SQLite4  Check-in [45e4472618]

  if( strcmp(azArg[0], "-")==0 ){
    pInput = stdin;
  }else{
    pClose = pInput = fopen(azArg[0], "r");
  }
  zDb = azArg[1];
  pEnv = tdb_lsm_env();
  rc = pEnv->xOpen(pEnv, zDb, 0, &pOut);
  if( rc!=LSM_OK ) return rc;

  while( feof(pInput)==0 ){
    char zLine[80];
    fgets(zLine, sizeof(zLine)-1, pInput);
    zLine[sizeof(zLine)-1] = '\0';

  lsm_env *pRealEnv = tdb_lsm_env();
  return pRealEnv->xFullpath(pRealEnv, zFile, zOut, pnOut);
}

static int testEnvOpen(
  lsm_env *pEnv,                  /* Environment for current LsmDb */
  const char *zFile,              /* Name of file to open */
  int flags,
  lsm_file **ppFile               /* OUT: New file handle object */
){
  lsm_env *pRealEnv = tdb_lsm_env();
  LsmDb *pDb = (LsmDb *)pEnv->pVfsCtx;
  int rc;                         /* Return Code */
  LsmFile *pRet;                  /* The new file handle */
  int nFile;                      /* Length of string zFile in bytes */

  nFile = strlen(zFile);
  pRet = (LsmFile *)testMalloc(sizeof(LsmFile));
  pRet->pDb = pDb;
  pRet->bLog = (nFile > 4 && 0==memcmp("-log", &zFile[nFile-4], 4));

  rc = pRealEnv->xOpen(pRealEnv, zFile, flags, &pRet->pReal);
  if( rc!=LSM_OK ){
    testFree(pRet);
    pRet = 0;
  }

  *ppFile = (lsm_file *)pRet;
  return rc;

  char *zFile = pDb->zName;
  char *zFree = 0;

  for(iFile=0; iFile<2; iFile++){
    lsm_file *pFile = 0;
    int i;

    pEnv->xOpen(pEnv, zFile, 0, &pFile);
    for(i=0; i<pDb->aFile[iFile].nSector; i++){
      u8 *aOld = pDb->aFile[iFile].aSector[i].aOld;
      if( aOld ){
        int iOpt = testPrngValue(iSeed++) % 3;
        switch( iOpt ){
          case 0:
            break;

typedef long long int lsm_i64;              /* 64-bit signed integer type */

/* Candidate values for the 3rd argument to lsm_env.xLock() */
#define LSM_LOCK_UNLOCK 0
#define LSM_LOCK_SHARED 1
#define LSM_LOCK_EXCL   2

/* Flags for lsm_env.xOpen() */
#define LSM_OPEN_READONLY 0x0001

/*
** CAPI: Database Runtime Environment
**
** Run-time environment used by LSM
*/
struct lsm_env {
  int nByte;                 /* Size of this structure in bytes */
  int iVersion;              /* Version number of this structure (1) */
  /****** file i/o ***********************************************/
  void *pVfsCtx;
  int (*xFullpath)(lsm_env*, const char *, char *, int *);
  int (*xOpen)(lsm_env*, const char *, int flags, lsm_file **);
  int (*xRead)(lsm_file *, lsm_i64, void *, int);
  int (*xWrite)(lsm_file *, lsm_i64, void *, int);
  int (*xTruncate)(lsm_file *, lsm_i64);
  int (*xSync)(lsm_file *);
  int (*xSectorSize)(lsm_file *);
  int (*xRemap)(lsm_file *, lsm_i64, void **, lsm_i64*);
  int (*xFileid)(lsm_file *, void *pBuf, int *pnBuf);

/*
** CAPI: LSM Error Codes
*/
#define LSM_OK         0
#define LSM_ERROR      1
#define LSM_BUSY       5
#define LSM_NOMEM      7
#define LSM_READONLY   8
#define LSM_IOERR     10
#define LSM_CORRUPT   11
#define LSM_FULL      13
#define LSM_CANTOPEN  14
#define LSM_PROTOCOL  15
#define LSM_MISUSE    21

#define LSM_MISMATCH  50


#define LSM_IOERR_NOENT (LSM_IOERR | (1<<8))

/* 
** CAPI: Creating and Destroying Database Connection Handles
**
** Open and close a database connection handle.
*/
int lsm_new(lsm_env*, lsm_db **ppDb);
int lsm_close(lsm_db *pDb);

** LSM_CONFIG_GET_COMPRESSION:
**   Query the compression methods used to compress and decompress database
**   content.
**
** LSM_CONFIG_SET_COMPRESSION_FACTORY:
**   Configure a factory method to be invoked in case of an LSM_MISMATCH
**   error.
**
** LSM_CONFIG_READONLY:
**   A read/write boolean parameter. This parameter may only be set before
**   lsm_open() is called.
*/
#define LSM_CONFIG_AUTOFLUSH                1
#define LSM_CONFIG_PAGE_SIZE                2
#define LSM_CONFIG_SAFETY                   3
#define LSM_CONFIG_BLOCK_SIZE               4
#define LSM_CONFIG_AUTOWORK                 5
#define LSM_CONFIG_MMAP                     7
#define LSM_CONFIG_USE_LOG                  8
#define LSM_CONFIG_AUTOMERGE                9
#define LSM_CONFIG_MAX_FREELIST            10
#define LSM_CONFIG_MULTIPLE_PROCESSES      11
#define LSM_CONFIG_AUTOCHECKPOINT          12
#define LSM_CONFIG_SET_COMPRESSION         13
#define LSM_CONFIG_GET_COMPRESSION         14
#define LSM_CONFIG_SET_COMPRESSION_FACTORY 15
#define LSM_CONFIG_READONLY                16

#define LSM_SAFETY_OFF    0
#define LSM_SAFETY_NORMAL 1
#define LSM_SAFETY_FULL   2

/*
** CAPI: Compression and/or Encryption Hooks

#ifdef LSM_DEBUG
int lsmErrorBkpt(int);
#else
# define lsmErrorBkpt(x) (x)
#endif

#define LSM_PROTOCOL_BKPT lsmErrorBkpt(LSM_PROTOCOL)
#define LSM_IOERR_BKPT    lsmErrorBkpt(LSM_IOERR)
#define LSM_NOMEM_BKPT    lsmErrorBkpt(LSM_NOMEM)
#define LSM_CORRUPT_BKPT  lsmErrorBkpt(LSM_CORRUPT)
#define LSM_MISUSE_BKPT   lsmErrorBkpt(LSM_MISUSE)

#define unused_parameter(x) (void)(x)
#define array_size(x) (sizeof(x)/sizeof(x[0]))


/* The size of each shared-memory chunk */
#define LSM_SHM_CHUNK_SIZE (32*1024)

/* The number of bytes reserved at the start of each shm chunk for MM. */
#define LSM_SHM_CHUNK_HDR  (sizeof(ShmChunk))

/* The number of available read locks. */
#define LSM_LOCK_NREADER   6

/* The number of available read-write client locks. */
#define LSM_LOCK_NRWCLIENT   16

/* Lock definitions. */
#define LSM_LOCK_DMS1         1
#define LSM_LOCK_DMS2         2

#define LSM_LOCK_WRITER       3
#define LSM_LOCK_WORKER       4
#define LSM_LOCK_CHECKPOINTER 5
#define LSM_LOCK_READER(i)    ((i) + LSM_LOCK_CHECKPOINTER + 1)
#define LSM_LOCK_RWCLIENT(i)  ((i) + LSM_LOCK_READER(LSM_LOCK_NREADER))

/*
** Hard limit on the number of free-list entries that may be stored in 
** a checkpoint (the remainder are stored as a system record in the LSM).
** See also LSM_CONFIG_MAX_FREELIST.

