
 908 extern int       jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
 909 extern int       jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
 910 extern int       jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
 911 extern int       jbd2_journal_dirty_data (handle_t *, struct buffer_head *);
 912 extern int       jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
 913 extern void      jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
 914 extern int       jbd2_journal_forget (handle_t *, struct buffer_head *);
 915 extern void      journal_sync_buffer (struct buffer_head *);
 916 extern void      jbd2_journal_invalidatepage(journal_t *,
 917                                 struct page *, unsigned long);
 918 extern int       jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
 919 extern int       jbd2_journal_stop(handle_t *);
 920 extern int       jbd2_journal_flush (journal_t *);
 921 extern void      jbd2_journal_lock_updates (journal_t *);
 922 extern void      jbd2_journal_unlock_updates (journal_t *);
 923
 924 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
 925                                 struct block_device *fs_dev,
 926                                 unsigned long long start, int len, int bsize);
 927 extern journal_t * jbd2_journal_init_inode (struct inode *);
 928 extern int         jbd2_journal_update_format (journal_t *);
 929 extern int         jbd2_journal_check_used_features
 930                    (journal_t *, unsigned long, unsigned long, unsigned long);
 931 extern int         jbd2_journal_check_available_features
 932                    (journal_t *, unsigned long, unsigned long, unsigned long);
 933 extern int         jbd2_journal_set_features
 934                    (journal_t *, unsigned long, unsigned long, unsigned long);
 935 extern int         jbd2_journal_create     (journal_t *);
 936 extern int         jbd2_journal_load       (journal_t *journal);
 937 extern void        jbd2_journal_destroy    (journal_t *);
 938 extern int         jbd2_journal_recover    (journal_t *journal);
 939 extern int         jbd2_journal_wipe       (journal_t *, int);
 940 extern int         jbd2_journal_skip_recovery   (journal_t *);
 941 extern void        jbd2_journal_update_superblock       (journal_t *, int);
 942 extern void        __jbd2_journal_abort_hard    (journal_t *);
 943 extern void        jbd2_journal_abort      (journal_t *, int);
 944 extern int         jbd2_journal_errno      (journal_t *);
 945 extern void        jbd2_journal_ack_err    (journal_t *);
 946 extern int         jbd2_journal_clear_err  (journal_t *);
 947 extern int         jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
 948 extern int         jbd2_journal_force_commit(journal_t *);
 949
 950 /*
 951  * journal_head management
 952  */
 953 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
 954 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
 955 void jbd2_journal_remove_journal_head(struct buffer_head *bh);
 956 void jbd2_journal_put_journal_head(struct journal_head *jh);
 957
 958 /*
 959  * handle management
 960  */
 961 extern struct kmem_cache *jbd2_handle_cache;
 962
 963 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
 964 {
 965         return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
 966 }
 967
 968 static inline void jbd2_free_handle(handle_t *handle)
 969 {
 970         kmem_cache_free(jbd2_handle_cache, handle);
 971 }
 972
 973 /* Primary revoke support */
 974 #define JOURNAL_REVOKE_DEFAULT_HASH 256
 975 extern int         jbd2_journal_init_revoke(journal_t *, int);
 976 extern void        jbd2_journal_destroy_revoke_caches(void);
 977 extern int         jbd2_journal_init_revoke_caches(void);
 978
 979 extern void        jbd2_journal_destroy_revoke(journal_t *);
 980 extern int         jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
 981 extern int         jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
 982 extern void        jbd2_journal_write_revoke_records(journal_t *, transaction_t *);
 983
 984 /* Recovery revoke support */
 985 extern int      jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
 986 extern int      jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
 987 extern void     jbd2_journal_clear_revoke(journal_t *);
 988 extern void     jbd2_journal_switch_revoke_table(journal_t *journal);
 989
 990 /*
 991  * The log thread user interface:
 992  *
 993  * Request space in the current transaction, and force transaction commit
 994  * transitions on demand.
