remove redefines for PAGER_MJ_PGNO to PAGER_SJ_PGNO
[sqlcipher.git] / src / pager.c
blob2545e6cc9cb352681970d10cf72c909417d31d15
1 /*
2 ** 2001 September 15
3 **
4 ** The author disclaims copyright to this source code. In place of
5 ** a legal notice, here is a blessing:
6 **
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 *************************************************************************
12 ** This is the implementation of the page cache subsystem or "pager".
13 **
14 ** The pager is used to access a database disk file. It implements
15 ** atomic commit and rollback through the use of a journal file that
16 ** is separate from the database file. The pager also implements file
17 ** locking to prevent two processes from writing the same database
18 ** file simultaneously, or one process from reading the database while
19 ** another is writing.
21 #ifndef SQLITE_OMIT_DISKIO
22 #include "sqliteInt.h"
23 #include "wal.h"
26 /******************* NOTES ON THE DESIGN OF THE PAGER ************************
28 ** This comment block describes invariants that hold when using a rollback
29 ** journal. These invariants do not apply for journal_mode=WAL,
30 ** journal_mode=MEMORY, or journal_mode=OFF.
32 ** Within this comment block, a page is deemed to have been synced
33 ** automatically as soon as it is written when PRAGMA synchronous=OFF.
34 ** Otherwise, the page is not synced until the xSync method of the VFS
35 ** is called successfully on the file containing the page.
37 ** Definition: A page of the database file is said to be "overwriteable" if
38 ** one or more of the following are true about the page:
39 **
40 ** (a) The original content of the page as it was at the beginning of
41 ** the transaction has been written into the rollback journal and
42 ** synced.
43 **
44 ** (b) The page was a freelist leaf page at the start of the transaction.
45 **
46 ** (c) The page number is greater than the largest page that existed in
47 ** the database file at the start of the transaction.
48 **
49 ** (1) A page of the database file is never overwritten unless one of the
50 ** following are true:
51 **
52 ** (a) The page and all other pages on the same sector are overwriteable.
53 **
54 ** (b) The atomic page write optimization is enabled, and the entire
55 ** transaction other than the update of the transaction sequence
56 ** number consists of a single page change.
57 **
58 ** (2) The content of a page written into the rollback journal exactly matches
59 ** both the content in the database when the rollback journal was written
60 ** and the content in the database at the beginning of the current
61 ** transaction.
62 **
63 ** (3) Writes to the database file are an integer multiple of the page size
64 ** in length and are aligned on a page boundary.
65 **
66 ** (4) Reads from the database file are either aligned on a page boundary and
67 ** an integer multiple of the page size in length or are taken from the
68 ** first 100 bytes of the database file.
69 **
70 ** (5) All writes to the database file are synced prior to the rollback journal
71 ** being deleted, truncated, or zeroed.
72 **
73 ** (6) If a super-journal file is used, then all writes to the database file
74 ** are synced prior to the super-journal being deleted.
75 **
76 ** Definition: Two databases (or the same database at two points it time)
77 ** are said to be "logically equivalent" if they give the same answer to
78 ** all queries. Note in particular the content of freelist leaf
79 ** pages can be changed arbitrarily without affecting the logical equivalence
80 ** of the database.
81 **
82 ** (7) At any time, if any subset, including the empty set and the total set,
83 ** of the unsynced changes to a rollback journal are removed and the
84 ** journal is rolled back, the resulting database file will be logically
85 ** equivalent to the database file at the beginning of the transaction.
86 **
87 ** (8) When a transaction is rolled back, the xTruncate method of the VFS
88 ** is called to restore the database file to the same size it was at
89 ** the beginning of the transaction. (In some VFSes, the xTruncate
90 ** method is a no-op, but that does not change the fact the SQLite will
91 ** invoke it.)
92 **
93 ** (9) Whenever the database file is modified, at least one bit in the range
94 ** of bytes from 24 through 39 inclusive will be changed prior to releasing
95 ** the EXCLUSIVE lock, thus signaling other connections on the same
96 ** database to flush their caches.
98 ** (10) The pattern of bits in bytes 24 through 39 shall not repeat in less
99 ** than one billion transactions.
101 ** (11) A database file is well-formed at the beginning and at the conclusion
102 ** of every transaction.
104 ** (12) An EXCLUSIVE lock is held on the database file when writing to
105 ** the database file.
