include/xfs_log_priv.h

   1 /*
   2  * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or modify it
   5  * under the terms of version 2 of the GNU General Public License as
   6  * published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it would be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  11  *
  12  * Further, this software is distributed without any warranty that it is
  13  * free of the rightful claim of any third person regarding infringement
  14  * or the like.  Any license provided herein, whether implied or
  15  * otherwise, applies only to this software file.  Patent licenses, if
  16  * any, provided herein do not apply to combinations of this program with
  17  * other software, or any other product whatsoever.
  18  *
  19  * You should have received a copy of the GNU General Public License along
  20  * with this program; if not, write the Free Software Foundation, Inc., 59
  21  * Temple Place - Suite 330, Boston MA 02111-1307, USA.
  22  *
  23  * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
  24  * Mountain View, CA  94043, or:
  25  *
  26  * http://www.sgi.com
  27  *
  28  * For further information regarding this notice, see:
  29  *
  30  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
  31  */
  32 #ifndef __XFS_LOG_PRIV_H__
  33 #define __XFS_LOG_PRIV_H__
  34
  35 #if defined(XFS_ALL_TRACE)
  36 #define XFS_LOG_TRACE
  37 #endif
  38
  39 #if !defined(DEBUG)
  40 #undef XFS_LOG_TRACE
  41 #endif
  42
  43 struct xfs_buf;
  44 struct ktrace;
  45 struct log;
  46 struct xfs_buf_cancel;
  47 struct xfs_mount;
  48
  49 /*
  50  * Macros, structures, prototypes for internal log manager use.
  51  */
  52
  53 #define XLOG_MIN_ICLOGS         2
  54 #define XLOG_MED_ICLOGS         4
  55 #define XLOG_MAX_ICLOGS         8
  56 #define XLOG_CALLBACK_SIZE      10
  57 #define XLOG_HEADER_MAGIC_NUM   0xFEEDbabe      /* Invalid cycle number */
  58 #define XLOG_VERSION_1          1
  59 #define XLOG_VERSION_2          2               /* Large IClogs, Log sunit */
  60 #define XLOG_VERSION_OKBITS     (XLOG_VERSION_1 | XLOG_VERSION_2)
  61 #define XLOG_RECORD_BSIZE       (16*1024)       /* eventually 32k */
  62 #define XLOG_BIG_RECORD_BSIZE   (32*1024)       /* 32k buffers */
  63 #define XLOG_MAX_RECORD_BSIZE   (256*1024)
  64 #define XLOG_HEADER_CYCLE_SIZE  (32*1024)       /* cycle data in header */
  65 #define XLOG_RECORD_BSHIFT      14              /* 16384 == 1 << 14 */
  66 #define XLOG_BIG_RECORD_BSHIFT  15              /* 32k == 1 << 15 */
  67 #define XLOG_MAX_RECORD_BSHIFT  18              /* 256k == 1 << 18 */
  68 #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_BTOLRBB)
  69 int xlog_btolrbb(int b);
  70 #define XLOG_BTOLRBB(b)         xlog_btolrbb(b)
  71 #else
  72 #define XLOG_BTOLRBB(b)         (((b)+XLOG_RECORD_BSIZE-1) >> XLOG_RECORD_BSHIFT)
  73 #endif
  74
  75 #define XLOG_HEADER_SIZE        512
  76
  77 #define XLOG_REC_SHIFT(log) \
  78         BTOBB(1 << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
  79          XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
  80 #define XLOG_TOTAL_REC_SHIFT(log) \
  81         BTOBB(XLOG_MAX_ICLOGS << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
  82          XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
  83
  84 /*
  85  *  set lsns
  86  */
  87
  88 #define ASSIGN_LSN_CYCLE(lsn,cycle,arch) \
  89     INT_SET(((uint *)&(lsn))[LSN_FIELD_CYCLE(arch)], arch, (cycle));
  90 #define ASSIGN_LSN_BLOCK(lsn,block,arch) \
  91     INT_SET(((uint *)&(lsn))[LSN_FIELD_BLOCK(arch)], arch, (block));
  92 #define ASSIGN_ANY_LSN(lsn,cycle,block,arch)  \
  93     { \
  94         ASSIGN_LSN_CYCLE(lsn,cycle,arch); \
  95         ASSIGN_LSN_BLOCK(lsn,block,arch); \
  96     }
  97 #define ASSIGN_LSN(lsn,log,arch) \
  98     ASSIGN_ANY_LSN(lsn,(log)->l_curr_cycle,(log)->l_curr_block,arch);
  99
 100 #define XLOG_SET(f,b)           (((f) & (b)) == (b))
 101
 102 #define GET_CYCLE(ptr, arch) \
 103     (INT_GET(*(uint *)(ptr), arch) == XLOG_HEADER_MAGIC_NUM ? \
 104          INT_GET(*((uint *)(ptr)+1), arch) : \
 105          INT_GET(*(uint *)(ptr), arch) \
 106     )
 107
 108 #define BLK_AVG(blk1, blk2)     ((blk1+blk2) >> 1)
 109
 110
 111 #ifdef __KERNEL__
 112 /*
 113  * get client id from packed copy.
