4 ** The author disclaims copyright to this source code. In place of
5 ** a legal notice, here is a blessing:
7 ** May you do good and not evil.
8 ** May you find forgiveness for yourself and forgive others.
9 ** May you share freely, never taking more than you give.
11 *************************************************************************
13 ** This file contains the implementation of a write-ahead log (WAL) used in
14 ** "journal_mode=WAL" mode.
16 ** WRITE-AHEAD LOG (WAL) FILE FORMAT
18 ** A WAL file consists of a header followed by zero or more "frames".
19 ** Each frame records the revised content of a single page from the
20 ** database file. All changes to the database are recorded by writing
21 ** frames into the WAL. Transactions commit when a frame is written that
22 ** contains a commit marker. A single WAL can and usually does record
23 ** multiple transactions. Periodically, the content of the WAL is
24 ** transferred back into the database file in an operation called a
27 ** A single WAL file can be used multiple times. In other words, the
28 ** WAL can fill up with frames and then be checkpointed and then new
29 ** frames can overwrite the old ones. A WAL always grows from beginning
30 ** toward the end. Checksums and counters attached to each frame are
31 ** used to determine which frames within the WAL are valid and which
32 ** are leftovers from prior checkpoints.
34 ** The WAL header is 32 bytes in size and consists of the following eight
35 ** big-endian 32-bit unsigned integer values:
37 ** 0: Magic number. 0x377f0682 or 0x377f0683
38 ** 4: File format version. Currently 3007000
39 ** 8: Database page size. Example: 1024
40 ** 12: Checkpoint sequence number
41 ** 16: Salt-1, random integer incremented with each checkpoint
42 ** 20: Salt-2, a different random integer changing with each ckpt
43 ** 24: Checksum-1 (first part of checksum for first 24 bytes of header).
44 ** 28: Checksum-2 (second part of checksum for first 24 bytes of header).
46 ** Immediately following the wal-header are zero or more frames. Each
47 ** frame consists of a 24-byte frame-header followed by a <page-size> bytes
48 ** of page data. The frame-header is six big-endian 32-bit unsigned
49 ** integer values, as follows:
52 ** 4: For commit records, the size of the database image in pages
53 ** after the commit. For all other records, zero.
54 ** 8: Salt-1 (copied from the header)
55 ** 12: Salt-2 (copied from the header)
59 ** A frame is considered valid if and only if the following conditions are
62 ** (1) The salt-1 and salt-2 values in the frame-header match
63 ** salt values in the wal-header
65 ** (2) The checksum values in the final 8 bytes of the frame-header
66 ** exactly match the checksum computed consecutively on the
67 ** WAL header and the first 8 bytes and the content of all frames
68 ** up to and including the current frame.
70 ** The checksum is computed using 32-bit big-endian integers if the
71 ** magic number in the first 4 bytes of the WAL is 0x377f0683 and it
72 ** is computed using little-endian if the magic number is 0x377f0682.
73 ** The checksum values are always stored in the frame header in a
74 ** big-endian format regardless of which byte order is used to compute
75 ** the checksum. The checksum is computed by interpreting the input as
76 ** an even number of unsigned 32-bit integers: x[0] through x[N]. The
77 ** algorithm used for the checksum is as follows:
79 ** for i from 0 to n-1 step 2:
84 ** Note that s0 and s1 are both weighted checksums using fibonacci weights
85 ** in reverse order (the largest fibonacci weight occurs on the first element
86 ** of the sequence being summed.) The s1 value spans all 32-bit
87 ** terms of the sequence whereas s0 omits the final term.
89 ** On a checkpoint, the WAL is first VFS.xSync-ed, then valid content of the
90 ** WAL is transferred into the database, then the database is VFS.xSync-ed.
91 ** The VFS.xSync operations serve as write barriers - all writes launched
92 ** before the xSync must complete before any write that launches after the
95 ** After each checkpoint, the salt-1 value is incremented and the salt-2
96 ** value is randomized. This prevents old and new frames in the WAL from
97 ** being considered valid at the same time and being checkpointing together
102 ** To read a page from the database (call it page number P), a reader
103 ** first checks the WAL to see if it contains page P. If so, then the
104 ** last valid instance of page P that is a followed by a commit frame
105 ** or is a commit frame itself becomes the value read. If the WAL
106 ** contains no copies of page P that are valid and which are a commit
107 ** frame or are followed by a commit frame, then page P is read from
108 ** the database file.
110 ** To start a read transaction, the reader records the index of the last
111 ** valid frame in the WAL. The reader uses this recorded "mxFrame" value
112 ** for all subsequent read operations. New transactions can be appended
113 ** to the WAL, but as long as the reader uses its original mxFrame value
114 ** and ignores the newly appended content, it will see a consistent snapshot
115 ** of the database from a single point in time. This technique allows
116 ** multiple concurrent readers to view different versions of the database
117 ** content simultaneously.
119 ** The reader algorithm in the previous paragraphs works correctly, but
120 ** because frames for page P can appear anywhere within the WAL, the
121 ** reader has to scan the entire WAL looking for page P frames. If the
122 ** WAL is large (multiple megabytes is typical) that scan can be slow,
123 ** and read performance suffers. To overcome this problem, a separate
124 ** data structure called the wal-index is maintained to expedite the
125 ** search for frames of a particular page.
129 ** Conceptually, the wal-index is shared memory, though VFS implementations
130 ** might choose to implement the wal-index using a mmapped file. Because
131 ** the wal-index is shared memory, SQLite does not support journal_mode=WAL
132 ** on a network filesystem. All users of the database must be able to
135 ** The wal-index is transient. After a crash, the wal-index can (and should
136 ** be) reconstructed from the original WAL file. In fact, the VFS is required
137 ** to either truncate or zero the header of the wal-index when the last
138 ** connection to it closes. Because the wal-index is transient, it can
139 ** use an architecture-specific format; it does not have to be cross-platform.
140 ** Hence, unlike the database and WAL file formats which store all values
141 ** as big endian, the wal-index can store multi-byte values in the native
142 ** byte order of the host computer.
144 ** The purpose of the wal-index is to answer this question quickly: Given
145 ** a page number P and a maximum frame index M, return the index of the
146 ** last frame in the wal before frame M for page P in the WAL, or return
147 ** NULL if there are no frames for page P in the WAL prior to M.
149 ** The wal-index consists of a header region, followed by an one or
150 ** more index blocks.
152 ** The wal-index header contains the total number of frames within the WAL
153 ** in the mxFrame field.
155 ** Each index block except for the first contains information on
156 ** HASHTABLE_NPAGE frames. The first index block contains information on
157 ** HASHTABLE_NPAGE_ONE frames. The values of HASHTABLE_NPAGE_ONE and
158 ** HASHTABLE_NPAGE are selected so that together the wal-index header and
159 ** first index block are the same size as all other index blocks in the
162 ** Each index block contains two sections, a page-mapping that contains the
163 ** database page number associated with each wal frame, and a hash-table
164 ** that allows readers to query an index block for a specific page number.
165 ** The page-mapping is an array of HASHTABLE_NPAGE (or HASHTABLE_NPAGE_ONE
166 ** for the first index block) 32-bit page numbers. The first entry in the
167 ** first index-block contains the database page number corresponding to the
168 ** first frame in the WAL file. The first entry in the second index block
169 ** in the WAL file corresponds to the (HASHTABLE_NPAGE_ONE+1)th frame in
170 ** the log, and so on.
172 ** The last index block in a wal-index usually contains less than the full
173 ** complement of HASHTABLE_NPAGE (or HASHTABLE_NPAGE_ONE) page-numbers,
174 ** depending on the contents of the WAL file. This does not change the
175 ** allocated size of the page-mapping array - the page-mapping array merely
176 ** contains unused entries.
178 ** Even without using the hash table, the last frame for page P
179 ** can be found by scanning the page-mapping sections of each index block
180 ** starting with the last index block and moving toward the first, and
181 ** within each index block, starting at the end and moving toward the
182 ** beginning. The first entry that equals P corresponds to the frame
183 ** holding the content for that page.
185 ** The hash table consists of HASHTABLE_NSLOT 16-bit unsigned integers.
186 ** HASHTABLE_NSLOT = 2*HASHTABLE_NPAGE, and there is one entry in the
187 ** hash table for each page number in the mapping section, so the hash
188 ** table is never more than half full. The expected number of collisions
189 ** prior to finding a match is 1. Each entry of the hash table is an
190 ** 1-based index of an entry in the mapping section of the same
191 ** index block. Let K be the 1-based index of the largest entry in
192 ** the mapping section. (For index blocks other than the last, K will
193 ** always be exactly HASHTABLE_NPAGE (4096) and for the last index block
194 ** K will be (mxFrame%HASHTABLE_NPAGE).) Unused slots of the hash table
195 ** contain a value of 0.
197 ** To look for page P in the hash table, first compute a hash iKey on
200 ** iKey = (P * 383) % HASHTABLE_NSLOT
202 ** Then start scanning entries of the hash table, starting with iKey
203 ** (wrapping around to the beginning when the end of the hash table is
204 ** reached) until an unused hash slot is found. Let the first unused slot
205 ** be at index iUnused. (iUnused might be less than iKey if there was
206 ** wrap-around.) Because the hash table is never more than half full,
207 ** the search is guaranteed to eventually hit an unused entry. Let
208 ** iMax be the value between iKey and iUnused, closest to iUnused,
209 ** where aHash[iMax]==P. If there is no iMax entry (if there exists
210 ** no hash slot such that aHash[i]==p) then page P is not in the
211 ** current index block. Otherwise the iMax-th mapping entry of the
212 ** current index block corresponds to the last entry that references
215 ** A hash search begins with the last index block and moves toward the
216 ** first index block, looking for entries corresponding to page P. On
217 ** average, only two or three slots in each index block need to be
218 ** examined in order to either find the last entry for page P, or to
219 ** establish that no such entry exists in the block. Each index block
220 ** holds over 4000 entries. So two or three index blocks are sufficient
221 ** to cover a typical 10 megabyte WAL file, assuming 1K pages. 8 or 10
222 ** comparisons (on average) suffice to either locate a frame in the
223 ** WAL or to establish that the frame does not exist in the WAL. This
224 ** is much faster than scanning the entire 10MB WAL.
226 ** Note that entries are added in order of increasing K. Hence, one
227 ** reader might be using some value K0 and a second reader that started
228 ** at a later time (after additional transactions were added to the WAL
229 ** and to the wal-index) might be using a different value K1, where K1>K0.
230 ** Both readers can use the same hash table and mapping section to get
231 ** the correct result. There may be entries in the hash table with
232 ** K>K0 but to the first reader, those entries will appear to be unused
233 ** slots in the hash table and so the first reader will get an answer as
234 ** if no values greater than K0 had ever been inserted into the hash table
235 ** in the first place - which is what reader one wants. Meanwhile, the
236 ** second reader using K1 will see additional values that were inserted
237 ** later, which is exactly what reader two wants.
239 ** When a rollback occurs, the value of K is decreased. Hash table entries
240 ** that correspond to frames greater than the new K value are removed
241 ** from the hash table at this point.
243 #ifndef SQLITE_OMIT_WAL
248 ** Trace output macros
250 #if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
251 int sqlite3WalTrace
= 0;
252 # define WALTRACE(X) if(sqlite3WalTrace) sqlite3DebugPrintf X
258 ** The maximum (and only) versions of the wal and wal-index formats
259 ** that may be interpreted by this version of SQLite.
261 ** If a client begins recovering a WAL file and finds that (a) the checksum
262 ** values in the wal-header are correct and (b) the version field is not
263 ** WAL_MAX_VERSION, recovery fails and SQLite returns SQLITE_CANTOPEN.
265 ** Similarly, if a client successfully reads a wal-index header (i.e. the
266 ** checksum test is successful) and finds that the version field is not
267 ** WALINDEX_MAX_VERSION, then no read-transaction is opened and SQLite
268 ** returns SQLITE_CANTOPEN.
270 #define WAL_MAX_VERSION 3007000
271 #define WALINDEX_MAX_VERSION 3007000
274 ** Indices of various locking bytes. WAL_NREADER is the number
275 ** of available reader locks and should be at least 3. The default
276 ** is SQLITE_SHM_NLOCK==8 and WAL_NREADER==5.
278 #define WAL_WRITE_LOCK 0
279 #define WAL_ALL_BUT_WRITE 1
280 #define WAL_CKPT_LOCK 1
281 #define WAL_RECOVER_LOCK 2
282 #define WAL_READ_LOCK(I) (3+(I))
283 #define WAL_NREADER (SQLITE_SHM_NLOCK-3)
286 /* Object declarations */
287 typedef struct WalIndexHdr WalIndexHdr
;
288 typedef struct WalIterator WalIterator
;
289 typedef struct WalCkptInfo WalCkptInfo
;
293 ** The following object holds a copy of the wal-index header content.
295 ** The actual header in the wal-index consists of two copies of this
296 ** object followed by one instance of the WalCkptInfo object.
297 ** For all versions of SQLite through 3.10.0 and probably beyond,
298 ** the locking bytes (WalCkptInfo.aLock) start at offset 120 and
299 ** the total header size is 136 bytes.
301 ** The szPage value can be any power of 2 between 512 and 32768, inclusive.
302 ** Or it can be 1 to represent a 65536-byte page. The latter case was
303 ** added in 3.7.1 when support for 64K pages was added.
306 u32 iVersion
; /* Wal-index version */
307 u32 unused
; /* Unused (padding) field */
308 u32 iChange
; /* Counter incremented each transaction */
309 u8 isInit
; /* 1 when initialized */
310 u8 bigEndCksum
; /* True if checksums in WAL are big-endian */
311 u16 szPage
; /* Database page size in bytes. 1==64K */
312 u32 mxFrame
; /* Index of last valid frame in the WAL */
313 u32 nPage
; /* Size of database in pages */
314 u32 aFrameCksum
[2]; /* Checksum of last frame in log */
315 u32 aSalt
[2]; /* Two salt values copied from WAL header */
316 u32 aCksum
[2]; /* Checksum over all prior fields */
320 ** A copy of the following object occurs in the wal-index immediately
321 ** following the second copy of the WalIndexHdr. This object stores
322 ** information used by checkpoint.
324 ** nBackfill is the number of frames in the WAL that have been written
325 ** back into the database. (We call the act of moving content from WAL to
326 ** database "backfilling".) The nBackfill number is never greater than
327 ** WalIndexHdr.mxFrame. nBackfill can only be increased by threads
328 ** holding the WAL_CKPT_LOCK lock (which includes a recovery thread).
329 ** However, a WAL_WRITE_LOCK thread can move the value of nBackfill from
330 ** mxFrame back to zero when the WAL is reset.
332 ** nBackfillAttempted is the largest value of nBackfill that a checkpoint
333 ** has attempted to achieve. Normally nBackfill==nBackfillAtempted, however
334 ** the nBackfillAttempted is set before any backfilling is done and the
335 ** nBackfill is only set after all backfilling completes. So if a checkpoint
336 ** crashes, nBackfillAttempted might be larger than nBackfill. The
337 ** WalIndexHdr.mxFrame must never be less than nBackfillAttempted.
339 ** The aLock[] field is a set of bytes used for locking. These bytes should
340 ** never be read or written.
342 ** There is one entry in aReadMark[] for each reader lock. If a reader
343 ** holds read-lock K, then the value in aReadMark[K] is no greater than
344 ** the mxFrame for that reader. The value READMARK_NOT_USED (0xffffffff)
345 ** for any aReadMark[] means that entry is unused. aReadMark[0] is
346 ** a special case; its value is never used and it exists as a place-holder
347 ** to avoid having to offset aReadMark[] indexs by one. Readers holding
348 ** WAL_READ_LOCK(0) always ignore the entire WAL and read all content
349 ** directly from the database.
351 ** The value of aReadMark[K] may only be changed by a thread that
352 ** is holding an exclusive lock on WAL_READ_LOCK(K). Thus, the value of
353 ** aReadMark[K] cannot changed while there is a reader is using that mark
354 ** since the reader will be holding a shared lock on WAL_READ_LOCK(K).
356 ** The checkpointer may only transfer frames from WAL to database where
357 ** the frame numbers are less than or equal to every aReadMark[] that is
358 ** in use (that is, every aReadMark[j] for which there is a corresponding
359 ** WAL_READ_LOCK(j)). New readers (usually) pick the aReadMark[] with the
360 ** largest value and will increase an unused aReadMark[] to mxFrame if there
361 ** is not already an aReadMark[] equal to mxFrame. The exception to the
362 ** previous sentence is when nBackfill equals mxFrame (meaning that everything
363 ** in the WAL has been backfilled into the database) then new readers
364 ** will choose aReadMark[0] which has value 0 and hence such reader will
365 ** get all their all content directly from the database file and ignore
368 ** Writers normally append new frames to the end of the WAL. However,
369 ** if nBackfill equals mxFrame (meaning that all WAL content has been
370 ** written back into the database) and if no readers are using the WAL
371 ** (in other words, if there are no WAL_READ_LOCK(i) where i>0) then
372 ** the writer will first "reset" the WAL back to the beginning and start
373 ** writing new content beginning at frame 1.
375 ** We assume that 32-bit loads are atomic and so no locks are needed in
376 ** order to read from any aReadMark[] entries.
379 u32 nBackfill
; /* Number of WAL frames backfilled into DB */
380 u32 aReadMark
[WAL_NREADER
]; /* Reader marks */
381 u8 aLock
[SQLITE_SHM_NLOCK
]; /* Reserved space for locks */
382 u32 nBackfillAttempted
; /* WAL frames perhaps written, or maybe not */
383 u32 notUsed0
; /* Available for future enhancements */
385 #define READMARK_NOT_USED 0xffffffff
388 /* A block of WALINDEX_LOCK_RESERVED bytes beginning at
389 ** WALINDEX_LOCK_OFFSET is reserved for locks. Since some systems
390 ** only support mandatory file-locks, we do not read or write data
391 ** from the region of the file on which locks are applied.
393 #define WALINDEX_LOCK_OFFSET (sizeof(WalIndexHdr)*2+offsetof(WalCkptInfo,aLock))
394 #define WALINDEX_HDR_SIZE (sizeof(WalIndexHdr)*2+sizeof(WalCkptInfo))
396 /* Size of header before each frame in wal */
397 #define WAL_FRAME_HDRSIZE 24
399 /* Size of write ahead log header, including checksum. */
400 /* #define WAL_HDRSIZE 24 */
401 #define WAL_HDRSIZE 32
403 /* WAL magic value. Either this value, or the same value with the least
404 ** significant bit also set (WAL_MAGIC | 0x00000001) is stored in 32-bit
405 ** big-endian format in the first 4 bytes of a WAL file.
407 ** If the LSB is set, then the checksums for each frame within the WAL
408 ** file are calculated by treating all data as an array of 32-bit
409 ** big-endian words. Otherwise, they are calculated by interpreting
410 ** all data as 32-bit little-endian words.
412 #define WAL_MAGIC 0x377f0682
415 ** Return the offset of frame iFrame in the write-ahead log file,
416 ** assuming a database page size of szPage bytes. The offset returned
417 ** is to the start of the write-ahead log frame-header.