  int bUseLog;                    /* Configured by LSM_CONFIG_USE_LOG */
  int nDfltPgsz;                  /* Configured by LSM_CONFIG_PAGE_SIZE */
  int nDfltBlksz;                 /* Configured by LSM_CONFIG_BLOCK_SIZE */
  int nMaxFreelist;               /* Configured by LSM_CONFIG_MAX_FREELIST */
  int bMmap;                      /* Configured by LSM_CONFIG_MMAP */
  i64 nAutockpt;                  /* Configured by LSM_CONFIG_AUTOCHECKPOINT */
  int bMultiProc;                 /* Configured by L_C_MULTIPLE_PROCESSES */
  int bReadonly;                  /* Configured by LSM_CONFIG_READONLY */
  lsm_compress compress;          /* Compression callbacks */
  lsm_compress_factory factory;   /* Compression callback factory */

  /* Sub-system handles */
  FileSystem *pFS;                /* On-disk portion of database */
  Database *pDatabase;            /* Database shared data */

  int iRwclient;                  /* Read-write client lock held (-1 == none) */

  /* Client transaction context */
  Snapshot *pClient;              /* Client snapshot */
  int iReader;                    /* Read lock held (-1 == unlocked) */
  int bRoTrans;                   /* True if a read-only db trans is open */
  MultiCursor *pCsr;              /* List of all open cursors */
  LogWriter *pLogWriter;          /* Context for writing to the log file */
  int nTransOpen;                 /* Number of opened write transactions */
  int nTransAlloc;                /* Allocated size of aTrans[] array */
  TransMark *aTrans;              /* Array of marks for transaction rollback */
  IntArray rollback;              /* List of tree-nodes to roll back */
  int bDiscardOld;                /* True if lsmTreeDiscardOld() was called */

int lsmMutexHeld(lsm_env *, lsm_mutex *);
int lsmMutexNotHeld(lsm_env *, lsm_mutex *);
#endif

/**************************************************************************
** Start of functions from "lsm_file.c".
*/
int lsmFsOpen(lsm_db *, const char *, int);
int lsmFsOpenLog(lsm_db *, int *);
void lsmFsCloseLog(lsm_db *);
void lsmFsClose(FileSystem *);

int lsmFsConfigure(lsm_db *db);

int lsmFsBlockSize(FileSystem *);
void lsmFsSetBlockSize(FileSystem *, int);

void lsmFsFlushWaiting(FileSystem *, int *);

/* Used by lsm_info(ARRAY_STRUCTURE) and lsm_config(MMAP) */
int lsmInfoArrayStructure(lsm_db *pDb, int bBlock, Pgno iFirst, char **pzOut);
int lsmInfoArrayPages(lsm_db *pDb, Pgno iFirst, char **pzOut);
int lsmConfigMmap(lsm_db *pDb, int *piParam);

int lsmEnvOpen(lsm_env *, const char *, int, lsm_file **);
int lsmEnvClose(lsm_env *pEnv, lsm_file *pFile);
int lsmEnvLock(lsm_env *pEnv, lsm_file *pFile, int iLock, int eLock);

int lsmEnvShmMap(lsm_env *, lsm_file *, int, int, void **); 
void lsmEnvShmBarrier(lsm_env *);
void lsmEnvShmUnmap(lsm_env *, lsm_file *, int);

        if( piRead ) *piRead = 2;
        return LSM_OK;
      }
    }

    lsmShmBarrier(pDb);
  }
  return LSM_PROTOCOL_BKPT;
}

int lsmInfoCompressionId(lsm_db *db, u32 *piCmpId){
  int rc;

  assert( db->pClient==0 && db->pWorker==0 );
  rc = lsmCheckpointLoad(db, 0);

  nInt2 = pShm->aSnap2[CKPT_HDR_NCKPT];
  if( nInt1!=nInt2 || memcmp(pShm->aSnap1, pShm->aSnap2, nInt2*sizeof(u32)) ){
    if( ckptChecksumOk(pShm->aSnap1) ){
      memcpy(pShm->aSnap2, pShm->aSnap1, sizeof(u32)*nInt1);
    }else if( ckptChecksumOk(pShm->aSnap2) ){
      memcpy(pShm->aSnap1, pShm->aSnap2, sizeof(u32)*nInt2);
    }else{
      return LSM_PROTOCOL_BKPT;
    }
  }

  rc = lsmCheckpointDeserialize(pDb, 1, pShm->aSnap1, &pDb->pWorker);
  if( pDb->pWorker ) pDb->pWorker->pDatabase = pDb->pDatabase;

  if( rc==LSM_OK ){

**     lsmEnvSync()
**     lsmEnvSectorSize()
**     lsmEnvClose()
**     lsmEnvTruncate()
**     lsmEnvUnlink()
**     lsmEnvRemap()
*/
int lsmEnvOpen(lsm_env *pEnv, const char *zFile, int flags, lsm_file **ppNew){
  return pEnv->xOpen(pEnv, zFile, flags, ppNew);
}
static int lsmEnvRead(
  lsm_env *pEnv, 
  lsm_file *pFile, 
  lsm_i64 iOff, 
  void *pRead, 
  int nRead

/*
** This is a helper function for lsmFsOpen(). It opens a single file on
** disk (either the database or log file).
*/
static lsm_file *fsOpenFile(
  FileSystem *pFS,                /* File system object */
  int bReadonly,                  /* True to open this file read-only */
  int bLog,                       /* True for log, false for db */
  int *pRc                        /* IN/OUT: Error code */
){
  lsm_file *pFile = 0;
  if( *pRc==LSM_OK ){
    int flags = (bReadonly ? LSM_OPEN_READONLY : 0);
    const char *zPath = (bLog ? pFS->zLog : pFS->zDb);

    *pRc = lsmEnvOpen(pFS->pEnv, zPath, flags, &pFile);
  }
  return pFile;
}

/*
** If it is not already open, this function opens the log file. It returns
** LSM_OK if successful (or if the log file was already open) or an LSM
** error code otherwise.
**
** The log file must be opened before any of the following may be called:
**
**     lsmFsWriteLog
**     lsmFsSyncLog
**     lsmFsReadLog
*/
int lsmFsOpenLog(lsm_db *db, int *pbOpen){
  int rc = LSM_OK;
  FileSystem *pFS = db->pFS;

  if( 0==pFS->fdLog ){ 
    pFS->fdLog = fsOpenFile(pFS, db->bReadonly, 1, &rc); 

    if( rc==LSM_IOERR_NOENT && db->bReadonly ){
      rc = LSM_OK;
    }
  }

  if( pbOpen ) *pbOpen = (pFS->fdLog!=0);
  return rc;
}

void lsmFsCloseLog(lsm_db *db){
  FileSystem *pFS = db->pFS;
  if( pFS->fdLog ){
    lsmEnvClose(pFS->pEnv, pFS->fdLog);
    pFS->fdLog = 0;
  }
}

/*
** Open a connection to a database stored within the file-system (the
** "system of files").
**
** If parameter bReadonly is true, then open a read-only file-descriptor
** on the database file. It is possible that bReadonly will be false even
** if the user requested that pDb be opened read-only. This is because the
** file-descriptor may later on be recycled by a read-write connection.
** If the db file can be opened for read-write access, it always is. Parameter
** bReadonly is only ever true if it has already been determined that the
** db can only be opened for read-only access.
*/
int lsmFsOpen(
  lsm_db *pDb,                    /* Database connection to open fd for */
  const char *zDb,                /* Full path to database file */
  int bReadonly                   /* True to open db file read-only */
){
  FileSystem *pFS;
  int rc = LSM_OK;
  int nDb = strlen(zDb);
  int nByte;

  assert( pDb->pFS==0 );
  assert( pDb->pWorker==0 && pDb->pClient==0 );

    if( pLsmFile ){
      pFS->pLsmFile = pLsmFile;
      pFS->fdDb = pLsmFile->pFile;
      memset(pLsmFile, 0, sizeof(LsmFile));
    }else{
      pFS->pLsmFile = lsmMallocZeroRc(pDb->pEnv, sizeof(LsmFile), &rc);
      if( rc==LSM_OK ){
        pFS->fdDb = fsOpenFile(pFS, bReadonly, 0, &rc);
      }
    }

    if( rc!=LSM_OK ){
      lsmFsClose(pFS);
      pFS = 0;
    }else{

  LogWriter *pNew;
  LogRegion *aReg;

  if( pDb->bUseLog==0 ) return LSM_OK;