 995  */
 996
 997 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
 998 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
 999 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1000 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1001 int jbd2_journal_force_commit_nested(journal_t *journal);
1002 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1003 int jbd2_log_do_checkpoint(journal_t *journal);
1004
1005 void __jbd2_log_wait_for_space(journal_t *journal);
1006 extern void     __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1007 extern int      jbd2_cleanup_journal_tail(journal_t *);
1008
1009 /* Debugging code only: */
1010
1011 #define jbd_ENOSYS() \
1012 do {                                                                       \
1013         printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1014         current->state = TASK_UNINTERRUPTIBLE;                             \
1015         schedule();                                                        \
1016 } while (1)
1017
1018 /*
1019  * is_journal_abort
1020  *
1021  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1022  * bit, when set, indicates that we have had a fatal error somewhere,
1023  * either inside the journaling layer or indicated to us by the client
1024  * (eg. ext3), and that we and should not commit any further
1025  * transactions.
1026  */
1027
1028 static inline int is_journal_aborted(journal_t *journal)
1029 {
1030         return journal->j_flags & JBD2_ABORT;
1031 }
1032
1033 static inline int is_handle_aborted(handle_t *handle)
1034 {
1035         if (handle->h_aborted)
1036                 return 1;
1037         return is_journal_aborted(handle->h_transaction->t_journal);
1038 }
1039
1040 static inline void jbd2_journal_abort_handle(handle_t *handle)
1041 {
1042         handle->h_aborted = 1;
1043 }
1044
1045 #endif /* __KERNEL__   */
1046
1047 /* Comparison functions for transaction IDs: perform comparisons using
1048  * modulo arithmetic so that they work over sequence number wraps. */
1049
1050 static inline int tid_gt(tid_t x, tid_t y)
1051 {
1052         int difference = (x - y);
1053         return (difference > 0);
1054 }
1055
1056 static inline int tid_geq(tid_t x, tid_t y)
1057 {
1058         int difference = (x - y);
1059         return (difference >= 0);
1060 }
1061
1062 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1063 extern size_t journal_tag_bytes(journal_t *journal);
1064
1065 /*
1066  * Return the minimum number of blocks which must be free in the journal
1067  * before a new transaction may be started.  Must be called under j_state_lock.
1068  */
1069 static inline int jbd_space_needed(journal_t *journal)
1070 {
1071         int nblocks = journal->j_max_transaction_buffers;
1072         if (journal->j_committing_transaction)
1073                 nblocks += journal->j_committing_transaction->
1074                                         t_outstanding_credits;
1075         return nblocks;
1076 }
1077
1078 /*
1079  * Definitions which augment the buffer_head layer
1080  */
1081
1082 /* journaling buffer types */
1083 #define BJ_None         0       /* Not journaled */
1084 #define BJ_SyncData     1       /* Normal data: flush before commit */
1085 #define BJ_Metadata     2       /* Normal journaled metadata */
1086 #define BJ_Forget       3       /* Buffer superseded by this transaction */
1087 #define BJ_IO           4       /* Buffer is for temporary IO use */
1088 #define BJ_Shadow       5       /* Buffer contents being shadowed to the log */
1089 #define BJ_LogCtl       6       /* Buffer contains log descriptors */
1090 #define BJ_Reserved     7       /* Buffer is reserved for access by journal */
1091 #define BJ_Locked       8       /* Locked for I/O during commit */
1092 #define BJ_Types        9
1093
1094 extern int jbd_blocks_per_page(struct inode *inode);
1095
1096 #ifdef __KERNEL__
1097
1098 #define buffer_trace_init(bh)   do {} while (0)
1099 #define print_buffer_fields(bh) do {} while (0)
1100 #define print_buffer_trace(bh)  do {} while (0)
1101 #define BUFFER_TRACE(bh, info)  do {} while (0)
1102 #define BUFFER_TRACE2(bh, bh2, info)    do {} while (0)
1103 #define JBUFFER_TRACE(jh, info) do {} while (0)
1104
1105 #endif  /* __KERNEL__ */
1106
1107 #endif  /* _LINUX_JBD2_H */