419 #define walFrameOffset(iFrame, szPage) ( \
420 WAL_HDRSIZE + ((iFrame)-1)*(i64)((szPage)+WAL_FRAME_HDRSIZE) \
424 ** An open write-ahead log file is represented by an instance of the
428 sqlite3_vfs
*pVfs
; /* The VFS used to create pDbFd */
429 sqlite3_file
*pDbFd
; /* File handle for the database file */
430 sqlite3_file
*pWalFd
; /* File handle for WAL file */
431 u32 iCallback
; /* Value to pass to log callback (or 0) */
432 i64 mxWalSize
; /* Truncate WAL to this size upon reset */
433 int nWiData
; /* Size of array apWiData */
434 int szFirstBlock
; /* Size of first block written to WAL file */
435 volatile u32
**apWiData
; /* Pointer to wal-index content in memory */
436 u32 szPage
; /* Database page size */
437 i16 readLock
; /* Which read lock is being held. -1 for none */
438 u8 syncFlags
; /* Flags to use to sync header writes */
439 u8 exclusiveMode
; /* Non-zero if connection is in exclusive mode */
440 u8 writeLock
; /* True if in a write transaction */
441 u8 ckptLock
; /* True if holding a checkpoint lock */
442 u8 readOnly
; /* WAL_RDWR, WAL_RDONLY, or WAL_SHM_RDONLY */
443 u8 truncateOnCommit
; /* True to truncate WAL file on commit */
444 u8 syncHeader
; /* Fsync the WAL header if true */
445 u8 padToSectorBoundary
; /* Pad transactions out to the next sector */
446 WalIndexHdr hdr
; /* Wal-index header for current transaction */
447 u32 minFrame
; /* Ignore wal frames before this one */
448 u32 iReCksum
; /* On commit, recalculate checksums from here */
449 const char *zWalName
; /* Name of WAL file */
450 u32 nCkpt
; /* Checkpoint sequence counter in the wal-header */
452 u8 lockError
; /* True if a locking error has occurred */
454 #ifdef SQLITE_ENABLE_SNAPSHOT
455 WalIndexHdr
*pSnapshot
; /* Start transaction here if not NULL */
460 ** Candidate values for Wal.exclusiveMode.
462 #define WAL_NORMAL_MODE 0
463 #define WAL_EXCLUSIVE_MODE 1
464 #define WAL_HEAPMEMORY_MODE 2
467 ** Possible values for WAL.readOnly
469 #define WAL_RDWR 0 /* Normal read/write connection */
470 #define WAL_RDONLY 1 /* The WAL file is readonly */
471 #define WAL_SHM_RDONLY 2 /* The SHM file is readonly */
474 ** Each page of the wal-index mapping contains a hash-table made up of
475 ** an array of HASHTABLE_NSLOT elements of the following type.
480 ** This structure is used to implement an iterator that loops through
481 ** all frames in the WAL in database page order. Where two or more frames
482 ** correspond to the same database page, the iterator visits only the
483 ** frame most recently written to the WAL (in other words, the frame with
484 ** the largest index).
486 ** The internals of this structure are only accessed by:
488 ** walIteratorInit() - Create a new iterator,
489 ** walIteratorNext() - Step an iterator,
490 ** walIteratorFree() - Free an iterator.
492 ** This functionality is used by the checkpoint code (see walCheckpoint()).
495 int iPrior
; /* Last result returned from the iterator */
496 int nSegment
; /* Number of entries in aSegment[] */
498 int iNext
; /* Next slot in aIndex[] not yet returned */
499 ht_slot
*aIndex
; /* i0, i1, i2... such that aPgno[iN] ascend */
500 u32
*aPgno
; /* Array of page numbers. */
501 int nEntry
; /* Nr. of entries in aPgno[] and aIndex[] */
502 int iZero
; /* Frame number associated with aPgno[0] */
503 } aSegment
[1]; /* One for every 32KB page in the wal-index */
507 ** Define the parameters of the hash tables in the wal-index file. There
508 ** is a hash-table following every HASHTABLE_NPAGE page numbers in the
511 ** Changing any of these constants will alter the wal-index format and
512 ** create incompatibilities.
514 #define HASHTABLE_NPAGE 4096 /* Must be power of 2 */
515 #define HASHTABLE_HASH_1 383 /* Should be prime */
516 #define HASHTABLE_NSLOT (HASHTABLE_NPAGE*2) /* Must be a power of 2 */
519 ** The block of page numbers associated with the first hash-table in a
520 ** wal-index is smaller than usual. This is so that there is a complete
521 ** hash-table on each aligned 32KB page of the wal-index.
523 #define HASHTABLE_NPAGE_ONE (HASHTABLE_NPAGE - (WALINDEX_HDR_SIZE/sizeof(u32)))
525 /* The wal-index is divided into pages of WALINDEX_PGSZ bytes each. */
526 #define WALINDEX_PGSZ ( \
527 sizeof(ht_slot)*HASHTABLE_NSLOT + HASHTABLE_NPAGE*sizeof(u32) \
531 ** Obtain a pointer to the iPage'th page of the wal-index. The wal-index
532 ** is broken into pages of WALINDEX_PGSZ bytes. Wal-index pages are
533 ** numbered from zero.
535 ** If this call is successful, *ppPage is set to point to the wal-index
536 ** page and SQLITE_OK is returned. If an error (an OOM or VFS error) occurs,
537 ** then an SQLite error code is returned and *ppPage is set to 0.
539 static int walIndexPage(Wal
*pWal
, int iPage
, volatile u32
**ppPage
){
542 /* Enlarge the pWal->apWiData[] array if required */
543 if( pWal
->nWiData
<=iPage
){
544 int nByte
= sizeof(u32
*)*(iPage
+1);
545 volatile u32
**apNew
;
546 apNew
= (volatile u32
**)sqlite3_realloc64((void *)pWal
->apWiData
, nByte
);
549 return SQLITE_NOMEM_BKPT
;
551 memset((void*)&apNew
[pWal
->nWiData
], 0,
552 sizeof(u32
*)*(iPage
+1-pWal
->nWiData
));
553 pWal
->apWiData
= apNew
;
554 pWal
->nWiData
= iPage
+1;
557 /* Request a pointer to the required page from the VFS */
558 if( pWal
->apWiData
[iPage
]==0 ){
559 if( pWal
->exclusiveMode
==WAL_HEAPMEMORY_MODE
){
560 pWal
->apWiData
[iPage
] = (u32
volatile *)sqlite3MallocZero(WALINDEX_PGSZ
);
561 if( !pWal
->apWiData
[iPage
] ) rc
= SQLITE_NOMEM_BKPT
;
563 rc
= sqlite3OsShmMap(pWal
->pDbFd
, iPage
, WALINDEX_PGSZ
,
564 pWal
->writeLock
, (void volatile **)&pWal
->apWiData
[iPage
]
566 if( rc
==SQLITE_READONLY
){
567 pWal
->readOnly
|= WAL_SHM_RDONLY
;
573 *ppPage
= pWal
->apWiData
[iPage
];
574 assert( iPage
==0 || *ppPage
|| rc
!=SQLITE_OK
);
579 ** Return a pointer to the WalCkptInfo structure in the wal-index.
581 static volatile WalCkptInfo
*walCkptInfo(Wal
*pWal
){
582 assert( pWal
->nWiData
>0 && pWal
->apWiData
[0] );
583 return (volatile WalCkptInfo
*)&(pWal
->apWiData
[0][sizeof(WalIndexHdr
)/2]);
587 ** Return a pointer to the WalIndexHdr structure in the wal-index.
589 static volatile WalIndexHdr
*walIndexHdr(Wal
*pWal
){
590 assert( pWal
->nWiData
>0 && pWal
->apWiData
[0] );
591 return (volatile WalIndexHdr
*)pWal
->apWiData
[0];
595 ** The argument to this macro must be of type u32. On a little-endian
596 ** architecture, it returns the u32 value that results from interpreting
597 ** the 4 bytes as a big-endian value. On a big-endian architecture, it
598 ** returns the value that would be produced by interpreting the 4 bytes
599 ** of the input value as a little-endian integer.
601 #define BYTESWAP32(x) ( \
602 (((x)&0x000000FF)<<24) + (((x)&0x0000FF00)<<8) \
603 + (((x)&0x00FF0000)>>8) + (((x)&0xFF000000)>>24) \
607 ** Generate or extend an 8 byte checksum based on the data in
608 ** array aByte[] and the initial values of aIn[0] and aIn[1] (or
609 ** initial values of 0 and 0 if aIn==NULL).
611 ** The checksum is written back into aOut[] before returning.
613 ** nByte must be a positive multiple of 8.
615 static void walChecksumBytes(
616 int nativeCksum
, /* True for native byte-order, false for non-native */
617 u8
*a
, /* Content to be checksummed */
618 int nByte
, /* Bytes of content in a[]. Must be a multiple of 8. */
619 const u32
*aIn
, /* Initial checksum value input */
620 u32
*aOut
/* OUT: Final checksum value output */
623 u32
*aData
= (u32
*)a
;
624 u32
*aEnd
= (u32
*)&a
[nByte
];
634 assert( (nByte
&0x00000007)==0 );
640 }while( aData
<aEnd
);
643 s1
+= BYTESWAP32(aData
[0]) + s2
;
644 s2
+= BYTESWAP32(aData
[1]) + s1
;
646 }while( aData
<aEnd
);
653 static void walShmBarrier(Wal
*pWal
){
654 if( pWal
->exclusiveMode
!=WAL_HEAPMEMORY_MODE
){
655 sqlite3OsShmBarrier(pWal
->pDbFd
);
660 ** Write the header information in pWal->hdr into the wal-index.
662 ** The checksum on pWal->hdr is updated before it is written.
664 static void walIndexWriteHdr(Wal
*pWal
){
665 volatile WalIndexHdr
*aHdr
= walIndexHdr(pWal
);
666 const int nCksum
= offsetof(WalIndexHdr
, aCksum
);
668 assert( pWal
->writeLock
);
669 pWal
->hdr
.isInit
= 1;
670 pWal
->hdr
.iVersion
= WALINDEX_MAX_VERSION
;
671 walChecksumBytes(1, (u8
*)&pWal
->hdr
, nCksum
, 0, pWal
->hdr
.aCksum
);
672 memcpy((void*)&aHdr
[1], (const void*)&pWal
->hdr
, sizeof(WalIndexHdr
));
674 memcpy((void*)&aHdr
[0], (const void*)&pWal
->hdr
, sizeof(WalIndexHdr
));
678 ** This function encodes a single frame header and writes it to a buffer
679 ** supplied by the caller. A frame-header is made up of a series of
680 ** 4-byte big-endian integers, as follows:
683 ** 4: For commit records, the size of the database image in pages
684 ** after the commit. For all other records, zero.
685 ** 8: Salt-1 (copied from the wal-header)
686 ** 12: Salt-2 (copied from the wal-header)
690 static void walEncodeFrame(
691 Wal
*pWal
, /* The write-ahead log */
692 u32 iPage
, /* Database page number for frame */
693 u32 nTruncate
, /* New db size (or 0 for non-commit frames) */
694 u8
*aData
, /* Pointer to page data */
695 u8
*aFrame
/* OUT: Write encoded frame here */
697 int nativeCksum
; /* True for native byte-order checksums */
698 u32
*aCksum
= pWal
->hdr
.aFrameCksum
;
699 assert( WAL_FRAME_HDRSIZE
==24 );
700 sqlite3Put4byte(&aFrame
[0], iPage
);
701 sqlite3Put4byte(&aFrame
[4], nTruncate
);
702 if( pWal
->iReCksum
==0 ){
703 memcpy(&aFrame
[8], pWal
->hdr
.aSalt
, 8);
705 nativeCksum
= (pWal
->hdr
.bigEndCksum
==SQLITE_BIGENDIAN
);
706 walChecksumBytes(nativeCksum
, aFrame
, 8, aCksum
, aCksum
);
707 walChecksumBytes(nativeCksum
, aData
, pWal
->szPage
, aCksum
, aCksum
);
709 sqlite3Put4byte(&aFrame
[16], aCksum
[0]);
710 sqlite3Put4byte(&aFrame
[20], aCksum
[1]);
712 memset(&aFrame
[8], 0, 16);
717 ** Check to see if the frame with header in aFrame[] and content
718 ** in aData[] is valid. If it is a valid frame, fill *piPage and
719 ** *pnTruncate and return true. Return if the frame is not valid.
721 static int walDecodeFrame(
722 Wal
*pWal
, /* The write-ahead log */
723 u32
*piPage
, /* OUT: Database page number for frame */
724 u32
*pnTruncate
, /* OUT: New db size (or 0 if not commit) */
725 u8
*aData
, /* Pointer to page data (for checksum) */
726 u8
*aFrame
/* Frame data */
728 int nativeCksum
; /* True for native byte-order checksums */
729 u32
*aCksum
= pWal
->hdr
.aFrameCksum
;
730 u32 pgno
; /* Page number of the frame */
731 assert( WAL_FRAME_HDRSIZE
==24 );
733 /* A frame is only valid if the salt values in the frame-header
734 ** match the salt values in the wal-header.
736 if( memcmp(&pWal
->hdr
.aSalt
, &aFrame
[8], 8)!=0 ){
740 /* A frame is only valid if the page number is creater than zero.
742 pgno
= sqlite3Get4byte(&aFrame
[0]);
747 /* A frame is only valid if a checksum of the WAL header,
748 ** all prior frams, the first 16 bytes of this frame-header,
749 ** and the frame-data matches the checksum in the last 8
750 ** bytes of this frame-header.
752 nativeCksum
= (pWal
->hdr
.bigEndCksum
==SQLITE_BIGENDIAN
);
753 walChecksumBytes(nativeCksum
, aFrame
, 8, aCksum
, aCksum
);
754 walChecksumBytes(nativeCksum
, aData
, pWal
->szPage
, aCksum
, aCksum
);
755 if( aCksum
[0]!=sqlite3Get4byte(&aFrame
[16])
756 || aCksum
[1]!=sqlite3Get4byte(&aFrame
[20])
758 /* Checksum failed. */
762 /* If we reach this point, the frame is valid. Return the page number
763 ** and the new database size.
766 *pnTruncate
= sqlite3Get4byte(&aFrame
[4]);
771 #if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
773 ** Names of locks. This routine is used to provide debugging output and is not
774 ** a part of an ordinary build.
776 static const char *walLockName(int lockIdx
){
777 if( lockIdx
==WAL_WRITE_LOCK
){
779 }else if( lockIdx
==WAL_CKPT_LOCK
){
781 }else if( lockIdx
==WAL_RECOVER_LOCK
){
782 return "RECOVER-LOCK";
784 static char zName
[15];
785 sqlite3_snprintf(sizeof(zName
), zName
, "READ-LOCK[%d]",
786 lockIdx
-WAL_READ_LOCK(0));
790 #endif /*defined(SQLITE_TEST) || defined(SQLITE_DEBUG) */
794 ** Set or release locks on the WAL. Locks are either shared or exclusive.
795 ** A lock cannot be moved directly between shared and exclusive - it must go
796 ** through the unlocked state first.
798 ** In locking_mode=EXCLUSIVE, all of these routines become no-ops.
800 static int walLockShared(Wal
*pWal
, int lockIdx
){
802 if( pWal
->exclusiveMode
) return SQLITE_OK
;
803 rc
= sqlite3OsShmLock(pWal
->pDbFd
, lockIdx
, 1,
804 SQLITE_SHM_LOCK
| SQLITE_SHM_SHARED
);
805 WALTRACE(("WAL%p: acquire SHARED-%s %s\n", pWal
,
806 walLockName(lockIdx
), rc
? "failed" : "ok"));
807 VVA_ONLY( pWal
->lockError
= (u8
)(rc
!=SQLITE_OK
&& rc
!=SQLITE_BUSY
); )
810 static void walUnlockShared(Wal
*pWal
, int lockIdx
){
811 if( pWal
->exclusiveMode
) return;
812 (void)sqlite3OsShmLock(pWal
->pDbFd
, lockIdx
, 1,
813 SQLITE_SHM_UNLOCK
| SQLITE_SHM_SHARED
);
814 WALTRACE(("WAL%p: release SHARED-%s\n", pWal
, walLockName(lockIdx
)));
816 static int walLockExclusive(Wal
*pWal
, int lockIdx
, int n
){
818 if( pWal
->exclusiveMode
) return SQLITE_OK
;
819 rc
= sqlite3OsShmLock(pWal
->pDbFd
, lockIdx
, n
,
820 SQLITE_SHM_LOCK
| SQLITE_SHM_EXCLUSIVE
);
821 WALTRACE(("WAL%p: acquire EXCLUSIVE-%s cnt=%d %s\n", pWal
,
822 walLockName(lockIdx
), n
, rc
? "failed" : "ok"));
823 VVA_ONLY( pWal
->lockError
= (u8
)(rc
!=SQLITE_OK
&& rc
!=SQLITE_BUSY
); )
826 static void walUnlockExclusive(Wal
*pWal
, int lockIdx
, int n
){
827 if( pWal
->exclusiveMode
) return;
828 (void)sqlite3OsShmLock(pWal
->pDbFd
, lockIdx
, n
,
829 SQLITE_SHM_UNLOCK
| SQLITE_SHM_EXCLUSIVE
);
830 WALTRACE(("WAL%p: release EXCLUSIVE-%s cnt=%d\n", pWal
,
831 walLockName(lockIdx
), n
));
835 ** Compute a hash on a page number. The resulting hash value must land
836 ** between 0 and (HASHTABLE_NSLOT-1). The walHashNext() function advances
837 ** the hash to the next value in the event of a collision.
839 static int walHash(u32 iPage
){
841 assert( (HASHTABLE_NSLOT
& (HASHTABLE_NSLOT
-1))==0 );
842 return (iPage
*HASHTABLE_HASH_1
) & (HASHTABLE_NSLOT
-1);
844 static int walNextHash(int iPriorHash
){
845 return (iPriorHash
+1)&(HASHTABLE_NSLOT
-1);
849 ** Return pointers to the hash table and page number array stored on
850 ** page iHash of the wal-index. The wal-index is broken into 32KB pages
851 ** numbered starting from 0.
853 ** Set output variable *paHash to point to the start of the hash table
854 ** in the wal-index file. Set *piZero to one less than the frame
855 ** number of the first frame indexed by this hash table. If a
856 ** slot in the hash table is set to N, it refers to frame number
857 ** (*piZero+N) in the log.
859 ** Finally, set *paPgno so that *paPgno[1] is the page number of the
860 ** first frame indexed by the hash table, frame (*piZero+1).
862 static int walHashGet(
863 Wal
*pWal
, /* WAL handle */
864 int iHash
, /* Find the iHash'th table */
865 volatile ht_slot
**paHash
, /* OUT: Pointer to hash index */
866 volatile u32
**paPgno
, /* OUT: Pointer to page number array */
867 u32
*piZero
/* OUT: Frame associated with *paPgno[0] */
869 int rc
; /* Return code */
872 rc
= walIndexPage(pWal
, iHash
, &aPgno
);
873 assert( rc
==SQLITE_OK
|| iHash
>0 );
877 volatile ht_slot
*aHash
;
879 aHash
= (volatile ht_slot
*)&aPgno
[HASHTABLE_NPAGE
];
881 aPgno
= &aPgno
[WALINDEX_HDR_SIZE
/sizeof(u32
)];
884 iZero
= HASHTABLE_NPAGE_ONE
+ (iHash
-1)*HASHTABLE_NPAGE
;
887 *paPgno
= &aPgno
[-1];
895 ** Return the number of the wal-index page that contains the hash-table
896 ** and page-number array that contain entries corresponding to WAL frame
897 ** iFrame. The wal-index is broken up into 32KB pages. Wal-index pages
898 ** are numbered starting from 0.