  /* If the log file has not yet been opened, open it now. Also allocate
  ** the LogWriter structure, if it has not already been allocated.  */
  rc = lsmFsOpenLog(pDb, 0);
  if( pDb->pLogWriter==0 ){
    pNew = lsmMallocZeroRc(pDb->pEnv, sizeof(LogWriter), &rc);
    if( pNew ){
      lsmStringInit(&pNew->buf, pDb->pEnv);
      rc = lsmStringExtend(&pNew->buf, 2);
    }
  }else{

  LsmString buf2;                 /* Value buffer */
  LogReader reader;               /* Log reader object */
  int rc = LSM_OK;                /* Return code */
  int nCommit = 0;                /* Number of transactions to recover */
  int iPass;
  int nJump = 0;                  /* Number of LSM_LOG_JUMP records in pass 0 */
  DbLog *pLog;
  int bOpen;

  rc = lsmFsOpenLog(pDb, &bOpen);
  if( rc!=LSM_OK ) return rc;

  rc = lsmTreeInit(pDb);
  if( rc!=LSM_OK ) return rc;

  pLog = &pDb->treehdr.log;
  lsmCheckpointLogoffset(pDb->pShmhdr->aSnap2, pLog);

  logReaderInit(pDb, pLog, 1, &reader);
  lsmStringInit(&buf1, pDb->pEnv);
  lsmStringInit(&buf2, pDb->pEnv);

  /* The outer for() loop runs at most twice. The first iteration is to 
  ** count the number of committed transactions in the log. The second 
  ** iterates through those transactions and updates the in-memory tree 
  ** structure with their contents.  */
  if( bOpen ){
    for(iPass=0; iPass<2 && rc==LSM_OK; iPass++){
      int bEof = 0;

      while( rc==LSM_OK && !bEof ){
        u8 eType = 0;
        logReaderByte(&reader, &eType, &rc);

        switch( eType ){
          case LSM_LOG_PAD1:
            break;

          case LSM_LOG_PAD2: {
            int nPad;
            logReaderVarint(&reader, &buf1, &nPad, &rc);
            logReaderBlob(&reader, &buf1, nPad, 0, &rc);
            break;
          }

          case LSM_LOG_WRITE:
          case LSM_LOG_WRITE_CKSUM: {
            int nKey;
            int nVal;
            u8 *aVal;
            logReaderVarint(&reader, &buf1, &nKey, &rc);
            logReaderVarint(&reader, &buf2, &nVal, &rc);

            if( eType==LSM_LOG_WRITE_CKSUM ){
              logReaderCksum(&reader, &buf1, &bEof, &rc);
            }else{
              bEof = logRequireCksum(&reader, nKey+nVal);
            }
            if( bEof ) break;

            logReaderBlob(&reader, &buf1, nKey, 0, &rc);
            logReaderBlob(&reader, &buf2, nVal, &aVal, &rc);
            if( iPass==1 && rc==LSM_OK ){ 
              rc = lsmTreeInsert(pDb, (u8 *)buf1.z, nKey, aVal, nVal);
            }
            break;
          }

          case LSM_LOG_DELETE:
          case LSM_LOG_DELETE_CKSUM: {
            int nKey; u8 *aKey;
            logReaderVarint(&reader, &buf1, &nKey, &rc);

            if( eType==LSM_LOG_DELETE_CKSUM ){
              logReaderCksum(&reader, &buf1, &bEof, &rc);
            }else{
              bEof = logRequireCksum(&reader, nKey);
            }
            if( bEof ) break;

            logReaderBlob(&reader, &buf1, nKey, &aKey, &rc);
            if( iPass==1 && rc==LSM_OK ){ 
              rc = lsmTreeInsert(pDb, aKey, nKey, NULL, -1);
            }
            break;
          }

          case LSM_LOG_COMMIT:
            logReaderCksum(&reader, &buf1, &bEof, &rc);
            if( bEof==0 ){
              nCommit++;
              assert( nCommit>0 || iPass==1 );
              if( nCommit==0 ) bEof = 1;
            }
            break;

          case LSM_LOG_JUMP: {
            int iOff = 0;
            logReaderVarint(&reader, &buf1, &iOff, &rc);
            if( rc==LSM_OK ){
              if( iPass==1 ){
                if( pLog->aRegion[2].iStart==0 ){
                  assert( pLog->aRegion[1].iStart==0 );
                  pLog->aRegion[1].iEnd = reader.iOff;
                }else{
                  assert( pLog->aRegion[0].iStart==0 );
                  pLog->aRegion[0].iStart = pLog->aRegion[2].iStart;
                  pLog->aRegion[0].iEnd = reader.iOff - reader.buf.n+reader.iBuf;
                }
                pLog->aRegion[2].iStart = iOff;
              }else{
                if( (nJump++)==2 ){
                  bEof = 1;
                }
              }

              reader.iOff = iOff;
              reader.buf.n = reader.iBuf;
            }
            break;
          }

          default:
            /* Including LSM_LOG_EOF */
            bEof = 1;
            break;
        }
      }

      if( rc==LSM_OK && iPass==0 ){
        if( nCommit==0 ){
          if( pLog->aRegion[2].iStart==0 ){
            iPass = 1;
          }else{
            pLog->aRegion[2].iStart = 0;
            iPass = -1;
            lsmCheckpointZeroLogoffset(pDb);
          }
        }
        logReaderInit(pDb, pLog, 0, &reader);
        nCommit = nCommit * -1;
      }
    }
  }

  /* Initialize DbLog object */
  if( rc==LSM_OK ){
    pLog->aRegion[2].iEnd = reader.iOff - reader.buf.n + reader.iBuf;
    pLog->cksum0 = reader.cksum0;
    pLog->cksum1 = reader.cksum1;
  }

  if( rc==LSM_OK ){
    rc = lsmFinishRecovery(pDb);
  }else{
    lsmFinishRecovery(pDb);
  }

  if( pDb->bRoTrans ){
    lsmFsCloseLog(pDb);
  }

  lsmStringClear(&buf1);
  lsmStringClear(&buf2);
  lsmStringClear(&reader.buf);
  return rc;
}

*/
static void assert_db_state(lsm_db *pDb){

  /* If there is at least one cursor or a write transaction open, the database
  ** handle must be holding a pointer to a client snapshot. And the reverse 
  ** - if there are no open cursors and no write transactions then there must 
  ** not be a client snapshot.  */
  
  assert( (pDb->pCsr!=0||pDb->nTransOpen>0)==(pDb->iReader>=0||pDb->bRoTrans) );

  assert( (pDb->iReader<0 && pDb->bRoTrans==0) || pDb->pClient!=0 );

  assert( pDb->nTransOpen>=0 );
}
#else
# define assert_db_state(x) 
#endif

    ** than one purpose - to open both the database and log files, and 
    ** perhaps to unlink the log file during disconnection. An absolute
    ** path is required to ensure that the correct files are operated
    ** on even if the application changes the cwd.  */
    rc = getFullpathname(pDb->pEnv, zFilename, &zFull);
    assert( rc==LSM_OK || zFull==0 );

    /* Connect to the database. */
    if( rc==LSM_OK ){
      rc = lsmDbDatabaseConnect(pDb, zFull);
    }

    if( pDb->bReadonly==0 ){
      /* Configure the file-system connection with the page-size and block-size
      ** of this database. Even if the database file is zero bytes in size
      ** on disk, these values have been set in shared-memory by now, and so 
      ** are guaranteed not to change during the lifetime of this connection.  
      */
      if( rc==LSM_OK && LSM_OK==(rc = lsmCheckpointLoad(pDb, 0)) ){
        lsmFsSetPageSize(pDb->pFS, lsmCheckpointPgsz(pDb->aSnapshot));
        lsmFsSetBlockSize(pDb->pFS, lsmCheckpointBlksz(pDb->aSnapshot));
      }
    }

    lsmFree(pDb->pEnv, zFull);
  }

  assert( pDb->bReadonly==0 || pDb->bReadonly==1 );
  assert( rc!=LSM_OK || (pDb->pShmhdr==0)==(pDb->bReadonly==1) );

  return rc;
}


int lsm_close(lsm_db *pDb){
  int rc = LSM_OK;
  if( pDb ){

        ** in multi-process mode.  */
        *piVal = lsmDbMultiProc(pDb);
      }else{
        pDb->bMultiProc = *piVal = (*piVal!=0);
      }
      break;
    }

    case LSM_CONFIG_READONLY: {
      int *piVal = va_arg(ap, int *);
      /* If lsm_open() has been called, this is a read-only parameter. */
      if( pDb->pDatabase==0 && *piVal>=0 ){
        pDb->bReadonly = *piVal = (*piVal!=0);
      }
      *piVal = pDb->bReadonly;
      break;
    }

    case LSM_CONFIG_SET_COMPRESSION: {
      lsm_compress *p = va_arg(ap, lsm_compress *);
      if( pDb->iReader>=0 && pDb->bInFactory==0 ){
        /* May not change compression schemes with an open transaction */
        rc = LSM_MISUSE_BKPT;
      }else{

