/ Check-in [8b1e0010]
Login

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

Overview
Comment:Take extra precautions to ensure access to unixInodeInfo.pUnused is protected by all necessary mutexes.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA3-256:8b1e0010b9e0b548a90087f4d25843d2b40f7e9551722ac587fa925d37b510c2
User & Date: drh 2018-08-13 20:46:18
Context
2018-08-13
22:50
Stop requiring the global VFS mutex to access the unixInodeInfo.pUnused field. The unixInodeInfo mutex is sufficient. check-in: e3ea43da user: drh tags: trunk
20:46
Take extra precautions to ensure access to unixInodeInfo.pUnused is protected by all necessary mutexes. check-in: 8b1e0010 user: drh tags: trunk
11:32
Fix an incorrect comment on the unix-nolock VFS object. No functional code changes. check-in: 90f7c193 user: drh tags: trunk
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/os_unix.c.

698
699
700
701
702
703
704











705
706
707

708
709
710

711
712
713
714
715
716
717
....
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121




1122
1123




1124
1125
1126
1127
1128
1129
1130
1131

1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
....
1143
1144
1145
1146
1147
1148
1149
1150















1151
1152
1153
1154
1155
1156
1157
....
1249
1250
1251
1252
1253
1254
1255



1256
1257
1258
1259
1260
1261
1262

1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273

1274
1275
1276
1277
1278
1279
1280
....
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
....
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
....
1822
1823
1824
1825
1826
1827
1828



1829
1830

1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
....
1847
1848
1849
1850
1851
1852
1853

1854
1855
1856
1857
1858
1859
1860
....
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997

1998






1999
2000
2001
2002
2003
2004
2005
....
2064
2065
2066
2067
2068
2069
2070

2071
2072
2073
2074
2075
2076
2077
....
2675
2676
2677
2678
2679
2680
2681

2682
2683
2684
2685
2686
2687
2688
....
3032
3033
3034
3035
3036
3037
3038

3039
3040
3041
3042
3043
3044
3045
....
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
....
4447
4448
4449
4450
4451
4452
4453

4454
4455
4456
4457
4458
4459
4460
....
4829
4830
4831
4832
4833
4834
4835

4836
4837
4838
4839
4840
4841
4842
....
4870
4871
4872
4873
4874
4875
4876

4877
4878
4879
4880
4881
4882
4883
....
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708


5709
5710
5711
5712
5713
5714

5715
5716
5717
5718
5719
5720
5721
** Function unixMutexHeld() is used to assert() that the global mutex 
** is held when required. This function is only used as part of assert() 
** statements. e.g.
**
**   unixEnterMutex()
**     assert( unixMutexHeld() );
**   unixEnterLeave()











*/
static sqlite3_mutex *unixBigLock = 0;
static void unixEnterMutex(void){

  sqlite3_mutex_enter(unixBigLock);
}
static void unixLeaveMutex(void){

  sqlite3_mutex_leave(unixBigLock);
}
#ifdef SQLITE_DEBUG
static int unixMutexHeld(void) {
  return sqlite3_mutex_held(unixBigLock);
}
#endif
................................................................................
**
** A single inode can have multiple file descriptors, so each unixFile
** structure contains a pointer to an instance of this object and this
** object keeps a count of the number of unixFile pointing to it.
**
** Mutex rules:
**
**  (1) The pLockMutex mutex must be held in order to read or write
**      any of the locking fields:
**          nShared, nLock, eFileLock, or bProcessLock
**
**  (2) When nRef>0, then the following fields are unchanging and can
**      be read (but not written) without holding any mutex:
**          fileId, pLockMutex
**




**  (3) With the exceptions above, all the fields may only be read
**      or written while holding the global unixBigLock mutex.