900 static int walFramePage(u32 iFrame
){
901 int iHash
= (iFrame
+HASHTABLE_NPAGE
-HASHTABLE_NPAGE_ONE
-1) / HASHTABLE_NPAGE
;
902 assert( (iHash
==0 || iFrame
>HASHTABLE_NPAGE_ONE
)
903 && (iHash
>=1 || iFrame
<=HASHTABLE_NPAGE_ONE
)
904 && (iHash
<=1 || iFrame
>(HASHTABLE_NPAGE_ONE
+HASHTABLE_NPAGE
))
905 && (iHash
>=2 || iFrame
<=HASHTABLE_NPAGE_ONE
+HASHTABLE_NPAGE
)
906 && (iHash
<=2 || iFrame
>(HASHTABLE_NPAGE_ONE
+2*HASHTABLE_NPAGE
))
912 ** Return the page number associated with frame iFrame in this WAL.
914 static u32
walFramePgno(Wal
*pWal
, u32 iFrame
){
915 int iHash
= walFramePage(iFrame
);
917 return pWal
->apWiData
[0][WALINDEX_HDR_SIZE
/sizeof(u32
) + iFrame
- 1];
919 return pWal
->apWiData
[iHash
][(iFrame
-1-HASHTABLE_NPAGE_ONE
)%HASHTABLE_NPAGE
];
923 ** Remove entries from the hash table that point to WAL slots greater
924 ** than pWal->hdr.mxFrame.
926 ** This function is called whenever pWal->hdr.mxFrame is decreased due
927 ** to a rollback or savepoint.
929 ** At most only the hash table containing pWal->hdr.mxFrame needs to be
930 ** updated. Any later hash tables will be automatically cleared when
931 ** pWal->hdr.mxFrame advances to the point where those hash tables are
934 static void walCleanupHash(Wal
*pWal
){
935 volatile ht_slot
*aHash
= 0; /* Pointer to hash table to clear */
936 volatile u32
*aPgno
= 0; /* Page number array for hash table */
937 u32 iZero
= 0; /* frame == (aHash[x]+iZero) */
938 int iLimit
= 0; /* Zero values greater than this */
939 int nByte
; /* Number of bytes to zero in aPgno[] */
940 int i
; /* Used to iterate through aHash[] */
942 assert( pWal
->writeLock
);
943 testcase( pWal
->hdr
.mxFrame
==HASHTABLE_NPAGE_ONE
-1 );
944 testcase( pWal
->hdr
.mxFrame
==HASHTABLE_NPAGE_ONE
);
945 testcase( pWal
->hdr
.mxFrame
==HASHTABLE_NPAGE_ONE
+1 );
947 if( pWal
->hdr
.mxFrame
==0 ) return;
949 /* Obtain pointers to the hash-table and page-number array containing
950 ** the entry that corresponds to frame pWal->hdr.mxFrame. It is guaranteed
951 ** that the page said hash-table and array reside on is already mapped.
953 assert( pWal
->nWiData
>walFramePage(pWal
->hdr
.mxFrame
) );
954 assert( pWal
->apWiData
[walFramePage(pWal
->hdr
.mxFrame
)] );
955 walHashGet(pWal
, walFramePage(pWal
->hdr
.mxFrame
), &aHash
, &aPgno
, &iZero
);
957 /* Zero all hash-table entries that correspond to frame numbers greater
958 ** than pWal->hdr.mxFrame.
960 iLimit
= pWal
->hdr
.mxFrame
- iZero
;
962 for(i
=0; i
<HASHTABLE_NSLOT
; i
++){
963 if( aHash
[i
]>iLimit
){
968 /* Zero the entries in the aPgno array that correspond to frames with
969 ** frame numbers greater than pWal->hdr.mxFrame.
971 nByte
= (int)((char *)aHash
- (char *)&aPgno
[iLimit
+1]);
972 memset((void *)&aPgno
[iLimit
+1], 0, nByte
);
974 #ifdef SQLITE_ENABLE_EXPENSIVE_ASSERT
975 /* Verify that the every entry in the mapping region is still reachable
976 ** via the hash table even after the cleanup.
979 int j
; /* Loop counter */
980 int iKey
; /* Hash key */
981 for(j
=1; j
<=iLimit
; j
++){
982 for(iKey
=walHash(aPgno
[j
]); aHash
[iKey
]; iKey
=walNextHash(iKey
)){
983 if( aHash
[iKey
]==j
) break;
985 assert( aHash
[iKey
]==j
);
988 #endif /* SQLITE_ENABLE_EXPENSIVE_ASSERT */
993 ** Set an entry in the wal-index that will map database page number
994 ** pPage into WAL frame iFrame.
996 static int walIndexAppend(Wal
*pWal
, u32 iFrame
, u32 iPage
){
997 int rc
; /* Return code */
998 u32 iZero
= 0; /* One less than frame number of aPgno[1] */
999 volatile u32
*aPgno
= 0; /* Page number array */
1000 volatile ht_slot
*aHash
= 0; /* Hash table */
1002 rc
= walHashGet(pWal
, walFramePage(iFrame
), &aHash
, &aPgno
, &iZero
);
1004 /* Assuming the wal-index file was successfully mapped, populate the
1005 ** page number array and hash table entry.
1007 if( rc
==SQLITE_OK
){
1008 int iKey
; /* Hash table key */
1009 int idx
; /* Value to write to hash-table slot */
1010 int nCollide
; /* Number of hash collisions */
1012 idx
= iFrame
- iZero
;
1013 assert( idx
<= HASHTABLE_NSLOT
/2 + 1 );
1015 /* If this is the first entry to be added to this hash-table, zero the
1016 ** entire hash table and aPgno[] array before proceeding.
1019 int nByte
= (int)((u8
*)&aHash
[HASHTABLE_NSLOT
] - (u8
*)&aPgno
[1]);
1020 memset((void*)&aPgno
[1], 0, nByte
);
1023 /* If the entry in aPgno[] is already set, then the previous writer
1024 ** must have exited unexpectedly in the middle of a transaction (after
1025 ** writing one or more dirty pages to the WAL to free up memory).
1026 ** Remove the remnants of that writers uncommitted transaction from
1027 ** the hash-table before writing any new entries.
1030 walCleanupHash(pWal
);
1031 assert( !aPgno
[idx
] );
1034 /* Write the aPgno[] array entry and the hash-table slot. */
1036 for(iKey
=walHash(iPage
); aHash
[iKey
]; iKey
=walNextHash(iKey
)){
1037 if( (nCollide
--)==0 ) return SQLITE_CORRUPT_BKPT
;
1040 aHash
[iKey
] = (ht_slot
)idx
;
1042 #ifdef SQLITE_ENABLE_EXPENSIVE_ASSERT
1043 /* Verify that the number of entries in the hash table exactly equals
1044 ** the number of entries in the mapping region.
1047 int i
; /* Loop counter */
1048 int nEntry
= 0; /* Number of entries in the hash table */
1049 for(i
=0; i
<HASHTABLE_NSLOT
; i
++){ if( aHash
[i
] ) nEntry
++; }
1050 assert( nEntry
==idx
);
1053 /* Verify that the every entry in the mapping region is reachable
1054 ** via the hash table. This turns out to be a really, really expensive
1055 ** thing to check, so only do this occasionally - not on every
1058 if( (idx
&0x3ff)==0 ){
1059 int i
; /* Loop counter */
1060 for(i
=1; i
<=idx
; i
++){
1061 for(iKey
=walHash(aPgno
[i
]); aHash
[iKey
]; iKey
=walNextHash(iKey
)){
1062 if( aHash
[iKey
]==i
) break;
1064 assert( aHash
[iKey
]==i
);
1067 #endif /* SQLITE_ENABLE_EXPENSIVE_ASSERT */
1076 ** Recover the wal-index by reading the write-ahead log file.
1078 ** This routine first tries to establish an exclusive lock on the
1079 ** wal-index to prevent other threads/processes from doing anything
1080 ** with the WAL or wal-index while recovery is running. The
1081 ** WAL_RECOVER_LOCK is also held so that other threads will know
1082 ** that this thread is running recovery. If unable to establish
1083 ** the necessary locks, this routine returns SQLITE_BUSY.
1085 static int walIndexRecover(Wal
*pWal
){
1086 int rc
; /* Return Code */
1087 i64 nSize
; /* Size of log file */
1088 u32 aFrameCksum
[2] = {0, 0};
1089 int iLock
; /* Lock offset to lock for checkpoint */
1090 int nLock
; /* Number of locks to hold */
1092 /* Obtain an exclusive lock on all byte in the locking range not already
1093 ** locked by the caller. The caller is guaranteed to have locked the
1094 ** WAL_WRITE_LOCK byte, and may have also locked the WAL_CKPT_LOCK byte.
1095 ** If successful, the same bytes that are locked here are unlocked before
1096 ** this function returns.
1098 assert( pWal
->ckptLock
==1 || pWal
->ckptLock
==0 );
1099 assert( WAL_ALL_BUT_WRITE
==WAL_WRITE_LOCK
+1 );
1100 assert( WAL_CKPT_LOCK
==WAL_ALL_BUT_WRITE
);
1101 assert( pWal
->writeLock
);
1102 iLock
= WAL_ALL_BUT_WRITE
+ pWal
->ckptLock
;
1103 nLock
= SQLITE_SHM_NLOCK
- iLock
;
1104 rc
= walLockExclusive(pWal
, iLock
, nLock
);
1108 WALTRACE(("WAL%p: recovery begin...\n", pWal
));
1110 memset(&pWal
->hdr
, 0, sizeof(WalIndexHdr
));
1112 rc
= sqlite3OsFileSize(pWal
->pWalFd
, &nSize
);
1113 if( rc
!=SQLITE_OK
){
1114 goto recovery_error
;
1117 if( nSize
>WAL_HDRSIZE
){
1118 u8 aBuf
[WAL_HDRSIZE
]; /* Buffer to load WAL header into */
1119 u8
*aFrame
= 0; /* Malloc'd buffer to load entire frame */
1120 int szFrame
; /* Number of bytes in buffer aFrame[] */
1121 u8
*aData
; /* Pointer to data part of aFrame buffer */
1122 int iFrame
; /* Index of last frame read */
1123 i64 iOffset
; /* Next offset to read from log file */
1124 int szPage
; /* Page size according to the log */
1125 u32 magic
; /* Magic value read from WAL header */
1126 u32 version
; /* Magic value read from WAL header */
1127 int isValid
; /* True if this frame is valid */
1129 /* Read in the WAL header. */
1130 rc
= sqlite3OsRead(pWal
->pWalFd
, aBuf
, WAL_HDRSIZE
, 0);
1131 if( rc
!=SQLITE_OK
){
1132 goto recovery_error
;
1135 /* If the database page size is not a power of two, or is greater than
1136 ** SQLITE_MAX_PAGE_SIZE, conclude that the WAL file contains no valid
1137 ** data. Similarly, if the 'magic' value is invalid, ignore the whole
1140 magic
= sqlite3Get4byte(&aBuf
[0]);
1141 szPage
= sqlite3Get4byte(&aBuf
[8]);
1142 if( (magic
&0xFFFFFFFE)!=WAL_MAGIC
1143 || szPage
&(szPage
-1)
1144 || szPage
>SQLITE_MAX_PAGE_SIZE
1149 pWal
->hdr
.bigEndCksum
= (u8
)(magic
&0x00000001);
1150 pWal
->szPage
= szPage
;
1151 pWal
->nCkpt
= sqlite3Get4byte(&aBuf
[12]);
1152 memcpy(&pWal
->hdr
.aSalt
, &aBuf
[16], 8);
1154 /* Verify that the WAL header checksum is correct */
1155 walChecksumBytes(pWal
->hdr
.bigEndCksum
==SQLITE_BIGENDIAN
,
1156 aBuf
, WAL_HDRSIZE
-2*4, 0, pWal
->hdr
.aFrameCksum
1158 if( pWal
->hdr
.aFrameCksum
[0]!=sqlite3Get4byte(&aBuf
[24])
1159 || pWal
->hdr
.aFrameCksum
[1]!=sqlite3Get4byte(&aBuf
[28])
1164 /* Verify that the version number on the WAL format is one that
1165 ** are able to understand */
1166 version
= sqlite3Get4byte(&aBuf
[4]);
1167 if( version
!=WAL_MAX_VERSION
){
1168 rc
= SQLITE_CANTOPEN_BKPT
;
1172 /* Malloc a buffer to read frames into. */
1173 szFrame
= szPage
+ WAL_FRAME_HDRSIZE
;
1174 aFrame
= (u8
*)sqlite3_malloc64(szFrame
);
1176 rc
= SQLITE_NOMEM_BKPT
;
1177 goto recovery_error
;
1179 aData
= &aFrame
[WAL_FRAME_HDRSIZE
];
1181 /* Read all frames from the log file. */
1183 for(iOffset
=WAL_HDRSIZE
; (iOffset
+szFrame
)<=nSize
; iOffset
+=szFrame
){
1184 u32 pgno
; /* Database page number for frame */
1185 u32 nTruncate
; /* dbsize field from frame header */
1187 /* Read and decode the next log frame. */
1189 rc
= sqlite3OsRead(pWal
->pWalFd
, aFrame
, szFrame
, iOffset
);
1190 if( rc
!=SQLITE_OK
) break;
1191 isValid
= walDecodeFrame(pWal
, &pgno
, &nTruncate
, aData
, aFrame
);
1192 if( !isValid
) break;
1193 rc
= walIndexAppend(pWal
, iFrame
, pgno
);
1194 if( rc
!=SQLITE_OK
) break;
1196 /* If nTruncate is non-zero, this is a commit record. */
1198 pWal
->hdr
.mxFrame
= iFrame
;
1199 pWal
->hdr
.nPage
= nTruncate
;
1200 pWal
->hdr
.szPage
= (u16
)((szPage
&0xff00) | (szPage
>>16));
1201 testcase( szPage
<=32768 );
1202 testcase( szPage
>=65536 );
1203 aFrameCksum
[0] = pWal
->hdr
.aFrameCksum
[0];
1204 aFrameCksum
[1] = pWal
->hdr
.aFrameCksum
[1];
1208 sqlite3_free(aFrame
);
1212 if( rc
==SQLITE_OK
){
1213 volatile WalCkptInfo
*pInfo
;
1215 pWal
->hdr
.aFrameCksum
[0] = aFrameCksum
[0];
1216 pWal
->hdr
.aFrameCksum
[1] = aFrameCksum
[1];
1217 walIndexWriteHdr(pWal
);
1219 /* Reset the checkpoint-header. This is safe because this thread is
1220 ** currently holding locks that exclude all other readers, writers and
1223 pInfo
= walCkptInfo(pWal
);
1224 pInfo
->nBackfill
= 0;
1225 pInfo
->nBackfillAttempted
= pWal
->hdr
.mxFrame
;
1226 pInfo
->aReadMark
[0] = 0;
1227 for(i
=1; i
<WAL_NREADER
; i
++) pInfo
->aReadMark
[i
] = READMARK_NOT_USED
;
1228 if( pWal
->hdr
.mxFrame
) pInfo
->aReadMark
[1] = pWal
->hdr
.mxFrame
;
1230 /* If more than one frame was recovered from the log file, report an
1231 ** event via sqlite3_log(). This is to help with identifying performance
1232 ** problems caused by applications routinely shutting down without
1233 ** checkpointing the log file.
1235 if( pWal
->hdr
.nPage
){
1236 sqlite3_log(SQLITE_NOTICE_RECOVER_WAL
,
1237 "recovered %d frames from WAL file %s",
1238 pWal
->hdr
.mxFrame
, pWal
->zWalName
1244 WALTRACE(("WAL%p: recovery %s\n", pWal
, rc
? "failed" : "ok"));
1245 walUnlockExclusive(pWal
, iLock
, nLock
);
1250 ** Close an open wal-index.
1252 static void walIndexClose(Wal
*pWal
, int isDelete
){
1253 if( pWal
->exclusiveMode
==WAL_HEAPMEMORY_MODE
){
1255 for(i
=0; i
<pWal
->nWiData
; i
++){
1256 sqlite3_free((void *)pWal
->apWiData
[i
]);
1257 pWal
->apWiData
[i
] = 0;
1260 sqlite3OsShmUnmap(pWal
->pDbFd
, isDelete
);
1265 ** Open a connection to the WAL file zWalName. The database file must
1266 ** already be opened on connection pDbFd. The buffer that zWalName points
1267 ** to must remain valid for the lifetime of the returned Wal* handle.
1269 ** A SHARED lock should be held on the database file when this function
1270 ** is called. The purpose of this SHARED lock is to prevent any other
1271 ** client from unlinking the WAL or wal-index file. If another process
1272 ** were to do this just after this client opened one of these files, the
1273 ** system would be badly broken.
1275 ** If the log file is successfully opened, SQLITE_OK is returned and
1276 ** *ppWal is set to point to a new WAL handle. If an error occurs,
1277 ** an SQLite error code is returned and *ppWal is left unmodified.
1280 sqlite3_vfs
*pVfs
, /* vfs module to open wal and wal-index */
1281 sqlite3_file
*pDbFd
, /* The open database file */
1282 const char *zWalName
, /* Name of the WAL file */
1283 int bNoShm
, /* True to run in heap-memory mode */
1284 i64 mxWalSize
, /* Truncate WAL to this size on reset */
1285 Wal
**ppWal
/* OUT: Allocated Wal handle */
1287 int rc
; /* Return Code */
1288 Wal
*pRet
; /* Object to allocate and return */
1289 int flags
; /* Flags passed to OsOpen() */
1291 assert( zWalName
&& zWalName
[0] );
1294 /* In the amalgamation, the os_unix.c and os_win.c source files come before
1295 ** this source file. Verify that the #defines of the locking byte offsets
1296 ** in os_unix.c and os_win.c agree with the WALINDEX_LOCK_OFFSET value.
1297 ** For that matter, if the lock offset ever changes from its initial design
1298 ** value of 120, we need to know that so there is an assert() to check it.
