/ Check-in [d1debe5d]
Login

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview
Comment:Simplifications to the SHM implementation in os_unix.c, taking advantage of the removal of the LinuxThreads mess.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA1: d1debe5def82a6bc72f11b8787176ac60259630f
User & Date: drh 2010-05-14 14:52:25
Context
2010-05-14
19:24
Make sure the value of an INTEGER PRIMARY KEY column supplied to triggers and especially to FK constraints really contains the ROWID and not the NULL that is stored in the column itself. Ticket [dd08e5a988d00dec]. check-in: 636f8609 user: drh tags: trunk
16:34
Pull in all the latest changes from the trunk. Update the win32 SHM methods to work with the new interface design. Closed-Leaf check-in: 4b69f2cd user: drh tags: wal-win32
14:52
Simplifications to the SHM implementation in os_unix.c, taking advantage of the removal of the LinuxThreads mess. check-in: d1debe5d user: drh tags: trunk
12:43
Simplify os_unix.c by removing support for LinuxThreads. Linux systems must either use NPTL or else not share database connections across threads. check-in: e294b696 user: drh tags: trunk
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/os_unix.c.

171
172
173
174
175
176
177
178
179
180


181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
...
711
712
713
714
715
716
717





718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746

747
748
749
750
751
752
753
...
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
...
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
....
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
....
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
....
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
....
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
....
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
....
3086
3087
3088
3089
3090
3091
3092
3093

3094
3095

3096
3097







3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153



3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
....
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
....
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
....
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
....
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431

3432
3433
3434
3435

3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
....
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537

3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586



3587
3588
3589
3590
3591
3592

3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
....
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639



3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
....
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701



3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
....
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
....
3769
3770
3771
3772
3773
3774
3775
3776
3777



3778
3779
3780
3781
3782
3783
3784
....
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
....
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507

/*
** Only set the lastErrno if the error code is a real error and not 
** a normal expected return code of SQLITE_BUSY or SQLITE_OK
*/
#define IS_LOCK_ERROR(x)  ((x != SQLITE_OK) && (x != SQLITE_BUSY))

/* Forward reference */
typedef struct unixShm unixShm;
typedef struct unixShmFile unixShmFile;



/*
** Sometimes, after a file handle is closed by SQLite, the file descriptor
** cannot be closed immediately. In these cases, instances of the following
** structure are used to store the file descriptor while waiting for an
** opportunity to either close or reuse it.
*/
typedef struct UnixUnusedFd UnixUnusedFd;
struct UnixUnusedFd {
  int fd;                   /* File descriptor to close */
  int flags;                /* Flags this file descriptor was opened with */
  UnixUnusedFd *pNext;      /* Next unused file descriptor on same file */
};

/*
** The unixFile structure is subclass of sqlite3_file specific to the unix
** VFS implementations.
*/
typedef struct unixFile unixFile;
struct unixFile {
  sqlite3_io_methods const *pMethod;  /* Always the first entry */
  struct unixInodeInfo *pInode;       /* Info about locks on this inode */
  int h;                              /* The file descriptor */
  int dirfd;                          /* File descriptor for the directory */
  unsigned char eFileLock;            /* The type of lock held on this fd */
  int lastErrno;                      /* The unix errno from last I/O error */
  void *lockingContext;               /* Locking style specific state */
  UnixUnusedFd *pUnused;              /* Pre-allocated UnixUnusedFd */
  int fileFlags;                      /* Miscellanous flags */
................................................................................
** object keeps a count of the number of unixFile pointing to it.
*/
struct unixInodeInfo {
  struct unixFileId fileId;       /* The lookup key */
  int nShared;                    /* Number of SHARED locks held */
  int eFileLock;                  /* One of SHARED_LOCK, RESERVED_LOCK etc. */
  int nRef;                       /* Number of pointers to this structure */





#if defined(SQLITE_ENABLE_LOCKING_STYLE)
  unsigned long long sharedByte;  /* for AFP simulated shared lock */
#endif
  int nLock;                      /* Number of outstanding file locks */
  UnixUnusedFd *pUnused;          /* Unused file descriptors to close */
#if OS_VXWORKS
  sem_t *pSem;                    /* Named POSIX semaphore */
  char aSemName[MAX_PATHNAME+2];  /* Name of that semaphore */
#endif
  struct unixInodeInfo *pNext;    /* List of all unixInodeInfo objects */
  struct unixInodeInfo *pPrev;    /*    .... doubly linked */
};

/*
** A lists of all unixInodeInfo objects.
*/
static struct unixInodeInfo *inodeList = 0;

/*
** Release a unixInodeInfo structure previously allocated by findInodeInfo().
**
** The mutex entered using the unixEnterMutex() function must be held
** when this function is called.
*/
static void releaseInodeInfo(struct unixInodeInfo *pInode){
  assert( unixMutexHeld() );
  if( pInode ){
    pInode->nRef--;
    if( pInode->nRef==0 ){

      if( pInode->pPrev ){
        assert( pInode->pPrev->pNext==pInode );
        pInode->pPrev->pNext = pInode->pNext;
      }else{
        assert( inodeList==pInode );
        inodeList = pInode->pNext;
      }
................................................................................
** The mutex entered using the unixEnterMutex() function must be held
** when this function is called.
**
** Return an appropriate error code.
*/
static int findInodeInfo(
  unixFile *pFile,               /* Unix file with file desc used in the key */
  struct unixInodeInfo **ppInode /* Return the unixInodeInfo object here */
){
  int rc;                        /* System call return code */
  int fd;                        /* The file descriptor for pFile */
  struct unixFileId fileId;          /* Lookup key for the unixInodeInfo */
  struct stat statbuf;               /* Low-level file information */
  struct unixInodeInfo *pInode = 0;  /* Candidate unixInodeInfo object */

  assert( unixMutexHeld() );

  /* Get low-level information about the file that we can used to
  ** create a unique name for the file.
  */
  fd = pFile->h;
................................................................................
  ** The reason a single byte cannot be used instead of the 'shared byte
  ** range' is that some versions of windows do not support read-locks. By
  ** locking a random byte from a range, concurrent SHARED locks may exist
  ** even if the locking primitive used is always a write-lock.
  */
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  struct unixInodeInfo *pInode = pFile->pInode;
  struct flock lock;
  int s = 0;
  int tErrno = 0;

  assert( pFile );
  OSTRACE(("LOCK    %d %s was %s(%s,%d) pid=%d (unix)\n", pFile->h,
      azFileLock(eFileLock), azFileLock(pFile->eFileLock),
................................................................................
**
** Otherwise, if an error occurs, then successfully closed file descriptor
** entries are removed from the list, and SQLITE_IOERR_CLOSE returned. 
** not deleted and SQLITE_IOERR_CLOSE returned.
*/ 
static int closePendingFds(unixFile *pFile){
  int rc = SQLITE_OK;
  struct unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *pError = 0;
  UnixUnusedFd *p;
  UnixUnusedFd *pNext;
  for(p=pInode->pUnused; p; p=pNext){
    pNext = p->pNext;
    if( close(p->fd) ){
      pFile->lastErrno = errno;
................................................................................
}

/*
** Add the file descriptor used by file handle pFile to the corresponding
** pUnused list.
*/
static void setPendingFd(unixFile *pFile){
  struct unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *p = pFile->pUnused;
  p->pNext = pInode->pUnused;
  pInode->pUnused = p;
  pFile->h = -1;
  pFile->pUnused = 0;
}

................................................................................
** the byte range is divided into 2 parts and the first part is unlocked then
** set to a read lock, then the other part is simply unlocked.  This works 
** around a bug in BSD NFS lockd (also seen on MacOSX 10.3+) that fails to 
** remove the write lock on a region when a read lock is set.
*/
static int _posixUnlock(sqlite3_file *id, int eFileLock, int handleNFSUnlock){
  unixFile *pFile = (unixFile*)id;
  struct unixInodeInfo *pInode;
  struct flock lock;
  int rc = SQLITE_OK;
  int h;
  int tErrno;                      /* Error code from system call errors */

  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (unix)\n", pFile->h, eFileLock,
................................................................................
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int afpLock(sqlite3_file *id, int eFileLock){
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  struct unixInodeInfo *pInode = pFile->pInode;
  afpLockingContext *context = (afpLockingContext *) pFile->lockingContext;
  
  assert( pFile );
  OSTRACE(("LOCK    %d %s was %s(%s,%d) pid=%d (afp)\n", pFile->h,
           azFileLock(eFileLock), azFileLock(pFile->eFileLock),
           azFileLock(pInode->eFileLock), pInode->nShared , getpid()));