 114  *
 115  * this hack is here because the xlog_pack code copies four bytes
 116  * of xlog_op_header containing the fields oh_clientid, oh_flags
 117  * and oh_res2 into the packed copy.
 118  *
 119  * later on this four byte chunk is treated as an int and the
 120  * client id is pulled out.
 121  *
 122  * this has endian issues, of course.
 123  */
 124
 125 #if __BYTE_ORDER == __LITTLE_ENDIAN
 126 #define GET_CLIENT_ID(i,arch) \
 127     ((i) & 0xff)
 128 #else
 129 #define GET_CLIENT_ID(i,arch) \
 130     ((i) >> 24)
 131 #endif
 132
 133 #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_GRANT_SUB_SPACE)
 134 void xlog_grant_sub_space(struct log *log, int bytes, int type);
 135 #define XLOG_GRANT_SUB_SPACE(log,bytes,type)    \
 136         xlog_grant_sub_space(log,bytes,type)
 137 #else
 138 #define XLOG_GRANT_SUB_SPACE(log,bytes,type)                            \
 139     {                                                                   \
 140         if (type == 'w') {                                              \
 141                 (log)->l_grant_write_bytes -= (bytes);                  \
 142                 if ((log)->l_grant_write_bytes < 0) {                   \
 143                         (log)->l_grant_write_bytes += (log)->l_logsize; \
 144                         (log)->l_grant_write_cycle--;                   \
 145                 }                                                       \
 146         } else {                                                        \
 147                 (log)->l_grant_reserve_bytes -= (bytes);                \
 148                 if ((log)->l_grant_reserve_bytes < 0) {                 \
 149                         (log)->l_grant_reserve_bytes += (log)->l_logsize;\
 150                         (log)->l_grant_reserve_cycle--;                 \
 151                 }                                                       \
 152          }                                                              \
 153     }
 154 #endif
 155 #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_GRANT_ADD_SPACE)
 156 void xlog_grant_add_space(struct log *log, int bytes, int type);
 157 #define XLOG_GRANT_ADD_SPACE(log,bytes,type)    \
 158         xlog_grant_add_space(log,bytes,type)
 159 #else
 160 #define XLOG_GRANT_ADD_SPACE(log,bytes,type)                            \
 161     {                                                                   \
 162         if (type == 'w') {                                              \
 163                 (log)->l_grant_write_bytes += (bytes);                  \
 164                 if ((log)->l_grant_write_bytes > (log)->l_logsize) {    \
 165                         (log)->l_grant_write_bytes -= (log)->l_logsize; \
 166                         (log)->l_grant_write_cycle++;                   \
 167                 }                                                       \
 168         } else {                                                        \
 169                 (log)->l_grant_reserve_bytes += (bytes);                \
 170                 if ((log)->l_grant_reserve_bytes > (log)->l_logsize) {  \
 171                         (log)->l_grant_reserve_bytes -= (log)->l_logsize;\
 172                         (log)->l_grant_reserve_cycle++;                 \
 173                 }                                                       \
 174          }                                                              \