1300 assert( 120==WALINDEX_LOCK_OFFSET
);
1301 assert( 136==WALINDEX_HDR_SIZE
);
1303 assert( WIN_SHM_BASE
==WALINDEX_LOCK_OFFSET
);
1305 #ifdef UNIX_SHM_BASE
1306 assert( UNIX_SHM_BASE
==WALINDEX_LOCK_OFFSET
);
1310 /* Allocate an instance of struct Wal to return. */
1312 pRet
= (Wal
*)sqlite3MallocZero(sizeof(Wal
) + pVfs
->szOsFile
);
1314 return SQLITE_NOMEM_BKPT
;
1318 pRet
->pWalFd
= (sqlite3_file
*)&pRet
[1];
1319 pRet
->pDbFd
= pDbFd
;
1320 pRet
->readLock
= -1;
1321 pRet
->mxWalSize
= mxWalSize
;
1322 pRet
->zWalName
= zWalName
;
1323 pRet
->syncHeader
= 1;
1324 pRet
->padToSectorBoundary
= 1;
1325 pRet
->exclusiveMode
= (bNoShm
? WAL_HEAPMEMORY_MODE
: WAL_NORMAL_MODE
);
1327 /* Open file handle on the write-ahead log file. */
1328 flags
= (SQLITE_OPEN_READWRITE
|SQLITE_OPEN_CREATE
|SQLITE_OPEN_WAL
);
1329 rc
= sqlite3OsOpen(pVfs
, zWalName
, pRet
->pWalFd
, flags
, &flags
);
1330 if( rc
==SQLITE_OK
&& flags
&SQLITE_OPEN_READONLY
){
1331 pRet
->readOnly
= WAL_RDONLY
;
1334 if( rc
!=SQLITE_OK
){
1335 walIndexClose(pRet
, 0);
1336 sqlite3OsClose(pRet
->pWalFd
);
1339 int iDC
= sqlite3OsDeviceCharacteristics(pDbFd
);
1340 if( iDC
& SQLITE_IOCAP_SEQUENTIAL
){ pRet
->syncHeader
= 0; }
1341 if( iDC
& SQLITE_IOCAP_POWERSAFE_OVERWRITE
){
1342 pRet
->padToSectorBoundary
= 0;
1345 WALTRACE(("WAL%d: opened\n", pRet
));
1351 ** Change the size to which the WAL file is trucated on each reset.
1353 void sqlite3WalLimit(Wal
*pWal
, i64 iLimit
){
1354 if( pWal
) pWal
->mxWalSize
= iLimit
;
1358 ** Find the smallest page number out of all pages held in the WAL that
1359 ** has not been returned by any prior invocation of this method on the
1360 ** same WalIterator object. Write into *piFrame the frame index where
1361 ** that page was last written into the WAL. Write into *piPage the page
1364 ** Return 0 on success. If there are no pages in the WAL with a page
1365 ** number larger than *piPage, then return 1.
1367 static int walIteratorNext(
1368 WalIterator
*p
, /* Iterator */
1369 u32
*piPage
, /* OUT: The page number of the next page */
1370 u32
*piFrame
/* OUT: Wal frame index of next page */
1372 u32 iMin
; /* Result pgno must be greater than iMin */
1373 u32 iRet
= 0xFFFFFFFF; /* 0xffffffff is never a valid page number */
1374 int i
; /* For looping through segments */
1377 assert( iMin
<0xffffffff );
1378 for(i
=p
->nSegment
-1; i
>=0; i
--){
1379 struct WalSegment
*pSegment
= &p
->aSegment
[i
];
1380 while( pSegment
->iNext
<pSegment
->nEntry
){
1381 u32 iPg
= pSegment
->aPgno
[pSegment
->aIndex
[pSegment
->iNext
]];
1385 *piFrame
= pSegment
->iZero
+ pSegment
->aIndex
[pSegment
->iNext
];
1393 *piPage
= p
->iPrior
= iRet
;
1394 return (iRet
==0xFFFFFFFF);
1398 ** This function merges two sorted lists into a single sorted list.
1400 ** aLeft[] and aRight[] are arrays of indices. The sort key is
1401 ** aContent[aLeft[]] and aContent[aRight[]]. Upon entry, the following
1402 ** is guaranteed for all J<K:
1404 ** aContent[aLeft[J]] < aContent[aLeft[K]]
1405 ** aContent[aRight[J]] < aContent[aRight[K]]
1407 ** This routine overwrites aRight[] with a new (probably longer) sequence
1408 ** of indices such that the aRight[] contains every index that appears in
1409 ** either aLeft[] or the old aRight[] and such that the second condition
1410 ** above is still met.
1412 ** The aContent[aLeft[X]] values will be unique for all X. And the
1413 ** aContent[aRight[X]] values will be unique too. But there might be
1414 ** one or more combinations of X and Y such that
1416 ** aLeft[X]!=aRight[Y] && aContent[aLeft[X]] == aContent[aRight[Y]]
1418 ** When that happens, omit the aLeft[X] and use the aRight[Y] index.
1420 static void walMerge(
1421 const u32
*aContent
, /* Pages in wal - keys for the sort */
1422 ht_slot
*aLeft
, /* IN: Left hand input list */
1423 int nLeft
, /* IN: Elements in array *paLeft */
1424 ht_slot
**paRight
, /* IN/OUT: Right hand input list */
1425 int *pnRight
, /* IN/OUT: Elements in *paRight */
1426 ht_slot
*aTmp
/* Temporary buffer */
1428 int iLeft
= 0; /* Current index in aLeft */
1429 int iRight
= 0; /* Current index in aRight */
1430 int iOut
= 0; /* Current index in output buffer */
1431 int nRight
= *pnRight
;
1432 ht_slot
*aRight
= *paRight
;
1434 assert( nLeft
>0 && nRight
>0 );
1435 while( iRight
<nRight
|| iLeft
<nLeft
){
1440 && (iRight
>=nRight
|| aContent
[aLeft
[iLeft
]]<aContent
[aRight
[iRight
]])
1442 logpage
= aLeft
[iLeft
++];
1444 logpage
= aRight
[iRight
++];
1446 dbpage
= aContent
[logpage
];
1448 aTmp
[iOut
++] = logpage
;
1449 if( iLeft
<nLeft
&& aContent
[aLeft
[iLeft
]]==dbpage
) iLeft
++;
1451 assert( iLeft
>=nLeft
|| aContent
[aLeft
[iLeft
]]>dbpage
);
1452 assert( iRight
>=nRight
|| aContent
[aRight
[iRight
]]>dbpage
);
1457 memcpy(aLeft
, aTmp
, sizeof(aTmp
[0])*iOut
);
1461 ** Sort the elements in list aList using aContent[] as the sort key.
1462 ** Remove elements with duplicate keys, preferring to keep the
1463 ** larger aList[] values.
1465 ** The aList[] entries are indices into aContent[]. The values in
1466 ** aList[] are to be sorted so that for all J<K:
1468 ** aContent[aList[J]] < aContent[aList[K]]
1470 ** For any X and Y such that
1472 ** aContent[aList[X]] == aContent[aList[Y]]
1474 ** Keep the larger of the two values aList[X] and aList[Y] and discard
1477 static void walMergesort(
1478 const u32
*aContent
, /* Pages in wal */
1479 ht_slot
*aBuffer
, /* Buffer of at least *pnList items to use */
1480 ht_slot
*aList
, /* IN/OUT: List to sort */
1481 int *pnList
/* IN/OUT: Number of elements in aList[] */
1484 int nList
; /* Number of elements in aList */
1485 ht_slot
*aList
; /* Pointer to sub-list content */
1488 const int nList
= *pnList
; /* Size of input list */
1489 int nMerge
= 0; /* Number of elements in list aMerge */
1490 ht_slot
*aMerge
= 0; /* List to be merged */
1491 int iList
; /* Index into input list */
1492 u32 iSub
= 0; /* Index into aSub array */
1493 struct Sublist aSub
[13]; /* Array of sub-lists */
1495 memset(aSub
, 0, sizeof(aSub
));
1496 assert( nList
<=HASHTABLE_NPAGE
&& nList
>0 );
1497 assert( HASHTABLE_NPAGE
==(1<<(ArraySize(aSub
)-1)) );
1499 for(iList
=0; iList
<nList
; iList
++){
1501 aMerge
= &aList
[iList
];
1502 for(iSub
=0; iList
& (1<<iSub
); iSub
++){
1504 assert( iSub
<ArraySize(aSub
) );
1506 assert( p
->aList
&& p
->nList
<=(1<<iSub
) );
1507 assert( p
->aList
==&aList
[iList
&~((2<<iSub
)-1)] );
1508 walMerge(aContent
, p
->aList
, p
->nList
, &aMerge
, &nMerge
, aBuffer
);
1510 aSub
[iSub
].aList
= aMerge
;
1511 aSub
[iSub
].nList
= nMerge
;
1514 for(iSub
++; iSub
<ArraySize(aSub
); iSub
++){
1515 if( nList
& (1<<iSub
) ){
1517 assert( iSub
<ArraySize(aSub
) );
1519 assert( p
->nList
<=(1<<iSub
) );
1520 assert( p
->aList
==&aList
[nList
&~((2<<iSub
)-1)] );
1521 walMerge(aContent
, p
->aList
, p
->nList
, &aMerge
, &nMerge
, aBuffer
);
1524 assert( aMerge
==aList
);
1530 for(i
=1; i
<*pnList
; i
++){
1531 assert( aContent
[aList
[i
]] > aContent
[aList
[i
-1]] );
1538 ** Free an iterator allocated by walIteratorInit().
1540 static void walIteratorFree(WalIterator
*p
){
1545 ** Construct a WalInterator object that can be used to loop over all
1546 ** pages in the WAL in ascending order. The caller must hold the checkpoint
1549 ** On success, make *pp point to the newly allocated WalInterator object
1550 ** return SQLITE_OK. Otherwise, return an error code. If this routine
1551 ** returns an error, the value of *pp is undefined.
1553 ** The calling routine should invoke walIteratorFree() to destroy the
1554 ** WalIterator object when it has finished with it.
1556 static int walIteratorInit(Wal
*pWal
, WalIterator
**pp
){
1557 WalIterator
*p
; /* Return value */
1558 int nSegment
; /* Number of segments to merge */
1559 u32 iLast
; /* Last frame in log */
1560 int nByte
; /* Number of bytes to allocate */
1561 int i
; /* Iterator variable */
1562 ht_slot
*aTmp
; /* Temp space used by merge-sort */
1563 int rc
= SQLITE_OK
; /* Return Code */
1565 /* This routine only runs while holding the checkpoint lock. And
1566 ** it only runs if there is actually content in the log (mxFrame>0).
1568 assert( pWal
->ckptLock
&& pWal
->hdr
.mxFrame
>0 );
1569 iLast
= pWal
->hdr
.mxFrame
;
1571 /* Allocate space for the WalIterator object. */
1572 nSegment
= walFramePage(iLast
) + 1;
1573 nByte
= sizeof(WalIterator
)
1574 + (nSegment
-1)*sizeof(struct WalSegment
)
1575 + iLast
*sizeof(ht_slot
);
1576 p
= (WalIterator
*)sqlite3_malloc64(nByte
);
1578 return SQLITE_NOMEM_BKPT
;
1580 memset(p
, 0, nByte
);
1581 p
->nSegment
= nSegment
;
1583 /* Allocate temporary space used by the merge-sort routine. This block
1584 ** of memory will be freed before this function returns.
1586 aTmp
= (ht_slot
*)sqlite3_malloc64(
1587 sizeof(ht_slot
) * (iLast
>HASHTABLE_NPAGE
?HASHTABLE_NPAGE
:iLast
)
1590 rc
= SQLITE_NOMEM_BKPT
;
1593 for(i
=0; rc
==SQLITE_OK
&& i
<nSegment
; i
++){
1594 volatile ht_slot
*aHash
;
1596 volatile u32
*aPgno
;
1598 rc
= walHashGet(pWal
, i
, &aHash
, &aPgno
, &iZero
);
1599 if( rc
==SQLITE_OK
){
1600 int j
; /* Counter variable */
1601 int nEntry
; /* Number of entries in this segment */
1602 ht_slot
*aIndex
; /* Sorted index for this segment */
1605 if( (i
+1)==nSegment
){
1606 nEntry
= (int)(iLast
- iZero
);
1608 nEntry
= (int)((u32
*)aHash
- (u32
*)aPgno
);
1610 aIndex
= &((ht_slot
*)&p
->aSegment
[p
->nSegment
])[iZero
];
1613 for(j
=0; j
<nEntry
; j
++){
1614 aIndex
[j
] = (ht_slot
)j
;
1616 walMergesort((u32
*)aPgno
, aTmp
, aIndex
, &nEntry
);
1617 p
->aSegment
[i
].iZero
= iZero
;
1618 p
->aSegment
[i
].nEntry
= nEntry
;
1619 p
->aSegment
[i
].aIndex
= aIndex
;
1620 p
->aSegment
[i
].aPgno
= (u32
*)aPgno
;
1625 if( rc
!=SQLITE_OK
){
1633 ** Attempt to obtain the exclusive WAL lock defined by parameters lockIdx and
1634 ** n. If the attempt fails and parameter xBusy is not NULL, then it is a
1635 ** busy-handler function. Invoke it and retry the lock until either the
1636 ** lock is successfully obtained or the busy-handler returns 0.
1638 static int walBusyLock(
1639 Wal
*pWal
, /* WAL connection */
1640 int (*xBusy
)(void*), /* Function to call when busy */
1641 void *pBusyArg
, /* Context argument for xBusyHandler */
1642 int lockIdx
, /* Offset of first byte to lock */
1643 int n
/* Number of bytes to lock */
1647 rc
= walLockExclusive(pWal
, lockIdx
, n
);
1648 }while( xBusy
&& rc
==SQLITE_BUSY
&& xBusy(pBusyArg
) );
1653 ** The cache of the wal-index header must be valid to call this function.
1654 ** Return the page-size in bytes used by the database.
1656 static int walPagesize(Wal
*pWal
){
1657 return (pWal
->hdr
.szPage
&0xfe00) + ((pWal
->hdr
.szPage
&0x0001)<<16);
1661 ** The following is guaranteed when this function is called:
1663 ** a) the WRITER lock is held,
1664 ** b) the entire log file has been checkpointed, and
1665 ** c) any existing readers are reading exclusively from the database
1666 ** file - there are no readers that may attempt to read a frame from
1669 ** This function updates the shared-memory structures so that the next
1670 ** client to write to the database (which may be this one) does so by
1671 ** writing frames into the start of the log file.
1673 ** The value of parameter salt1 is used as the aSalt[1] value in the
1674 ** new wal-index header. It should be passed a pseudo-random value (i.e.
1675 ** one obtained from sqlite3_randomness()).
1677 static void walRestartHdr(Wal
*pWal
, u32 salt1
){
1678 volatile WalCkptInfo
*pInfo
= walCkptInfo(pWal
);
1679 int i
; /* Loop counter */
1680 u32
*aSalt
= pWal
->hdr
.aSalt
; /* Big-endian salt values */
1682 pWal
->hdr
.mxFrame
= 0;
1683 sqlite3Put4byte((u8
*)&aSalt
[0], 1 + sqlite3Get4byte((u8
*)&aSalt
[0]));
1684 memcpy(&pWal
->hdr
.aSalt
[1], &salt1
, 4);
1685 walIndexWriteHdr(pWal
);
1686 pInfo
->nBackfill
= 0;
1687 pInfo
->nBackfillAttempted
= 0;
1688 pInfo
->aReadMark
[1] = 0;
1689 for(i
=2; i
<WAL_NREADER
; i
++) pInfo
->aReadMark
[i
] = READMARK_NOT_USED
;
1690 assert( pInfo
->aReadMark
[0]==0 );
1694 ** Copy as much content as we can from the WAL back into the database file
1695 ** in response to an sqlite3_wal_checkpoint() request or the equivalent.
1697 ** The amount of information copies from WAL to database might be limited
1698 ** by active readers. This routine will never overwrite a database page
1699 ** that a concurrent reader might be using.
1701 ** All I/O barrier operations (a.k.a fsyncs) occur in this routine when
1702 ** SQLite is in WAL-mode in synchronous=NORMAL. That means that if
1703 ** checkpoints are always run by a background thread or background
1704 ** process, foreground threads will never block on a lengthy fsync call.
1706 ** Fsync is called on the WAL before writing content out of the WAL and
1707 ** into the database. This ensures that if the new content is persistent
1708 ** in the WAL and can be recovered following a power-loss or hard reset.
1710 ** Fsync is also called on the database file if (and only if) the entire
1711 ** WAL content is copied into the database file. This second fsync makes
1712 ** it safe to delete the WAL since the new content will persist in the
1715 ** This routine uses and updates the nBackfill field of the wal-index header.
1716 ** This is the only routine that will increase the value of nBackfill.
1717 ** (A WAL reset or recovery will revert nBackfill to zero, but not increase
1720 ** The caller must be holding sufficient locks to ensure that no other
1721 ** checkpoint is running (in any other thread or process) at the same
1724 static int walCheckpoint(
1725 Wal
*pWal
, /* Wal connection */
1726 sqlite3
*db
, /* Check for interrupts on this handle */
1727 int eMode
, /* One of PASSIVE, FULL or RESTART */
1728 int (*xBusy
)(void*), /* Function to call when busy */
1729 void *pBusyArg
, /* Context argument for xBusyHandler */
1730 int sync_flags
, /* Flags for OsSync() (or 0) */
1731 u8
*zBuf
/* Temporary buffer to use */
1733 int rc
= SQLITE_OK
; /* Return code */
1734 int szPage
; /* Database page-size */
1735 WalIterator
*pIter
= 0; /* Wal iterator context */
1736 u32 iDbpage
= 0; /* Next database page to write */
1737 u32 iFrame
= 0; /* Wal frame containing data for iDbpage */
1738 u32 mxSafeFrame
; /* Max frame that can be backfilled */
1739 u32 mxPage
; /* Max database page to write */
1740 int i
; /* Loop counter */
1741 volatile WalCkptInfo
*pInfo
; /* The checkpoint status information */
1743 szPage
= walPagesize(pWal
);
1744 testcase( szPage
<=32768 );
1745 testcase( szPage
>=65536 );
1746 pInfo
= walCkptInfo(pWal
);
1747 if( pInfo
->nBackfill
<pWal
->hdr
.mxFrame
){
1749 /* Allocate the iterator */
1750 rc
= walIteratorInit(pWal
, &pIter
);
1751 if( rc
!=SQLITE_OK
){
1756 /* EVIDENCE-OF: R-62920-47450 The busy-handler callback is never invoked
1757 ** in the SQLITE_CHECKPOINT_PASSIVE mode. */
1758 assert( eMode
!=SQLITE_CHECKPOINT_PASSIVE
|| xBusy
==0 );
1760 /* Compute in mxSafeFrame the index of the last frame of the WAL that is
1761 ** safe to write into the database. Frames beyond mxSafeFrame might
1762 ** overwrite database pages that are in use by active readers and thus
1763 ** cannot be backfilled from the WAL.
1765 mxSafeFrame
= pWal
->hdr
.mxFrame
;
1766 mxPage
= pWal
->hdr
.nPage
;
1767 for(i
=1; i
<WAL_NREADER
; i
++){
1768 /* Thread-sanitizer reports that the following is an unsafe read,
1769 ** as some other thread may be in the process of updating the value
1770 ** of the aReadMark[] slot. The assumption here is that if that is
1771 ** happening, the other client may only be increasing the value,
1772 ** not decreasing it. So assuming either that either the "old" or
1773 ** "new" version of the value is read, and not some arbitrary value
1774 ** that would never be written by a real client, things are still
1776 u32 y
= pInfo
->aReadMark
[i
];
1777 if( mxSafeFrame
>y
){
1778 assert( y
<=pWal
->hdr
.mxFrame
);
1779 rc
= walBusyLock(pWal
, xBusy
, pBusyArg
, WAL_READ_LOCK(i
), 1);
1780 if( rc
==SQLITE_OK
){
1781 pInfo
->aReadMark
[i
] = (i
==1 ? mxSafeFrame
: READMARK_NOT_USED
);
1782 walUnlockExclusive(pWal
, WAL_READ_LOCK(i
), 1);
1783 }else if( rc
==SQLITE_BUSY
){
1787 goto walcheckpoint_out
;
1792 if( pInfo
->nBackfill
<mxSafeFrame
1793 && (rc
= walBusyLock(pWal
, xBusy
, pBusyArg
, WAL_READ_LOCK(0),1))==SQLITE_OK
1795 i64 nSize
; /* Current size of database file */
1796 u32 nBackfill
= pInfo
->nBackfill
;
1798 pInfo
->nBackfillAttempted
= mxSafeFrame
;
1800 /* Sync the WAL to disk */
1802 rc
= sqlite3OsSync(pWal
->pWalFd
, sync_flags
);
1805 /* If the database may grow as a result of this checkpoint, hint
1806 ** about the eventual size of the db file to the VFS layer.