................................................................................
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int afpUnlock(sqlite3_file *id, int eFileLock) {
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  struct unixInodeInfo *pInode;
  afpLockingContext *context = (afpLockingContext *) pFile->lockingContext;
  int skipShared = 0;
#ifdef SQLITE_TEST
  int h = pFile->h;
#endif

  assert( pFile );
................................................................................
  return 0;
}

#ifndef SQLITE_OMIT_WAL


/*
** Object used to represent a single file opened and mmapped to provide

** shared memory.  When multiple threads all reference the same
** log-summary, each thread has its own unixFile object, but they all

** point to a single instance of this object.  In other words, each
** log-summary is opened only once per process.







**
** unixMutexHeld() must be true when creating or destroying
** this object or while reading or writing the following fields:
**
**      nRef
**      pNext 
**
** The following fields are read-only after the object is created:
** 
**      fid
**      zFilename
**
** Either unixShmFile.mutex must be held or unixShmFile.nRef==0 and
** unixMutexHeld() is true when reading or writing any other field
** in this structure.
**
** To avoid deadlocks, mutex and mutexBuf are always released in the
** reverse order that they are acquired.  mutexBuf is always acquired
** first and released last.  This invariant is check by asserting
** sqlite3_mutex_notheld() on mutex whenever mutexBuf is acquired or
** released.
*/
struct unixShmFile {
  struct unixFileId fid;     /* Unique file identifier */
  sqlite3_mutex *mutex;      /* Mutex to access this object */
  sqlite3_mutex *mutexBuf;   /* Mutex to access zBuf[] */
  char *zFilename;           /* Name of the mmapped file */
  int h;                     /* Open file descriptor */
  int szMap;                 /* Size of the mapping of file into memory */
  char *pMMapBuf;            /* Where currently mmapped().  NULL if unmapped */
  int nRef;                  /* Number of unixShm objects pointing to this */
  unixShm *pFirst;           /* All unixShm objects pointing to this */
  unixShmFile *pNext;        /* Next in list of all unixShmFile objects */
#ifdef SQLITE_DEBUG
  u8 exclMask;               /* Mask of exclusive locks held */
  u8 sharedMask;             /* Mask of shared locks held */
  u8 nextShmId;              /* Next available unixShm.id value */
#endif
};

/*
** A global array of all unixShmFile objects.
**
** The unixMutexHeld() must be true while reading or writing this list.
*/
static unixShmFile *unixShmFileList = 0;

/*
** Structure used internally by this VFS to record the state of an
** open shared memory connection.
**
** unixShm.pFile->mutex must be held while reading or writing the
** unixShm.pNext and unixShm.locks[] elements.
**
** The unixShm.pFile element is initialized when the object is created
** and is read-only thereafter.



*/
struct unixShm {
  unixShmFile *pFile;        /* The underlying unixShmFile object */
  unixShm *pNext;            /* Next unixShm with the same unixShmFile */
  u8 lockState;              /* Current lock state */
  u8 hasMutex;               /* True if holding the unixShmFile mutex */
  u8 hasMutexBuf;            /* True if holding pFile->mutexBuf */
  u8 sharedMask;             /* Mask of shared locks held */
  u8 exclMask;               /* Mask of exclusive locks held */
#ifdef SQLITE_DEBUG
  u8 id;                     /* Id of this connection with its unixShmFile */
#endif
};

/*
** Size increment by which shared memory grows
*/
#define SQLITE_UNIX_SHM_INCR  4096
................................................................................
** lockMask might contain multiple bits but all bits are guaranteed
** to be contiguous.
**
** Locks block if the mask is exactly UNIX_SHM_C and are non-blocking
** otherwise.
*/
static int unixShmSystemLock(
  unixShmFile *pFile,   /* Apply locks to this open shared-memory segment */
  int lockType,         /* F_UNLCK, F_RDLCK, or F_WRLCK */
  u8 lockMask           /* Which bytes to lock or unlock */
){
  struct flock f;       /* The posix advisory locking structure */
  int lockOp;           /* The opcode for fcntl() */
  int i;                /* Offset into the locking byte range */
  int rc;               /* Result code form fcntl() */
  u8 mask;              /* Mask of bits in lockMask */

  /* Access to the unixShmFile object is serialized by the caller */
  assert( sqlite3_mutex_held(pFile->mutex) || pFile->nRef==0 );

  /* Initialize the locking parameters */
  memset(&f, 0, sizeof(f));
  f.l_type = lockType;
  f.l_whence = SEEK_SET;
  if( lockMask==UNIX_SHM_C && lockType!=F_UNLCK ){
    lockOp = F_SETLKW;
................................................................................
    mask <<= 1;
  }

  /* Verify that all bits set in lockMask are contiguous */
  assert( mask==0 || (lockMask & ~(mask | (mask-1)))==0 );

  /* Acquire the system-level lock */
  rc = fcntl(pFile->h, lockOp, &f);
  rc = (rc!=(-1)) ? SQLITE_OK : SQLITE_BUSY;

  /* Update the global lock state and do debug tracing */
#ifdef SQLITE_DEBUG
  OSTRACE(("SHM-LOCK "));
  if( rc==SQLITE_OK ){
    if( lockType==F_UNLCK ){
      OSTRACE(("unlock ok"));
      pFile->exclMask &= ~lockMask;
      pFile->sharedMask &= ~lockMask;
    }else if( lockType==F_RDLCK ){
      OSTRACE(("read-lock ok"));
      pFile->exclMask &= ~lockMask;
      pFile->sharedMask |= lockMask;
    }else{
      assert( lockType==F_WRLCK );
      OSTRACE(("write-lock ok"));
      pFile->exclMask |= lockMask;
      pFile->sharedMask &= ~lockMask;
    }
  }else{
    if( lockType==F_UNLCK ){
      OSTRACE(("unlock failed"));
    }else if( lockType==F_RDLCK ){
      OSTRACE(("read-lock failed"));
    }else{
      assert( lockType==F_WRLCK );
      OSTRACE(("write-lock failed"));
    }
  }
  OSTRACE((" - change requested %s - afterwards %s:%s\n",
           unixShmLockString(lockMask),
           unixShmLockString(pFile->sharedMask),
           unixShmLockString(pFile->exclMask)));
#endif

  return rc;        
}

/*
** For connection p, unlock all of the locks identified by the unlockMask
** parameter.
*/
static int unixShmUnlock(
  unixShmFile *pFile,   /* The underlying shared-memory file */
  unixShm *p,           /* The connection to be unlocked */
  u8 unlockMask         /* Mask of locks to be unlocked */
){
  int rc;      /* Result code */
  unixShm *pX; /* For looping over all sibling connections */
  u8 allMask;  /* Union of locks held by connections other than "p" */

  /* Access to the unixShmFile object is serialized by the caller */
  assert( sqlite3_mutex_held(pFile->mutex) );

  /* Compute locks held by sibling connections */
  allMask = 0;
  for(pX=pFile->pFirst; pX; pX=pX->pNext){
    if( pX==p ) continue;
    assert( (pX->exclMask & (p->exclMask|p->sharedMask))==0 );
    allMask |= pX->sharedMask;
  }

  /* Unlock the system-level locks */
  if( (unlockMask & allMask)!=unlockMask ){
    rc = unixShmSystemLock(pFile, F_UNLCK, unlockMask & ~allMask);
  }else{
    rc = SQLITE_OK;
  }

  /* Undo the local locks */
  if( rc==SQLITE_OK ){
    p->exclMask &= ~unlockMask;
................................................................................
  return rc;
}

/*
** Get reader locks for connection p on all locks in the readMask parameter.
*/
static int unixShmSharedLock(
  unixShmFile *pFile,   /* The underlying shared-memory file */
  unixShm *p,           /* The connection to get the shared locks */
  u8 readMask           /* Mask of shared locks to be acquired */
){
  int rc;        /* Result code */
  unixShm *pX;   /* For looping over all sibling connections */
  u8 allShared;  /* Union of locks held by connections other than "p" */

  /* Access to the unixShmFile object is serialized by the caller */
  assert( sqlite3_mutex_held(pFile->mutex) );

  /* Find out which shared locks are already held by sibling connections.
  ** If any sibling already holds an exclusive lock, go ahead and return
  ** SQLITE_BUSY.
  */
  allShared = 0;
  for(pX=pFile->pFirst; pX; pX=pX->pNext){
    if( pX==p ) continue;
    if( (pX->exclMask & readMask)!=0 ) return SQLITE_BUSY;
    allShared |= pX->sharedMask;
  }