*/
int lsm_csr_open(lsm_db *pDb, lsm_cursor **ppCsr){
  int rc;                         /* Return code */
  MultiCursor *pCsr = 0;          /* New cursor object */

  /* Open a read transaction if one is not already open. */
  assert_db_state(pDb);

  if( pDb->pShmhdr==0 ){
    assert( pDb->bReadonly );
    rc = lsmBeginRoTrans(pDb);
  }else{
    assert( pDb->bRoTrans==0 );
    rc = lsmBeginReadTrans(pDb);
  }

  /* Allocate the multi-cursor. */
  if( rc==LSM_OK ) rc = lsmMCursorNew(pDb, &pCsr);

  /* If an error has occured, set the output to NULL and delete any partially
  ** allocated cursor. If this means there are no open cursors, release the
  ** client snapshot.  */

    va_end(ap2);
    pDb->xLog(pDb->pLogCtx, rc, s.z);
    lsmStringClear(&s);
  }
}

int lsm_begin(lsm_db *pDb, int iLevel){
  int rc;

  assert_db_state( pDb );
  rc = (pDb->bReadonly ? LSM_READONLY : LSM_OK);

  /* A value less than zero means open one more transaction. */
  if( iLevel<0 ) iLevel = pDb->nTransOpen + 1;

  if( iLevel>pDb->nTransOpen ){
    int i;

    /* Extend the pDb->aTrans[] array if required. */
    if( rc==LSM_OK && pDb->nTransAlloc<iLevel ){
      TransMark *aNew;            /* New allocation */
      int nByte = sizeof(TransMark) * (iLevel+1);

  /* Protected by the global mutex (enterGlobalMutex/leaveGlobalMutex): */
  char *zName;                    /* Canonical path to database file */
  int nName;                      /* strlen(zName) */
  int nDbRef;                     /* Number of associated lsm_db handles */
  Database *pDbNext;              /* Next Database structure in global list */

  /* Protected by the local mutex (pClientMutex) */
  int bReadonly;                  /* True if Database.pFile is read-only */
  int bMultiProc;                 /* True if running in multi-process mode */
  lsm_file *pFile;                /* Used for locks/shm in multi-proc mode */
  LsmFile *pLsmFile;              /* List of deferred closes */
  lsm_mutex *pClientMutex;        /* Protects the apShmChunk[] and pConn */
  int nShmChunk;                  /* Number of entries in apShmChunk[] array */
  void **apShmChunk;              /* Array of "shared" memory regions */
  lsm_db *pConn;                  /* List of connections to this db. */

        break;
      }
    }
  }
  lsmShmLock(pDb, LSM_LOCK_DMS1, LSM_LOCK_UNLOCK, 0);
  return rc;
}

static int dbOpenSharedFd(lsm_env *pEnv, Database *p, int bRoOk){
  int rc;

  rc = lsmEnvOpen(pEnv, p->zName, 0, &p->pFile);
  if( rc==LSM_IOERR && bRoOk ){
    rc = lsmEnvOpen(pEnv, p->zName, LSM_OPEN_READONLY, &p->pFile);
    p->bReadonly = 1;
  }

  return rc;
}

/*
** Return a reference to the shared Database handle for the database 
** identified by canonical path zName. If this is the first connection to
** the named database, a new Database object is allocated. Otherwise, a
** pointer to an existing object is returned.
**

        rc = lsmMutexNew(pEnv, &p->pClientMutex);
      }

      /* If nothing has gone wrong so far, open the shared fd. And if that
      ** succeeds and this connection requested single-process mode, 
      ** attempt to take the exclusive lock on DMS2.  */
      if( rc==LSM_OK ){
        int bReadonly = (pDb->bReadonly && pDb->bMultiProc);
        rc = dbOpenSharedFd(pDb->pEnv, p, bReadonly);
      }

      if( rc==LSM_OK && p->bMultiProc==0 ){
        assert( p->bReadonly==0 );
        rc = lsmEnvLock(pDb->pEnv, p->pFile, LSM_LOCK_DMS2, LSM_LOCK_EXCL);
      }

      if( rc==LSM_OK ){
        p->pDbNext = gShared.pDatabase;
        gShared.pDatabase = p;
      }else{

      lsmMutexLeave(pDb->pEnv, p->pClientMutex);
    }
  }

  pDb->pDatabase = p;
  if( rc==LSM_OK ){
    assert( p );
    rc = lsmFsOpen(pDb, zName, p->bReadonly);
  }

  /* If the db handle is read-write, then connect to the system now. Run
  ** recovery as necessary. Or, if this is a read-only database handle,
  ** defer attempting to connect to the system until a read-transaction
  ** is opened.  */
  if( pDb->bReadonly==0 ){
    if( rc==LSM_OK ){
      rc = doDbConnect(pDb);
    }
    if( rc==LSM_OK ){
      rc = lsmFsConfigure(pDb);
    }
  }

  return rc;
}

static void dbDeferClose(lsm_db *pDb){
  if( pDb->pFS ){

  if( p ){
    lsmSortedFreeLevel(pEnv, p->pLevel);
    lsmFree(pEnv, p->freelist.aEntry);
    lsmFree(pEnv, p->redirect.a);
    lsmFree(pEnv, p);
  }
}

/*
** Attempt to populate one of the read-lock slots to contain lock values
** iLsm/iShm. Or, if such a slot exists already, this function is a no-op.
**
** It is not an error if no slot can be populated because the write-lock
** cannot be obtained. If any other error occurs, return an LSM error code.
** Otherwise, LSM_OK.
**
** This function is called at various points to try to ensure that there
** always exists at least one read-lock slot that can be used by a read-only
** client. And so that, in the usual case, there is an "exact match" available
** whenever a read transaction is opened by any client. At present this
** function is called when:
**
**    * A write transaction that called lsmTreeDiscardOld() is committed, and
**    * Whenever the working snapshot is updated (i.e. lsmFinishWork()).
*/
static int dbSetReadLock(lsm_db *db, i64 iLsm, u32 iShm){
  int rc = LSM_OK;
  ShmHeader *pShm = db->pShmhdr;
  int i;

  /* Check if there is already a slot containing the required values. */
  for(i=0; i<LSM_LOCK_NREADER; i++){
    ShmReader *p = &pShm->aReader[i];
    if( p->iLsmId==iLsm && p->iTreeId==iShm ) return LSM_OK;
  }

  /* Iterate through all read-lock slots, attempting to take a write-lock
  ** on each of them. If a write-lock succeeds, populate the locked slot
  ** with the required values and break out of the loop.  */
  for(i=0; rc==LSM_OK && i<LSM_LOCK_NREADER; i++){
    rc = lsmShmLock(db, LSM_LOCK_READER(i), LSM_LOCK_EXCL, 0);
    if( rc==LSM_BUSY ){
      rc = LSM_OK;
    }else{
      ShmReader *p = &pShm->aReader[i];
      p->iLsmId = iLsm;
      p->iTreeId = iShm;
      lsmShmLock(db, LSM_LOCK_READER(i), LSM_LOCK_UNLOCK, 0);
      break;
    }
  }

  return rc;
}


/*
** Argument bFlush is true if the contents of the in-memory tree has just
** been flushed to disk. The significance of this is that once the snapshot
** created to hold the updated state of the database is synced to disk, log
** file space can be recycled.
*/
void lsmFinishWork(lsm_db *pDb, int bFlush, int *pRc){
  int rc = *pRc;
  assert( rc!=0 || pDb->pWorker );
  if( pDb->pWorker ){
    /* If no error has occurred, serialize the worker snapshot and write
    ** it to shared memory.  */
    if( rc==LSM_OK ){
      rc = lsmSaveWorker(pDb, bFlush);
    }