1808 if( rc
==SQLITE_OK
){
1809 i64 nReq
= ((i64
)mxPage
* szPage
);
1810 rc
= sqlite3OsFileSize(pWal
->pDbFd
, &nSize
);
1811 if( rc
==SQLITE_OK
&& nSize
<nReq
){
1812 sqlite3OsFileControlHint(pWal
->pDbFd
, SQLITE_FCNTL_SIZE_HINT
, &nReq
);
1817 /* Iterate through the contents of the WAL, copying data to the db file */
1818 while( rc
==SQLITE_OK
&& 0==walIteratorNext(pIter
, &iDbpage
, &iFrame
) ){
1820 assert( walFramePgno(pWal
, iFrame
)==iDbpage
);
1821 if( db
->u1
.isInterrupted
){
1822 rc
= db
->mallocFailed
? SQLITE_NOMEM_BKPT
: SQLITE_INTERRUPT
;
1825 if( iFrame
<=nBackfill
|| iFrame
>mxSafeFrame
|| iDbpage
>mxPage
){
1828 iOffset
= walFrameOffset(iFrame
, szPage
) + WAL_FRAME_HDRSIZE
;
1829 /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL file */
1830 rc
= sqlite3OsRead(pWal
->pWalFd
, zBuf
, szPage
, iOffset
);
1831 if( rc
!=SQLITE_OK
) break;
1832 iOffset
= (iDbpage
-1)*(i64
)szPage
;
1833 testcase( IS_BIG_INT(iOffset
) );
1834 rc
= sqlite3OsWrite(pWal
->pDbFd
, zBuf
, szPage
, iOffset
);
1835 if( rc
!=SQLITE_OK
) break;
1838 /* If work was actually accomplished... */
1839 if( rc
==SQLITE_OK
){
1840 if( mxSafeFrame
==walIndexHdr(pWal
)->mxFrame
){
1841 i64 szDb
= pWal
->hdr
.nPage
*(i64
)szPage
;
1842 testcase( IS_BIG_INT(szDb
) );
1843 rc
= sqlite3OsTruncate(pWal
->pDbFd
, szDb
);
1844 if( rc
==SQLITE_OK
&& sync_flags
){
1845 rc
= sqlite3OsSync(pWal
->pDbFd
, sync_flags
);
1848 if( rc
==SQLITE_OK
){
1849 pInfo
->nBackfill
= mxSafeFrame
;
1853 /* Release the reader lock held while backfilling */
1854 walUnlockExclusive(pWal
, WAL_READ_LOCK(0), 1);
1857 if( rc
==SQLITE_BUSY
){
1858 /* Reset the return code so as not to report a checkpoint failure
1859 ** just because there are active readers. */
1864 /* If this is an SQLITE_CHECKPOINT_RESTART or TRUNCATE operation, and the
1865 ** entire wal file has been copied into the database file, then block
1866 ** until all readers have finished using the wal file. This ensures that
1867 ** the next process to write to the database restarts the wal file.
1869 if( rc
==SQLITE_OK
&& eMode
!=SQLITE_CHECKPOINT_PASSIVE
){
1870 assert( pWal
->writeLock
);
1871 if( pInfo
->nBackfill
<pWal
->hdr
.mxFrame
){
1873 }else if( eMode
>=SQLITE_CHECKPOINT_RESTART
){
1875 sqlite3_randomness(4, &salt1
);
1876 assert( pInfo
->nBackfill
==pWal
->hdr
.mxFrame
);
1877 rc
= walBusyLock(pWal
, xBusy
, pBusyArg
, WAL_READ_LOCK(1), WAL_NREADER
-1);
1878 if( rc
==SQLITE_OK
){
1879 if( eMode
==SQLITE_CHECKPOINT_TRUNCATE
){
1880 /* IMPLEMENTATION-OF: R-44699-57140 This mode works the same way as
1881 ** SQLITE_CHECKPOINT_RESTART with the addition that it also
1882 ** truncates the log file to zero bytes just prior to a
1883 ** successful return.
1885 ** In theory, it might be safe to do this without updating the
1886 ** wal-index header in shared memory, as all subsequent reader or
1887 ** writer clients should see that the entire log file has been
1888 ** checkpointed and behave accordingly. This seems unsafe though,
1889 ** as it would leave the system in a state where the contents of
1890 ** the wal-index header do not match the contents of the
1891 ** file-system. To avoid this, update the wal-index header to
1892 ** indicate that the log file contains zero valid frames. */
1893 walRestartHdr(pWal
, salt1
);
1894 rc
= sqlite3OsTruncate(pWal
->pWalFd
, 0);
1896 walUnlockExclusive(pWal
, WAL_READ_LOCK(1), WAL_NREADER
-1);
1902 walIteratorFree(pIter
);
1907 ** If the WAL file is currently larger than nMax bytes in size, truncate
1908 ** it to exactly nMax bytes. If an error occurs while doing so, ignore it.
1910 static void walLimitSize(Wal
*pWal
, i64 nMax
){
1913 sqlite3BeginBenignMalloc();
1914 rx
= sqlite3OsFileSize(pWal
->pWalFd
, &sz
);
1915 if( rx
==SQLITE_OK
&& (sz
> nMax
) ){
1916 rx
= sqlite3OsTruncate(pWal
->pWalFd
, nMax
);
1918 sqlite3EndBenignMalloc();
1920 sqlite3_log(rx
, "cannot limit WAL size: %s", pWal
->zWalName
);
1925 ** Close a connection to a log file.
1927 int sqlite3WalClose(
1928 Wal
*pWal
, /* Wal to close */
1929 sqlite3
*db
, /* For interrupt flag */
1930 int sync_flags
, /* Flags to pass to OsSync() (or 0) */
1932 u8
*zBuf
/* Buffer of at least nBuf bytes */
1936 int isDelete
= 0; /* True to unlink wal and wal-index files */
1938 /* If an EXCLUSIVE lock can be obtained on the database file (using the
1939 ** ordinary, rollback-mode locking methods, this guarantees that the
1940 ** connection associated with this log file is the only connection to
1941 ** the database. In this case checkpoint the database and unlink both
1942 ** the wal and wal-index files.
1944 ** The EXCLUSIVE lock is not released before returning.
1947 && SQLITE_OK
==(rc
= sqlite3OsLock(pWal
->pDbFd
, SQLITE_LOCK_EXCLUSIVE
))
1949 if( pWal
->exclusiveMode
==WAL_NORMAL_MODE
){
1950 pWal
->exclusiveMode
= WAL_EXCLUSIVE_MODE
;
1952 rc
= sqlite3WalCheckpoint(pWal
, db
,
1953 SQLITE_CHECKPOINT_PASSIVE
, 0, 0, sync_flags
, nBuf
, zBuf
, 0, 0
1955 if( rc
==SQLITE_OK
){
1957 sqlite3OsFileControlHint(
1958 pWal
->pDbFd
, SQLITE_FCNTL_PERSIST_WAL
, &bPersist
1961 /* Try to delete the WAL file if the checkpoint completed and
1962 ** fsyned (rc==SQLITE_OK) and if we are not in persistent-wal
1963 ** mode (!bPersist) */
1965 }else if( pWal
->mxWalSize
>=0 ){
1966 /* Try to truncate the WAL file to zero bytes if the checkpoint
1967 ** completed and fsynced (rc==SQLITE_OK) and we are in persistent
1968 ** WAL mode (bPersist) and if the PRAGMA journal_size_limit is a
1969 ** non-negative value (pWal->mxWalSize>=0). Note that we truncate
1970 ** to zero bytes as truncating to the journal_size_limit might
1971 ** leave a corrupt WAL file on disk. */
1972 walLimitSize(pWal
, 0);
1977 walIndexClose(pWal
, isDelete
);
1978 sqlite3OsClose(pWal
->pWalFd
);
1980 sqlite3BeginBenignMalloc();
1981 sqlite3OsDelete(pWal
->pVfs
, pWal
->zWalName
, 0);
1982 sqlite3EndBenignMalloc();
1984 WALTRACE(("WAL%p: closed\n", pWal
));
1985 sqlite3_free((void *)pWal
->apWiData
);
1992 ** Try to read the wal-index header. Return 0 on success and 1 if
1993 ** there is a problem.
1995 ** The wal-index is in shared memory. Another thread or process might
1996 ** be writing the header at the same time this procedure is trying to
1997 ** read it, which might result in inconsistency. A dirty read is detected
1998 ** by verifying that both copies of the header are the same and also by
1999 ** a checksum on the header.
2001 ** If and only if the read is consistent and the header is different from
2002 ** pWal->hdr, then pWal->hdr is updated to the content of the new header
2003 ** and *pChanged is set to 1.
2005 ** If the checksum cannot be verified return non-zero. If the header
2006 ** is read successfully and the checksum verified, return zero.
2008 static int walIndexTryHdr(Wal
*pWal
, int *pChanged
){
2009 u32 aCksum
[2]; /* Checksum on the header content */
2010 WalIndexHdr h1
, h2
; /* Two copies of the header content */
2011 WalIndexHdr
volatile *aHdr
; /* Header in shared memory */
2013 /* The first page of the wal-index must be mapped at this point. */
2014 assert( pWal
->nWiData
>0 && pWal
->apWiData
[0] );
2016 /* Read the header. This might happen concurrently with a write to the
2017 ** same area of shared memory on a different CPU in a SMP,
2018 ** meaning it is possible that an inconsistent snapshot is read
2019 ** from the file. If this happens, return non-zero.
2021 ** There are two copies of the header at the beginning of the wal-index.
2022 ** When reading, read [0] first then [1]. Writes are in the reverse order.
2023 ** Memory barriers are used to prevent the compiler or the hardware from
2024 ** reordering the reads and writes.
2026 aHdr
= walIndexHdr(pWal
);
2027 memcpy(&h1
, (void *)&aHdr
[0], sizeof(h1
));
2028 walShmBarrier(pWal
);
2029 memcpy(&h2
, (void *)&aHdr
[1], sizeof(h2
));
2031 if( memcmp(&h1
, &h2
, sizeof(h1
))!=0 ){
2032 return 1; /* Dirty read */
2035 return 1; /* Malformed header - probably all zeros */
2037 walChecksumBytes(1, (u8
*)&h1
, sizeof(h1
)-sizeof(h1
.aCksum
), 0, aCksum
);
2038 if( aCksum
[0]!=h1
.aCksum
[0] || aCksum
[1]!=h1
.aCksum
[1] ){
2039 return 1; /* Checksum does not match */
2042 if( memcmp(&pWal
->hdr
, &h1
, sizeof(WalIndexHdr
)) ){
2044 memcpy(&pWal
->hdr
, &h1
, sizeof(WalIndexHdr
));
2045 pWal
->szPage
= (pWal
->hdr
.szPage
&0xfe00) + ((pWal
->hdr
.szPage
&0x0001)<<16);
2046 testcase( pWal
->szPage
<=32768 );
2047 testcase( pWal
->szPage
>=65536 );
2050 /* The header was successfully read. Return zero. */
2055 ** Read the wal-index header from the wal-index and into pWal->hdr.
2056 ** If the wal-header appears to be corrupt, try to reconstruct the
2057 ** wal-index from the WAL before returning.
2059 ** Set *pChanged to 1 if the wal-index header value in pWal->hdr is
2060 ** changed by this operation. If pWal->hdr is unchanged, set *pChanged
2063 ** If the wal-index header is successfully read, return SQLITE_OK.
2064 ** Otherwise an SQLite error code.
2066 static int walIndexReadHdr(Wal
*pWal
, int *pChanged
){
2067 int rc
; /* Return code */
2068 int badHdr
; /* True if a header read failed */
2069 volatile u32
*page0
; /* Chunk of wal-index containing header */
2071 /* Ensure that page 0 of the wal-index (the page that contains the
2072 ** wal-index header) is mapped. Return early if an error occurs here.
2075 rc
= walIndexPage(pWal
, 0, &page0
);
2076 if( rc
!=SQLITE_OK
){
2079 assert( page0
|| pWal
->writeLock
==0 );
2081 /* If the first page of the wal-index has been mapped, try to read the
2082 ** wal-index header immediately, without holding any lock. This usually
2083 ** works, but may fail if the wal-index header is corrupt or currently
2084 ** being modified by another thread or process.
2086 badHdr
= (page0
? walIndexTryHdr(pWal
, pChanged
) : 1);
2088 /* If the first attempt failed, it might have been due to a race
2089 ** with a writer. So get a WRITE lock and try again.
2091 assert( badHdr
==0 || pWal
->writeLock
==0 );
2093 if( pWal
->readOnly
& WAL_SHM_RDONLY
){
2094 if( SQLITE_OK
==(rc
= walLockShared(pWal
, WAL_WRITE_LOCK
)) ){
2095 walUnlockShared(pWal
, WAL_WRITE_LOCK
);
2096 rc
= SQLITE_READONLY_RECOVERY
;
2098 }else if( SQLITE_OK
==(rc
= walLockExclusive(pWal
, WAL_WRITE_LOCK
, 1)) ){
2099 pWal
->writeLock
= 1;
2100 if( SQLITE_OK
==(rc
= walIndexPage(pWal
, 0, &page0
)) ){
2101 badHdr
= walIndexTryHdr(pWal
, pChanged
);
2103 /* If the wal-index header is still malformed even while holding
2104 ** a WRITE lock, it can only mean that the header is corrupted and
2105 ** needs to be reconstructed. So run recovery to do exactly that.
2107 rc
= walIndexRecover(pWal
);
2111 pWal
->writeLock
= 0;
2112 walUnlockExclusive(pWal
, WAL_WRITE_LOCK
, 1);
2116 /* If the header is read successfully, check the version number to make
2117 ** sure the wal-index was not constructed with some future format that
2118 ** this version of SQLite cannot understand.
2120 if( badHdr
==0 && pWal
->hdr
.iVersion
!=WALINDEX_MAX_VERSION
){
2121 rc
= SQLITE_CANTOPEN_BKPT
;
2128 ** This is the value that walTryBeginRead returns when it needs to
2131 #define WAL_RETRY (-1)
2134 ** Attempt to start a read transaction. This might fail due to a race or
2135 ** other transient condition. When that happens, it returns WAL_RETRY to
2136 ** indicate to the caller that it is safe to retry immediately.
2138 ** On success return SQLITE_OK. On a permanent failure (such an
2139 ** I/O error or an SQLITE_BUSY because another process is running
2140 ** recovery) return a positive error code.
2142 ** The useWal parameter is true to force the use of the WAL and disable
2143 ** the case where the WAL is bypassed because it has been completely
2144 ** checkpointed. If useWal==0 then this routine calls walIndexReadHdr()
2145 ** to make a copy of the wal-index header into pWal->hdr. If the
2146 ** wal-index header has changed, *pChanged is set to 1 (as an indication
2147 ** to the caller that the local paget cache is obsolete and needs to be
2148 ** flushed.) When useWal==1, the wal-index header is assumed to already
2149 ** be loaded and the pChanged parameter is unused.
2151 ** The caller must set the cnt parameter to the number of prior calls to
2152 ** this routine during the current read attempt that returned WAL_RETRY.
2153 ** This routine will start taking more aggressive measures to clear the
2154 ** race conditions after multiple WAL_RETRY returns, and after an excessive
2155 ** number of errors will ultimately return SQLITE_PROTOCOL. The
2156 ** SQLITE_PROTOCOL return indicates that some other process has gone rogue
2157 ** and is not honoring the locking protocol. There is a vanishingly small
2158 ** chance that SQLITE_PROTOCOL could be returned because of a run of really
2159 ** bad luck when there is lots of contention for the wal-index, but that
2160 ** possibility is so small that it can be safely neglected, we believe.
2162 ** On success, this routine obtains a read lock on
2163 ** WAL_READ_LOCK(pWal->readLock). The pWal->readLock integer is
2164 ** in the range 0 <= pWal->readLock < WAL_NREADER. If pWal->readLock==(-1)
2165 ** that means the Wal does not hold any read lock. The reader must not
2166 ** access any database page that is modified by a WAL frame up to and
2167 ** including frame number aReadMark[pWal->readLock]. The reader will
2168 ** use WAL frames up to and including pWal->hdr.mxFrame if pWal->readLock>0
2169 ** Or if pWal->readLock==0, then the reader will ignore the WAL
2170 ** completely and get all content directly from the database file.
2171 ** If the useWal parameter is 1 then the WAL will never be ignored and
2172 ** this routine will always set pWal->readLock>0 on success.
2173 ** When the read transaction is completed, the caller must release the
2174 ** lock on WAL_READ_LOCK(pWal->readLock) and set pWal->readLock to -1.
2176 ** This routine uses the nBackfill and aReadMark[] fields of the header
2177 ** to select a particular WAL_READ_LOCK() that strives to let the
2178 ** checkpoint process do as much work as possible. This routine might
2179 ** update values of the aReadMark[] array in the header, but if it does
2180 ** so it takes care to hold an exclusive lock on the corresponding
2181 ** WAL_READ_LOCK() while changing values.
2183 static int walTryBeginRead(Wal
*pWal
, int *pChanged
, int useWal
, int cnt
){
2184 volatile WalCkptInfo
*pInfo
; /* Checkpoint information in wal-index */
2185 u32 mxReadMark
; /* Largest aReadMark[] value */
2186 int mxI
; /* Index of largest aReadMark[] value */
2187 int i
; /* Loop counter */
2188 int rc
= SQLITE_OK
; /* Return code */
2189 u32 mxFrame
; /* Wal frame to lock to */
2191 assert( pWal
->readLock
<0 ); /* Not currently locked */
2193 /* Take steps to avoid spinning forever if there is a protocol error.
2195 ** Circumstances that cause a RETRY should only last for the briefest
2196 ** instances of time. No I/O or other system calls are done while the
2197 ** locks are held, so the locks should not be held for very long. But
2198 ** if we are unlucky, another process that is holding a lock might get
2199 ** paged out or take a page-fault that is time-consuming to resolve,
2200 ** during the few nanoseconds that it is holding the lock. In that case,
2201 ** it might take longer than normal for the lock to free.
2203 ** After 5 RETRYs, we begin calling sqlite3OsSleep(). The first few
2204 ** calls to sqlite3OsSleep() have a delay of 1 microsecond. Really this
2205 ** is more of a scheduler yield than an actual delay. But on the 10th
2206 ** an subsequent retries, the delays start becoming longer and longer,
2207 ** so that on the 100th (and last) RETRY we delay for 323 milliseconds.
2208 ** The total delay time before giving up is less than 10 seconds.
