/ Check-in [90760e72]
Login

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

Overview
Comment:When searching the wal file for a frame, do not search that part that was already checkpointed when the snapshot being read was at the head of the wal file.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | wal-read-change
Files: files | file ages | folders
SHA1: 90760e72327eb0593cbfa6d7058b554198cd8044
User & Date: dan 2015-08-13 20:23:46
Context
2015-08-28
15:50
Merge latest trunk into this branch. Closed-Leaf check-in: ab93024d user: dan tags: wal-read-change
2015-08-13
20:23
When searching the wal file for a frame, do not search that part that was already checkpointed when the snapshot being read was at the head of the wal file. check-in: 90760e72 user: dan tags: wal-read-change
2015-08-12
19:42
Avoid reading frames that have already been checkpointed from the wal file. check-in: 5669ac4a user: dan tags: wal-read-change
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/wal.c.

424
425
426
427
428
429
430

431
432
433
434
435
436
437
....
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303














2304

2305
2306
2307
2308
2309
2310
2311
....
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
....
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
....
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
  u8 writeLock;              /* True if in a write transaction */
  u8 ckptLock;               /* True if holding a checkpoint lock */
  u8 readOnly;               /* WAL_RDWR, WAL_RDONLY, or WAL_SHM_RDONLY */
  u8 truncateOnCommit;       /* True to truncate WAL file on commit */
  u8 syncHeader;             /* Fsync the WAL header if true */
  u8 padToSectorBoundary;    /* Pad transactions out to the next sector */
  WalIndexHdr hdr;           /* Wal-index header for current transaction */

  const char *zWalName;      /* Name of WAL file */
  u32 nCkpt;                 /* Checkpoint sequence counter in the wal-header */
#ifdef SQLITE_DEBUG
  u8 lockError;              /* True if a locking error has occurred */
#endif
};

................................................................................
    ** that the log file may have been wrapped by a writer, or that frames
    ** that occur later in the log than pWal->hdr.mxFrame may have been
    ** copied into the database by a checkpointer. If either of these things
    ** happened, then reading the database with the current value of
    ** pWal->hdr.mxFrame risks reading a corrupted snapshot. So, retry
    ** instead.
    **
    ** This does not guarantee that the copy of the wal-index header is up to
    ** date before proceeding. That would not be possible without somehow
    ** blocking writers. It only guarantees that a dangerous checkpoint or 
    ** log-wrap (either of which would require an exclusive lock on
    ** WAL_READ_LOCK(mxI)) has not occurred since the snapshot was valid.














    */

    walShmBarrier(pWal);
    if( pInfo->aReadMark[mxI]!=mxReadMark
     || memcmp((void *)walIndexHdr(pWal), &pWal->hdr, sizeof(WalIndexHdr))
    ){
      walUnlockShared(pWal, WAL_READ_LOCK(mxI));
      return WAL_RETRY;
    }else{
................................................................................
  Wal *pWal,                      /* WAL handle */
  Pgno pgno,                      /* Database page number to read data for */
  u32 *piRead                     /* OUT: Frame number (or zero) */
){
  u32 iRead = 0;                  /* If !=0, WAL frame to return data from */
  u32 iLast = pWal->hdr.mxFrame;  /* Last page in WAL for this reader */
  int iHash;                      /* Used to loop through N hash tables */
  u32 iFirst;
  int iMinHash;

  /* This routine is only be called from within a read transaction. */
  assert( pWal->readLock>=0 || pWal->lockError );

  /* If the "last page" field of the wal-index header snapshot is 0, then
  ** no data will be read from the wal under any circumstances. Return early
  ** in this case as an optimization.  Likewise, if pWal->readLock==0, 
  ** then the WAL is ignored by the reader so return early, as if the 
  ** WAL were empty.
  */
  if( iLast==0 
   || pWal->readLock==0 
   || iLast==(iFirst = walCkptInfo(pWal)->nBackfill)
  ){
    *piRead = 0;
    return SQLITE_OK;
  }

  /* Search the hash table or tables for an entry matching page number
  ** pgno. Each iteration of the following for() loop searches one
  ** hash table (each hash table indexes up to HASHTABLE_NPAGE frames).
................................................................................
  **   (aPgno[iFrame]==pgno): 
  **     This condition filters out normal hash-table collisions.
  **
  **   (iFrame<=iLast): 
  **     This condition filters out entries that were added to the hash
  **     table after the current read-transaction had started.
  */
  iMinHash = walFramePage(iFirst);
  for(iHash=walFramePage(iLast); iHash>=iMinHash && iRead==0; iHash--){
    volatile ht_slot *aHash;      /* Pointer to hash table */
    volatile u32 *aPgno;          /* Pointer to array of page numbers */
    u32 iZero;                    /* Frame number corresponding to aPgno[0] */
    int iKey;                     /* Hash slot index */
    int nCollide;                 /* Number of hash collisions remaining */
    int rc;                       /* Error code */
................................................................................
    rc = walHashGet(pWal, iHash, &aHash, &aPgno, &iZero);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    nCollide = HASHTABLE_NSLOT;
    for(iKey=walHash(pgno); aHash[iKey]; iKey=walNextHash(iKey)){
      u32 iFrame = aHash[iKey] + iZero;
      if( iFrame<=iLast && iFrame>iFirst && aPgno[aHash[iKey]]==pgno ){
        assert( iFrame>iRead || CORRUPT_DB );
        iRead = iFrame;
      }
      if( (nCollide--)==0 ){
        return SQLITE_CORRUPT_BKPT;
      }
    }