  /* Get shared locks at the system level, if necessary */
  if( (~allShared) & readMask ){
    rc = unixShmSystemLock(pFile, F_RDLCK, readMask);
  }else{
    rc = SQLITE_OK;
  }

  /* Get the local shared locks */
  if( rc==SQLITE_OK ){
    p->sharedMask |= readMask;
................................................................................
}

/*
** For connection p, get an exclusive lock on all locks identified in
** the writeMask parameter.
*/
static int unixShmExclusiveLock(
  unixShmFile *pFile,    /* The underlying shared-memory file */
  unixShm *p,            /* The connection to get the exclusive locks */
  u8 writeMask           /* Mask of exclusive locks to be acquired */
){
  int rc;        /* Result code */
  unixShm *pX;   /* For looping over all sibling connections */

  /* Access to the unixShmFile object is serialized by the caller */
  assert( sqlite3_mutex_held(pFile->mutex) );

  /* Make sure no sibling connections hold locks that will block this
  ** lock.  If any do, return SQLITE_BUSY right away.
  */
  for(pX=pFile->pFirst; pX; pX=pX->pNext){
    if( pX==p ) continue;
    if( (pX->exclMask & writeMask)!=0 ) return SQLITE_BUSY;
    if( (pX->sharedMask & writeMask)!=0 ) return SQLITE_BUSY;
  }

  /* Get the exclusive locks at the system level.  Then if successful
  ** also mark the local connection as being locked.
  */
  rc = unixShmSystemLock(pFile, F_WRLCK, writeMask);
  if( rc==SQLITE_OK ){
    p->sharedMask &= ~writeMask;
    p->exclMask |= writeMask;
  }
  return rc;
}

/*
** Purge the unixShmFileList list of all entries with unixShmFile.nRef==0.
**
** This is not a VFS shared-memory method; it is a utility function called
** by VFS shared-memory methods.
*/
static void unixShmPurge(void){
  unixShmFile **pp;
  unixShmFile *p;
  assert( unixMutexHeld() );
  pp = &unixShmFileList;
  while( (p = *pp)!=0 ){
    if( p->nRef==0 ){

      if( p->mutex ) sqlite3_mutex_free(p->mutex);
      if( p->mutexBuf ) sqlite3_mutex_free(p->mutexBuf);
      if( p->h>=0 ) close(p->h);
      *pp = p->pNext;

      sqlite3_free(p);
    }else{
      pp = &p->pNext;
    }
  }
}

/*
** Open a shared-memory area.  This particular implementation uses
** mmapped files.
**
................................................................................
** file are currently open, in this process or in other processes, then
** the file must be truncated to zero length or have its header cleared.
*/
static int unixShmOpen(
  sqlite3_file *fd      /* The file descriptor of the associated database */
){
  struct unixShm *p = 0;             /* The connection to be opened */
  struct unixShmFile *pFile = 0;     /* The underlying mmapped file */
  int rc;                            /* Result code */
  struct unixFileId fid;             /* Unix file identifier */
  struct unixShmFile *pNew;          /* Newly allocated pFile */
  struct stat sStat;                 /* Result from stat() an fstat() */
  struct unixFile *pDbFd;            /* Underlying database file */
  int nPath;                         /* Size of pDbFd->zPath in bytes */

  /* Allocate space for the new sqlite3_shm object.  Also speculatively
  ** allocate space for a new unixShmFile and filename.
  */
  p = sqlite3_malloc( sizeof(*p) );
  if( p==0 ) return SQLITE_NOMEM;
  memset(p, 0, sizeof(*p));
  pDbFd = (struct unixFile*)fd;
  assert( pDbFd->pShm==0 );
  nPath = strlen(pDbFd->zPath);
  pNew = sqlite3_malloc( sizeof(*pFile) + nPath + 15 );
  if( pNew==0 ){
    sqlite3_free(p);
    return SQLITE_NOMEM;
  }
  memset(pNew, 0, sizeof(*pNew));
  pNew->zFilename = (char*)&pNew[1];
  sqlite3_snprintf(nPath+15, pNew->zFilename, "%s-wal-index", pDbFd->zPath);

  /* Look to see if there is an existing unixShmFile that can be used.
  ** If no matching unixShmFile currently exists, create a new one.
  */
  unixEnterMutex();
  rc = stat(pNew->zFilename, &sStat);
  if( rc==0 ){
    memset(&fid, 0, sizeof(fid));
    fid.dev = sStat.st_dev;
    fid.ino = sStat.st_ino;
    for(pFile = unixShmFileList; pFile; pFile=pFile->pNext){
      if( memcmp(&pFile->fid, &fid, sizeof(fid))==0 ) break;
    }
  }
  if( pFile ){
    sqlite3_free(pNew);
  }else{
    pFile = pNew;
    pNew = 0;
    pFile->h = -1;
    pFile->pNext = unixShmFileList;
    unixShmFileList = pFile;

    pFile->mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
    if( pFile->mutex==0 ){
      rc = SQLITE_NOMEM;
      goto shm_open_err;
    }
    pFile->mutexBuf = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
    if( pFile->mutexBuf==0 ){
      rc = SQLITE_NOMEM;
      goto shm_open_err;
    }

    pFile->h = open(pFile->zFilename, O_RDWR|O_CREAT, 0664);
    if( pFile->h<0 ){
      rc = SQLITE_CANTOPEN_BKPT;
      goto shm_open_err;
    }

    rc = fstat(pFile->h, &sStat);
    if( rc ){
      rc = SQLITE_CANTOPEN_BKPT;
      goto shm_open_err;
    }
    pFile->fid.dev = sStat.st_dev;
    pFile->fid.ino = sStat.st_ino;

    /* Check to see if another process is holding the dead-man switch.
    ** If not, truncate the file to zero length. 
    */

    if( unixShmSystemLock(pFile, F_WRLCK, UNIX_SHM_DMS)==SQLITE_OK ){
      if( ftruncate(pFile->h, 0) ){
        rc = SQLITE_IOERR;
      }
    }
    if( rc==SQLITE_OK ){
      rc = unixShmSystemLock(pFile, F_RDLCK, UNIX_SHM_DMS);
    }
    if( rc ) goto shm_open_err;
  }

  /* Make the new connection a child of the unixShmFile */
  p->pFile = pFile;
  p->pNext = pFile->pFirst;
#ifdef SQLITE_DEBUG
  p->id = pFile->nextShmId++;
#endif
  pFile->pFirst = p;
  pFile->nRef++;
  pDbFd->pShm = p;
  unixLeaveMutex();
  return SQLITE_OK;

  /* Jump here on any error */
shm_open_err:
  unixShmPurge();                 /* This call frees pFile if required */
  sqlite3_free(p);
  sqlite3_free(pNew);
  unixLeaveMutex();
  return rc;
}

/*
** Close a connection to shared-memory.  Delete the underlying 
** storage if deleteFlag is true.
*/
static int unixShmClose(
  sqlite3_file *fd,          /* The underlying database file */
  int deleteFlag             /* Delete shared-memory if true */
){
  unixShm *p;            /* The connection to be closed */
  unixShmFile *pFile;    /* The underlying shared-memory file */
  unixShm **pp;          /* For looping over sibling connections */
  unixFile *pDbFd;       /* The underlying database file */

  pDbFd = (unixFile*)fd;
  p = pDbFd->pShm;
  if( p==0 ) return SQLITE_OK;
  pFile = p->pFile;




  /* Verify that the connection being closed holds no locks */
  assert( p->exclMask==0 );
  assert( p->sharedMask==0 );

  /* Remove connection p from the set of connections associated with pFile */

  sqlite3_mutex_enter(pFile->mutex);
  for(pp=&pFile->pFirst; (*pp)!=p; pp = &(*pp)->pNext){}
  *pp = p->pNext;

  /* Free the connection p */
  sqlite3_free(p);
  pDbFd->pShm = 0;
  sqlite3_mutex_leave(pFile->mutex);

  /* If pFile->nRef has reached 0, then close the underlying
  ** shared-memory file, too */
  unixEnterMutex();
  assert( pFile->nRef>0 );
  pFile->nRef--;
  if( pFile->nRef==0 ){
    if( deleteFlag ) unlink(pFile->zFilename);
    unixShmPurge();
  }
  unixLeaveMutex();

  return SQLITE_OK;
}

/*
................................................................................
static int unixShmSize(
  sqlite3_file *fd,         /* The open database file holding SHM */
  int reqSize,              /* Requested size.  -1 for query only */
  int *pNewSize             /* Write new size here */
){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;
  unixShmFile *pFile = p->pFile;
  int rc = SQLITE_OK;
  struct stat sStat;