2211 int nDelay
= 1; /* Pause time in microseconds */
2213 VVA_ONLY( pWal
->lockError
= 1; )
2214 return SQLITE_PROTOCOL
;
2216 if( cnt
>=10 ) nDelay
= (cnt
-9)*(cnt
-9)*39;
2217 sqlite3OsSleep(pWal
->pVfs
, nDelay
);
2221 rc
= walIndexReadHdr(pWal
, pChanged
);
2222 if( rc
==SQLITE_BUSY
){
2223 /* If there is not a recovery running in another thread or process
2224 ** then convert BUSY errors to WAL_RETRY. If recovery is known to
2225 ** be running, convert BUSY to BUSY_RECOVERY. There is a race here
2226 ** which might cause WAL_RETRY to be returned even if BUSY_RECOVERY
2227 ** would be technically correct. But the race is benign since with
2228 ** WAL_RETRY this routine will be called again and will probably be
2229 ** right on the second iteration.
2231 if( pWal
->apWiData
[0]==0 ){
2232 /* This branch is taken when the xShmMap() method returns SQLITE_BUSY.
2233 ** We assume this is a transient condition, so return WAL_RETRY. The
2234 ** xShmMap() implementation used by the default unix and win32 VFS
2235 ** modules may return SQLITE_BUSY due to a race condition in the
2236 ** code that determines whether or not the shared-memory region
2237 ** must be zeroed before the requested page is returned.
2240 }else if( SQLITE_OK
==(rc
= walLockShared(pWal
, WAL_RECOVER_LOCK
)) ){
2241 walUnlockShared(pWal
, WAL_RECOVER_LOCK
);
2243 }else if( rc
==SQLITE_BUSY
){
2244 rc
= SQLITE_BUSY_RECOVERY
;
2247 if( rc
!=SQLITE_OK
){
2252 pInfo
= walCkptInfo(pWal
);
2253 if( !useWal
&& pInfo
->nBackfill
==pWal
->hdr
.mxFrame
2254 #ifdef SQLITE_ENABLE_SNAPSHOT
2255 && (pWal
->pSnapshot
==0 || pWal
->hdr
.mxFrame
==0
2256 || 0==memcmp(&pWal
->hdr
, pWal
->pSnapshot
, sizeof(WalIndexHdr
)))
2259 /* The WAL has been completely backfilled (or it is empty).
2260 ** and can be safely ignored.
2262 rc
= walLockShared(pWal
, WAL_READ_LOCK(0));
2263 walShmBarrier(pWal
);
2264 if( rc
==SQLITE_OK
){
2265 if( memcmp((void *)walIndexHdr(pWal
), &pWal
->hdr
, sizeof(WalIndexHdr
)) ){
2266 /* It is not safe to allow the reader to continue here if frames
2267 ** may have been appended to the log before READ_LOCK(0) was obtained.
2268 ** When holding READ_LOCK(0), the reader ignores the entire log file,
2269 ** which implies that the database file contains a trustworthy
2270 ** snapshot. Since holding READ_LOCK(0) prevents a checkpoint from
2271 ** happening, this is usually correct.
2273 ** However, if frames have been appended to the log (or if the log
2274 ** is wrapped and written for that matter) before the READ_LOCK(0)
2275 ** is obtained, that is not necessarily true. A checkpointer may
2276 ** have started to backfill the appended frames but crashed before
2277 ** it finished. Leaving a corrupt image in the database file.
2279 walUnlockShared(pWal
, WAL_READ_LOCK(0));
2284 }else if( rc
!=SQLITE_BUSY
){
2289 /* If we get this far, it means that the reader will want to use
2290 ** the WAL to get at content from recent commits. The job now is
2291 ** to select one of the aReadMark[] entries that is closest to
2292 ** but not exceeding pWal->hdr.mxFrame and lock that entry.
2296 mxFrame
= pWal
->hdr
.mxFrame
;
2297 #ifdef SQLITE_ENABLE_SNAPSHOT
2298 if( pWal
->pSnapshot
&& pWal
->pSnapshot
->mxFrame
<mxFrame
){
2299 mxFrame
= pWal
->pSnapshot
->mxFrame
;
2302 for(i
=1; i
<WAL_NREADER
; i
++){
2303 u32 thisMark
= pInfo
->aReadMark
[i
];
2304 if( mxReadMark
<=thisMark
&& thisMark
<=mxFrame
){
2305 assert( thisMark
!=READMARK_NOT_USED
);
2306 mxReadMark
= thisMark
;
2310 if( (pWal
->readOnly
& WAL_SHM_RDONLY
)==0
2311 && (mxReadMark
<mxFrame
|| mxI
==0)
2313 for(i
=1; i
<WAL_NREADER
; i
++){
2314 rc
= walLockExclusive(pWal
, WAL_READ_LOCK(i
), 1);
2315 if( rc
==SQLITE_OK
){
2316 mxReadMark
= pInfo
->aReadMark
[i
] = mxFrame
;
2318 walUnlockExclusive(pWal
, WAL_READ_LOCK(i
), 1);
2320 }else if( rc
!=SQLITE_BUSY
){
2326 assert( rc
==SQLITE_BUSY
|| (pWal
->readOnly
& WAL_SHM_RDONLY
)!=0 );
2327 return rc
==SQLITE_BUSY
? WAL_RETRY
: SQLITE_READONLY_CANTLOCK
;
2330 rc
= walLockShared(pWal
, WAL_READ_LOCK(mxI
));
2332 return rc
==SQLITE_BUSY
? WAL_RETRY
: rc
;
2334 /* Now that the read-lock has been obtained, check that neither the
2335 ** value in the aReadMark[] array or the contents of the wal-index
2336 ** header have changed.
2338 ** It is necessary to check that the wal-index header did not change
2339 ** between the time it was read and when the shared-lock was obtained
2340 ** on WAL_READ_LOCK(mxI) was obtained to account for the possibility
2341 ** that the log file may have been wrapped by a writer, or that frames
2342 ** that occur later in the log than pWal->hdr.mxFrame may have been
2343 ** copied into the database by a checkpointer. If either of these things
2344 ** happened, then reading the database with the current value of
2345 ** pWal->hdr.mxFrame risks reading a corrupted snapshot. So, retry
2348 ** Before checking that the live wal-index header has not changed
2349 ** since it was read, set Wal.minFrame to the first frame in the wal
2350 ** file that has not yet been checkpointed. This client will not need
2351 ** to read any frames earlier than minFrame from the wal file - they
2352 ** can be safely read directly from the database file.
2354 ** Because a ShmBarrier() call is made between taking the copy of
2355 ** nBackfill and checking that the wal-header in shared-memory still
2356 ** matches the one cached in pWal->hdr, it is guaranteed that the
2357 ** checkpointer that set nBackfill was not working with a wal-index
2358 ** header newer than that cached in pWal->hdr. If it were, that could
2359 ** cause a problem. The checkpointer could omit to checkpoint
2360 ** a version of page X that lies before pWal->minFrame (call that version
2361 ** A) on the basis that there is a newer version (version B) of the same
2362 ** page later in the wal file. But if version B happens to like past
2363 ** frame pWal->hdr.mxFrame - then the client would incorrectly assume
2364 ** that it can read version A from the database file. However, since
2365 ** we can guarantee that the checkpointer that set nBackfill could not
2366 ** see any pages past pWal->hdr.mxFrame, this problem does not come up.
2368 pWal
->minFrame
= pInfo
->nBackfill
+1;
2369 walShmBarrier(pWal
);
2370 if( pInfo
->aReadMark
[mxI
]!=mxReadMark
2371 || memcmp((void *)walIndexHdr(pWal
), &pWal
->hdr
, sizeof(WalIndexHdr
))
2373 walUnlockShared(pWal
, WAL_READ_LOCK(mxI
));
2376 assert( mxReadMark
<=pWal
->hdr
.mxFrame
);
2377 pWal
->readLock
= (i16
)mxI
;
2382 #ifdef SQLITE_ENABLE_SNAPSHOT
2384 ** Attempt to reduce the value of the WalCkptInfo.nBackfillAttempted
2385 ** variable so that older snapshots can be accessed. To do this, loop
2386 ** through all wal frames from nBackfillAttempted to (nBackfill+1),
2387 ** comparing their content to the corresponding page with the database
2388 ** file, if any. Set nBackfillAttempted to the frame number of the
2389 ** first frame for which the wal file content matches the db file.
2391 ** This is only really safe if the file-system is such that any page
2392 ** writes made by earlier checkpointers were atomic operations, which
2393 ** is not always true. It is also possible that nBackfillAttempted
2394 ** may be left set to a value larger than expected, if a wal frame
2395 ** contains content that duplicate of an earlier version of the same
2398 ** SQLITE_OK is returned if successful, or an SQLite error code if an
2399 ** error occurs. It is not an error if nBackfillAttempted cannot be
2400 ** decreased at all.
2402 int sqlite3WalSnapshotRecover(Wal
*pWal
){
2405 assert( pWal
->readLock
>=0 );
2406 rc
= walLockExclusive(pWal
, WAL_CKPT_LOCK
, 1);
2407 if( rc
==SQLITE_OK
){
2408 volatile WalCkptInfo
*pInfo
= walCkptInfo(pWal
);
2409 int szPage
= (int)pWal
->szPage
;
2410 i64 szDb
; /* Size of db file in bytes */
2412 rc
= sqlite3OsFileSize(pWal
->pDbFd
, &szDb
);
2413 if( rc
==SQLITE_OK
){
2414 void *pBuf1
= sqlite3_malloc(szPage
);
2415 void *pBuf2
= sqlite3_malloc(szPage
);
2416 if( pBuf1
==0 || pBuf2
==0 ){
2419 u32 i
= pInfo
->nBackfillAttempted
;
2420 for(i
=pInfo
->nBackfillAttempted
; i
>pInfo
->nBackfill
; i
--){
2421 volatile ht_slot
*dummy
;
2422 volatile u32
*aPgno
; /* Array of page numbers */
2423 u32 iZero
; /* Frame corresponding to aPgno[0] */
2424 u32 pgno
; /* Page number in db file */
2425 i64 iDbOff
; /* Offset of db file entry */
2426 i64 iWalOff
; /* Offset of wal file entry */
2428 rc
= walHashGet(pWal
, walFramePage(i
), &dummy
, &aPgno
, &iZero
);
2429 if( rc
!=SQLITE_OK
) break;
2430 pgno
= aPgno
[i
-iZero
];
2431 iDbOff
= (i64
)(pgno
-1) * szPage
;
2433 if( iDbOff
+szPage
<=szDb
){
2434 iWalOff
= walFrameOffset(i
, szPage
) + WAL_FRAME_HDRSIZE
;
2435 rc
= sqlite3OsRead(pWal
->pWalFd
, pBuf1
, szPage
, iWalOff
);
2437 if( rc
==SQLITE_OK
){
2438 rc
= sqlite3OsRead(pWal
->pDbFd
, pBuf2
, szPage
, iDbOff
);
2441 if( rc
!=SQLITE_OK
|| 0==memcmp(pBuf1
, pBuf2
, szPage
) ){
2446 pInfo
->nBackfillAttempted
= i
-1;
2450 sqlite3_free(pBuf1
);
2451 sqlite3_free(pBuf2
);
2453 walUnlockExclusive(pWal
, WAL_CKPT_LOCK
, 1);
2458 #endif /* SQLITE_ENABLE_SNAPSHOT */
2461 ** Begin a read transaction on the database.
2463 ** This routine used to be called sqlite3OpenSnapshot() and with good reason:
2464 ** it takes a snapshot of the state of the WAL and wal-index for the current
2465 ** instant in time. The current thread will continue to use this snapshot.
2466 ** Other threads might append new content to the WAL and wal-index but
2467 ** that extra content is ignored by the current thread.
2469 ** If the database contents have changes since the previous read
2470 ** transaction, then *pChanged is set to 1 before returning. The
2471 ** Pager layer will use this to know that is cache is stale and
2472 ** needs to be flushed.
2474 int sqlite3WalBeginReadTransaction(Wal
*pWal
, int *pChanged
){
2475 int rc
; /* Return code */
2476 int cnt
= 0; /* Number of TryBeginRead attempts */
2478 #ifdef SQLITE_ENABLE_SNAPSHOT
2480 WalIndexHdr
*pSnapshot
= pWal
->pSnapshot
;
2481 if( pSnapshot
&& memcmp(pSnapshot
, &pWal
->hdr
, sizeof(WalIndexHdr
))!=0 ){
2487 rc
= walTryBeginRead(pWal
, pChanged
, 0, ++cnt
);
2488 }while( rc
==WAL_RETRY
);
2489 testcase( (rc
&0xff)==SQLITE_BUSY
);
2490 testcase( (rc
&0xff)==SQLITE_IOERR
);
2491 testcase( rc
==SQLITE_PROTOCOL
);
2492 testcase( rc
==SQLITE_OK
);
2494 #ifdef SQLITE_ENABLE_SNAPSHOT
2495 if( rc
==SQLITE_OK
){
2496 if( pSnapshot
&& memcmp(pSnapshot
, &pWal
->hdr
, sizeof(WalIndexHdr
))!=0 ){
2497 /* At this point the client has a lock on an aReadMark[] slot holding
2498 ** a value equal to or smaller than pSnapshot->mxFrame, but pWal->hdr
2499 ** is populated with the wal-index header corresponding to the head
2500 ** of the wal file. Verify that pSnapshot is still valid before
2501 ** continuing. Reasons why pSnapshot might no longer be valid:
2503 ** (1) The WAL file has been reset since the snapshot was taken.
2504 ** In this case, the salt will have changed.
2506 ** (2) A checkpoint as been attempted that wrote frames past
2507 ** pSnapshot->mxFrame into the database file. Note that the
2508 ** checkpoint need not have completed for this to cause problems.
2510 volatile WalCkptInfo
*pInfo
= walCkptInfo(pWal
);
2512 assert( pWal
->readLock
>0 || pWal
->hdr
.mxFrame
==0 );
2513 assert( pInfo
->aReadMark
[pWal
->readLock
]<=pSnapshot
->mxFrame
);
2515 /* It is possible that there is a checkpointer thread running
2516 ** concurrent with this code. If this is the case, it may be that the
2517 ** checkpointer has already determined that it will checkpoint
2518 ** snapshot X, where X is later in the wal file than pSnapshot, but
2519 ** has not yet set the pInfo->nBackfillAttempted variable to indicate
2520 ** its intent. To avoid the race condition this leads to, ensure that
2521 ** there is no checkpointer process by taking a shared CKPT lock
2522 ** before checking pInfo->nBackfillAttempted.
2524 ** TODO: Does the aReadMark[] lock prevent a checkpointer from doing
2527 rc
= walLockShared(pWal
, WAL_CKPT_LOCK
);
2529 if( rc
==SQLITE_OK
){
2530 /* Check that the wal file has not been wrapped. Assuming that it has
2531 ** not, also check that no checkpointer has attempted to checkpoint any
2532 ** frames beyond pSnapshot->mxFrame. If either of these conditions are
2533 ** true, return SQLITE_BUSY_SNAPSHOT. Otherwise, overwrite pWal->hdr
2534 ** with *pSnapshot and set *pChanged as appropriate for opening the
2536 if( !memcmp(pSnapshot
->aSalt
, pWal
->hdr
.aSalt
, sizeof(pWal
->hdr
.aSalt
))
2537 && pSnapshot
->mxFrame
>=pInfo
->nBackfillAttempted
2539 assert( pWal
->readLock
>0 );
2540 memcpy(&pWal
->hdr
, pSnapshot
, sizeof(WalIndexHdr
));
2541 *pChanged
= bChanged
;
2543 rc
= SQLITE_BUSY_SNAPSHOT
;
2546 /* Release the shared CKPT lock obtained above. */
2547 walUnlockShared(pWal
, WAL_CKPT_LOCK
);
2551 if( rc
!=SQLITE_OK
){
2552 sqlite3WalEndReadTransaction(pWal
);
2561 ** Finish with a read transaction. All this does is release the
2564 void sqlite3WalEndReadTransaction(Wal
*pWal
){
2565 sqlite3WalEndWriteTransaction(pWal
);
2566 if( pWal
->readLock
>=0 ){
2567 walUnlockShared(pWal
, WAL_READ_LOCK(pWal
->readLock
));
2568 pWal
->readLock
= -1;
2573 ** Search the wal file for page pgno. If found, set *piRead to the frame that
2574 ** contains the page. Otherwise, if pgno is not in the wal file, set *piRead
2577 ** Return SQLITE_OK if successful, or an error code if an error occurs. If an
2578 ** error does occur, the final value of *piRead is undefined.
2580 int sqlite3WalFindFrame(
2581 Wal
*pWal
, /* WAL handle */
2582 Pgno pgno
, /* Database page number to read data for */
2583 u32
*piRead
/* OUT: Frame number (or zero) */
2585 u32 iRead
= 0; /* If !=0, WAL frame to return data from */
2586 u32 iLast
= pWal
->hdr
.mxFrame
; /* Last page in WAL for this reader */
2587 int iHash
; /* Used to loop through N hash tables */
2590 /* This routine is only be called from within a read transaction. */
2591 assert( pWal
->readLock
>=0 || pWal
->lockError
);
2593 /* If the "last page" field of the wal-index header snapshot is 0, then
2594 ** no data will be read from the wal under any circumstances. Return early
2595 ** in this case as an optimization. Likewise, if pWal->readLock==0,
2596 ** then the WAL is ignored by the reader so return early, as if the
2599 if( iLast
==0 || pWal
->readLock
==0 ){
2604 /* Search the hash table or tables for an entry matching page number
2605 ** pgno. Each iteration of the following for() loop searches one
2606 ** hash table (each hash table indexes up to HASHTABLE_NPAGE frames).
2608 ** This code might run concurrently to the code in walIndexAppend()
2609 ** that adds entries to the wal-index (and possibly to this hash
2610 ** table). This means the value just read from the hash
2611 ** slot (aHash[iKey]) may have been added before or after the
2612 ** current read transaction was opened. Values added after the
2613 ** read transaction was opened may have been written incorrectly -
2614 ** i.e. these slots may contain garbage data. However, we assume
2615 ** that any slots written before the current read transaction was
2616 ** opened remain unmodified.
2618 ** For the reasons above, the if(...) condition featured in the inner
2619 ** loop of the following block is more stringent that would be required
2620 ** if we had exclusive access to the hash-table:
2622 ** (aPgno[iFrame]==pgno):
2623 ** This condition filters out normal hash-table collisions.
2626 ** This condition filters out entries that were added to the hash
2627 ** table after the current read-transaction had started.