107 ** (13) A SHARED lock is held on the database file while reading any
108 ** content out of the database file.
110 ******************************************************************************/
113 ** Macros for troubleshooting. Normally turned off
115 #if 0
116 int sqlite3PagerTrace=1; /* True to enable tracing */
117 #define sqlite3DebugPrintf printf
118 #define PAGERTRACE(X) if( sqlite3PagerTrace ){ sqlite3DebugPrintf X; }
119 #else
120 #define PAGERTRACE(X)
121 #endif
124 ** The following two macros are used within the PAGERTRACE() macros above
125 ** to print out file-descriptors.
127 ** PAGERID() takes a pointer to a Pager struct as its argument. The
128 ** associated file-descriptor is returned. FILEHANDLEID() takes an sqlite3_file
129 ** struct as its argument.
131 #define PAGERID(p) (SQLITE_PTR_TO_INT(p->fd))
132 #define FILEHANDLEID(fd) (SQLITE_PTR_TO_INT(fd))
135 ** The Pager.eState variable stores the current 'state' of a pager. A
136 ** pager may be in any one of the seven states shown in the following
137 ** state diagram.
139 ** OPEN <------+------+
140 ** | | |
141 ** V | |
142 ** +---------> READER-------+ |
143 ** | | |
144 ** | V |
145 ** |<-------WRITER_LOCKED------> ERROR
146 ** | | ^
147 ** | V |
148 ** |<------WRITER_CACHEMOD-------->|
149 ** | | |
150 ** | V |
151 ** |<-------WRITER_DBMOD---------->|
152 ** | | |
153 ** | V |
154 ** +<------WRITER_FINISHED-------->+
157 ** List of state transitions and the C [function] that performs each:
159 ** OPEN -> READER [sqlite3PagerSharedLock]
160 ** READER -> OPEN [pager_unlock]
162 ** READER -> WRITER_LOCKED [sqlite3PagerBegin]
163 ** WRITER_LOCKED -> WRITER_CACHEMOD [pager_open_journal]
164 ** WRITER_CACHEMOD -> WRITER_DBMOD [syncJournal]
165 ** WRITER_DBMOD -> WRITER_FINISHED [sqlite3PagerCommitPhaseOne]
166 ** WRITER_*** -> READER [pager_end_transaction]
168 ** WRITER_*** -> ERROR [pager_error]
169 ** ERROR -> OPEN [pager_unlock]
172 ** OPEN:
174 ** The pager starts up in this state. Nothing is guaranteed in this
175 ** state - the file may or may not be locked and the database size is
176 ** unknown. The database may not be read or written.
178 ** * No read or write transaction is active.
179 ** * Any lock, or no lock at all, may be held on the database file.
180 ** * The dbSize, dbOrigSize and dbFileSize variables may not be trusted.
182 ** READER:
184 ** In this state all the requirements for reading the database in
185 ** rollback (non-WAL) mode are met. Unless the pager is (or recently
186 ** was) in exclusive-locking mode, a user-level read transaction is
187 ** open. The database size is known in this state.
189 ** A connection running with locking_mode=normal enters this state when
190 ** it opens a read-transaction on the database and returns to state
191 ** OPEN after the read-transaction is completed. However a connection
192 ** running in locking_mode=exclusive (including temp databases) remains in
193 ** this state even after the read-transaction is closed. The only way
194 ** a locking_mode=exclusive connection can transition from READER to OPEN
195 ** is via the ERROR state (see below).
197 ** * A read transaction may be active (but a write-transaction cannot).
198 ** * A SHARED or greater lock is held on the database file.
199 ** * The dbSize variable may be trusted (even if a user-level read
200 ** transaction is not active). The dbOrigSize and dbFileSize variables
201 ** may not be trusted at this point.
202 ** * If the database is a WAL database, then the WAL connection is open.
203 ** * Even if a read-transaction is not open, it is guaranteed that
204 ** there is no hot-journal in the file-system.
206 ** WRITER_LOCKED:
208 ** The pager moves to this state from READER when a write-transaction
209 ** is first opened on the database. In WRITER_LOCKED state, all locks
210 ** required to start a write-transaction are held, but no actual
211 ** modifications to the cache or database have taken place.