*/
struct unixInodeInfo {
  struct unixFileId fileId;       /* The lookup key */
  sqlite3_mutex *pLockMutex;      /* Hold this mutex for... */
  int nShared;                      /* Number of SHARED locks held */
  int nLock;                        /* Number of outstanding file locks */
  unsigned char eFileLock;          /* One of SHARED_LOCK, RESERVED_LOCK etc. */
  unsigned char bProcessLock;       /* An exclusive process lock is held */

  int nRef;                       /* Number of pointers to this structure */
  unixShmNode *pShmNode;          /* Shared memory associated with this inode */
  UnixUnusedFd *pUnused;          /* Unused file descriptors to close */
  unixInodeInfo *pNext;           /* List of all unixInodeInfo objects */
  unixInodeInfo *pPrev;           /*    .... doubly linked */
#if SQLITE_ENABLE_LOCKING_STYLE
  unsigned long long sharedByte;  /* for AFP simulated shared lock */
#endif
#if OS_VXWORKS
  sem_t *pSem;                    /* Named POSIX semaphore */
................................................................................
#endif
};

/*
** A lists of all unixInodeInfo objects.
*/
static unixInodeInfo *inodeList = 0;  /* All unixInodeInfo objects */
static unsigned int nUnusedFd = 0;    /* Total unused file descriptors */
















/*
**
** This function - unixLogErrorAtLine(), is only ever called via the macro
** unixLogError().
**
** It is invoked after an error occurs in an OS function and errno has been
................................................................................
/*
** Close all file descriptors accumuated in the unixInodeInfo->pUnused list.
*/ 
static void closePendingFds(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *p;
  UnixUnusedFd *pNext;



  for(p=pInode->pUnused; p; p=pNext){
    pNext = p->pNext;
    robust_close(pFile, p->fd, __LINE__);
    sqlite3_free(p);
    nUnusedFd--;
  }
  pInode->pUnused = 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(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  assert( unixMutexHeld() );

  if( ALWAYS(pInode) ){
    pInode->nRef--;
    if( pInode->nRef==0 ){
      assert( pInode->pShmNode==0 );
      closePendingFds(pFile);
      if( pInode->pPrev ){
        assert( pInode->pPrev->pNext==pInode );
................................................................................
        assert( pInode->pNext->pPrev==pInode );
        pInode->pNext->pPrev = pInode->pPrev;
      }
      sqlite3_mutex_free(pInode->pLockMutex);
      sqlite3_free(pInode);
    }
  }
  assert( inodeList!=0 || nUnusedFd==0 );
}

/*
** Given a file descriptor, locate the unixInodeInfo object that
** describes that file descriptor.  Create a new one if necessary.  The
** return value might be uninitialized if an error occurs.
**
................................................................................
  memset(&fileId, 0, sizeof(fileId));
  fileId.dev = statbuf.st_dev;
#if OS_VXWORKS
  fileId.pId = pFile->pId;
#else
  fileId.ino = (u64)statbuf.st_ino;
#endif
  assert( inodeList!=0 || nUnusedFd==0 );
  pInode = inodeList;
  while( pInode && memcmp(&fileId, &pInode->fileId, sizeof(fileId)) ){
    pInode = pInode->pNext;
  }
  if( pInode==0 ){
    pInode = sqlite3_malloc64( sizeof(*pInode) );
    if( pInode==0 ){
................................................................................
/*
** 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->pPreallocatedUnused;



  p->pNext = pInode->pUnused;
  pInode->pUnused = p;

  pFile->h = -1;
  pFile->pPreallocatedUnused = 0;
  nUnusedFd++;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
................................................................................
** 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;


  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (unix)\n", pFile->h, eFileLock,
      pFile->eFileLock, pFile->pInode->eFileLock, pFile->pInode->nShared,
      osGetpid(0)));

  assert( eFileLock<=SHARED_LOCK );
................................................................................

    /* Decrement the count of locks against this same file.  When the
    ** count reaches zero, close any other file descriptors whose close
    ** was deferred because of outstanding locks.
    */
    pInode->nLock--;
    assert( pInode->nLock>=0 );
    if( pInode->nLock==0 ){
      closePendingFds(pFile);
    }
  }

end_unlock:
  sqlite3_mutex_leave(pInode->pLockMutex);

  if( rc==SQLITE_OK ) pFile->eFileLock = eFileLock;






  return rc;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
................................................................................
** Close a file.
*/
static int unixClose(sqlite3_file *id){
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile *)id;
  verifyDbFile(pFile);
  unixUnlock(id, NO_LOCK);

  unixEnterMutex();