2629 iMinHash
= walFramePage(pWal
->minFrame
);
2630 for(iHash
=walFramePage(iLast
); iHash
>=iMinHash
&& iRead
==0; iHash
--){
2631 volatile ht_slot
*aHash
; /* Pointer to hash table */
2632 volatile u32
*aPgno
; /* Pointer to array of page numbers */
2633 u32 iZero
; /* Frame number corresponding to aPgno[0] */
2634 int iKey
; /* Hash slot index */
2635 int nCollide
; /* Number of hash collisions remaining */
2636 int rc
; /* Error code */
2638 rc
= walHashGet(pWal
, iHash
, &aHash
, &aPgno
, &iZero
);
2639 if( rc
!=SQLITE_OK
){
2642 nCollide
= HASHTABLE_NSLOT
;
2643 for(iKey
=walHash(pgno
); aHash
[iKey
]; iKey
=walNextHash(iKey
)){
2644 u32 iFrame
= aHash
[iKey
] + iZero
;
2645 if( iFrame
<=iLast
&& iFrame
>=pWal
->minFrame
&& aPgno
[aHash
[iKey
]]==pgno
){
2646 assert( iFrame
>iRead
|| CORRUPT_DB
);
2649 if( (nCollide
--)==0 ){
2650 return SQLITE_CORRUPT_BKPT
;
2655 #ifdef SQLITE_ENABLE_EXPENSIVE_ASSERT
2656 /* If expensive assert() statements are available, do a linear search
2657 ** of the wal-index file content. Make sure the results agree with the
2658 ** result obtained using the hash indexes above. */
2662 assert( pWal
->minFrame
>0 );
2663 for(iTest
=iLast
; iTest
>=pWal
->minFrame
; iTest
--){
2664 if( walFramePgno(pWal
, iTest
)==pgno
){
2669 assert( iRead
==iRead2
);
2678 ** Read the contents of frame iRead from the wal file into buffer pOut
2679 ** (which is nOut bytes in size). Return SQLITE_OK if successful, or an
2680 ** error code otherwise.
2682 int sqlite3WalReadFrame(
2683 Wal
*pWal
, /* WAL handle */
2684 u32 iRead
, /* Frame to read */
2685 int nOut
, /* Size of buffer pOut in bytes */
2686 u8
*pOut
/* Buffer to write page data to */
2690 sz
= pWal
->hdr
.szPage
;
2691 sz
= (sz
&0xfe00) + ((sz
&0x0001)<<16);
2692 testcase( sz
<=32768 );
2693 testcase( sz
>=65536 );
2694 iOffset
= walFrameOffset(iRead
, sz
) + WAL_FRAME_HDRSIZE
;
2695 /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL */
2696 return sqlite3OsRead(pWal
->pWalFd
, pOut
, (nOut
>sz
? sz
: nOut
), iOffset
);
2700 ** Return the size of the database in pages (or zero, if unknown).
2702 Pgno
sqlite3WalDbsize(Wal
*pWal
){
2703 if( pWal
&& ALWAYS(pWal
->readLock
>=0) ){
2704 return pWal
->hdr
.nPage
;
2711 ** This function starts a write transaction on the WAL.
2713 ** A read transaction must have already been started by a prior call
2714 ** to sqlite3WalBeginReadTransaction().
2716 ** If another thread or process has written into the database since
2717 ** the read transaction was started, then it is not possible for this
2718 ** thread to write as doing so would cause a fork. So this routine
2719 ** returns SQLITE_BUSY in that case and no write transaction is started.
2721 ** There can only be a single writer active at a time.
2723 int sqlite3WalBeginWriteTransaction(Wal
*pWal
){
2726 /* Cannot start a write transaction without first holding a read
2728 assert( pWal
->readLock
>=0 );
2729 assert( pWal
->writeLock
==0 && pWal
->iReCksum
==0 );
2731 if( pWal
->readOnly
){
2732 return SQLITE_READONLY
;
2735 /* Only one writer allowed at a time. Get the write lock. Return
2736 ** SQLITE_BUSY if unable.
2738 rc
= walLockExclusive(pWal
, WAL_WRITE_LOCK
, 1);
2742 pWal
->writeLock
= 1;
2744 /* If another connection has written to the database file since the
2745 ** time the read transaction on this connection was started, then
2746 ** the write is disallowed.
2748 if( memcmp(&pWal
->hdr
, (void *)walIndexHdr(pWal
), sizeof(WalIndexHdr
))!=0 ){
2749 walUnlockExclusive(pWal
, WAL_WRITE_LOCK
, 1);
2750 pWal
->writeLock
= 0;
2751 rc
= SQLITE_BUSY_SNAPSHOT
;
2758 ** End a write transaction. The commit has already been done. This
2759 ** routine merely releases the lock.
2761 int sqlite3WalEndWriteTransaction(Wal
*pWal
){
2762 if( pWal
->writeLock
){
2763 walUnlockExclusive(pWal
, WAL_WRITE_LOCK
, 1);
2764 pWal
->writeLock
= 0;
2766 pWal
->truncateOnCommit
= 0;
2772 ** If any data has been written (but not committed) to the log file, this
2773 ** function moves the write-pointer back to the start of the transaction.
2775 ** Additionally, the callback function is invoked for each frame written
2776 ** to the WAL since the start of the transaction. If the callback returns
2777 ** other than SQLITE_OK, it is not invoked again and the error code is
2778 ** returned to the caller.
2780 ** Otherwise, if the callback function does not return an error, this
2781 ** function returns SQLITE_OK.
2783 int sqlite3WalUndo(Wal
*pWal
, int (*xUndo
)(void *, Pgno
), void *pUndoCtx
){
2785 if( ALWAYS(pWal
->writeLock
) ){
2786 Pgno iMax
= pWal
->hdr
.mxFrame
;
2789 /* Restore the clients cache of the wal-index header to the state it
2790 ** was in before the client began writing to the database.
2792 memcpy(&pWal
->hdr
, (void *)walIndexHdr(pWal
), sizeof(WalIndexHdr
));
2794 for(iFrame
=pWal
->hdr
.mxFrame
+1;
2795 ALWAYS(rc
==SQLITE_OK
) && iFrame
<=iMax
;
2798 /* This call cannot fail. Unless the page for which the page number
2799 ** is passed as the second argument is (a) in the cache and
2800 ** (b) has an outstanding reference, then xUndo is either a no-op
2801 ** (if (a) is false) or simply expels the page from the cache (if (b)
2804 ** If the upper layer is doing a rollback, it is guaranteed that there
2805 ** are no outstanding references to any page other than page 1. And
2806 ** page 1 is never written to the log until the transaction is
2807 ** committed. As a result, the call to xUndo may not fail.
2809 assert( walFramePgno(pWal
, iFrame
)!=1 );
2810 rc
= xUndo(pUndoCtx
, walFramePgno(pWal
, iFrame
));
2812 if( iMax
!=pWal
->hdr
.mxFrame
) walCleanupHash(pWal
);
2818 ** Argument aWalData must point to an array of WAL_SAVEPOINT_NDATA u32
2819 ** values. This function populates the array with values required to
2820 ** "rollback" the write position of the WAL handle back to the current
2821 ** point in the event of a savepoint rollback (via WalSavepointUndo()).
2823 void sqlite3WalSavepoint(Wal
*pWal
, u32
*aWalData
){
2824 assert( pWal
->writeLock
);
2825 aWalData
[0] = pWal
->hdr
.mxFrame
;
2826 aWalData
[1] = pWal
->hdr
.aFrameCksum
[0];
2827 aWalData
[2] = pWal
->hdr
.aFrameCksum
[1];
2828 aWalData
[3] = pWal
->nCkpt
;
2832 ** Move the write position of the WAL back to the point identified by
2833 ** the values in the aWalData[] array. aWalData must point to an array
2834 ** of WAL_SAVEPOINT_NDATA u32 values that has been previously populated
2835 ** by a call to WalSavepoint().
2837 int sqlite3WalSavepointUndo(Wal
*pWal
, u32
*aWalData
){
2840 assert( pWal
->writeLock
);
2841 assert( aWalData
[3]!=pWal
->nCkpt
|| aWalData
[0]<=pWal
->hdr
.mxFrame
);
2843 if( aWalData
[3]!=pWal
->nCkpt
){
2844 /* This savepoint was opened immediately after the write-transaction
2845 ** was started. Right after that, the writer decided to wrap around
2846 ** to the start of the log. Update the savepoint values to match.
2849 aWalData
[3] = pWal
->nCkpt
;
2852 if( aWalData
[0]<pWal
->hdr
.mxFrame
){
2853 pWal
->hdr
.mxFrame
= aWalData
[0];
2854 pWal
->hdr
.aFrameCksum
[0] = aWalData
[1];
2855 pWal
->hdr
.aFrameCksum
[1] = aWalData
[2];
2856 walCleanupHash(pWal
);
2863 ** This function is called just before writing a set of frames to the log
2864 ** file (see sqlite3WalFrames()). It checks to see if, instead of appending
2865 ** to the current log file, it is possible to overwrite the start of the
2866 ** existing log file with the new frames (i.e. "reset" the log). If so,
2867 ** it sets pWal->hdr.mxFrame to 0. Otherwise, pWal->hdr.mxFrame is left
2870 ** SQLITE_OK is returned if no error is encountered (regardless of whether
2871 ** or not pWal->hdr.mxFrame is modified). An SQLite error code is returned
2872 ** if an error occurs.
2874 static int walRestartLog(Wal
*pWal
){
2878 if( pWal
->readLock
==0 ){
2879 volatile WalCkptInfo
*pInfo
= walCkptInfo(pWal
);
2880 assert( pInfo
->nBackfill
==pWal
->hdr
.mxFrame
);
2881 if( pInfo
->nBackfill
>0 ){
2883 sqlite3_randomness(4, &salt1
);
2884 rc
= walLockExclusive(pWal
, WAL_READ_LOCK(1), WAL_NREADER
-1);
2885 if( rc
==SQLITE_OK
){
2886 /* If all readers are using WAL_READ_LOCK(0) (in other words if no
2887 ** readers are currently using the WAL), then the transactions
2888 ** frames will overwrite the start of the existing log. Update the
2889 ** wal-index header to reflect this.
2891 ** In theory it would be Ok to update the cache of the header only
2892 ** at this point. But updating the actual wal-index header is also
2893 ** safe and means there is no special case for sqlite3WalUndo()
2894 ** to handle if this transaction is rolled back. */
2895 walRestartHdr(pWal
, salt1
);
2896 walUnlockExclusive(pWal
, WAL_READ_LOCK(1), WAL_NREADER
-1);
2897 }else if( rc
!=SQLITE_BUSY
){
2901 walUnlockShared(pWal
, WAL_READ_LOCK(0));
2902 pWal
->readLock
= -1;
2906 rc
= walTryBeginRead(pWal
, ¬Used
, 1, ++cnt
);
2907 }while( rc
==WAL_RETRY
);
2908 assert( (rc
&0xff)!=SQLITE_BUSY
); /* BUSY not possible when useWal==1 */
2909 testcase( (rc
&0xff)==SQLITE_IOERR
);
2910 testcase( rc
==SQLITE_PROTOCOL
);
2911 testcase( rc
==SQLITE_OK
);
2917 ** Information about the current state of the WAL file and where
2918 ** the next fsync should occur - passed from sqlite3WalFrames() into
2921 typedef struct WalWriter
{
2922 Wal
*pWal
; /* The complete WAL information */
2923 sqlite3_file
*pFd
; /* The WAL file to which we write */
2924 sqlite3_int64 iSyncPoint
; /* Fsync at this offset */
2925 int syncFlags
; /* Flags for the fsync */
2926 int szPage
; /* Size of one page */
2930 ** Write iAmt bytes of content into the WAL file beginning at iOffset.
2931 ** Do a sync when crossing the p->iSyncPoint boundary.
2933 ** In other words, if iSyncPoint is in between iOffset and iOffset+iAmt,
2934 ** first write the part before iSyncPoint, then sync, then write the
2937 static int walWriteToLog(
2938 WalWriter
*p
, /* WAL to write to */
2939 void *pContent
, /* Content to be written */
2940 int iAmt
, /* Number of bytes to write */
2941 sqlite3_int64 iOffset
/* Start writing at this offset */
2944 if( iOffset
<p
->iSyncPoint
&& iOffset
+iAmt
>=p
->iSyncPoint
){
2945 int iFirstAmt
= (int)(p
->iSyncPoint
- iOffset
);
2946 rc
= sqlite3OsWrite(p
->pFd
, pContent
, iFirstAmt
, iOffset
);
2948 iOffset
+= iFirstAmt
;
2950 pContent
= (void*)(iFirstAmt
+ (char*)pContent
);
2951 assert( p
->syncFlags
& (SQLITE_SYNC_NORMAL
|SQLITE_SYNC_FULL
) );
2952 rc
= sqlite3OsSync(p
->pFd
, p
->syncFlags
& SQLITE_SYNC_MASK
);
2953 if( iAmt
==0 || rc
) return rc
;
2955 rc
= sqlite3OsWrite(p
->pFd
, pContent
, iAmt
, iOffset
);
2960 ** Write out a single frame of the WAL
2962 static int walWriteOneFrame(
2963 WalWriter
*p
, /* Where to write the frame */
2964 PgHdr
*pPage
, /* The page of the frame to be written */
2965 int nTruncate
, /* The commit flag. Usually 0. >0 for commit */
2966 sqlite3_int64 iOffset
/* Byte offset at which to write */
2968 int rc
; /* Result code from subfunctions */
2969 void *pData
; /* Data actually written */
2970 u8 aFrame
[WAL_FRAME_HDRSIZE
]; /* Buffer to assemble frame-header in */
2971 #if defined(SQLITE_HAS_CODEC)
2972 if( (pData
= sqlite3PagerCodec(pPage
))==0 ) return SQLITE_NOMEM_BKPT
;
2974 pData
= pPage
->pData
;
2976 walEncodeFrame(p
->pWal
, pPage
->pgno
, nTruncate
, pData
, aFrame
);
2977 rc
= walWriteToLog(p
, aFrame
, sizeof(aFrame
), iOffset
);
2979 /* Write the page data */
2980 rc
= walWriteToLog(p
, pData
, p
->szPage
, iOffset
+sizeof(aFrame
));
2985 ** This function is called as part of committing a transaction within which
2986 ** one or more frames have been overwritten. It updates the checksums for
2987 ** all frames written to the wal file by the current transaction starting
2988 ** with the earliest to have been overwritten.
2990 ** SQLITE_OK is returned if successful, or an SQLite error code otherwise.
2992 static int walRewriteChecksums(Wal
*pWal
, u32 iLast
){
2993 const int szPage
= pWal
->szPage
;/* Database page size */
2994 int rc
= SQLITE_OK
; /* Return code */
2995 u8
*aBuf
; /* Buffer to load data from wal file into */
2996 u8 aFrame
[WAL_FRAME_HDRSIZE
]; /* Buffer to assemble frame-headers in */
2997 u32 iRead
; /* Next frame to read from wal file */
3000 aBuf
= sqlite3_malloc(szPage
+ WAL_FRAME_HDRSIZE
);
3001 if( aBuf
==0 ) return SQLITE_NOMEM_BKPT
;
3003 /* Find the checksum values to use as input for the recalculating the
3004 ** first checksum. If the first frame is frame 1 (implying that the current
3005 ** transaction restarted the wal file), these values must be read from the
3006 ** wal-file header. Otherwise, read them from the frame header of the
3007 ** previous frame. */
3008 assert( pWal
->iReCksum
>0 );
3009 if( pWal
->iReCksum
==1 ){
3012 iCksumOff
= walFrameOffset(pWal
->iReCksum
-1, szPage
) + 16;
3014 rc
= sqlite3OsRead(pWal
->pWalFd
, aBuf
, sizeof(u32
)*2, iCksumOff
);
3015 pWal
->hdr
.aFrameCksum
[0] = sqlite3Get4byte(aBuf
);
3016 pWal
->hdr
.aFrameCksum
[1] = sqlite3Get4byte(&aBuf
[sizeof(u32
)]);
3018 iRead
= pWal
->iReCksum
;
3020 for(; rc
==SQLITE_OK
&& iRead
<=iLast
; iRead
++){
3021 i64 iOff
= walFrameOffset(iRead
, szPage
);
3022 rc
= sqlite3OsRead(pWal
->pWalFd
, aBuf
, szPage
+WAL_FRAME_HDRSIZE
, iOff
);
3023 if( rc
==SQLITE_OK
){
3025 iPgno
= sqlite3Get4byte(aBuf
);
3026 nDbSize
= sqlite3Get4byte(&aBuf
[4]);
3028 walEncodeFrame(pWal
, iPgno
, nDbSize
, &aBuf
[WAL_FRAME_HDRSIZE
], aFrame
);
3029 rc
= sqlite3OsWrite(pWal
->pWalFd
, aFrame
, sizeof(aFrame
), iOff
);
3038 ** Write a set of frames to the log. The caller must hold the write-lock
3039 ** on the log file (obtained using sqlite3WalBeginWriteTransaction()).
3041 int sqlite3WalFrames(
3042 Wal
*pWal
, /* Wal handle to write to */
3043 int szPage
, /* Database page-size in bytes */
3044 PgHdr
*pList
, /* List of dirty pages to write */
3045 Pgno nTruncate
, /* Database size after this commit */
3046 int isCommit
, /* True if this is a commit */
3047 int sync_flags
/* Flags to pass to OsSync() (or 0) */
3049 int rc
; /* Used to catch return codes */
3050 u32 iFrame
; /* Next frame address */
3051 PgHdr
*p
; /* Iterator to run through pList with. */
3052 PgHdr
*pLast
= 0; /* Last frame in list */
3053 int nExtra
= 0; /* Number of extra copies of last page */
3054 int szFrame
; /* The size of a single frame */
3055 i64 iOffset
; /* Next byte to write in WAL file */
3056 WalWriter w
; /* The writer */
3057 u32 iFirst
= 0; /* First frame that may be overwritten */
3058 WalIndexHdr
*pLive
; /* Pointer to shared header */
3061 assert( pWal
->writeLock
);
3063 /* If this frame set completes a transaction, then nTruncate>0. If
3064 ** nTruncate==0 then this frame set does not complete the transaction. */
3065 assert( (isCommit
!=0)==(nTruncate
!=0) );
3067 #if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
3068 { int cnt
; for(cnt
=0, p
=pList
; p
; p
=p
->pDirty
, cnt
++){}
3069 WALTRACE(("WAL%p: frame write begin. %d frames. mxFrame=%d. %s\n",
3070 pWal
, cnt
, pWal
->hdr
.mxFrame
, isCommit
? "Commit" : "Spill"));
3074 pLive
= (WalIndexHdr
*)walIndexHdr(pWal
);
3075 if( memcmp(&pWal
->hdr
, (void *)pLive
, sizeof(WalIndexHdr
))!=0 ){
3076 iFirst
= pLive
->mxFrame
+1;
3079 /* See if it is possible to write these frames into the start of the
3080 ** log file, instead of appending to it at pWal->hdr.mxFrame.
3082 if( SQLITE_OK
!=(rc
= walRestartLog(pWal
)) ){
3086 /* If this is the first frame written into the log, write the WAL
3087 ** header to the start of the WAL file. See comments at the top of
3088 ** this source file for a description of the WAL header format.