  /* On a query, this loop runs once.  When reqSize>=0, the loop potentially
  ** runs twice, except if the actual size is already greater than or equal
  ** to the requested size, reqSize is set to -1 on the first iteration and
  ** the loop only runs once.
  */
  while( 1 ){
    if( fstat(pFile->h, &sStat)==0 ){
      *pNewSize = (int)sStat.st_size;
      if( reqSize>=0 && reqSize<=(int)sStat.st_size ) break;
    }else{
      *pNewSize = 0;
      rc = SQLITE_IOERR;
      break;
    }
    if( reqSize<0 ) break;
    reqSize = (reqSize + SQLITE_UNIX_SHM_INCR - 1)/SQLITE_UNIX_SHM_INCR;
    reqSize *= SQLITE_UNIX_SHM_INCR;
    rc = ftruncate(pFile->h, reqSize);
    reqSize = -1;
  }
  return rc;
}


/*
................................................................................
  sqlite3_file *fd,        /* Database file holding shared memory */
  int reqMapSize,          /* Requested size of mapping. -1 means don't care */
  int *pNewMapSize,        /* Write new size of mapping here */
  void **ppBuf             /* Write mapping buffer origin here */
){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;
  unixShmFile *pFile = p->pFile;
  int rc = SQLITE_OK;




  if( p->lockState!=SQLITE_SHM_CHECKPOINT && p->hasMutexBuf==0 ){
    assert( sqlite3_mutex_notheld(pFile->mutex) );
    sqlite3_mutex_enter(pFile->mutexBuf);
    p->hasMutexBuf = 1;
  }
  sqlite3_mutex_enter(pFile->mutex);
  if( pFile->szMap==0 || reqMapSize>pFile->szMap ){
    int actualSize;
    if( unixShmSize(fd, -1, &actualSize)==SQLITE_OK
     && reqMapSize<actualSize
    ){
      reqMapSize = actualSize;
    }
    if( pFile->pMMapBuf ){
      munmap(pFile->pMMapBuf, pFile->szMap);
    }
    pFile->pMMapBuf = mmap(0, reqMapSize, PROT_READ|PROT_WRITE, MAP_SHARED,
                           pFile->h, 0);
    pFile->szMap = pFile->pMMapBuf ? reqMapSize : 0;
  }
  *pNewMapSize = pFile->szMap;
  *ppBuf = pFile->pMMapBuf;
  sqlite3_mutex_leave(pFile->mutex);
  return rc;
}

/*
** Release the lock held on the shared memory segment to that other
** threads are free to resize it if necessary.
**
................................................................................
** is a no-op.
*/
static int unixShmRelease(sqlite3_file *fd){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;

  if( p->hasMutexBuf && p->lockState!=SQLITE_SHM_RECOVER ){
    assert( sqlite3_mutex_notheld(p->pFile->mutex) );
    sqlite3_mutex_leave(p->pFile->mutexBuf);
    p->hasMutexBuf = 0;
  }
  return SQLITE_OK;
}

/*
** Symbolic names for LOCK states used for debugging.
................................................................................
static int unixShmLock(
  sqlite3_file *fd,          /* Database file holding the shared memory */
  int desiredLock,           /* One of SQLITE_SHM_xxxxx locking states */
  int *pGotLock              /* The lock you actually got */
){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;
  unixShmFile *pFile = p->pFile;
  int rc = SQLITE_PROTOCOL;




  /* Note that SQLITE_SHM_READ_FULL and SQLITE_SHM_PENDING are never
  ** directly requested; they are side effects from requesting
  ** SQLITE_SHM_READ and SQLITE_SHM_CHECKPOINT, respectively.
  */
  assert( desiredLock==SQLITE_SHM_UNLOCK
       || desiredLock==SQLITE_SHM_READ
................................................................................
    return SQLITE_OK;
  }

  OSTRACE(("SHM-LOCK shmid-%d, pid-%d request %s->%s\n",
            p->id, getpid(), azLkName[p->lockState], azLkName[desiredLock]));
  
  if( desiredLock==SQLITE_SHM_RECOVER && !p->hasMutexBuf ){
    assert( sqlite3_mutex_notheld(pFile->mutex) );
    sqlite3_mutex_enter(pFile->mutexBuf);
    p->hasMutexBuf = 1;
  }
  sqlite3_mutex_enter(pFile->mutex);
  switch( desiredLock ){
    case SQLITE_SHM_UNLOCK: {
      assert( p->lockState!=SQLITE_SHM_RECOVER );
      unixShmUnlock(pFile, p, UNIX_SHM_A|UNIX_SHM_B|UNIX_SHM_C|UNIX_SHM_D);
      rc = SQLITE_OK;
      p->lockState = SQLITE_SHM_UNLOCK;
      break;
    }
    case SQLITE_SHM_READ: {
      if( p->lockState==SQLITE_SHM_UNLOCK ){
        int nAttempt;
        rc = SQLITE_BUSY;
        assert( p->lockState==SQLITE_SHM_UNLOCK );
        for(nAttempt=0; nAttempt<5 && rc==SQLITE_BUSY; nAttempt++){
          rc = unixShmSharedLock(pFile, p, UNIX_SHM_A|UNIX_SHM_B);
          if( rc==SQLITE_BUSY ){
            rc = unixShmSharedLock(pFile, p, UNIX_SHM_D);
            if( rc==SQLITE_OK ){
              p->lockState = SQLITE_SHM_READ_FULL;
            }
          }else{
            unixShmUnlock(pFile, p, UNIX_SHM_B);
            p->lockState = SQLITE_SHM_READ;
          }
        }
      }else{
       assert( p->lockState==SQLITE_SHM_WRITE
               || p->lockState==SQLITE_SHM_RECOVER );
        rc = unixShmSharedLock(pFile, p, UNIX_SHM_A);
        unixShmUnlock(pFile, p, UNIX_SHM_C|UNIX_SHM_D);
        p->lockState = SQLITE_SHM_READ;
      }
      break;
    }
    case SQLITE_SHM_WRITE: {
      assert( p->lockState==SQLITE_SHM_READ 
              || p->lockState==SQLITE_SHM_READ_FULL );
      rc = unixShmExclusiveLock(pFile, p, UNIX_SHM_C|UNIX_SHM_D);
      if( rc==SQLITE_OK ){
        p->lockState = SQLITE_SHM_WRITE;
      }
      break;
    }
    case SQLITE_SHM_CHECKPOINT: {
      assert( p->lockState==SQLITE_SHM_UNLOCK
           || p->lockState==SQLITE_SHM_PENDING
      );
      if( p->lockState==SQLITE_SHM_UNLOCK ){
        rc = unixShmExclusiveLock(pFile, p, UNIX_SHM_B|UNIX_SHM_C);
        if( rc==SQLITE_OK ){
          p->lockState = SQLITE_SHM_PENDING;
        }
      }
      if( p->lockState==SQLITE_SHM_PENDING ){
        rc = unixShmExclusiveLock(pFile, p, UNIX_SHM_A);
        if( rc==SQLITE_OK ){
          p->lockState = SQLITE_SHM_CHECKPOINT;
        }
      }
      break;
    }
    default: {
      assert( desiredLock==SQLITE_SHM_RECOVER );
      assert( p->lockState==SQLITE_SHM_READ
           || p->lockState==SQLITE_SHM_READ_FULL
      );
      assert( sqlite3_mutex_held(pFile->mutexBuf) );
      rc = unixShmExclusiveLock(pFile, p, UNIX_SHM_C);
      if( rc==SQLITE_OK ){
        p->lockState = SQLITE_SHM_RECOVER;
      }
      break;
    }
  }
  sqlite3_mutex_leave(pFile->mutex);
  OSTRACE(("SHM-LOCK shmid-%d, pid-%d got %s\n",
           p->id, getpid(), azLkName[p->lockState]));
  if( pGotLock ) *pGotLock = p->lockState;
  return rc;
}

#else
................................................................................
  ** For this reason, if an error occurs in the stat() call here, it is
  ** ignored and -1 is returned. The caller will try to open a new file
  ** descriptor on the same path, fail, and return an error to SQLite.
  **
  ** Even if a subsequent open() call does succeed, the consequences of
  ** not searching for a resusable file descriptor are not dire.  */
  if( 0==stat(zPath, &sStat) ){
    struct unixInodeInfo *pInode;

    unixEnterMutex();
    pInode = inodeList;
    while( pInode && (pInode->fileId.dev!=sStat.st_dev
                     || pInode->fileId.ino!=sStat.st_ino) ){
       pInode = pInode->pNext;
    }







|
|
|
>
>







<













|







 