 175     }
 176 #endif
 177 #define XLOG_INS_TICKETQ(q,tic)                         \
 178     {                                                   \
 179         if (q) {                                        \
 180                 (tic)->t_next       = (q);              \
 181                 (tic)->t_prev       = (q)->t_prev;      \
 182                 (q)->t_prev->t_next = (tic);            \
 183                 (q)->t_prev         = (tic);            \
 184         } else {                                        \
 185                 (tic)->t_prev = (tic)->t_next = (tic);  \
 186                 (q) = (tic);                            \
 187         }                                               \
 188         (tic)->t_flags |= XLOG_TIC_IN_Q;                \
 189     }
 190 #define XLOG_DEL_TICKETQ(q,tic)                         \
 191     {                                                   \
 192         if ((tic) == (tic)->t_next) {                   \
 193                 (q) = NULL;                             \
 194         } else {                                        \
 195                 (q) = (tic)->t_next;                    \
 196                 (tic)->t_next->t_prev = (tic)->t_prev;  \
 197                 (tic)->t_prev->t_next = (tic)->t_next;  \
 198         }                                               \
 199         (tic)->t_next = (tic)->t_prev = NULL;           \
 200         (tic)->t_flags &= ~XLOG_TIC_IN_Q;               \
 201     }
 202
 203
 204 #define GRANT_LOCK(log)         mutex_spinlock(&(log)->l_grant_lock)
 205 #define GRANT_UNLOCK(log, s)    mutex_spinunlock(&(log)->l_grant_lock, s)
 206 #define LOG_LOCK(log)           mutex_spinlock(&(log)->l_icloglock)
 207 #define LOG_UNLOCK(log, s)      mutex_spinunlock(&(log)->l_icloglock, s)
 208
 209 #define xlog_panic(args...)     cmn_err(CE_PANIC, ## args)
 210 #define xlog_exit(args...)      cmn_err(CE_PANIC, ## args)
 211 #define xlog_warn(args...)      cmn_err(CE_WARN, ## args)
 212
 213 /*
 214  * In core log state
 215  */
 216 #define XLOG_STATE_ACTIVE    0x0001 /* Current IC log being written to */
 217 #define XLOG_STATE_WANT_SYNC 0x0002 /* Want to sync this iclog; no more writes */
 218 #define XLOG_STATE_SYNCING   0x0004 /* This IC log is syncing */
 219 #define XLOG_STATE_DONE_SYNC 0x0008 /* Done syncing to disk */
 220 #define XLOG_STATE_DO_CALLBACK \
 221                              0x0010 /* Process callback functions */
 222 #define XLOG_STATE_CALLBACK  0x0020 /* Callback functions now */
 223 #define XLOG_STATE_DIRTY     0x0040 /* Dirty IC log, not ready for ACTIVE status*/
 224 #define XLOG_STATE_IOERROR   0x0080 /* IO error happened in sync'ing log */
 225 #define XLOG_STATE_ALL       0x7FFF /* All possible valid flags */
 226 #define XLOG_STATE_NOTUSED   0x8000 /* This IC log not being used */
 227 #endif  /* __KERNEL__ */
 228
 229 /*
 230  * Flags to log operation header
 231  *
 232  * The first write of a new transaction will be preceded with a start
 233  * record, XLOG_START_TRANS.  Once a transaction is committed, a commit
 234  * record is written, XLOG_COMMIT_TRANS.  If a single region can not fit into
 235  * the remainder of the current active in-core log, it is split up into
 236  * multiple regions.  Each partial region will be marked with a
 237  * XLOG_CONTINUE_TRANS until the last one, which gets marked with XLOG_END_TRANS.