3090 iFrame
= pWal
->hdr
.mxFrame
;
3092 u8 aWalHdr
[WAL_HDRSIZE
]; /* Buffer to assemble wal-header in */
3093 u32 aCksum
[2]; /* Checksum for wal-header */
3095 sqlite3Put4byte(&aWalHdr
[0], (WAL_MAGIC
| SQLITE_BIGENDIAN
));
3096 sqlite3Put4byte(&aWalHdr
[4], WAL_MAX_VERSION
);
3097 sqlite3Put4byte(&aWalHdr
[8], szPage
);
3098 sqlite3Put4byte(&aWalHdr
[12], pWal
->nCkpt
);
3099 if( pWal
->nCkpt
==0 ) sqlite3_randomness(8, pWal
->hdr
.aSalt
);
3100 memcpy(&aWalHdr
[16], pWal
->hdr
.aSalt
, 8);
3101 walChecksumBytes(1, aWalHdr
, WAL_HDRSIZE
-2*4, 0, aCksum
);
3102 sqlite3Put4byte(&aWalHdr
[24], aCksum
[0]);
3103 sqlite3Put4byte(&aWalHdr
[28], aCksum
[1]);
3105 pWal
->szPage
= szPage
;
3106 pWal
->hdr
.bigEndCksum
= SQLITE_BIGENDIAN
;
3107 pWal
->hdr
.aFrameCksum
[0] = aCksum
[0];
3108 pWal
->hdr
.aFrameCksum
[1] = aCksum
[1];
3109 pWal
->truncateOnCommit
= 1;
3111 rc
= sqlite3OsWrite(pWal
->pWalFd
, aWalHdr
, sizeof(aWalHdr
), 0);
3112 WALTRACE(("WAL%p: wal-header write %s\n", pWal
, rc
? "failed" : "ok"));
3113 if( rc
!=SQLITE_OK
){
3117 /* Sync the header (unless SQLITE_IOCAP_SEQUENTIAL is true or unless
3118 ** all syncing is turned off by PRAGMA synchronous=OFF). Otherwise
3119 ** an out-of-order write following a WAL restart could result in
3120 ** database corruption. See the ticket:
3122 ** http://localhost:591/sqlite/info/ff5be73dee
3124 if( pWal
->syncHeader
&& sync_flags
){
3125 rc
= sqlite3OsSync(pWal
->pWalFd
, sync_flags
& SQLITE_SYNC_MASK
);
3129 assert( (int)pWal
->szPage
==szPage
);
3131 /* Setup information needed to write frames into the WAL */
3133 w
.pFd
= pWal
->pWalFd
;
3135 w
.syncFlags
= sync_flags
;
3137 iOffset
= walFrameOffset(iFrame
+1, szPage
);
3138 szFrame
= szPage
+ WAL_FRAME_HDRSIZE
;
3140 /* Write all frames into the log file exactly once */
3141 for(p
=pList
; p
; p
=p
->pDirty
){
3142 int nDbSize
; /* 0 normally. Positive == commit flag */
3144 /* Check if this page has already been written into the wal file by
3145 ** the current transaction. If so, overwrite the existing frame and
3146 ** set Wal.writeLock to WAL_WRITELOCK_RECKSUM - indicating that
3147 ** checksums must be recomputed when the transaction is committed. */
3148 if( iFirst
&& (p
->pDirty
|| isCommit
==0) ){
3150 VVA_ONLY(rc
=) sqlite3WalFindFrame(pWal
, p
->pgno
, &iWrite
);
3151 assert( rc
==SQLITE_OK
|| iWrite
==0 );
3152 if( iWrite
>=iFirst
){
3153 i64 iOff
= walFrameOffset(iWrite
, szPage
) + WAL_FRAME_HDRSIZE
;
3155 if( pWal
->iReCksum
==0 || iWrite
<pWal
->iReCksum
){
3156 pWal
->iReCksum
= iWrite
;
3158 #if defined(SQLITE_HAS_CODEC)
3159 if( (pData
= sqlite3PagerCodec(p
))==0 ) return SQLITE_NOMEM
;
3163 rc
= sqlite3OsWrite(pWal
->pWalFd
, pData
, szPage
, iOff
);
3165 p
->flags
&= ~PGHDR_WAL_APPEND
;
3171 assert( iOffset
==walFrameOffset(iFrame
, szPage
) );
3172 nDbSize
= (isCommit
&& p
->pDirty
==0) ? nTruncate
: 0;
3173 rc
= walWriteOneFrame(&w
, p
, nDbSize
, iOffset
);
3177 p
->flags
|= PGHDR_WAL_APPEND
;
3180 /* Recalculate checksums within the wal file if required. */
3181 if( isCommit
&& pWal
->iReCksum
){
3182 rc
= walRewriteChecksums(pWal
, iFrame
);
3186 /* If this is the end of a transaction, then we might need to pad
3187 ** the transaction and/or sync the WAL file.
3189 ** Padding and syncing only occur if this set of frames complete a
3190 ** transaction and if PRAGMA synchronous=FULL. If synchronous==NORMAL
3191 ** or synchronous==OFF, then no padding or syncing are needed.
3193 ** If SQLITE_IOCAP_POWERSAFE_OVERWRITE is defined, then padding is not
3194 ** needed and only the sync is done. If padding is needed, then the
3195 ** final frame is repeated (with its commit mark) until the next sector
3196 ** boundary is crossed. Only the part of the WAL prior to the last
3197 ** sector boundary is synced; the part of the last frame that extends
3198 ** past the sector boundary is written after the sync.
3200 if( isCommit
&& (sync_flags
& WAL_SYNC_TRANSACTIONS
)!=0 ){
3202 if( pWal
->padToSectorBoundary
){
3203 int sectorSize
= sqlite3SectorSize(pWal
->pWalFd
);
3204 w
.iSyncPoint
= ((iOffset
+sectorSize
-1)/sectorSize
)*sectorSize
;
3205 bSync
= (w
.iSyncPoint
==iOffset
);
3207 while( iOffset
<w
.iSyncPoint
){
3208 rc
= walWriteOneFrame(&w
, pLast
, nTruncate
, iOffset
);
3215 assert( rc
==SQLITE_OK
);
3216 rc
= sqlite3OsSync(w
.pFd
, sync_flags
& SQLITE_SYNC_MASK
);
3220 /* If this frame set completes the first transaction in the WAL and
3221 ** if PRAGMA journal_size_limit is set, then truncate the WAL to the
3222 ** journal size limit, if possible.
3224 if( isCommit
&& pWal
->truncateOnCommit
&& pWal
->mxWalSize
>=0 ){
3225 i64 sz
= pWal
->mxWalSize
;
3226 if( walFrameOffset(iFrame
+nExtra
+1, szPage
)>pWal
->mxWalSize
){
3227 sz
= walFrameOffset(iFrame
+nExtra
+1, szPage
);
3229 walLimitSize(pWal
, sz
);
3230 pWal
->truncateOnCommit
= 0;
3233 /* Append data to the wal-index. It is not necessary to lock the
3234 ** wal-index to do this as the SQLITE_SHM_WRITE lock held on the wal-index
3235 ** guarantees that there are no other writers, and no data that may
3236 ** be in use by existing readers is being overwritten.
3238 iFrame
= pWal
->hdr
.mxFrame
;
3239 for(p
=pList
; p
&& rc
==SQLITE_OK
; p
=p
->pDirty
){
3240 if( (p
->flags
& PGHDR_WAL_APPEND
)==0 ) continue;
3242 rc
= walIndexAppend(pWal
, iFrame
, p
->pgno
);
3244 while( rc
==SQLITE_OK
&& nExtra
>0 ){
3247 rc
= walIndexAppend(pWal
, iFrame
, pLast
->pgno
);
3250 if( rc
==SQLITE_OK
){
3251 /* Update the private copy of the header. */
3252 pWal
->hdr
.szPage
= (u16
)((szPage
&0xff00) | (szPage
>>16));
3253 testcase( szPage
<=32768 );
3254 testcase( szPage
>=65536 );
3255 pWal
->hdr
.mxFrame
= iFrame
;
3257 pWal
->hdr
.iChange
++;
3258 pWal
->hdr
.nPage
= nTruncate
;
3260 /* If this is a commit, update the wal-index header too. */
3262 walIndexWriteHdr(pWal
);
3263 pWal
->iCallback
= iFrame
;
3267 WALTRACE(("WAL%p: frame write %s\n", pWal
, rc
? "failed" : "ok"));
3272 ** This routine is called to implement sqlite3_wal_checkpoint() and
3273 ** related interfaces.
3275 ** Obtain a CHECKPOINT lock and then backfill as much information as
3276 ** we can from WAL into the database.
3278 ** If parameter xBusy is not NULL, it is a pointer to a busy-handler
3279 ** callback. In this case this function runs a blocking checkpoint.
3281 int sqlite3WalCheckpoint(
3282 Wal
*pWal
, /* Wal connection */
3283 sqlite3
*db
, /* Check this handle's interrupt flag */
3284 int eMode
, /* PASSIVE, FULL, RESTART, or TRUNCATE */
3285 int (*xBusy
)(void*), /* Function to call when busy */
3286 void *pBusyArg
, /* Context argument for xBusyHandler */
3287 int sync_flags
, /* Flags to sync db file with (or 0) */
3288 int nBuf
, /* Size of temporary buffer */
3289 u8
*zBuf
, /* Temporary buffer to use */
3290 int *pnLog
, /* OUT: Number of frames in WAL */
3291 int *pnCkpt
/* OUT: Number of backfilled frames in WAL */
3293 int rc
; /* Return code */
3294 int isChanged
= 0; /* True if a new wal-index header is loaded */
3295 int eMode2
= eMode
; /* Mode to pass to walCheckpoint() */
3296 int (*xBusy2
)(void*) = xBusy
; /* Busy handler for eMode2 */
3298 assert( pWal
->ckptLock
==0 );
3299 assert( pWal
->writeLock
==0 );
3301 /* EVIDENCE-OF: R-62920-47450 The busy-handler callback is never invoked
3302 ** in the SQLITE_CHECKPOINT_PASSIVE mode. */
3303 assert( eMode
!=SQLITE_CHECKPOINT_PASSIVE
|| xBusy
==0 );
3305 if( pWal
->readOnly
) return SQLITE_READONLY
;
3306 WALTRACE(("WAL%p: checkpoint begins\n", pWal
));
3308 /* IMPLEMENTATION-OF: R-62028-47212 All calls obtain an exclusive
3309 ** "checkpoint" lock on the database file. */
3310 rc
= walLockExclusive(pWal
, WAL_CKPT_LOCK
, 1);
3312 /* EVIDENCE-OF: R-10421-19736 If any other process is running a
3313 ** checkpoint operation at the same time, the lock cannot be obtained and
3314 ** SQLITE_BUSY is returned.
3315 ** EVIDENCE-OF: R-53820-33897 Even if there is a busy-handler configured,
3316 ** it will not be invoked in this case.
3318 testcase( rc
==SQLITE_BUSY
);
3319 testcase( xBusy
!=0 );
3324 /* IMPLEMENTATION-OF: R-59782-36818 The SQLITE_CHECKPOINT_FULL, RESTART and
3325 ** TRUNCATE modes also obtain the exclusive "writer" lock on the database
3328 ** EVIDENCE-OF: R-60642-04082 If the writer lock cannot be obtained
3329 ** immediately, and a busy-handler is configured, it is invoked and the
3330 ** writer lock retried until either the busy-handler returns 0 or the
3331 ** lock is successfully obtained.
3333 if( eMode
!=SQLITE_CHECKPOINT_PASSIVE
){
3334 rc
= walBusyLock(pWal
, xBusy
, pBusyArg
, WAL_WRITE_LOCK
, 1);
3335 if( rc
==SQLITE_OK
){
3336 pWal
->writeLock
= 1;
3337 }else if( rc
==SQLITE_BUSY
){
3338 eMode2
= SQLITE_CHECKPOINT_PASSIVE
;
3344 /* Read the wal-index header. */
3345 if( rc
==SQLITE_OK
){
3346 rc
= walIndexReadHdr(pWal
, &isChanged
);
3347 if( isChanged
&& pWal
->pDbFd
->pMethods
->iVersion
>=3 ){
3348 sqlite3OsUnfetch(pWal
->pDbFd
, 0, 0);
3352 /* Copy data from the log to the database file. */
3353 if( rc
==SQLITE_OK
){
3355 if( pWal
->hdr
.mxFrame
&& walPagesize(pWal
)!=nBuf
){
3356 rc
= SQLITE_CORRUPT_BKPT
;
3358 rc
= walCheckpoint(pWal
, db
, eMode2
, xBusy2
, pBusyArg
, sync_flags
, zBuf
);
3361 /* If no error occurred, set the output variables. */
3362 if( rc
==SQLITE_OK
|| rc
==SQLITE_BUSY
){
3363 if( pnLog
) *pnLog
= (int)pWal
->hdr
.mxFrame
;
3364 if( pnCkpt
) *pnCkpt
= (int)(walCkptInfo(pWal
)->nBackfill
);
3369 /* If a new wal-index header was loaded before the checkpoint was
3370 ** performed, then the pager-cache associated with pWal is now
3371 ** out of date. So zero the cached wal-index header to ensure that
3372 ** next time the pager opens a snapshot on this database it knows that
3373 ** the cache needs to be reset.
3375 memset(&pWal
->hdr
, 0, sizeof(WalIndexHdr
));
3378 /* Release the locks. */
3379 sqlite3WalEndWriteTransaction(pWal
);
3380 walUnlockExclusive(pWal
, WAL_CKPT_LOCK
, 1);
3382 WALTRACE(("WAL%p: checkpoint %s\n", pWal
, rc
? "failed" : "ok"));
3383 return (rc
==SQLITE_OK
&& eMode
!=eMode2
? SQLITE_BUSY
: rc
);
3386 /* Return the value to pass to a sqlite3_wal_hook callback, the
3387 ** number of frames in the WAL at the point of the last commit since
3388 ** sqlite3WalCallback() was called. If no commits have occurred since
3389 ** the last call, then return 0.
3391 int sqlite3WalCallback(Wal
*pWal
){
3394 ret
= pWal
->iCallback
;
3395 pWal
->iCallback
= 0;
3401 ** This function is called to change the WAL subsystem into or out
3402 ** of locking_mode=EXCLUSIVE.
3404 ** If op is zero, then attempt to change from locking_mode=EXCLUSIVE
3405 ** into locking_mode=NORMAL. This means that we must acquire a lock
3406 ** on the pWal->readLock byte. If the WAL is already in locking_mode=NORMAL
3407 ** or if the acquisition of the lock fails, then return 0. If the
3408 ** transition out of exclusive-mode is successful, return 1. This
3409 ** operation must occur while the pager is still holding the exclusive
3410 ** lock on the main database file.
3412 ** If op is one, then change from locking_mode=NORMAL into
3413 ** locking_mode=EXCLUSIVE. This means that the pWal->readLock must
3414 ** be released. Return 1 if the transition is made and 0 if the
3415 ** WAL is already in exclusive-locking mode - meaning that this
3416 ** routine is a no-op. The pager must already hold the exclusive lock
3417 ** on the main database file before invoking this operation.
3419 ** If op is negative, then do a dry-run of the op==1 case but do
3420 ** not actually change anything. The pager uses this to see if it
3421 ** should acquire the database exclusive lock prior to invoking
3424 int sqlite3WalExclusiveMode(Wal
*pWal
, int op
){
3426 assert( pWal
->writeLock
==0 );
3427 assert( pWal
->exclusiveMode
!=WAL_HEAPMEMORY_MODE
|| op
==-1 );
3429 /* pWal->readLock is usually set, but might be -1 if there was a
3430 ** prior error while attempting to acquire are read-lock. This cannot
3431 ** happen if the connection is actually in exclusive mode (as no xShmLock
3432 ** locks are taken in this case). Nor should the pager attempt to
3433 ** upgrade to exclusive-mode following such an error.
3435 assert( pWal
->readLock
>=0 || pWal
->lockError
);
3436 assert( pWal
->readLock
>=0 || (op
<=0 && pWal
->exclusiveMode
==0) );
3439 if( pWal
->exclusiveMode
){
3440 pWal
->exclusiveMode
= 0;
3441 if( walLockShared(pWal
, WAL_READ_LOCK(pWal
->readLock
))!=SQLITE_OK
){
3442 pWal
->exclusiveMode
= 1;
3444 rc
= pWal
->exclusiveMode
==0;
3446 /* Already in locking_mode=NORMAL */
3450 assert( pWal
->exclusiveMode
==0 );
3451 assert( pWal
->readLock
>=0 );
3452 walUnlockShared(pWal
, WAL_READ_LOCK(pWal
->readLock
));
3453 pWal
->exclusiveMode
= 1;
3456 rc
= pWal
->exclusiveMode
==0;
3462 ** Return true if the argument is non-NULL and the WAL module is using
3463 ** heap-memory for the wal-index. Otherwise, if the argument is NULL or the
3464 ** WAL module is using shared-memory, return false.
3466 int sqlite3WalHeapMemory(Wal
*pWal
){
3467 return (pWal
&& pWal
->exclusiveMode
==WAL_HEAPMEMORY_MODE
);
3470 #ifdef SQLITE_ENABLE_SNAPSHOT
3471 /* Create a snapshot object. The content of a snapshot is opaque to
3472 ** every other subsystem, so the WAL module can put whatever it needs
3475 int sqlite3WalSnapshotGet(Wal
*pWal
, sqlite3_snapshot
**ppSnapshot
){
3478 static const u32 aZero
[4] = { 0, 0, 0, 0 };
3480 assert( pWal
->readLock
>=0 && pWal
->writeLock
==0 );
3482 if( memcmp(&pWal
->hdr
.aFrameCksum
[0],aZero
,16)==0 ){
3484 return SQLITE_ERROR
;
3486 pRet
= (WalIndexHdr
*)sqlite3_malloc(sizeof(WalIndexHdr
));
3488 rc
= SQLITE_NOMEM_BKPT
;
3490 memcpy(pRet
, &pWal
->hdr
, sizeof(WalIndexHdr
));
3491 *ppSnapshot
= (sqlite3_snapshot
*)pRet
;
3497 /* Try to open on pSnapshot when the next read-transaction starts
3499 void sqlite3WalSnapshotOpen(Wal
*pWal
, sqlite3_snapshot
*pSnapshot
){
3500 pWal
->pSnapshot
= (WalIndexHdr
*)pSnapshot
;
3504 ** Return a +ve value if snapshot p1 is newer than p2. A -ve value if
3505 ** p1 is older than p2 and zero if p1 and p2 are the same snapshot.
3507 int sqlite3_snapshot_cmp(sqlite3_snapshot
*p1
, sqlite3_snapshot
*p2
){
3508 WalIndexHdr
*pHdr1
= (WalIndexHdr
*)p1
;
3509 WalIndexHdr
*pHdr2
= (WalIndexHdr
*)p2
;
3511 /* aSalt[0] is a copy of the value stored in the wal file header. It
3512 ** is incremented each time the wal file is restarted. */
3513 if( pHdr1
->aSalt
[0]<pHdr2
->aSalt
[0] ) return -1;
3514 if( pHdr1
->aSalt
[0]>pHdr2
->aSalt
[0] ) return +1;
3515 if( pHdr1
->mxFrame
<pHdr2
->mxFrame
) return -1;
3516 if( pHdr1
->mxFrame
>pHdr2
->mxFrame
) return +1;
3519 #endif /* SQLITE_ENABLE_SNAPSHOT */
3521 #ifdef SQLITE_ENABLE_ZIPVFS
3523 ** If the argument is not NULL, it points to a Wal object that holds a
3524 ** read-lock. This function returns the database page-size if it is known,
3525 ** or zero if it is not (or if pWal is NULL).
3527 int sqlite3WalFramesize(Wal
*pWal
){
3528 assert( pWal
==0 || pWal
->readLock
>=0 );
3529 return (pWal
? pWal
->szPage
: 0);
3533 /* Return the sqlite3_file object for the WAL file
3535 sqlite3_file
*sqlite3WalFile(Wal
*pWal
){
3536 return pWal
->pWalFd
;
3539 #endif /* #ifndef SQLITE_OMIT_WAL */