>
>
>
>
>



<
<




<
<





|







|




>







 







|



|
|
|







 







|







 







|







 







|







 







|







 







|







 







|







 







|
>
|
<
>
|
|
>
>
>
>
>
>
>





<






|









|
|




|



<







<
<
<
<
<
<
<




|
|

|
|
>
>
>


|
|

|




|







 







|
|
|







|
|







 







|








|
|


|
|



|
|













|
|










|
|
|





|
|



|







|







 







|
|
|





|
|






|







|







 







|
|
|




|
|




|








|








|




|
|
<

<
<
|
>
|
|
|
<
>
|
<
<
<







 







|

<
<
<



|
<






|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<
<
<
<
<
<
<
<
<
<
<
<




|
|




<
<
<
<
<
<
<
<



>
|
|




|




|
|
|

|

|
|






|

<













|






|
>
>
>





|
>
|
|





|

|


|
|
|
|
|







 







|


>
>
>







|










|







 







|


>
>
>

|
|


|
|






|
|

|
|
|

|
|
|







 







|
|







 







|

>
>
>







 







|
|


|



|










|

|




|






|
|







|










|





|











|
|






|







 







|







171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189

190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
...
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726


727
728
729
730


731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
...
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
...
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
....
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
....
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
....
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
....
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
....
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
....
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098

3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113

3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139

3140
3141
3142
3143
3144
3145
3146







3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
....
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
....
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
....
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
....
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431

3432


3433
3434
3435
3436
3437

3438
3439



3440
3441
3442
3443
3444
3445
3446
....
3455
3456
3457
3458
3459
3460
3461
3462
3463



3464
3465
3466
3467

3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500












3501
3502
3503
3504
3505
3506
3507
3508
3509
3510








3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541

3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
....
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
....
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
....
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
....
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
....
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
....
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496

/*
** Only set the lastErrno if the error code is a real error and not 
** a normal expected return code of SQLITE_BUSY or SQLITE_OK
*/
#define IS_LOCK_ERROR(x)  ((x != SQLITE_OK) && (x != SQLITE_BUSY))

/* Forward references */
typedef struct unixShm unixShm;               /* Connection shared memory */
typedef struct unixShmNode unixShmNode;       /* Shared memory instance */
typedef struct unixInodeInfo unixInodeInfo;   /* An i-node */
typedef struct UnixUnusedFd UnixUnusedFd;     /* An unused file descriptor */

/*
** Sometimes, after a file handle is closed by SQLite, the file descriptor
** cannot be closed immediately. In these cases, instances of the following
** structure are used to store the file descriptor while waiting for an
** opportunity to either close or reuse it.
*/

struct UnixUnusedFd {
  int fd;                   /* File descriptor to close */
  int flags;                /* Flags this file descriptor was opened with */
  UnixUnusedFd *pNext;      /* Next unused file descriptor on same file */
};

/*
** The unixFile structure is subclass of sqlite3_file specific to the unix
** VFS implementations.
*/
typedef struct unixFile unixFile;
struct unixFile {
  sqlite3_io_methods const *pMethod;  /* Always the first entry */
  unixInodeInfo *pInode;              /* Info about locks on this inode */
  int h;                              /* The file descriptor */
  int dirfd;                          /* File descriptor for the directory */
  unsigned char eFileLock;            /* The type of lock held on this fd */
  int lastErrno;                      /* The unix errno from last I/O error */
  void *lockingContext;               /* Locking style specific state */
  UnixUnusedFd *pUnused;              /* Pre-allocated UnixUnusedFd */
  int fileFlags;                      /* Miscellanous flags */
................................................................................
** object keeps a count of the number of unixFile pointing to it.
*/
struct unixInodeInfo {
  struct unixFileId fileId;       /* The lookup key */
  int nShared;                    /* Number of SHARED locks held */
  int eFileLock;                  /* One of SHARED_LOCK, RESERVED_LOCK etc. */
  int nRef;                       /* Number of pointers to this structure */
  unixShmNode *pShmNode;          /* Shared memory associated with this inode */
  int nLock;                      /* Number of outstanding file locks */
  UnixUnusedFd *pUnused;          /* Unused file descriptors to close */
  unixInodeInfo *pNext;           /* List of all unixInodeInfo objects */
  unixInodeInfo *pPrev;           /*    .... doubly linked */
#if defined(SQLITE_ENABLE_LOCKING_STYLE)
  unsigned long long sharedByte;  /* for AFP simulated shared lock */
#endif


#if OS_VXWORKS
  sem_t *pSem;                    /* Named POSIX semaphore */
  char aSemName[MAX_PATHNAME+2];  /* Name of that semaphore */
#endif


};

/*
** A lists of all unixInodeInfo objects.
*/
static unixInodeInfo *inodeList = 0;

/*
** Release a unixInodeInfo structure previously allocated by findInodeInfo().
**
** The mutex entered using the unixEnterMutex() function must be held
** when this function is called.
*/
static void releaseInodeInfo(unixInodeInfo *pInode){
  assert( unixMutexHeld() );
  if( pInode ){
    pInode->nRef--;
    if( pInode->nRef==0 ){
      assert( pInode->pShmNode==0 );
      if( pInode->pPrev ){
        assert( pInode->pPrev->pNext==pInode );
        pInode->pPrev->pNext = pInode->pNext;
      }else{
        assert( inodeList==pInode );
        inodeList = pInode->pNext;
      }
................................................................................
** The mutex entered using the unixEnterMutex() function must be held
** when this function is called.
**
** Return an appropriate error code.
*/
static int findInodeInfo(
  unixFile *pFile,               /* Unix file with file desc used in the key */
  unixInodeInfo **ppInode        /* Return the unixInodeInfo object here */
){
  int rc;                        /* System call return code */
  int fd;                        /* The file descriptor for pFile */
  struct unixFileId fileId;      /* Lookup key for the unixInodeInfo */
  struct stat statbuf;           /* Low-level file information */
  unixInodeInfo *pInode = 0;     /* Candidate unixInodeInfo object */

  assert( unixMutexHeld() );

  /* Get low-level information about the file that we can used to
  ** create a unique name for the file.
  */
  fd = pFile->h;
................................................................................
  ** The reason a single byte cannot be used instead of the 'shared byte
  ** range' is that some versions of windows do not support read-locks. By
  ** locking a random byte from a range, concurrent SHARED locks may exist
  ** even if the locking primitive used is always a write-lock.
  */
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode = pFile->pInode;
  struct flock lock;
  int s = 0;
  int tErrno = 0;

  assert( pFile );
  OSTRACE(("LOCK    %d %s was %s(%s,%d) pid=%d (unix)\n", pFile->h,
      azFileLock(eFileLock), azFileLock(pFile->eFileLock),
................................................................................
**
** Otherwise, if an error occurs, then successfully closed file descriptor
** entries are removed from the list, and SQLITE_IOERR_CLOSE returned. 
** not deleted and SQLITE_IOERR_CLOSE returned.
*/ 
static int closePendingFds(unixFile *pFile){
  int rc = SQLITE_OK;
  unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *pError = 0;
  UnixUnusedFd *p;
  UnixUnusedFd *pNext;
  for(p=pInode->pUnused; p; p=pNext){
    pNext = p->pNext;
    if( close(p->fd) ){
      pFile->lastErrno = errno;
................................................................................
}

/*
** Add the file descriptor used by file handle pFile to the corresponding
** pUnused list.
*/
static void setPendingFd(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *p = pFile->pUnused;
  p->pNext = pInode->pUnused;
  pInode->pUnused = p;
  pFile->h = -1;
  pFile->pUnused = 0;
}

................................................................................
** the byte range is divided into 2 parts and the first part is unlocked then
** set to a read lock, then the other part is simply unlocked.  This works 
** around a bug in BSD NFS lockd (also seen on MacOSX 10.3+) that fails to 
** remove the write lock on a region when a read lock is set.
*/
static int _posixUnlock(sqlite3_file *id, int eFileLock, int handleNFSUnlock){
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode;
  struct flock lock;
  int rc = SQLITE_OK;
  int h;
  int tErrno;                      /* Error code from system call errors */

  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (unix)\n", pFile->h, eFileLock,
................................................................................
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int afpLock(sqlite3_file *id, int eFileLock){
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode = pFile->pInode;
  afpLockingContext *context = (afpLockingContext *) pFile->lockingContext;
  
  assert( pFile );
  OSTRACE(("LOCK    %d %s was %s(%s,%d) pid=%d (afp)\n", pFile->h,
           azFileLock(eFileLock), azFileLock(pFile->eFileLock),
           azFileLock(pInode->eFileLock), pInode->nShared , getpid()));