213 ** In rollback mode, a RESERVED or (if the transaction was opened with
214 ** BEGIN EXCLUSIVE) EXCLUSIVE lock is obtained on the database file when
215 ** moving to this state, but the journal file is not written to or opened
216 ** to in this state. If the transaction is committed or rolled back while
217 ** in WRITER_LOCKED state, all that is required is to unlock the database
218 ** file.
220 ** IN WAL mode, WalBeginWriteTransaction() is called to lock the log file.
221 ** If the connection is running with locking_mode=exclusive, an attempt
222 ** is made to obtain an EXCLUSIVE lock on the database file.
224 ** * A write transaction is active.
225 ** * If the connection is open in rollback-mode, a RESERVED or greater
226 ** lock is held on the database file.
227 ** * If the connection is open in WAL-mode, a WAL write transaction
228 ** is open (i.e. sqlite3WalBeginWriteTransaction() has been successfully
229 ** called).
230 ** * The dbSize, dbOrigSize and dbFileSize variables are all valid.
231 ** * The contents of the pager cache have not been modified.
232 ** * The journal file may or may not be open.
233 ** * Nothing (not even the first header) has been written to the journal.
235 ** WRITER_CACHEMOD:
237 ** A pager moves from WRITER_LOCKED state to this state when a page is
238 ** first modified by the upper layer. In rollback mode the journal file
239 ** is opened (if it is not already open) and a header written to the
240 ** start of it. The database file on disk has not been modified.
242 ** * A write transaction is active.
243 ** * A RESERVED or greater lock is held on the database file.
244 ** * The journal file is open and the first header has been written
245 ** to it, but the header has not been synced to disk.
246 ** * The contents of the page cache have been modified.
248 ** WRITER_DBMOD:
250 ** The pager transitions from WRITER_CACHEMOD into WRITER_DBMOD state
251 ** when it modifies the contents of the database file. WAL connections
252 ** never enter this state (since they do not modify the database file,
253 ** just the log file).
255 ** * A write transaction is active.
256 ** * An EXCLUSIVE or greater lock is held on the database file.
257 ** * The journal file is open and the first header has been written
258 ** and synced to disk.
259 ** * The contents of the page cache have been modified (and possibly
260 ** written to disk).
262 ** WRITER_FINISHED:
264 ** It is not possible for a WAL connection to enter this state.
266 ** A rollback-mode pager changes to WRITER_FINISHED state from WRITER_DBMOD
267 ** state after the entire transaction has been successfully written into the
268 ** database file. In this state the transaction may be committed simply
269 ** by finalizing the journal file. Once in WRITER_FINISHED state, it is
270 ** not possible to modify the database further. At this point, the upper
271 ** layer must either commit or rollback the transaction.
273 ** * A write transaction is active.
274 ** * An EXCLUSIVE or greater lock is held on the database file.
275 ** * All writing and syncing of journal and database data has finished.
276 ** If no error occurred, all that remains is to finalize the journal to
277 ** commit the transaction. If an error did occur, the caller will need
278 ** to rollback the transaction.
280 ** ERROR:
282 ** The ERROR state is entered when an IO or disk-full error (including
283 ** SQLITE_IOERR_NOMEM) occurs at a point in the code that makes it
284 ** difficult to be sure that the in-memory pager state (cache contents,
285 ** db size etc.) are consistent with the contents of the file-system.
287 ** Temporary pager files may enter the ERROR state, but in-memory pagers
288 ** cannot.
290 ** For example, if an IO error occurs while performing a rollback,
291 ** the contents of the page-cache may be left in an inconsistent state.
292 ** At this point it would be dangerous to change back to READER state
293 ** (as usually happens after a rollback). Any subsequent readers might
294 ** report database corruption (due to the inconsistent cache), and if
295 ** they upgrade to writers, they may inadvertently corrupt the database
296 ** file. To avoid this hazard, the pager switches into the ERROR state
297 ** instead of READER following such an error.
299 ** Once it has entered the ERROR state, any attempt to use the pager
300 ** to read or write data returns an error. Eventually, once all
301 ** outstanding transactions have been abandoned, the pager is able to
302 ** transition back to OPEN state, discarding the contents of the
303 ** page-cache and any other in-memory state at the same time. Everything
304 ** is reloaded from disk (and, if necessary, hot-journal rollback peformed)
305 ** when a read-transaction is next opened on the pager (transitioning
306 ** the pager into READER state). At that point the system has recovered
307 ** from the error.