>







 







|
|
|
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>

>







 







<











<
|
<
<







 







|







 







|







424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
....
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
....
2384
2385
2386
2387
2388
2389
2390

2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401

2402


2403
2404
2405
2406
2407
2408
2409
....
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
....
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
  u8 writeLock;              /* True if in a write transaction */
  u8 ckptLock;               /* True if holding a checkpoint lock */
  u8 readOnly;               /* WAL_RDWR, WAL_RDONLY, or WAL_SHM_RDONLY */
  u8 truncateOnCommit;       /* True to truncate WAL file on commit */
  u8 syncHeader;             /* Fsync the WAL header if true */
  u8 padToSectorBoundary;    /* Pad transactions out to the next sector */
  WalIndexHdr hdr;           /* Wal-index header for current transaction */
  u32 minFrame;              /* Ignore wal frames before this one */
  const char *zWalName;      /* Name of WAL file */
  u32 nCkpt;                 /* Checkpoint sequence counter in the wal-header */
#ifdef SQLITE_DEBUG
  u8 lockError;              /* True if a locking error has occurred */
#endif
};

................................................................................
    ** that the log file may have been wrapped by a writer, or that frames
    ** that occur later in the log than pWal->hdr.mxFrame may have been
    ** copied into the database by a checkpointer. If either of these things
    ** happened, then reading the database with the current value of
    ** pWal->hdr.mxFrame risks reading a corrupted snapshot. So, retry
    ** instead.
    **
    ** Before checking that the live wal-index header has not changed
    ** since it was read, set Wal.minFrame to the first frame in the wal
    ** file that has not yet been checkpointed. This client will not need
    ** to read any frames earlier than minFrame from the wal file - they
    ** can be safely read directly from the database file.
    **
    ** Because a ShmBarrier() call is made between taking the copy of 
    ** nBackfill and checking that the wal-header in shared-memory still
    ** matches the one cached in pWal->hdr, it is guaranteed that the 
    ** checkpointer that set nBackfill was not working with a wal-index
    ** header newer than that cached in pWal->hdr. If it were, that could
    ** cause a problem. The checkpointer could omit to checkpoint
    ** a version of page X that lies before pWal->minFrame (call that version
    ** A) on the basis that there is a newer version (version B) of the same
    ** page later in the wal file. But if version B happens to like past
    ** frame pWal->hdr.mxFrame - then the client would incorrectly assume
    ** that it can read version A from the database file. However, since
    ** we can guarantee that the checkpointer that set nBackfill could not
    ** see any pages past pWal->hdr.mxFrame, this problem does not come up.
    */
    pWal->minFrame = pInfo->nBackfill+1;
    walShmBarrier(pWal);
    if( pInfo->aReadMark[mxI]!=mxReadMark
     || memcmp((void *)walIndexHdr(pWal), &pWal->hdr, sizeof(WalIndexHdr))
    ){
      walUnlockShared(pWal, WAL_READ_LOCK(mxI));
      return WAL_RETRY;
    }else{
................................................................................
  Wal *pWal,                      /* WAL handle */
  Pgno pgno,                      /* Database page number to read data for */
  u32 *piRead                     /* OUT: Frame number (or zero) */
){
  u32 iRead = 0;                  /* If !=0, WAL frame to return data from */
  u32 iLast = pWal->hdr.mxFrame;  /* Last page in WAL for this reader */
  int iHash;                      /* Used to loop through N hash tables */

  int iMinHash;

  /* This routine is only be called from within a read transaction. */
  assert( pWal->readLock>=0 || pWal->lockError );

  /* If the "last page" field of the wal-index header snapshot is 0, then
  ** no data will be read from the wal under any circumstances. Return early
  ** in this case as an optimization.  Likewise, if pWal->readLock==0, 
  ** then the WAL is ignored by the reader so return early, as if the 
  ** WAL were empty.
  */

  if( iLast==0 || pWal->readLock==0 ){


    *piRead = 0;
    return SQLITE_OK;
  }

  /* Search the hash table or tables for an entry matching page number
  ** pgno. Each iteration of the following for() loop searches one
  ** hash table (each hash table indexes up to HASHTABLE_NPAGE frames).
................................................................................
  **   (aPgno[iFrame]==pgno): 
  **     This condition filters out normal hash-table collisions.
  **
  **   (iFrame<=iLast): 
  **     This condition filters out entries that were added to the hash
  **     table after the current read-transaction had started.
  */
  iMinHash = walFramePage(pWal->minFrame);
  for(iHash=walFramePage(iLast); iHash>=iMinHash && iRead==0; iHash--){
    volatile ht_slot *aHash;      /* Pointer to hash table */
    volatile u32 *aPgno;          /* Pointer to array of page numbers */
    u32 iZero;                    /* Frame number corresponding to aPgno[0] */
    int iKey;                     /* Hash slot index */
    int nCollide;                 /* Number of hash collisions remaining */
    int rc;                       /* Error code */
................................................................................
    rc = walHashGet(pWal, iHash, &aHash, &aPgno, &iZero);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    nCollide = HASHTABLE_NSLOT;
    for(iKey=walHash(pgno); aHash[iKey]; iKey=walNextHash(iKey)){
      u32 iFrame = aHash[iKey] + iZero;
      if( iFrame<=iLast && iFrame>=pWal->minFrame && aPgno[aHash[iKey]]==pgno ){
        assert( iFrame>iRead || CORRUPT_DB );
        iRead = iFrame;
      }
      if( (nCollide--)==0 ){
        return SQLITE_CORRUPT_BKPT;
      }
    }