    /* Assuming no error has occurred, update a read lock slot with the
    ** new snapshot id (see comments above function dbSetReadLock()).  */
    if( rc==LSM_OK ){
      if( pDb->iReader<0 ){
        rc = lsmTreeLoadHeader(pDb, 0);
      }
      if( rc==LSM_OK ){
        rc = dbSetReadLock(pDb, pDb->pWorker->iId, pDb->treehdr.iUsedShmid);
      }
    }

    /* Free the snapshot object. */
    lsmFreeSnapshot(pDb->pEnv, pDb->pWorker);
    pDb->pWorker = 0;
  }

/*
** Begin a read transaction. This function is a no-op if the connection
** passed as the only argument already has an open read transaction.
*/
int lsmBeginReadTrans(lsm_db *pDb){
  const int MAX_READLOCK_ATTEMPTS = 10;
  const int nMaxAttempt = (pDb->bRoTrans ? 1 : MAX_READLOCK_ATTEMPTS);

  int rc = LSM_OK;                /* Return code */
  int iAttempt = 0;


  assert( pDb->pWorker==0 );

  while( rc==LSM_OK && pDb->iReader<0 && (iAttempt++)<nMaxAttempt ){
    int iTreehdr = 0;
    int iSnap = 0;
    assert( pDb->pCsr==0 && pDb->nTransOpen==0 );

    /* Load the in-memory tree header. */
    rc = lsmTreeLoadHeader(pDb, &iTreehdr);

          ** checkpoint just loaded. TODO: This will be removed after 
          ** lsm_sorted.c is changed to work directly from the serialized
          ** version of the snapshot.  */
          if( pDb->pClient==0 ){
            rc = lsmCheckpointDeserialize(pDb, 0, pDb->aSnapshot,&pDb->pClient);
          }
          assert( (rc==LSM_OK)==(pDb->pClient!=0) );
          assert( pDb->iReader>=0 || pDb->bRoTrans );

          /* Check that the client has the right compression hooks loaded.
          ** If not, set rc to LSM_MISMATCH.  */
          if( rc==LSM_OK ){
            rc = lsmCheckCompressionId(pDb, pDb->pClient->iCmpId);
          }
        }else{

  }
  if( rc!=LSM_OK ){
    lsmReleaseReadlock(pDb);
  }
  if( pDb->pClient==0 && rc==LSM_OK ) rc = LSM_BUSY;
  return rc;
}

/*
** db is a read-only database handle in the disconnected state. This function
** attempts to open a read-transaction on the database. This may involve
** connecting to the database system (opening shared memory etc.).
*/
int lsmBeginRoTrans(lsm_db *db){
  int rc = LSM_OK;

  assert( db->bReadonly && db->pShmhdr==0 );
  assert( db->iReader<0 );

  if( db->bRoTrans==0 ){
    if( 1 ){
      rc = lsmShmLock(db, LSM_LOCK_CHECKPOINTER, LSM_LOCK_SHARED, 0);
      if( rc==LSM_OK ){
        db->bRoTrans = 1;
        rc = lsmShmCacheChunks(db, 1);
        if( rc==LSM_OK ){
          db->pShmhdr = (ShmHeader *)db->apShm[0];
          memset(db->pShmhdr, 0, sizeof(ShmHeader));
          rc = lsmCheckpointRecover(db);
          if( rc==LSM_OK ){
            rc = lsmLogRecover(db);
          }
        }
      }
    }else{
      /* lock(DMS2, SHARED) etc. */
    }

    if( rc==LSM_OK ){
      rc = lsmBeginReadTrans(db);
    }
  }

  return rc;
}

/*
** Close the currently open read transaction.
*/
void lsmFinishReadTrans(lsm_db *pDb){

  /* Worker connections should not be closing read transactions. And
  ** read transactions should only be closed after all cursors and write
  ** transactions have been closed. Finally pClient should be non-NULL
  ** only iff pDb->iReader>=0.  */
  assert( pDb->pWorker==0 );
  assert( pDb->pCsr==0 && pDb->nTransOpen==0 );

  lsmReleaseReadlock(pDb);
  if( pDb->bRoTrans ){
    int i;
    for(i=0; i<pDb->nShm; i++){
      lsmFree(pDb->pEnv, pDb->apShm[i]);

    }
    lsmFree(pDb->pEnv, pDb->apShm);
    pDb->apShm = 0;
    pDb->nShm = 0;
    pDb->pShmhdr = 0;

    lsmShmLock(pDb, LSM_LOCK_CHECKPOINTER, LSM_LOCK_UNLOCK, 0);
    pDb->bRoTrans = 0;




  }


}

/*
** Open a write transaction.
*/
int lsmBeginWriteTrans(lsm_db *pDb){
  int rc;                         /* Return code */

  }
  lsmTreeEndTransaction(pDb, bCommit);

  if( rc==LSM_OK ){
    if( bFlush && pDb->bAutowork ){
      rc = lsmSortedAutoWork(pDb, 1);
    }else if( bCommit && pDb->bDiscardOld ){
      rc = dbSetReadLock(pDb, pDb->pClient->iId, pDb->treehdr.iUsedShmid);
    }
  }
  pDb->bDiscardOld = 0;
  lsmShmLock(pDb, LSM_LOCK_WRITER, LSM_LOCK_UNLOCK, 0);

  if( bFlush && pDb->bAutowork==0 && pDb->xWork ){
    pDb->xWork(pDb, pDb->pWorkCtx);

  return( 
      p->iLsmId && p->iLsmId<=iLsm 
      && shm_sequence_ge(iShmMax, p->iTreeId)
      && shm_sequence_ge(p->iTreeId, iShmMin)
  );
}
















































/*
** Obtain a read-lock on database version identified by the combination
** of snapshot iLsm and tree iTree. Return LSM_OK if successful, or
** an LSM error code otherwise.
*/
int lsmReadlock(lsm_db *db, i64 iLsm, u32 iShmMin, u32 iShmMax){
  int rc = LSM_OK;
  ShmHeader *pShm = db->pShmhdr;
  int i;

  assert( db->iReader<0 );
  assert( shm_sequence_ge(iShmMax, iShmMin) );

  /* This is a no-op if the read-only transaction flag is set. */
  if( db->bRoTrans ) return LSM_OK;

  /* Search for an exact match. */
  for(i=0; db->iReader<0 && rc==LSM_OK && i<LSM_LOCK_NREADER; i++){
    ShmReader *p = &pShm->aReader[i];
    if( p->iLsmId==iLsm && p->iTreeId==iShmMax ){
      rc = lsmShmLock(db, LSM_LOCK_READER(i), LSM_LOCK_SHARED, 0);
      if( rc==LSM_OK && p->iLsmId==iLsm && p->iTreeId==iShmMax ){

      void **apShm;
      nAlloc += NINCR;
      apShm = lsmRealloc(pEnv, db->apShm, sizeof(void*)*nAlloc);
      if( !apShm ) return LSM_NOMEM_BKPT;
      db->apShm = apShm;
    }

    if( db->bRoTrans ){
      for(i=db->nShm; rc==LSM_OK && i<nChunk; i++){
        db->apShm[i] = lsmMallocZeroRc(pEnv, LSM_SHM_CHUNK_SIZE, &rc);
        db->nShm++;
      }

    }else{

      /* Enter the client mutex */
      lsmMutexEnter(pEnv, p->pClientMutex);