309 ** Specifically, the pager jumps into the ERROR state if:
311 ** 1. An error occurs while attempting a rollback. This happens in
312 ** function sqlite3PagerRollback().
314 ** 2. An error occurs while attempting to finalize a journal file
315 ** following a commit in function sqlite3PagerCommitPhaseTwo().
317 ** 3. An error occurs while attempting to write to the journal or
318 ** database file in function pagerStress() in order to free up
319 ** memory.
321 ** In other cases, the error is returned to the b-tree layer. The b-tree
322 ** layer then attempts a rollback operation. If the error condition
323 ** persists, the pager enters the ERROR state via condition (1) above.
325 ** Condition (3) is necessary because it can be triggered by a read-only
326 ** statement executed within a transaction. In this case, if the error
327 ** code were simply returned to the user, the b-tree layer would not
328 ** automatically attempt a rollback, as it assumes that an error in a
329 ** read-only statement cannot leave the pager in an internally inconsistent
330 ** state.
332 ** * The Pager.errCode variable is set to something other than SQLITE_OK.
333 ** * There are one or more outstanding references to pages (after the
334 ** last reference is dropped the pager should move back to OPEN state).
335 ** * The pager is not an in-memory pager.
338 ** Notes:
340 ** * A pager is never in WRITER_DBMOD or WRITER_FINISHED state if the
341 ** connection is open in WAL mode. A WAL connection is always in one
342 ** of the first four states.
344 ** * Normally, a connection open in exclusive mode is never in PAGER_OPEN
345 ** state. There are two exceptions: immediately after exclusive-mode has
346 ** been turned on (and before any read or write transactions are
347 ** executed), and when the pager is leaving the "error state".
349 ** * See also: assert_pager_state().
351 #define PAGER_OPEN 0
352 #define PAGER_READER 1
353 #define PAGER_WRITER_LOCKED 2
354 #define PAGER_WRITER_CACHEMOD 3
355 #define PAGER_WRITER_DBMOD 4
356 #define PAGER_WRITER_FINISHED 5
357 #define PAGER_ERROR 6
360 ** The Pager.eLock variable is almost always set to one of the
361 ** following locking-states, according to the lock currently held on
362 ** the database file: NO_LOCK, SHARED_LOCK, RESERVED_LOCK or EXCLUSIVE_LOCK.
363 ** This variable is kept up to date as locks are taken and released by
364 ** the pagerLockDb() and pagerUnlockDb() wrappers.
366 ** If the VFS xLock() or xUnlock() returns an error other than SQLITE_BUSY
367 ** (i.e. one of the SQLITE_IOERR subtypes), it is not clear whether or not
368 ** the operation was successful. In these circumstances pagerLockDb() and
369 ** pagerUnlockDb() take a conservative approach - eLock is always updated
370 ** when unlocking the file, and only updated when locking the file if the
371 ** VFS call is successful. This way, the Pager.eLock variable may be set
372 ** to a less exclusive (lower) value than the lock that is actually held
373 ** at the system level, but it is never set to a more exclusive value.
375 ** This is usually safe. If an xUnlock fails or appears to fail, there may
376 ** be a few redundant xLock() calls or a lock may be held for longer than
377 ** required, but nothing really goes wrong.
379 ** The exception is when the database file is unlocked as the pager moves
380 ** from ERROR to OPEN state. At this point there may be a hot-journal file
381 ** in the file-system that needs to be rolled back (as part of an OPEN->SHARED
382 ** transition, by the same pager or any other). If the call to xUnlock()
383 ** fails at this point and the pager is left holding an EXCLUSIVE lock, this
384 ** can confuse the call to xCheckReservedLock() call made later as part
385 ** of hot-journal detection.
387 ** xCheckReservedLock() is defined as returning true "if there is a RESERVED
388 ** lock held by this process or any others". So xCheckReservedLock may
389 ** return true because the caller itself is holding an EXCLUSIVE lock (but
390 ** doesn't know it because of a previous error in xUnlock). If this happens
391 ** a hot-journal may be mistaken for a journal being created by an active
392 ** transaction in another process, causing SQLite to read from the database
393 ** without rolling it back.