 238  *
 239  */
 240 #define XLOG_START_TRANS        0x01    /* Start a new transaction */
 241 #define XLOG_COMMIT_TRANS       0x02    /* Commit this transaction */
 242 #define XLOG_CONTINUE_TRANS     0x04    /* Cont this trans into new region */
 243 #define XLOG_WAS_CONT_TRANS     0x08    /* Cont this trans into new region */
 244 #define XLOG_END_TRANS          0x10    /* End a continued transaction */
 245 #define XLOG_UNMOUNT_TRANS      0x20    /* Unmount a filesystem transaction */
 246 #define XLOG_SKIP_TRANS         (XLOG_COMMIT_TRANS | XLOG_CONTINUE_TRANS | \
 247                                  XLOG_WAS_CONT_TRANS | XLOG_END_TRANS | \
 248                                  XLOG_UNMOUNT_TRANS)
 249
 250 #ifdef __KERNEL__
 251 /*
 252  * Flags to log ticket
 253  */
 254 #define XLOG_TIC_INITED         0x1     /* has been initialized */
 255 #define XLOG_TIC_PERM_RESERV    0x2     /* permanent reservation */
 256 #define XLOG_TIC_IN_Q           0x4
 257 #endif  /* __KERNEL__ */
 258
 259 #define XLOG_UNMOUNT_TYPE       0x556e  /* Un for Unmount */
 260
 261 /*
 262  * Flags for log structure
 263  */
 264 #define XLOG_CHKSUM_MISMATCH    0x1     /* used only during recovery */
 265 #define XLOG_ACTIVE_RECOVERY    0x2     /* in the middle of recovery */
 266 #define XLOG_RECOVERY_NEEDED    0x4     /* log was recovered */
 267 #define XLOG_IO_ERROR           0x8     /* log hit an I/O error, and being
 268                                            shutdown */
 269 typedef __uint32_t xlog_tid_t;
 270
 271
 272 #ifdef __KERNEL__
 273 /*
 274  * Below are states for covering allocation transactions.
 275  * By covering, we mean changing the h_tail_lsn in the last on-disk
 276  * log write such that no allocation transactions will be re-done during
 277  * recovery after a system crash. Recovery starts at the last on-disk
 278  * log write.
 279  *
 280  * These states are used to insert dummy log entries to cover
 281  * space allocation transactions which can undo non-transactional changes
 282  * after a crash. Writes to a file with space
 283  * already allocated do not result in any transactions. Allocations
 284  * might include space beyond the EOF. So if we just push the EOF a
 285  * little, the last transaction for the file could contain the wrong
 286  * size. If there is no file system activity, after an allocation
 287  * transaction, and the system crashes, the allocation transaction
 288  * will get replayed and the file will be truncated. This could
 289  * be hours/days/... after the allocation occurred.
 290  *
 291  * The fix for this is to do two dummy transactions when the
 292  * system is idle. We need two dummy transaction because the h_tail_lsn
 293  * in the log record header needs to point beyond the last possible
 294  * non-dummy transaction. The first dummy changes the h_tail_lsn to
 295  * the first transaction before the dummy. The second dummy causes
 296  * h_tail_lsn to point to the first dummy. Recovery starts at h_tail_lsn.
 297  *
 298  * These dummy transactions get committed when everything
 299  * is idle (after there has been some activity).
 300  *
 301  * There are 5 states used to control this.
 302  *
 303  *  IDLE -- no logging has been done on the file system or
 304  *              we are done covering previous transactions.
 305  *  NEED -- logging has occurred and we need a dummy transaction
 306  *              when the log becomes idle.
 307  *  DONE -- we were in the NEED state and have committed a dummy
 308  *              transaction.
 309  *  NEED2 -- we detected that a dummy transaction has gone to the
 310  *              on disk log with no other transactions.
 311  *  DONE2 -- we committed a dummy transaction when in the NEED2 state.
 312  *
 313  * There are two places where we switch states:
 314  *
 315  * 1.) In xfs_sync, when we detect an idle log and are in NEED or NEED2.
 316  *      We commit the dummy transaction and switch to DONE or DONE2,
 317  *      respectively. In all other states, we don't do anything.
 318  *
 319  * 2.) When we finish writing the on-disk log (xlog_state_clean_log).
 320  *
 321  *      No matter what state we are in, if this isn't the dummy
 322  *      transaction going out, the next state is NEED.
 323  *      So, if we aren't in the DONE or DONE2 states, the next state
 324  *      is NEED. We can't be finishing a write of the dummy record
 325  *      unless it was committed and the state switched to DONE or DONE2.
 326  *
 327  *      If we are in the DONE state and this was a write of the
 328  *              dummy transaction, we move to NEED2.
 329  *
 330  *      If we are in the DONE2 state and this was a write of the
 331  *              dummy transaction, we move to IDLE.
 332  *
 333  *
 334  * Writing only one dummy transaction can get appended to
 335  * one file space allocation. When this happens, the log recovery
 336  * code replays the space allocation and a file could be truncated.