  /* unixFile.pInode is always valid here. Otherwise, a different close
  ** routine (e.g. nolockClose()) would be called instead.
  */
  assert( pFile->pInode->nLock>0 || pFile->pInode->bProcessLock==0 );
  if( ALWAYS(pFile->pInode) && pFile->pInode->nLock ){
................................................................................
 ** Close a file.
 */
static int semXClose(sqlite3_file *id) {
  if( id ){
    unixFile *pFile = (unixFile*)id;
    semXUnlock(id, NO_LOCK);
    assert( pFile );

    unixEnterMutex();
    releaseInodeInfo(pFile);
    unixLeaveMutex();
    closeUnixFile(id);
  }
  return SQLITE_OK;
}
................................................................................
*/
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 );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (afp)\n", pFile->h, eFileLock,
           pFile->eFileLock, pFile->pInode->eFileLock, pFile->pInode->nShared,
................................................................................
        pInode->eFileLock = NO_LOCK;
        pFile->eFileLock = NO_LOCK;
      }
    }
    if( rc==SQLITE_OK ){
      pInode->nLock--;
      assert( pInode->nLock>=0 );
      if( pInode->nLock==0 ){
        closePendingFds(pFile);
      }
    }
  }
  
  sqlite3_mutex_leave(pInode->pLockMutex);

  if( rc==SQLITE_OK ) pFile->eFileLock = eFileLock;






  return rc;
}

/*
** Close a file & cleanup AFP specific locking context 
*/
static int afpClose(sqlite3_file *id) {
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  assert( id!=0 );
  afpUnlock(id, NO_LOCK);

  unixEnterMutex();
  if( pFile->pInode && pFile->pInode->nLock ){
    /* If there are outstanding locks, do not actually close the file just
    ** yet because that would clear those locks.  Instead, add the file
    ** descriptor to pInode->aPending.  It will be automatically closed when
    ** the last lock is cleared.
    */
................................................................................
  if( p==0 ) return SQLITE_NOMEM_BKPT;
  memset(p, 0, sizeof(*p));
  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();
  pInode = pDbFd->pInode;
  pShmNode = pInode->pShmNode;
  if( pShmNode==0 ){
    struct stat sStat;                 /* fstat() info for database file */
#ifndef SQLITE_SHM_DIRECTORY
    const char *zBasePath = pDbFd->zPath;
................................................................................
** any load or store begun after the barrier.
*/
static void unixShmBarrier(
  sqlite3_file *fd                /* Database file holding the shared memory */
){
  UNUSED_PARAMETER(fd);
  sqlite3MemoryBarrier();         /* compiler-defined memory barrier */

  unixEnterMutex();               /* Also mutex, for redundancy */
  unixLeaveMutex();
}

/*
** Close a connection to shared-memory.  Delete the underlying 
** storage if deleteFlag is true.
................................................................................
  /* 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 && pShmNode->h>=0 ){
      osUnlink(pShmNode->zFilename);
    }
................................................................................
  ** almost certain that an open() call on the same path will also fail.
  ** 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 reusable file descriptor are not dire.  */
  if( nUnusedFd>0 && 0==osStat(zPath, &sStat) ){
    unixInodeInfo *pInode;

    pInode = inodeList;
    while( pInode && (pInode->fileId.dev!=sStat.st_dev
                     || pInode->fileId.ino!=(u64)sStat.st_ino) ){
       pInode = pInode->pNext;
    }
    if( pInode ){
      UnixUnusedFd **pp;


      for(pp=&pInode->pUnused; *pp && (*pp)->flags!=flags; pp=&((*pp)->pNext));
      pUnused = *pp;
      if( pUnused ){
        nUnusedFd--;
        *pp = pUnused->pNext;
      }

    }
  }
  unixLeaveMutex();
#endif    /* if !OS_VXWORKS */
  return pUnused;
}








>
>
>
>
>
>
>
>
>
>
>



>



>







 







|







>
>
>
>
|

>
>
>
>








>


<







 







<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







 







>
>
>




<


>











>







 







<







 







<







 







>
>
>


>


<







 







>







 







|
|





>
|
>
>
>
>
>
>







 







>







 







>







 







>







 







|
<
<




>
|
>
>
>
>
>
>











>







 







>







 







>







 







>







 







|









>
>



<


>







698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
....
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155

1156
1157
1158
1159
1160
1161
1162
....
1164
1165
1166
1167
1168
1169
1170

1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
....
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297

1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
....
1326
1327
1328
1329
1330
1331
1332

1333
1334
1335
1336
1337
1338
1339
....
1395
1396
1397
1398
1399
1400
1401

1402
1403
1404
1405
1406
1407
1408
....
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873