395 ** To work around this, if a call to xUnlock() fails when unlocking the
396 ** database in the ERROR state, Pager.eLock is set to UNKNOWN_LOCK. It
397 ** is only changed back to a real locking state after a successful call
398 ** to xLock(EXCLUSIVE). Also, the code to do the OPEN->SHARED state transition
399 ** omits the check for a hot-journal if Pager.eLock is set to UNKNOWN_LOCK
400 ** lock. Instead, it assumes a hot-journal exists and obtains an EXCLUSIVE
401 ** lock on the database file before attempting to roll it back. See function
402 ** PagerSharedLock() for more detail.
404 ** Pager.eLock may only be set to UNKNOWN_LOCK when the pager is in
405 ** PAGER_OPEN state.
407 #define UNKNOWN_LOCK (EXCLUSIVE_LOCK+1)
410 ** A macro used for invoking the codec if there is one
412 /* BEGIN SQLCIPHER */
413 #ifdef SQLITE_HAS_CODEC
414 # define CODEC1(P,D,N,X,E) \
415 if( P->xCodec && P->xCodec(P->pCodec,D,N,X)==0 ){ E; }
416 # define CODEC2(P,D,N,X,E,O) \
417 if( P->xCodec==0 ){ O=(char*)D; }else \
418 if( (O=(char*)(P->xCodec(P->pCodec,D,N,X)))==0 ){ E; }
419 #else
420 # define CODEC1(P,D,N,X,E) /* NO-OP */
421 # define CODEC2(P,D,N,X,E,O) O=(char*)D
422 #endif
423 /* END SQLCIPHER */
426 ** The maximum allowed sector size. 64KiB. If the xSectorsize() method
427 ** returns a value larger than this, then MAX_SECTOR_SIZE is used instead.
428 ** This could conceivably cause corruption following a power failure on
429 ** such a system. This is currently an undocumented limit.
431 #define MAX_SECTOR_SIZE 0x10000
435 ** An instance of the following structure is allocated for each active
436 ** savepoint and statement transaction in the system. All such structures
437 ** are stored in the Pager.aSavepoint[] array, which is allocated and
438 ** resized using sqlite3Realloc().
440 ** When a savepoint is created, the PagerSavepoint.iHdrOffset field is
441 ** set to 0. If a journal-header is written into the main journal while
442 ** the savepoint is active, then iHdrOffset is set to the byte offset
443 ** immediately following the last journal record written into the main
444 ** journal before the journal-header. This is required during savepoint
445 ** rollback (see pagerPlaybackSavepoint()).
447 typedef struct PagerSavepoint PagerSavepoint;
448 struct PagerSavepoint {
449 i64 iOffset; /* Starting offset in main journal */
450 i64 iHdrOffset; /* See above */
451 Bitvec *pInSavepoint; /* Set of pages in this savepoint */
452 Pgno nOrig; /* Original number of pages in file */
453 Pgno iSubRec; /* Index of first record in sub-journal */
454 int bTruncateOnRelease; /* If stmt journal may be truncated on RELEASE */
455 #ifndef SQLITE_OMIT_WAL
456 u32 aWalData[WAL_SAVEPOINT_NDATA]; /* WAL savepoint context */
457 #endif
461 ** Bits of the Pager.doNotSpill flag. See further description below.
463 #define SPILLFLAG_OFF 0x01 /* Never spill cache. Set via pragma */
464 #define SPILLFLAG_ROLLBACK 0x02 /* Current rolling back, so do not spill */
465 #define SPILLFLAG_NOSYNC 0x04 /* Spill is ok, but do not sync */
468 ** An open page cache is an instance of struct Pager. A description of
469 ** some of the more important member variables follows:
471 ** eState
473 ** The current 'state' of the pager object. See the comment and state
474 ** diagram above for a description of the pager state.
476 ** eLock
478 ** For a real on-disk database, the current lock held on the database file -
479 ** NO_LOCK, SHARED_LOCK, RESERVED_LOCK or EXCLUSIVE_LOCK.
481 ** For a temporary or in-memory database (neither of which require any
482 ** locks), this variable is always set to EXCLUSIVE_LOCK. Since such
483 ** databases always have Pager.exclusiveMode==1, this tricks the pager
484 ** logic into thinking that it already has all the locks it will ever
485 ** need (and no reason to release them).