 337  * This is why we have the NEED2 and DONE2 states before going idle.
 338  */
 339
 340 #define XLOG_STATE_COVER_IDLE   0
 341 #define XLOG_STATE_COVER_NEED   1
 342 #define XLOG_STATE_COVER_DONE   2
 343 #define XLOG_STATE_COVER_NEED2  3
 344 #define XLOG_STATE_COVER_DONE2  4
 345
 346 #define XLOG_COVER_OPS          5
 347
 348 typedef struct xlog_ticket {
 349         sv_t               t_sema;       /* sleep on this semaphore      :20 */
 350         struct xlog_ticket *t_next;      /*                              : 4 */
 351         struct xlog_ticket *t_prev;      /*                              : 4 */
 352         xlog_tid_t         t_tid;        /* transaction identifier       : 4 */
 353         int                t_curr_res;   /* current reservation in bytes : 4 */
 354         int                t_unit_res;   /* unit reservation in bytes    : 4 */
 355         __uint8_t          t_ocnt;       /* original count               : 1 */
 356         __uint8_t          t_cnt;        /* current count                : 1 */
 357         __uint8_t          t_clientid;   /* who does this belong to;     : 1 */
 358         __uint8_t          t_flags;      /* properties of reservation    : 1 */
 359 } xlog_ticket_t;
 360 #endif
 361
 362
 363 typedef struct xlog_op_header {
 364         xlog_tid_t oh_tid;      /* transaction id of operation  :  4 b */
 365         int        oh_len;      /* bytes in data region         :  4 b */
 366         __uint8_t  oh_clientid; /* who sent me this             :  1 b */
 367         __uint8_t  oh_flags;    /*                              :  1 b */
 368         ushort     oh_res2;     /* 32 bit align                 :  2 b */
 369 } xlog_op_header_t;
 370
 371
 372 /* valid values for h_fmt */
 373 #define XLOG_FMT_UNKNOWN  0
 374 #define XLOG_FMT_LINUX_LE 1
 375 #define XLOG_FMT_LINUX_BE 2
 376 #define XLOG_FMT_IRIX_BE  3
 377
 378 /* our fmt */
 379 #if __BYTE_ORDER == __LITTLE_ENDIAN
 380 #define XLOG_FMT XLOG_FMT_LINUX_LE
 381 #else
 382 #if __BYTE_ORDER == __BIG_ENDIAN
 383 #define XLOG_FMT XLOG_FMT_LINUX_BE
 384 #else
 385 #error unknown byte order
 386 #endif
 387 #endif
 388
 389 typedef struct xlog_rec_header {
 390         uint      h_magicno;    /* log record (LR) identifier           :  4 */
 391         uint      h_cycle;      /* write cycle of log                   :  4 */
 392         int       h_version;    /* LR version                           :  4 */
 393         int       h_len;        /* len in bytes; should be 64-bit aligned: 4 */
 394         xfs_lsn_t h_lsn;        /* lsn of this LR                       :  8 */
 395         xfs_lsn_t h_tail_lsn;   /* lsn of 1st LR w/ buffers not committed: 8 */
 396         uint      h_chksum;     /* may not be used; non-zero if used    :  4 */
 397         int       h_prev_block; /* block number to previous LR          :  4 */
 398         int       h_num_logops; /* number of log operations in this LR  :  4 */
 399         uint      h_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE];
 400         /* new fields */
 401         int       h_fmt;        /* format of log record                 :  4 */
 402         uuid_t    h_fs_uuid;    /* uuid of FS                           : 16 */
 403         int       h_size;       /* iclog size                           :  4 */
 404 } xlog_rec_header_t;
 405
 406 typedef struct xlog_rec_ext_header {
 407         uint      xh_cycle;     /* write cycle of log                   : 4 */
 408         uint      xh_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; /*    : 256 */
 409 } xlog_rec_ext_header_t;
 410
 411 #ifdef __KERNEL__
 412 /*
 413  * - A log record header is 512 bytes.  There is plenty of room to grow the
 414  *      xlog_rec_header_t into the reserved space.
 415  * - ic_data follows, so a write to disk can start at the beginning of
 416  *      the iclog.
 417  * - ic_forcesema is used to implement synchronous forcing of the iclog to disk.