................................................................................
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int afpUnlock(sqlite3_file *id, int eFileLock) {
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode;
  afpLockingContext *context = (afpLockingContext *) pFile->lockingContext;
  int skipShared = 0;
#ifdef SQLITE_TEST
  int h = pFile->h;
#endif

  assert( pFile );
................................................................................
  return 0;
}

#ifndef SQLITE_OMIT_WAL


/*
** Object used to represent an shared memory buffer.  
**
** When multiple threads all reference the same wal-index, each thread

** has its own unixShm object, but they all point to a single instance
** of this unixShmNode object.  In other words, each wal-index is opened
** only once per process.
**
** Each unixShmNode object is connected to a single unixInodeInfo object.
** We could coalesce this object into unixInodeInfo, but that would mean
** every open file that does not use shared memory (in other words, most
** open files) would have to carry around this extra information.  So
** the unixInodeInfo object contains a pointer to this unixShmNode object
** and the unixShmNode object is created only when needed.
**
** unixMutexHeld() must be true when creating or destroying
** this object or while reading or writing the following fields:
**
**      nRef

**
** The following fields are read-only after the object is created:
** 
**      fid
**      zFilename
**
** Either unixShmNode.mutex must be held or unixShmNode.nRef==0 and
** unixMutexHeld() is true when reading or writing any other field
** in this structure.
**
** To avoid deadlocks, mutex and mutexBuf are always released in the
** reverse order that they are acquired.  mutexBuf is always acquired
** first and released last.  This invariant is check by asserting
** sqlite3_mutex_notheld() on mutex whenever mutexBuf is acquired or
** released.
*/
struct unixShmNode {
  unixInodeInfo *pInode;     /* unixInodeInfo that owns this SHM node */
  sqlite3_mutex *mutex;      /* Mutex to access this object */
  sqlite3_mutex *mutexBuf;   /* Mutex to access zBuf[] */
  char *zFilename;           /* Name of the mmapped file */
  int h;                     /* Open file descriptor */
  int szMap;                 /* Size of the mapping into memory */
  char *pMMapBuf;            /* Where currently mmapped().  NULL if unmapped */
  int nRef;                  /* Number of unixShm objects pointing to this */
  unixShm *pFirst;           /* All unixShm objects pointing to this */

#ifdef SQLITE_DEBUG
  u8 exclMask;               /* Mask of exclusive locks held */
  u8 sharedMask;             /* Mask of shared locks held */
  u8 nextShmId;              /* Next available unixShm.id value */
#endif
};








/*
** Structure used internally by this VFS to record the state of an
** open shared memory connection.
**
** The following fields are initialized when this object is created and
** are read-only thereafter:
**
**    unixShm.pFile
**    unixShm.id
**
** All other fields are read/write.  The unixShm.pFile->mutex must be held
** while accessing any read/write fields.
*/
struct unixShm {
  unixShmNode *pShmNode;     /* The underlying unixShmNode object */
  unixShm *pNext;            /* Next unixShm with the same unixShmNode */
  u8 lockState;              /* Current lock state */
  u8 hasMutex;               /* True if holding the unixShmNode mutex */
  u8 hasMutexBuf;            /* True if holding pFile->mutexBuf */
  u8 sharedMask;             /* Mask of shared locks held */
  u8 exclMask;               /* Mask of exclusive locks held */
#ifdef SQLITE_DEBUG
  u8 id;                     /* Id of this connection within its unixShmNode */
#endif
};

/*
** Size increment by which shared memory grows
*/
#define SQLITE_UNIX_SHM_INCR  4096
................................................................................
** lockMask might contain multiple bits but all bits are guaranteed
** to be contiguous.
**
** Locks block if the mask is exactly UNIX_SHM_C and are non-blocking
** otherwise.
*/
static int unixShmSystemLock(
  unixShmNode *pShmNode, /* Apply locks to this open shared-memory segment */
  int lockType,          /* F_UNLCK, F_RDLCK, or F_WRLCK */
  u8 lockMask            /* Which bytes to lock or unlock */
){
  struct flock f;       /* The posix advisory locking structure */
  int lockOp;           /* The opcode for fcntl() */
  int i;                /* Offset into the locking byte range */
  int rc;               /* Result code form fcntl() */
  u8 mask;              /* Mask of bits in lockMask */

  /* Access to the unixShmNode object is serialized by the caller */
  assert( sqlite3_mutex_held(pShmNode->mutex) || pShmNode->nRef==0 );

  /* Initialize the locking parameters */
  memset(&f, 0, sizeof(f));
  f.l_type = lockType;
  f.l_whence = SEEK_SET;
  if( lockMask==UNIX_SHM_C && lockType!=F_UNLCK ){
    lockOp = F_SETLKW;
................................................................................
    mask <<= 1;
  }

  /* Verify that all bits set in lockMask are contiguous */
  assert( mask==0 || (lockMask & ~(mask | (mask-1)))==0 );

  /* Acquire the system-level lock */
  rc = fcntl(pShmNode->h, lockOp, &f);
  rc = (rc!=(-1)) ? SQLITE_OK : SQLITE_BUSY;

  /* Update the global lock state and do debug tracing */
#ifdef SQLITE_DEBUG
  OSTRACE(("SHM-LOCK "));
  if( rc==SQLITE_OK ){
    if( lockType==F_UNLCK ){
      OSTRACE(("unlock ok"));
      pShmNode->exclMask &= ~lockMask;
      pShmNode->sharedMask &= ~lockMask;
    }else if( lockType==F_RDLCK ){
      OSTRACE(("read-lock ok"));
      pShmNode->exclMask &= ~lockMask;
      pShmNode->sharedMask |= lockMask;
    }else{
      assert( lockType==F_WRLCK );
      OSTRACE(("write-lock ok"));
      pShmNode->exclMask |= lockMask;
      pShmNode->sharedMask &= ~lockMask;
    }
  }else{
    if( lockType==F_UNLCK ){
      OSTRACE(("unlock failed"));
    }else if( lockType==F_RDLCK ){
      OSTRACE(("read-lock failed"));
    }else{
      assert( lockType==F_WRLCK );
      OSTRACE(("write-lock failed"));
    }
  }
  OSTRACE((" - change requested %s - afterwards %s:%s\n",
           unixShmLockString(lockMask),
           unixShmLockString(pShmNode->sharedMask),
           unixShmLockString(pShmNode->exclMask)));
#endif

  return rc;        
}

/*
** For connection p, unlock all of the locks identified by the unlockMask
** parameter.
*/
static int unixShmUnlock(
  unixShmNode *pShmNode,   /* The underlying shared-memory file */
  unixShm *p,              /* The connection to be unlocked */
  u8 unlockMask            /* Mask of locks to be unlocked */
){
  int rc;      /* Result code */
  unixShm *pX; /* For looping over all sibling connections */
  u8 allMask;  /* Union of locks held by connections other than "p" */

  /* Access to the unixShmNode object is serialized by the caller */
  assert( sqlite3_mutex_held(pShmNode->mutex) );

  /* Compute locks held by sibling connections */
  allMask = 0;
  for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
    if( pX==p ) continue;
    assert( (pX->exclMask & (p->exclMask|p->sharedMask))==0 );
    allMask |= pX->sharedMask;
  }

  /* Unlock the system-level locks */
  if( (unlockMask & allMask)!=unlockMask ){
    rc = unixShmSystemLock(pShmNode, F_UNLCK, unlockMask & ~allMask);
  }else{
    rc = SQLITE_OK;
  }

  /* Undo the local locks */
  if( rc==SQLITE_OK ){
    p->exclMask &= ~unlockMask;
................................................................................
  return rc;
}

/*
** Get reader locks for connection p on all locks in the readMask parameter.
*/
static int unixShmSharedLock(
  unixShmNode *pShmNode,   /* The underlying shared-memory file */
  unixShm *p,              /* The connection to get the shared locks */
  u8 readMask              /* Mask of shared locks to be acquired */
){
  int rc;        /* Result code */
  unixShm *pX;   /* For looping over all sibling connections */
  u8 allShared;  /* Union of locks held by connections other than "p" */

  /* Access to the unixShmNode object is serialized by the caller */
  assert( sqlite3_mutex_held(pShmNode->mutex) );

  /* Find out which shared locks are already held by sibling connections.
  ** If any sibling already holds an exclusive lock, go ahead and return
  ** SQLITE_BUSY.
  */
  allShared = 0;
  for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
    if( pX==p ) continue;
    if( (pX->exclMask & readMask)!=0 ) return SQLITE_BUSY;
    allShared |= pX->sharedMask;
  }