1874
1875
1876
1877
1878
1879
1880
....
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
....
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
....
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
....
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
....
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
....
3166
3167
3168
3169
3170
3171
3172
3173


3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
....
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
....
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
....
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
....
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773

5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
** Function unixMutexHeld() is used to assert() that the global mutex 
** is held when required. This function is only used as part of assert() 
** statements. e.g.
**
**   unixEnterMutex()
**     assert( unixMutexHeld() );
**   unixEnterLeave()
**
** To prevent deadlock, the global unixBigLock must must be acquired
** before the unixInodeInfo.pLockMutex mutex, if both are held.  It is
** OK to get the pLockMutex without holding unixBigLock first, but if
** that happens, the unixBigLock mutex must not be acquired until after
** pLockMutex is released.
**
**      OK:     enter(unixBigLock),  enter(pLockInfo)
**      OK:     enter(unixBigLock)
**      OK:     enter(pLockInfo)
**   ERROR:     enter(pLockInfo), enter(unixBigLock)
*/
static sqlite3_mutex *unixBigLock = 0;
static void unixEnterMutex(void){
  assert( sqlite3_mutex_notheld(unixBigLock) );  /* Not a recursive mutex */
  sqlite3_mutex_enter(unixBigLock);
}
static void unixLeaveMutex(void){
  assert( sqlite3_mutex_held(unixBigLock) );
  sqlite3_mutex_leave(unixBigLock);
}
#ifdef SQLITE_DEBUG
static int unixMutexHeld(void) {
  return sqlite3_mutex_held(unixBigLock);
}
#endif
................................................................................
**
** A single inode can have multiple file descriptors, so each unixFile
** structure contains a pointer to an instance of this object and this
** object keeps a count of the number of unixFile pointing to it.
**
** Mutex rules:
**
**  (1) Only the pLockMutex mutex must be held in order to read or write
**      any of the locking fields:
**          nShared, nLock, eFileLock, or bProcessLock
**
**  (2) When nRef>0, then the following fields are unchanging and can
**      be read (but not written) without holding any mutex:
**          fileId, pLockMutex
**
**  (3) The pUnused field may only be changed while holding bo the
**      pLockMutex and the bigUnixLock mutex.  But it may be read
**      while holding either.
**
**  (4) With the exceptions above, all the fields may only be read
**      or written while holding the global unixBigLock mutex.
**
** Deadlock prevention:  The global unixBigLock mutex may not
** be acquired while holding the pLockMutex mutex.  If both unixBigLock
** and pLockMutex are needed, then unixBigLock must be acquired first.
*/
struct unixInodeInfo {
  struct unixFileId fileId;       /* The lookup key */
  sqlite3_mutex *pLockMutex;      /* Hold this mutex for... */
  int nShared;                      /* Number of SHARED locks held */
  int nLock;                        /* Number of outstanding file locks */
  unsigned char eFileLock;          /* One of SHARED_LOCK, RESERVED_LOCK etc. */
  unsigned char bProcessLock;       /* An exclusive process lock is held */
  UnixUnusedFd *pUnused;          /* Unused file descriptors to close */
  int nRef;                       /* Number of pointers to this structure */
  unixShmNode *pShmNode;          /* Shared memory associated with this inode */

  unixInodeInfo *pNext;           /* List of all unixInodeInfo objects */
  unixInodeInfo *pPrev;           /*    .... doubly linked */
#if SQLITE_ENABLE_LOCKING_STYLE
  unsigned long long sharedByte;  /* for AFP simulated shared lock */
#endif
#if OS_VXWORKS
  sem_t *pSem;                    /* Named POSIX semaphore */
................................................................................
#endif
};

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


#ifdef SQLITE_DEBUG
/*
** True if the inode mutex is held, or not.  Used only within assert()
** to help verify correct mutex usage.
*/
int unixFileMutexHeld(unixFile *pFile){
  assert( pFile->pInode );
  return sqlite3_mutex_held(pFile->pInode->pLockMutex);
}
int unixFileMutexNotheld(unixFile *pFile){
  assert( pFile->pInode );
  return sqlite3_mutex_notheld(pFile->pInode->pLockMutex);
}
#endif