 418  * - ic_next is the pointer to the next iclog in the ring.
 419  * - ic_bp is a pointer to the buffer used to write this incore log to disk.
 420  * - ic_log is a pointer back to the global log structure.
 421  * - ic_callback is a linked list of callback function/argument pairs to be
 422  *      called after an iclog finishes writing.
 423  * - ic_size is the full size of the header plus data.
 424  * - ic_offset is the current number of bytes written to in this iclog.
 425  * - ic_refcnt is bumped when someone is writing to the log.
 426  * - ic_state is the state of the iclog.
 427  */
 428 typedef struct xlog_iclog_fields {
 429         sv_t                    ic_forcesema;
 430         sv_t                    ic_writesema;
 431         struct xlog_in_core     *ic_next;
 432         struct xlog_in_core     *ic_prev;
 433         struct xfs_buf          *ic_bp;
 434         struct log              *ic_log;
 435         xfs_log_callback_t      *ic_callback;
 436         xfs_log_callback_t      **ic_callback_tail;
 437 #ifdef DEBUG
 438         struct ktrace           *ic_trace;
 439 #endif
 440         int                     ic_size;
 441         int                     ic_offset;
 442         int                     ic_refcnt;
 443         int                     ic_roundoff;
 444         int                     ic_bwritecnt;
 445         ushort_t                ic_state;
 446         char                    *ic_datap;      /* pointer to iclog data */
 447 } xlog_iclog_fields_t;
 448
 449 typedef union xlog_in_core2 {
 450         xlog_rec_header_t       hic_header;
 451         xlog_rec_ext_header_t   hic_xheader;
 452         char                    hic_sector[XLOG_HEADER_SIZE];
 453 } xlog_in_core_2_t;
 454
 455 typedef struct xlog_in_core {
 456         xlog_iclog_fields_t     hic_fields;
 457         xlog_in_core_2_t        *hic_data;
 458 } xlog_in_core_t;
 459
 460 /*
 461  * Defines to save our code from this glop.
 462  */
 463 #define ic_forcesema    hic_fields.ic_forcesema
 464 #define ic_writesema    hic_fields.ic_writesema
 465 #define ic_next         hic_fields.ic_next
 466 #define ic_prev         hic_fields.ic_prev
 467 #define ic_bp           hic_fields.ic_bp
 468 #define ic_log          hic_fields.ic_log
 469 #define ic_callback     hic_fields.ic_callback
 470 #define ic_callback_tail hic_fields.ic_callback_tail
 471 #define ic_trace        hic_fields.ic_trace
 472 #define ic_size         hic_fields.ic_size
 473 #define ic_offset       hic_fields.ic_offset
 474 #define ic_refcnt       hic_fields.ic_refcnt
 475 #define ic_roundoff     hic_fields.ic_roundoff
 476 #define ic_bwritecnt    hic_fields.ic_bwritecnt
 477 #define ic_state        hic_fields.ic_state
 478 #define ic_datap        hic_fields.ic_datap
 479 #define ic_header       hic_data->hic_header
 480
 481 /*
 482  * The reservation head lsn is not made up of a cycle number and block number.
 483  * Instead, it uses a cycle number and byte number.  Logs don't expect to
 484  * overflow 31 bits worth of byte offset, so using a byte number will mean
 485  * that round off problems won't occur when releasing partial reservations.