  /* Get shared locks at the system level, if necessary */
  if( (~allShared) & readMask ){
    rc = unixShmSystemLock(pShmNode, F_RDLCK, readMask);
  }else{
    rc = SQLITE_OK;
  }

  /* Get the local shared locks */
  if( rc==SQLITE_OK ){
    p->sharedMask |= readMask;
................................................................................
}

/*
** For connection p, get an exclusive lock on all locks identified in
** the writeMask parameter.
*/
static int unixShmExclusiveLock(
  unixShmNode *pShmNode,    /* The underlying shared-memory file */
  unixShm *p,               /* The connection to get the exclusive locks */
  u8 writeMask              /* Mask of exclusive locks to be acquired */
){
  int rc;        /* Result code */
  unixShm *pX;   /* For looping over all sibling connections */

  /* Access to the unixShmNode object is serialized by the caller */
  assert( sqlite3_mutex_held(pShmNode->mutex) );

  /* Make sure no sibling connections hold locks that will block this
  ** lock.  If any do, return SQLITE_BUSY right away.
  */
  for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
    if( pX==p ) continue;
    if( (pX->exclMask & writeMask)!=0 ) return SQLITE_BUSY;
    if( (pX->sharedMask & writeMask)!=0 ) return SQLITE_BUSY;
  }

  /* Get the exclusive locks at the system level.  Then if successful
  ** also mark the local connection as being locked.
  */
  rc = unixShmSystemLock(pShmNode, F_WRLCK, writeMask);
  if( rc==SQLITE_OK ){
    p->sharedMask &= ~writeMask;
    p->exclMask |= writeMask;
  }
  return rc;
}

/*
** Purge the unixShmNodeList list of all entries with unixShmNode.nRef==0.
**
** This is not a VFS shared-memory method; it is a utility function called
** by VFS shared-memory methods.
*/
static void unixShmPurge(unixFile *pFd){
  unixShmNode *p = pFd->pInode->pShmNode;

  assert( unixMutexHeld() );


  if( p && p->nRef==0 ){
    assert( p->pInode==pFd->pInode );
    if( p->mutex ) sqlite3_mutex_free(p->mutex);
    if( p->mutexBuf ) sqlite3_mutex_free(p->mutexBuf);
    if( p->h>=0 ) close(p->h);

    p->pInode->pShmNode = 0;
    sqlite3_free(p);



  }
}

/*
** Open a shared-memory area.  This particular implementation uses
** mmapped files.
**
................................................................................
** file are currently open, in this process or in other processes, then
** the file must be truncated to zero length or have its header cleared.
*/
static int unixShmOpen(
  sqlite3_file *fd      /* The file descriptor of the associated database */
){
  struct unixShm *p = 0;             /* The connection to be opened */
  struct unixShmNode *pShmNode = 0;  /* The underlying mmapped file */
  int rc;                            /* Result code */



  struct unixFile *pDbFd;            /* Underlying database file */
  int nPath;                         /* Size of pDbFd->zPath in bytes */

  /* Allocate space for the new sqlite3_shm object.

  */
  p = sqlite3_malloc( sizeof(*p) );
  if( p==0 ) return SQLITE_NOMEM;
  memset(p, 0, sizeof(*p));
  pDbFd = (struct unixFile*)fd;
  assert( pDbFd->pShm==0 );

  /* Check to see if a unixShmNode object already exists.  Reuse an existing
  ** one if present.  Create a new one if necessary.
  */
  unixEnterMutex();
  pShmNode = pDbFd->pInode->pShmNode;
  if( pShmNode==0 ){
    nPath = strlen(pDbFd->zPath);
    pShmNode = sqlite3_malloc( sizeof(*pShmNode) + nPath + 15 );
    if( pShmNode==0 ){
      rc = SQLITE_NOMEM;
      goto shm_open_err;
    }
    memset(pShmNode, 0, sizeof(*pShmNode));
    pShmNode->zFilename = (char*)&pShmNode[1];
    sqlite3_snprintf(nPath+15, pShmNode->zFilename,
                     "%s-wal-index", pDbFd->zPath);
    pShmNode->h = -1;
    pDbFd->pInode->pShmNode = pShmNode;
    pShmNode->pInode = pDbFd->pInode;
    pShmNode->mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
    if( pShmNode->mutex==0 ){
      rc = SQLITE_NOMEM;
      goto shm_open_err;
    }
    pShmNode->mutexBuf = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
    if( pShmNode->mutexBuf==0 ){












      rc = SQLITE_NOMEM;
      goto shm_open_err;
    }

    pShmNode->h = open(pShmNode->zFilename, O_RDWR|O_CREAT, 0664);
    if( pShmNode->h<0 ){
      rc = SQLITE_CANTOPEN_BKPT;
      goto shm_open_err;
    }









    /* Check to see if another process is holding the dead-man switch.
    ** If not, truncate the file to zero length. 
    */
    rc = SQLITE_OK;
    if( unixShmSystemLock(pShmNode, F_WRLCK, UNIX_SHM_DMS)==SQLITE_OK ){
      if( ftruncate(pShmNode->h, 0) ){
        rc = SQLITE_IOERR;
      }
    }
    if( rc==SQLITE_OK ){
      rc = unixShmSystemLock(pShmNode, F_RDLCK, UNIX_SHM_DMS);
    }
    if( rc ) goto shm_open_err;
  }

  /* Make the new connection a child of the unixShmNode */
  p->pShmNode = pShmNode;
  p->pNext = pShmNode->pFirst;
#ifdef SQLITE_DEBUG
  p->id = pShmNode->nextShmId++;
#endif
  pShmNode->pFirst = p;
  pShmNode->nRef++;
  pDbFd->pShm = p;
  unixLeaveMutex();
  return SQLITE_OK;

  /* Jump here on any error */
shm_open_err:
  unixShmPurge(pDbFd);       /* This call frees pShmNode if required */
  sqlite3_free(p);

  unixLeaveMutex();
  return rc;
}

/*
** Close a connection to shared-memory.  Delete the underlying 
** storage if deleteFlag is true.
*/
static int unixShmClose(
  sqlite3_file *fd,          /* The underlying database file */
  int deleteFlag             /* Delete shared-memory if true */
){
  unixShm *p;            /* The connection to be closed */
  unixShmNode *pShmNode; /* The underlying shared-memory file */
  unixShm **pp;          /* For looping over sibling connections */
  unixFile *pDbFd;       /* The underlying database file */

  pDbFd = (unixFile*)fd;
  p = pDbFd->pShm;
  if( p==0 ) return SQLITE_OK;
  pShmNode = p->pShmNode;

  assert( pShmNode==pDbFd->pInode->pShmNode );
  assert( pShmNode->pInode==pDbFd->pInode );

  /* Verify that the connection being closed holds no locks */
  assert( p->exclMask==0 );
  assert( p->sharedMask==0 );

  /* Remove connection p from the set of connections associated
  ** with pShmNode */
  sqlite3_mutex_enter(pShmNode->mutex);
  for(pp=&pShmNode->pFirst; (*pp)!=p; pp = &(*pp)->pNext){}
  *pp = p->pNext;

  /* Free the connection p */
  sqlite3_free(p);
  pDbFd->pShm = 0;
  sqlite3_mutex_leave(pShmNode->mutex);

  /* If pShmNode->nRef has reached 0, then close the underlying
  ** shared-memory file, too */
  unixEnterMutex();
  assert( pShmNode->nRef>0 );
  pShmNode->nRef--;
  if( pShmNode->nRef==0 ){
    if( deleteFlag ) unlink(pShmNode->zFilename);
    unixShmPurge(pDbFd);
  }
  unixLeaveMutex();

  return SQLITE_OK;
}

/*
................................................................................
static int unixShmSize(
  sqlite3_file *fd,         /* The open database file holding SHM */
  int reqSize,              /* Requested size.  -1 for query only */
  int *pNewSize             /* Write new size here */
){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;
  unixShmNode *pShmNode = p->pShmNode;
  int rc = SQLITE_OK;
  struct stat sStat;

  assert( pShmNode==pDbFd->pInode->pShmNode );
  assert( pShmNode->pInode==pDbFd->pInode );