/*
**
** This function - unixLogErrorAtLine(), is only ever called via the macro
** unixLogError().
**
** It is invoked after an error occurs in an OS function and errno has been
................................................................................
/*
** Close all file descriptors accumuated in the unixInodeInfo->pUnused list.
*/ 
static void closePendingFds(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *p;
  UnixUnusedFd *pNext;
  assert( unixMutexHeld() );
  assert( unixFileMutexNotheld(pFile) );
  sqlite3_mutex_enter(pInode->pLockMutex);
  for(p=pInode->pUnused; p; p=pNext){
    pNext = p->pNext;
    robust_close(pFile, p->fd, __LINE__);
    sqlite3_free(p);

  }
  pInode->pUnused = 0;
  sqlite3_mutex_leave(pInode->pLockMutex);
}

/*
** 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(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  assert( unixMutexHeld() );
  assert( unixFileMutexNotheld(pFile) );
  if( ALWAYS(pInode) ){
    pInode->nRef--;
    if( pInode->nRef==0 ){
      assert( pInode->pShmNode==0 );
      closePendingFds(pFile);
      if( pInode->pPrev ){
        assert( pInode->pPrev->pNext==pInode );
................................................................................
        assert( pInode->pNext->pPrev==pInode );
        pInode->pNext->pPrev = pInode->pPrev;
      }
      sqlite3_mutex_free(pInode->pLockMutex);
      sqlite3_free(pInode);
    }
  }

}

/*
** Given a file descriptor, locate the unixInodeInfo object that
** describes that file descriptor.  Create a new one if necessary.  The
** return value might be uninitialized if an error occurs.
**
................................................................................
  memset(&fileId, 0, sizeof(fileId));
  fileId.dev = statbuf.st_dev;
#if OS_VXWORKS
  fileId.pId = pFile->pId;
#else
  fileId.ino = (u64)statbuf.st_ino;
#endif

  pInode = inodeList;
  while( pInode && memcmp(&fileId, &pInode->fileId, sizeof(fileId)) ){
    pInode = pInode->pNext;
  }
  if( pInode==0 ){
    pInode = sqlite3_malloc64( sizeof(*pInode) );
    if( pInode==0 ){
................................................................................
/*
** 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->pPreallocatedUnused;
  assert( unixMutexHeld() );
  assert( unixFileMutexNotheld(pFile) );
  sqlite3_mutex_enter(pInode->pLockMutex);
  p->pNext = pInode->pUnused;
  pInode->pUnused = p;
  sqlite3_mutex_leave(pInode->pLockMutex);
  pFile->h = -1;
  pFile->pPreallocatedUnused = 0;

}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
................................................................................
** 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 wantToClosePending = 0;  /* True to try to close file old descriptors */

  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (unix)\n", pFile->h, eFileLock,
      pFile->eFileLock, pFile->pInode->eFileLock, pFile->pInode->nShared,
      osGetpid(0)));

  assert( eFileLock<=SHARED_LOCK );
................................................................................

    /* Decrement the count of locks against this same file.  When the
    ** count reaches zero, close any other file descriptors whose close
    ** was deferred because of outstanding locks.
    */
    pInode->nLock--;
    assert( pInode->nLock>=0 );
    if( pInode->nLock==0 && pInode->pUnused!=0 ){
      wantToClosePending = 1;
    }
  }

end_unlock:
  sqlite3_mutex_leave(pInode->pLockMutex);
  if( rc==SQLITE_OK ){
    pFile->eFileLock = eFileLock;
    if( wantToClosePending ){
      unixEnterMutex();
      if( pInode->nLock==0 ) closePendingFds(pFile);
      unixLeaveMutex();
    }
  }
  return rc;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
................................................................................
** Close a file.
*/
static int unixClose(sqlite3_file *id){
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile *)id;
  verifyDbFile(pFile);
  unixUnlock(id, NO_LOCK);
  assert( unixFileMutexNotheld(pFile) );
  unixEnterMutex();

  /* unixFile.pInode is always valid here. Otherwise, a different close
  ** routine (e.g. nolockClose()) would be called instead.
  */
  assert( pFile->pInode->nLock>0 || pFile->pInode->bProcessLock==0 );
  if( ALWAYS(pFile->pInode) && pFile->pInode->nLock ){
................................................................................
 ** Close a file.
 */
static int semXClose(sqlite3_file *id) {
  if( id ){
    unixFile *pFile = (unixFile*)id;
    semXUnlock(id, NO_LOCK);
    assert( pFile );
    assert( unixFileMutexNotheld(pFile) );
    unixEnterMutex();
    releaseInodeInfo(pFile);
    unixLeaveMutex();
    closeUnixFile(id);
  }
  return SQLITE_OK;
}
................................................................................
*/
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;
  int wantToClosePending = 0;
#ifdef SQLITE_TEST
  int h = pFile->h;
#endif