 486  */
 487 typedef struct log {
 488         /* The following block of fields are changed while holding icloglock */
 489         sema_t                  l_flushsema;    /* iclog flushing semaphore */
 490         int                     l_flushcnt;     /* # of procs waiting on this
 491                                                  * sema */
 492         int                     l_ticket_cnt;   /* free ticket count */
 493         int                     l_ticket_tcnt;  /* total ticket count */
 494         int                     l_covered_state;/* state of "covering disk
 495                                                  * log entries" */
 496         xlog_ticket_t           *l_freelist;    /* free list of tickets */
 497         xlog_ticket_t           *l_unmount_free;/* kmem_free these addresses */
 498         xlog_ticket_t           *l_tail;        /* free list of tickets */
 499         xlog_in_core_t          *l_iclog;       /* head log queue       */
 500         lock_t                  l_icloglock;    /* grab to change iclog state */
 501         xfs_lsn_t               l_tail_lsn;     /* lsn of 1st LR with unflushed
 502                                                  * buffers */
 503         xfs_lsn_t               l_last_sync_lsn;/* lsn of last LR on disk */
 504         struct xfs_mount        *l_mp;          /* mount point */
 505         struct xfs_buf          *l_xbuf;        /* extra buffer for log
 506                                                  * wrapping */
 507         dev_t                   l_dev;          /* dev_t of log */
 508         xfs_daddr_t             l_logBBstart;   /* start block of log */
 509         int                     l_logsize;      /* size of log in bytes */
 510         int                     l_logBBsize;    /* size of log in BB chunks */
 511         int                     l_roundoff;     /* round off error of iclogs */
 512         int                     l_curr_cycle;   /* Cycle number of log writes */
 513         int                     l_prev_cycle;   /* Cycle number before last
 514                                                  * block increment */
 515         int                     l_curr_block;   /* current logical log block */
 516         int                     l_prev_block;   /* previous logical log block */
 517         int                     l_iclog_size;   /* size of log in bytes */
 518         int                     l_iclog_size_log; /* log power size of log */
 519         int                     l_iclog_bufs;   /* number of iclog buffers */
 520
 521         /* The following field are used for debugging; need to hold icloglock */
 522         char                    *l_iclog_bak[XLOG_MAX_ICLOGS];
 523
 524         /* The following block of fields are changed while holding grant_lock */
 525         lock_t                  l_grant_lock;
 526         xlog_ticket_t           *l_reserve_headq;
 527         xlog_ticket_t           *l_write_headq;
 528         int                     l_grant_reserve_cycle;
 529         int                     l_grant_reserve_bytes;
 530         int                     l_grant_write_cycle;
 531         int                     l_grant_write_bytes;
 532
 533         /* The following fields don't need locking */
 534 #ifdef DEBUG
 535         struct ktrace           *l_trace;
 536         struct ktrace           *l_grant_trace;
 537 #endif
 538         uint                    l_flags;
 539         uint                    l_quotaoffs_flag; /* XFS_DQ_*, for QUOTAOFFs */
 540         struct xfs_buf_cancel   **l_buf_cancel_table;
 541         int                     l_stripemask;   /* log stripe mask */
 542         int                     l_iclog_hsize;  /* size of iclog header */
 543         int                     l_iclog_heads;  /* # of iclog header sectors */
 544         uint                    l_sectbb_log;   /* log2 of sector size in BBs */
 545         uint                    l_sectbb_mask;  /* sector size (in BBs)
 546                                                  * alignment mask */
 547 } xlog_t;
 548
 549
 550 /* common routines */
 551 extern xfs_lsn_t xlog_assign_tail_lsn(struct xfs_mount *mp);
 552 extern int       xlog_find_head(xlog_t *log, xfs_daddr_t *head_blk);
 553 extern int       xlog_find_tail(xlog_t  *log,
 554                                 xfs_daddr_t *head_blk,
 555                                 xfs_daddr_t *tail_blk,
 556                                 int readonly);
 557 extern int       xlog_print_find_oldest(xlog_t *log, xfs_daddr_t *last_blk);
 558 extern int       xlog_recover(xlog_t *log, int readonly);
 559 extern int       xlog_recover_finish(xlog_t *log, int mfsi_flags);
 560 extern void      xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog);
 561 extern void      xlog_recover_process_iunlinks(xlog_t *log);
 562
 563 extern struct xfs_buf *xlog_get_bp(xlog_t *, int);
 564 extern void      xlog_put_bp(struct xfs_buf *);
 565 extern int       xlog_bread(xlog_t *, xfs_daddr_t, int, struct xfs_buf *);
 566 extern xfs_caddr_t xlog_align(xlog_t *, xfs_daddr_t, int, struct xfs_buf *);
 567
 568 #define XLOG_TRACE_GRAB_FLUSH  1
 569 #define XLOG_TRACE_REL_FLUSH   2
 570 #define XLOG_TRACE_SLEEP_FLUSH 3
 571 #define XLOG_TRACE_WAKE_FLUSH  4
 572
 573 #endif  /* __KERNEL__ */
 574
 575 #endif  /* __XFS_LOG_PRIV_H__ */