      /* Extend the Database objects apShmChunk[] array if necessary. Using the
       ** same pattern as for the lsm_db.apShm[] array above.  */
      nAlloc = ((p->nShmChunk + NINCR - 1) / NINCR) * NINCR;
      while( nChunk>=nAlloc ){
        void **apShm;
        nAlloc +=  NINCR;
        apShm = lsmRealloc(pEnv, p->apShmChunk, sizeof(void*)*nAlloc);
        if( !apShm ){
          rc = LSM_NOMEM_BKPT;
          break;
        }
        p->apShmChunk = apShm;
      }

      for(i=db->nShm; rc==LSM_OK && i<nChunk; i++){
        if( i>=p->nShmChunk ){
          void *pChunk = 0;
          if( p->bMultiProc==0 ){
            /* Single process mode */
            pChunk = lsmMallocZeroRc(pEnv, LSM_SHM_CHUNK_SIZE, &rc);
          }else{
            /* Multi-process mode */
            rc = lsmEnvShmMap(pEnv, p->pFile, i, LSM_SHM_CHUNK_SIZE, &pChunk);
          }
          if( rc==LSM_OK ){
            p->apShmChunk[i] = pChunk;
            p->nShmChunk++;
          }
        }
        if( rc==LSM_OK ){
          db->apShm[i] = p->apShmChunk[i];
          db->nShm++;
        }
      }

      /* Release the client mutex */
      lsmMutexLeave(pEnv, p->pClientMutex);
    }
  }

  return rc;
}

static int lockSharedFile(lsm_env *pEnv, Database *p, int iLock, int eOp){
  int rc = LSM_OK;

){
  lsm_db *pIter;
  const u64 me = ((u64)1 << (iLock-1));
  const u64 ms = ((u64)1 << (iLock+32-1));
  int rc = LSM_OK;
  Database *p = db->pDatabase;

  assert( eOp!=LSM_LOCK_EXCL || db->bReadonly==0 );
  assert( iLock>=1 && iLock<=LSM_LOCK_RWCLIENT(LSM_LOCK_NRWCLIENT-1) );
  assert( LSM_LOCK_RWCLIENT(LSM_LOCK_NRWCLIENT-1)<=32 );
  assert( eOp==LSM_LOCK_UNLOCK || eOp==LSM_LOCK_SHARED || eOp==LSM_LOCK_EXCL );

  /* Check for a no-op. Proceed only if this is not one of those. */
  if( (eOp==LSM_LOCK_UNLOCK && (db->mLock & (me|ms))!=0)
   || (eOp==LSM_LOCK_SHARED && (db->mLock & (me|ms))!=ms)

    if( treeHeaderChecksumOk(&pDb->treehdr) ){
      if( piRead ) *piRead = 2;
      return LSM_OK;
    }

    lsmShmBarrier(pDb);
  }
  return LSM_PROTOCOL_BKPT;
}

int lsmTreeLoadHeaderOk(lsm_db *pDb, int iRead){
  TreeHeader *p = (iRead==1) ? &pDb->pShmhdr->hdr1 : &pDb->pShmhdr->hdr2;
  assert( iRead==1 || iRead==2 );
  return (0==memcmp(pDb->treehdr.aCksum, p->aCksum, sizeof(u32)*2));
}

  int shmfd;                      /* Shared memory file-descriptor */
  void *pMap;                     /* Pointer to mapping of file fd */
  off_t nMap;                     /* Size of mapping at pMap in bytes */
  int nShm;                       /* Number of entries in array apShm[] */
  void **apShm;                   /* Array of 32K shared memory segments */
};



static char *posixShmFile(PosixFile *p){
  char *zShm;
  int nName = strlen(p->zName);
  zShm = (char *)lsmMalloc(p->pEnv, nName+4+1);
  if( zShm ){
    memcpy(zShm, p->zName, nName);
    memcpy(&zShm[nName], "-shm", 5);
  }
  return zShm;
}

static int lsmPosixOsOpen(
  lsm_env *pEnv,
  const char *zFile,
  int flags,
  lsm_file **ppFile
){
  int rc = LSM_OK;
  PosixFile *p;

  p = lsm_malloc(pEnv, sizeof(PosixFile));
  if( p==0 ){
    rc = LSM_NOMEM;
  }else{
    int bReadonly = (flags & LSM_OPEN_READONLY);
    int oflags = (bReadonly ? O_RDONLY : (O_RDWR|O_CREAT));
    memset(p, 0, sizeof(PosixFile));
    p->zName = zFile;
    p->pEnv = pEnv;
    p->fd = open(zFile, oflags, 0644);
    if( p->fd<0 ){
      lsm_free(pEnv, p);
      p = 0;
      if( errno==ENOENT ){
        rc = lsmErrorBkpt(LSM_IOERR_NOENT);
      }else{
        rc = LSM_IOERR_BKPT;
      }
    }
  }

  *ppFile = (lsm_file *)p;
  return rc;
}

static int lsmPosixOsWrite(
  lsm_file *pFile,                /* File to write to */
  lsm_i64 iOff,                   /* Offset to write to */
  void *pData,                    /* Write data from this buffer */
  int nData                       /* Bytes of data to write */
){
  int rc = LSM_OK;
  PosixFile *p = (PosixFile *)pFile;
  off_t offset;

  offset = lseek(p->fd, (off_t)iOff, SEEK_SET);
  if( offset!=iOff ){
    rc = LSM_IOERR_BKPT;
  }else{
    ssize_t prc = write(p->fd, pData, (size_t)nData);
    if( prc<0 ) rc = LSM_IOERR_BKPT;
  }

  return rc;
}

static int lsmPosixOsTruncate(
  lsm_file *pFile,                /* File to write to */
  lsm_i64 nSize                   /* Size to truncate file to */
){
  PosixFile *p = (PosixFile *)pFile;
  int rc = LSM_OK;                /* Return code */
  int prc;                        /* Posix Return Code */
  struct stat sStat;              /* Result of fstat() invocation */
  
  prc = fstat(p->fd, &sStat);
  if( prc==0 && sStat.st_size>nSize ){
    prc = ftruncate(p->fd, (off_t)nSize);
  }
  if( prc<0 ) rc = LSM_IOERR_BKPT;

  return rc;
}

static int lsmPosixOsRead(
  lsm_file *pFile,                /* File to read from */
  lsm_i64 iOff,                   /* Offset to read from */
  void *pData,                    /* Read data into this buffer */
  int nData                       /* Bytes of data to read */
){
  int rc = LSM_OK;
  PosixFile *p = (PosixFile *)pFile;
  off_t offset;

  offset = lseek(p->fd, (off_t)iOff, SEEK_SET);
  if( offset!=iOff ){
    rc = LSM_IOERR_BKPT;
  }else{
    ssize_t prc = read(p->fd, pData, (size_t)nData);
    if( prc<0 ){ 
      rc = LSM_IOERR_BKPT;
    }else if( prc<nData ){
      memset(&((u8 *)pData)[prc], 0, nData - prc);
    }

  }

  return rc;
}

static int lsmPosixOsSync(lsm_file *pFile){
  int rc = LSM_OK;

#ifndef LSM_NO_SYNC
  PosixFile *p = (PosixFile *)pFile;
  int prc = 0;

  if( p->pMap ){
    prc = msync(p->pMap, p->nMap, MS_SYNC);
  }
  if( prc==0 ) prc = fdatasync(p->fd);
  if( prc<0 ) rc = LSM_IOERR_BKPT;
#else
  (void)pFile;
#endif

  return rc;
}

  lock.l_start = (4096-iLock);

  if( fcntl(p->fd, F_SETLK, &lock) ){
    int e = errno;
    if( e==EACCES || e==EAGAIN ){
      rc = LSM_BUSY;
    }else{
      rc = LSM_IOERR_BKPT;
    }
  }

  return rc;
}

int lsmPosixOsShmMap(lsm_file *pFile, int iChunk, int sz, void **ppShm){

    p->nShm = nNew;
  }

  if( p->apShm[iChunk]==0 ){
    p->apShm[iChunk] = mmap(0, LSM_SHM_CHUNK_SIZE, 
        PROT_READ|PROT_WRITE, MAP_SHARED, p->shmfd, iChunk*LSM_SHM_CHUNK_SIZE
    );
    if( p->apShm[iChunk]==0 ) return LSM_IOERR_BKPT;
  }

  *ppShm = p->apShm[iChunk];
  return LSM_OK;
}

void lsmPosixOsShmBarrier(void){

# The focus of this file is testing the LSM library. 
#

set testdir [file dirname $argv0]
source $testdir/tester.tcl
set testprefix lsm5
db close

proc db_fetch {db key} {
  db csr_open csr
  csr seek $key eq
  set ret [csr value]
  csr close
  set ret
}