  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (afp)\n", pFile->h, eFileLock,
           pFile->eFileLock, pFile->pInode->eFileLock, pFile->pInode->nShared,
................................................................................
        pInode->eFileLock = NO_LOCK;
        pFile->eFileLock = NO_LOCK;
      }
    }
    if( rc==SQLITE_OK ){
      pInode->nLock--;
      assert( pInode->nLock>=0 );
      if( pInode->nLock==0 && pInode->pUnused!=0 ) wantToClosePending = 1;


    }
  }
  
  sqlite3_mutex_leave(pInode->pLockMutex);
  if( rc==SQLITE_OK ){
    pFile->eFileLock = eFileLock;
    if( wantToClosePending ){
      unixEnterMutex();
      if( pInode->nLock==0 ) closePendingFds(pFile);
      unixLeaveMutex();
    }
  }
  return rc;
}

/*
** Close a file & cleanup AFP specific locking context 
*/
static int afpClose(sqlite3_file *id) {
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  assert( id!=0 );
  afpUnlock(id, NO_LOCK);
  assert( unixFileMutexNotheld(pFile) );
  unixEnterMutex();
  if( pFile->pInode && pFile->pInode->nLock ){
    /* If there are outstanding locks, do not actually close the file just
    ** yet because that would clear those locks.  Instead, add the file
    ** descriptor to pInode->aPending.  It will be automatically closed when
    ** the last lock is cleared.
    */
................................................................................
  if( p==0 ) return SQLITE_NOMEM_BKPT;
  memset(p, 0, sizeof(*p));
  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.
  */
  assert( unixFileMutexNotheld(pDbFd) );
  unixEnterMutex();
  pInode = pDbFd->pInode;
  pShmNode = pInode->pShmNode;
  if( pShmNode==0 ){
    struct stat sStat;                 /* fstat() info for database file */
#ifndef SQLITE_SHM_DIRECTORY
    const char *zBasePath = pDbFd->zPath;
................................................................................
** any load or store begun after the barrier.
*/
static void unixShmBarrier(
  sqlite3_file *fd                /* Database file holding the shared memory */
){
  UNUSED_PARAMETER(fd);
  sqlite3MemoryBarrier();         /* compiler-defined memory barrier */
  assert( unixFileMutexNotheld((unixFile*)fd) );
  unixEnterMutex();               /* Also mutex, for redundancy */
  unixLeaveMutex();
}

/*
** Close a connection to shared-memory.  Delete the underlying 
** storage if deleteFlag is true.
................................................................................
  /* 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 */
  assert( unixFileMutexNotheld(pDbFd) );
  unixEnterMutex();
  assert( pShmNode->nRef>0 );
  pShmNode->nRef--;
  if( pShmNode->nRef==0 ){
    if( deleteFlag && pShmNode->h>=0 ){
      osUnlink(pShmNode->zFilename);
    }
................................................................................
  ** almost certain that an open() call on the same path will also fail.
  ** 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 reusable file descriptor are not dire.  */
  if( inodeList!=0 && 0==osStat(zPath, &sStat) ){
    unixInodeInfo *pInode;

    pInode = inodeList;
    while( pInode && (pInode->fileId.dev!=sStat.st_dev
                     || pInode->fileId.ino!=(u64)sStat.st_ino) ){
       pInode = pInode->pNext;
    }
    if( pInode ){
      UnixUnusedFd **pp;
      assert( sqlite3_mutex_notheld(pInode->pLockMutex) );
      sqlite3_mutex_enter(pInode->pLockMutex);
      for(pp=&pInode->pUnused; *pp && (*pp)->flags!=flags; pp=&((*pp)->pNext));
      pUnused = *pp;
      if( pUnused ){

        *pp = pUnused->pNext;
      }
      sqlite3_mutex_leave(pInode->pLockMutex);
    }
  }
  unixLeaveMutex();
#endif    /* if !OS_VXWORKS */
  return pUnused;
}