  /* On a query, this loop runs once.  When reqSize>=0, the loop potentially
  ** runs twice, except if the actual size is already greater than or equal
  ** to the requested size, reqSize is set to -1 on the first iteration and
  ** the loop only runs once.
  */
  while( 1 ){
    if( fstat(pShmNode->h, &sStat)==0 ){
      *pNewSize = (int)sStat.st_size;
      if( reqSize>=0 && reqSize<=(int)sStat.st_size ) break;
    }else{
      *pNewSize = 0;
      rc = SQLITE_IOERR;
      break;
    }
    if( reqSize<0 ) break;
    reqSize = (reqSize + SQLITE_UNIX_SHM_INCR - 1)/SQLITE_UNIX_SHM_INCR;
    reqSize *= SQLITE_UNIX_SHM_INCR;
    rc = ftruncate(pShmNode->h, reqSize);
    reqSize = -1;
  }
  return rc;
}


/*
................................................................................
  sqlite3_file *fd,        /* Database file holding shared memory */
  int reqMapSize,          /* Requested size of mapping. -1 means don't care */
  int *pNewMapSize,        /* Write new size of mapping here */
  void **ppBuf             /* Write mapping buffer origin here */
){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;
  unixShmNode *pShmNode = p->pShmNode;
  int rc = SQLITE_OK;

  assert( pShmNode==pDbFd->pInode->pShmNode );
  assert( pShmNode->pInode==pDbFd->pInode );

  if( p->lockState!=SQLITE_SHM_CHECKPOINT && p->hasMutexBuf==0 ){
    assert( sqlite3_mutex_notheld(pShmNode->mutex) );
    sqlite3_mutex_enter(pShmNode->mutexBuf);
    p->hasMutexBuf = 1;
  }
  sqlite3_mutex_enter(pShmNode->mutex);
  if( pShmNode->szMap==0 || reqMapSize>pShmNode->szMap ){
    int actualSize;
    if( unixShmSize(fd, -1, &actualSize)==SQLITE_OK
     && reqMapSize<actualSize
    ){
      reqMapSize = actualSize;
    }
    if( pShmNode->pMMapBuf ){
      munmap(pShmNode->pMMapBuf, pShmNode->szMap);
    }
    pShmNode->pMMapBuf = mmap(0, reqMapSize, PROT_READ|PROT_WRITE, MAP_SHARED,
                           pShmNode->h, 0);
    pShmNode->szMap = pShmNode->pMMapBuf ? reqMapSize : 0;
  }
  *pNewMapSize = pShmNode->szMap;
  *ppBuf = pShmNode->pMMapBuf;
  sqlite3_mutex_leave(pShmNode->mutex);
  return rc;
}

/*
** Release the lock held on the shared memory segment to that other
** threads are free to resize it if necessary.
**
................................................................................
** is a no-op.
*/
static int unixShmRelease(sqlite3_file *fd){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;

  if( p->hasMutexBuf && p->lockState!=SQLITE_SHM_RECOVER ){
    assert( sqlite3_mutex_notheld(p->pShmNode->mutex) );
    sqlite3_mutex_leave(p->pShmNode->mutexBuf);
    p->hasMutexBuf = 0;
  }
  return SQLITE_OK;
}

/*
** Symbolic names for LOCK states used for debugging.
................................................................................
static int unixShmLock(
  sqlite3_file *fd,          /* Database file holding the shared memory */
  int desiredLock,           /* One of SQLITE_SHM_xxxxx locking states */
  int *pGotLock              /* The lock you actually got */
){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p = pDbFd->pShm;
  unixShmNode *pShmNode = p->pShmNode;
  int rc = SQLITE_PROTOCOL;

  assert( pShmNode==pDbFd->pInode->pShmNode );
  assert( pShmNode->pInode==pDbFd->pInode );

  /* Note that SQLITE_SHM_READ_FULL and SQLITE_SHM_PENDING are never
  ** directly requested; they are side effects from requesting
  ** SQLITE_SHM_READ and SQLITE_SHM_CHECKPOINT, respectively.
  */
  assert( desiredLock==SQLITE_SHM_UNLOCK
       || desiredLock==SQLITE_SHM_READ
................................................................................
    return SQLITE_OK;
  }

  OSTRACE(("SHM-LOCK shmid-%d, pid-%d request %s->%s\n",
            p->id, getpid(), azLkName[p->lockState], azLkName[desiredLock]));
  
  if( desiredLock==SQLITE_SHM_RECOVER && !p->hasMutexBuf ){
    assert( sqlite3_mutex_notheld(pShmNode->mutex) );
    sqlite3_mutex_enter(pShmNode->mutexBuf);
    p->hasMutexBuf = 1;
  }
  sqlite3_mutex_enter(pShmNode->mutex);
  switch( desiredLock ){
    case SQLITE_SHM_UNLOCK: {
      assert( p->lockState!=SQLITE_SHM_RECOVER );
      unixShmUnlock(pShmNode, p, UNIX_SHM_A|UNIX_SHM_B|UNIX_SHM_C|UNIX_SHM_D);
      rc = SQLITE_OK;
      p->lockState = SQLITE_SHM_UNLOCK;
      break;
    }
    case SQLITE_SHM_READ: {
      if( p->lockState==SQLITE_SHM_UNLOCK ){
        int nAttempt;
        rc = SQLITE_BUSY;
        assert( p->lockState==SQLITE_SHM_UNLOCK );
        for(nAttempt=0; nAttempt<5 && rc==SQLITE_BUSY; nAttempt++){
          rc = unixShmSharedLock(pShmNode, p, UNIX_SHM_A|UNIX_SHM_B);
          if( rc==SQLITE_BUSY ){
            rc = unixShmSharedLock(pShmNode, p, UNIX_SHM_D);
            if( rc==SQLITE_OK ){
              p->lockState = SQLITE_SHM_READ_FULL;
            }
          }else{
            unixShmUnlock(pShmNode, p, UNIX_SHM_B);
            p->lockState = SQLITE_SHM_READ;
          }
        }
      }else{
       assert( p->lockState==SQLITE_SHM_WRITE
               || p->lockState==SQLITE_SHM_RECOVER );
        rc = unixShmSharedLock(pShmNode, p, UNIX_SHM_A);
        unixShmUnlock(pShmNode, p, UNIX_SHM_C|UNIX_SHM_D);
        p->lockState = SQLITE_SHM_READ;
      }
      break;
    }
    case SQLITE_SHM_WRITE: {
      assert( p->lockState==SQLITE_SHM_READ 
              || p->lockState==SQLITE_SHM_READ_FULL );
      rc = unixShmExclusiveLock(pShmNode, p, UNIX_SHM_C|UNIX_SHM_D);
      if( rc==SQLITE_OK ){
        p->lockState = SQLITE_SHM_WRITE;
      }
      break;
    }
    case SQLITE_SHM_CHECKPOINT: {
      assert( p->lockState==SQLITE_SHM_UNLOCK
           || p->lockState==SQLITE_SHM_PENDING
      );
      if( p->lockState==SQLITE_SHM_UNLOCK ){
        rc = unixShmExclusiveLock(pShmNode, p, UNIX_SHM_B|UNIX_SHM_C);
        if( rc==SQLITE_OK ){
          p->lockState = SQLITE_SHM_PENDING;
        }
      }
      if( p->lockState==SQLITE_SHM_PENDING ){
        rc = unixShmExclusiveLock(pShmNode, p, UNIX_SHM_A);
        if( rc==SQLITE_OK ){
          p->lockState = SQLITE_SHM_CHECKPOINT;
        }
      }
      break;
    }
    default: {
      assert( desiredLock==SQLITE_SHM_RECOVER );
      assert( p->lockState==SQLITE_SHM_READ
           || p->lockState==SQLITE_SHM_READ_FULL
      );
      assert( sqlite3_mutex_held(pShmNode->mutexBuf) );
      rc = unixShmExclusiveLock(pShmNode, p, UNIX_SHM_C);
      if( rc==SQLITE_OK ){
        p->lockState = SQLITE_SHM_RECOVER;
      }
      break;
    }
  }
  sqlite3_mutex_leave(pShmNode->mutex);
  OSTRACE(("SHM-LOCK shmid-%d, pid-%d got %s\n",
           p->id, getpid(), azLkName[p->lockState]));
  if( pGotLock ) *pGotLock = p->lockState;
  return rc;
}

#else
................................................................................
  ** For this reason, if an error occurs in the stat() call here, it is
  ** ignored and -1 is returned. The caller will try to open a new file
  ** descriptor on the same path, fail, and return an error to SQLite.
  **
  ** Even if a subsequent open() call does succeed, the consequences of
  ** not searching for a resusable file descriptor are not dire.  */
  if( 0==stat(zPath, &sStat) ){
    unixInodeInfo *pInode;

    unixEnterMutex();
    pInode = inodeList;
    while( pInode && (pInode->fileId.dev!=sStat.st_dev
                     || pInode->fileId.ino!=sStat.st_ino) ){
       pInode = pInode->pNext;
    }