# Create a new database with file name $file.
#
proc create_abc_db {file} {
  forcedelete $file
  lsm_open db $file
  db write a alpha
  db write b bravo
  db write c charlie
  db close
}

#-------------------------------------------------------------------------
# When the database system is shut down (i.e. when the last connection
# disconnects), an attempt is made to truncate the database file to the
# minimum number of blocks required.
# 
# This test case checks that this process does not actually cause the

  db write 1 one
  db write 2 two
  db close
} {}
do_test 1.3 {
  expr [file size test.db] < (64*1024)
} 1

#-------------------------------------------------------------------------
# Test that if an attempt is made to open a read-write connection to a 
# database that the client does not have permission to write to is attempted
# an error is reported. In order to open a read-write connection to a
# database, the client requires:
#
#   * read-write access to the db file,
#   * read-write access to the log file,
#   * for multi-process mode, read-write access to the shm file.
#
# In the above, "read-write access" includes the ability to create the db,
# log or shm file if it does not exist.
#
# These tests verify that the lsm_open() command returns LSM_IOERR. At some
# point in the future this will be improved. Likely when sqlite4 level tests 
# for opening read-only databases are added.
# 

foreach {tn filename setup} {

  1 test.dir/test.db {
    # Create a directory "test.dir".
    forcedelete test.dir
    file mkdir test.dir

    # Create a database within test.dir
    create_abc_db test.dir/test.db
  
    # Now make the db and its directory read-only.
    file attr test.dir/test.db -perm r--r--r--
    file attr test.dir -perm r-xr-xr-x
  }

  2 test.db {
    # Create a database test.db and set its permissions to read-only
    create_abc_db test.db
    file attr test.db -perm r--r--r--
  }

  3 test.dir/test.db {
    # Create a directory "test.dir".
    forcedelete test.dir
    file mkdir test.dir

    # Create a database within test.dir
    create_abc_db test.dir/test.db
  
    # Now make test.dir read-only.
    file attr test.dir -perm r-xr-xr-x
  }

} {

  do_test 2.$tn.1 {
    eval $setup
    set rc [catch {lsm_open db $filename} msg]
    list $rc $msg
  } {1 {error in lsm_open() - 10}}

  do_test 2.$tn.2 {
    eval $setup
    lsm_open db $filename {readonly 1}
    set res [list [db_fetch db a] [db_fetch db b] [db_fetch db c]]
    db close
    set res
  } {alpha bravo charlie}

}


finish_test

    { "mmap",                    LSM_CONFIG_MMAP },
    { "use_log",                 LSM_CONFIG_USE_LOG },
    { "automerge",               LSM_CONFIG_AUTOMERGE },
    { "max_freelist",            LSM_CONFIG_MAX_FREELIST },
    { "multi_proc",              LSM_CONFIG_MULTIPLE_PROCESSES },
    { "set_compression",         LSM_CONFIG_SET_COMPRESSION },
    { "set_compression_factory", LSM_CONFIG_SET_COMPRESSION_FACTORY },
    { "readonly",                LSM_CONFIG_READONLY },
    { 0, 0 }
  };
  int nElem;
  int i;
  Tcl_Obj **apElem;
  int rc;

<title>LSM Design Overview</title>
<nowiki>


<div id=start_of_toc></div>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#summary style=text-decoration:none>1. Summary </a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#data_structures style=text-decoration:none>2. Data Structures </a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#locks style=text-decoration:none>2.1. Locks</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#database_file style=text-decoration:none>2.2. Database file</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#sorted_runs style=text-decoration:none>2.2.1. Sorted Runs</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#levels style=text-decoration:none>2.2.2. Levels</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#snapshots style=text-decoration:none>2.2.3. Snapshots</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#in-memory_tree style=text-decoration:none>2.3. In-Memory Tree</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#memory_allocation style=text-decoration:none>2.3.1. Memory Allocation</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#header_fields style=text-decoration:none>2.3.2. Header Fields</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#other_shared-memory_fields style=text-decoration:none>2.4. Other Shared-Memory Fields</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#log_file style=text-decoration:none>2.5. Log file</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#database_recovery_and_shutdown style=text-decoration:none>3. Database Recovery and Shutdown</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#read-write_clients style=text-decoration:none>3.1. Read-write clients</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#database_operations style=text-decoration:none>4. Database Operations </a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#reading style=text-decoration:none>4.1. Reading</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#writing style=text-decoration:none>4.2. Writing</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#flushing_the_in-memory_tree_to_disk style=text-decoration:none>4.2.1. Flushing the in-memory tree to disk</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#shared-memory_management style=text-decoration:none>4.2.2. Shared-memory management</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#log_file_management style=text-decoration:none>4.2.3. Log file management</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#working style=text-decoration:none>4.3. Working</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#free-block_list_management style=text-decoration:none>4.3.1. Free-block list management</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#checkpoint_operations style=text-decoration:none>4.4. Checkpoint Operations</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a href=#scheduling_policies style=text-decoration:none>5. Scheduling Policies</a><br>

<div id=end_of_toc></div>

<h1 id=summary>1. Summary </h1>

The LSM embedded database software stores data in three distinct data
structures:

<ul>
  <li> <p>The <b>shared-memory region</b>. This may actually be allocated in
       either shared or heap memory, depending on whether LSM is running in

<p>
When an application writes to the database, the new data is written to the
in-memory tree. Once the in-memory tree has grown large enough, its contents
are written into the database file as a new sorted run. To reduce the number
of sorted runs in the database file, chronologically adjacent sorted runs 
may be merged together into a single run, either automatically or on demand.

<h1 id=data_structures>2. Data Structures </h1>

<h2 id=locks>2.1. Locks</h2>
<p>
Read/write (shared/exclusive) file locks are used to control concurrent 
access. LSM uses the following file-locks:

<ul>
  <li> <p>The <b>DMS1</b>, <b>DMS2</b> locking regions. These are used to
       implement the "dead-man-switch" mechanism copied from SQLite's WAL
       mode for safely connecting to and disconnecting from a database. 
       See "Database Recovery and Shutdown" below.

  <li> <p>Several (say 3) <b>READER</b> locking regions. Database clients 
       hold a SHARED lock one of the READER locking regions while reading the
       database. As in SQLite WAL mode, each reader lock is paired with a 

<p>
In the following sections, "the WRITER lock", refers to an exclusive lock
on the WRITER locking region. For example "holding the WRITER lock" is
equivalent to "holding an exclusive lock on the WRITER locking region".
Similar interpretations apply to "the WORKER lock" and "the CHECKPOINTER 
lock".

<h2 id=database_file>2.2. Database file</h2>

<p>
This section summarizes the contents of the database file informally. A 
detailed description is found in the header comments for source code files 
<a href="../src/lsm_file.c">lsm_file.c</a> (blocks, pages etc.), 
<a href="../src/lsm_sorted.c">lsm_sorted.c</a> (sorted run format) and 
<a href="../src/lsm_ckpt.c">lsm_ckpt.c</a> (database snapshot format). 

<p>
As with an SQLite database file, each page in the database may be addressed 
by its 32-bit page number. This means the maximum database size is roughly
(pgsz * 2^32) bytes. The first and last pages in each block are 4 bytes
smaller than the others. This is to make room for a single page-number.

<h3 id=sorted_runs>2.2.1. Sorted Runs</h3>

<p>
A single sorted run is spread across one or more database pages (each page
is a part of at most one sorted run). Given the page number of a page in a
sorted run the following statements are true:

<ul>

In other words, given the page numbers of the first and last pages of a 
sorted run and the page number of the root page for the embedded b-tree,
it is possible to traverse the entire run in either direction or query for
arbitrary values.

<p><span style="color:red"> TODO: Embedded pointers.  </span>

<h3 id=levels>2.2.2. Levels</h3>

<p>
Each sorted run is assigned to a "level". Normally, a level consists of a
single sorted run. However, a level may also consist of a set of sorted runs
being incrementally merged into a single run.

<p>

time for all entries.






<h3 id=snapshots>2.2.3. Snapshots</h3>

<p>
Each meta page may contain a database <b>snapshot</b>. A snapshot contains all 
the information required to interpret the remainder of the database file (the
sorted runs and free space). Specifically, it contains:

<ul>

       Recovery and Shutdown" below).
</ul>

<p>
A more detailed description is available in the header comments in
source code file <a href="../src/lsm_ckpt.c">lsm_ckpt.c</a>

<h2 id=in-memory_tree>2.3. In-Memory Tree</h2>

<p>
The in-memory tree is an append-only b-tree of order 4 (each node may have 
up to 4 children), which is more or less equivalent to a red-black tree. 
An append-only tree is convenient, as it naturally supports the 
single-writer/many-readers MVCC concurrency model. 

<p>
The implementation includes some optimizations to reduce the number of 
interior nodes that are updated when a leaf node is written that are not 
described here. See header comments in source code file 
<a href=../src/lsm_tree.c>lsm_tree.c</a> for details.

<h3 id=memory_allocation>2.3.1. Memory Allocation</h3>

<p>
More than one in-memory 
tree may exist in shared-memory at any time. For example in the following 
scenario:

<ol>

but the values that connect the linked list together are not. The writer 
that detects the failure must scan the entire shared-memory region to 
reconstruct the linked list. Any sequence ids assigned by the failed 
writer are reverted (perhaps not to their original values, but to values
that put them at the start of the linked list - before those chunks that
may still be in use by existing readers).

<h3 id=header_fields>2.3.2. Header Fields</h3>
<p>
As well as the in-memory tree data, the following fixed-size fields 
stored in well-known locations in shared-memory are part of the in-memory
tree. Like the in-memory tree data, outside of recovery these fields are only
ever written to by clients holding the WRITER lock.

<ul>
  <li> Two copies of a data structure called a "tree-header". Tree-header-1
       and tree-header 2. A tree-header structure contains all the information
       required to read or write to a particular version of the append only
       b-tree. It also contains a 64-bit checksum.

  <li> A boolean flag set to true at the beginning of every write 
       transaction and cleared after that transaction is successfully 
       concluded - the "writer flag". This is used to detect failures that
       occur mid-transaction. It is only ever read (or written) by clients
       that hold the WRITER lock.
</ul>


<h2 id=other_shared-memory_fields>2.4. Other Shared-Memory Fields</h2>

<ul>
  <li> Snapshot 1.
  <li> Snapshot 2.
  <li> The meta-page pointer. This value is either 1 or 2. It indicates which
       of the two meta-pages contains the most recent database snapshot.
  <li> READER lock values.
</ul>

<h2 id=log_file>2.5. Log file</h2>

<a href=../src/lsm_log.c>lsm_log.c</a>.

<h1 id=database_recovery_and_shutdown>3. Database Recovery and Shutdown</h1>

<h2 id=read-write_clients>3.1. Read-write clients</h2>

<p>
Exclusive locks on locking region DMS1 are used to serialize all connect and
disconnect operations performed by read-write clients. 

<p>When an LSM database connection is opened (i.e. lsm_open() is called):

<pre>
  lock(DMS1, EXCLUSIVE)           # Block until successful
    lock(DMS2, EXCLUSIVE)         # Abandon if not immediately successful
    if( DMS2 successfully locked ){

      ...TODO...
      delete *-shm file (or equivalent)
    }
    unlock(DMS2)
  unlock(DMS1)
</pre>

<h2 id=read-only_clients>3.1. Read-only clients</h2>

<p>It is assumed that read-only clients may take SHARED locks only. And
that a read-only client may not run database recovery when a db is opened 
in multi-process mode.

<p>

<h1 id=database_operations>4. Database Operations </h1>

<h2 id=reading>4.1. Reading</h2>

<p>
Opening a read transaction:

<ol>
  <li> <p>Load the current tree-header from shared-memory.

Once a read transaction is opened, the reader may continue to read the 
versions of the in-memory tree and database file for as long as the READER
lock is held.
<p>
To close a read transaction all that is required is to drop the SHARED lock
held on the READER slot. 

<h2 id=writing>4.2. Writing</h2>

<p>
To open a write transaction:

<ol>
  <li> <p>Open a read transaction, if one is not already open.
  <li> <p>Obtain the WRITER lock.

  <li> Sweep the shared-memory area to rebuild the linked list of chunks so
       that it is consistent with the current tree-header.

  <li> Clear the writer flag.
</ol>

<h3 id=flushing_the_in-memory_tree_to_disk>4.2.1. Flushing the in-memory tree to disk</h3>

<p>
For the purposes of writing, the database file and the in-memory tree are
largely independent. Processes holding the WRITER lock write to the in-memory
tree, and processes holding the WORKER lock write to the database file.

<ol>

  <li> Update the private copy of the tree-header to reflect a new, empty tree.

  <li> Commit the write transaction, writing the new, empty tree to
       shared-memory.
</ol>

<h3 id=shared-memory_management>4.2.2. Shared-memory management</h3>

<p>
A writer client may have to allocate new shared-memory chunks. This can be
done either by extending the shared-memory region or by recycling the first
chunk in the linked-list. To check if the first chunk in the linked-list may
be reused, the writer must check that:

<ul>
  <li> The chunk is not part of the current in-memory tree (the one being
       appended to by the writer). A writer can check this by examining its
       private copy of the tree-header.

  <li> The chunk is not part of an in-memory tree being used by an existing
       reader. A writer checks this by scanning (and possibly updating) the
       values associated with the READER locks - similar to the way SQLite 
       does in WAL mode.
</ul>

<h3 id=log_file_management>4.2.3. Log file management</h3>

<p>
A writer client also writes to the log file. All information required to write
to the log file  (the offset to write to and the initial checksum values) is
embedded in the tree-header. Except, in order to reuse log file space (wrap
around to the start of the log file), a writer needs to know that the space
being recycled will not be required by any recovery process in the future.
In other words, that the information contained in the transactions being
overwritten has been written into the database file and is part of the
snapshot written into the database file by a checkpointer (see "Checkpoint
Operations" below).

<p>
To determine whether or not the log file can be wrapped, the writer requires
access to information stored in the newest snapshot written into the database
header. Their exists a shared-memory variable indicating which of the two
meta-pages contain this snapshot, but the writer process still has to read the
snapshot data and verify its checksum from disk.

<h2 id=working>4.3. Working</h2>

<p>
Working is similar to writing. The difference is that a "writer" modifies
the in-memory tree. A "worker" modifies the contents of the database file.

<ol>
  <li> <p>Take the WORKER lock.

  <li> <p>Invoke xShmBarrier().

  <li> <p>Update snapshot-1 in shared-memory.

  <li> <p>Release the WORKER lock.
</ol>

<h3 id=free-block_list_management>4.3.1. Free-block list management</h3>

<p>
Worker clients occasionally need to allocate new database blocks or move
existing blocks to the free-block list. Along with the block number of each
free block, the free-block list contains the snapshot-id of the first 
snapshot created after the block was moved to the free list. The free-block
list is always stored in order of snapshot-id, so that the first block in

       header. This is done by reading (and verifying the checksum) of the
       snapshot currently stored in the database meta-page indicated by the
       shared-memory variable.
</ul>



<h2 id=checkpoint_operations>4.4. Checkpoint Operations</h2>

<ol>
  <li> Take CHECKPOINTER lock.

  <li> Load snapshot-1 from shared-memory. If the checksum does not match
       the content here, release the CHECKPOINTER lock and abandon the 
       attempt to checkpoint the database.

  <li> Update the shared-memory variable to indicate the meta-page written in
       step 5.

  <li> Drop the CHECKPOINTER lock.
</ol>

<h1 id=scheduling_policies>5. Scheduling Policies</h1>

<p>
When a client writes to a database, the in-memory tree and log file are
updated by the client itself before the lsm_write() call returns. Eventually, 
once sufficient writes have accumulated in memory, the client marks the 
current tree as "old", and subsequent writes are accumulated in a new tree.

       write both the old and new tree to a new database level.

      <p> If the WORKER lock cannot be obtained immediately, block until it
       can be
</ul>

<p><b> Auto work </b>