487 ** In some (obscure) circumstances, this variable may also be set to
488 ** UNKNOWN_LOCK. See the comment above the #define of UNKNOWN_LOCK for
489 ** details.
491 ** changeCountDone
493 ** This boolean variable is used to make sure that the change-counter
494 ** (the 4-byte header field at byte offset 24 of the database file) is
495 ** not updated more often than necessary.
497 ** It is set to true when the change-counter field is updated, which
498 ** can only happen if an exclusive lock is held on the database file.
499 ** It is cleared (set to false) whenever an exclusive lock is
500 ** relinquished on the database file. Each time a transaction is committed,
501 ** The changeCountDone flag is inspected. If it is true, the work of
502 ** updating the change-counter is omitted for the current transaction.
504 ** This mechanism means that when running in exclusive mode, a connection
505 ** need only update the change-counter once, for the first transaction
506 ** committed.
508 ** setSuper
510 ** When PagerCommitPhaseOne() is called to commit a transaction, it may
511 ** (or may not) specify a super-journal name to be written into the
512 ** journal file before it is synced to disk.
514 ** Whether or not a journal file contains a super-journal pointer affects
515 ** the way in which the journal file is finalized after the transaction is
516 ** committed or rolled back when running in "journal_mode=PERSIST" mode.
517 ** If a journal file does not contain a super-journal pointer, it is
518 ** finalized by overwriting the first journal header with zeroes. If
519 ** it does contain a super-journal pointer the journal file is finalized
520 ** by truncating it to zero bytes, just as if the connection were
521 ** running in "journal_mode=truncate" mode.
523 ** Journal files that contain super-journal pointers cannot be finalized
524 ** simply by overwriting the first journal-header with zeroes, as the
525 ** super-journal pointer could interfere with hot-journal rollback of any
526 ** subsequently interrupted transaction that reuses the journal file.
528 ** The flag is cleared as soon as the journal file is finalized (either
529 ** by PagerCommitPhaseTwo or PagerRollback). If an IO error prevents the
530 ** journal file from being successfully finalized, the setSuper flag
531 ** is cleared anyway (and the pager will move to ERROR state).
533 ** doNotSpill
535 ** This variables control the behavior of cache-spills (calls made by
536 ** the pcache module to the pagerStress() routine to write cached data
537 ** to the file-system in order to free up memory).
539 ** When bits SPILLFLAG_OFF or SPILLFLAG_ROLLBACK of doNotSpill are set,
540 ** writing to the database from pagerStress() is disabled altogether.
541 ** The SPILLFLAG_ROLLBACK case is done in a very obscure case that
542 ** comes up during savepoint rollback that requires the pcache module
543 ** to allocate a new page to prevent the journal file from being written
544 ** while it is being traversed by code in pager_playback(). The SPILLFLAG_OFF
545 ** case is a user preference.
547 ** If the SPILLFLAG_NOSYNC bit is set, writing to the database from
548 ** pagerStress() is permitted, but syncing the journal file is not.
549 ** This flag is set by sqlite3PagerWrite() when the file-system sector-size
550 ** is larger than the database page-size in order to prevent a journal sync
551 ** from happening in between the journalling of two pages on the same sector.
553 ** subjInMemory
555 ** This is a boolean variable. If true, then any required sub-journal
556 ** is opened as an in-memory journal file. If false, then in-memory
557 ** sub-journals are only used for in-memory pager files.
559 ** This variable is updated by the upper layer each time a new
560 ** write-transaction is opened.
562 ** dbSize, dbOrigSize, dbFileSize
564 ** Variable dbSize is set to the number of pages in the database file.
565 ** It is valid in PAGER_READER and higher states (all states except for
566 ** OPEN and ERROR).
568 ** dbSize is set based on the size of the database file, which may be
569 ** larger than the size of the database (the value stored at offset
570 ** 28 of the database header by the btree). If the size of the file
571 ** is not an integer multiple of the page-size, the value stored in
572 ** dbSize is rounded down (i.e. a 5KB file with 2K page-size has dbSize==2).
573 ** Except, any file that is greater than 0 bytes in size is considered
574 ** to have at least one page. (i.e. a 1KB file with 2K page-size leads
575 ** to dbSize==1).
577 ** During a write-transaction, if pages with page-numbers greater than
578 ** dbSize are modified in the cache, dbSize is updated accordingly.
579 ** Similarly, if the database is truncated using PagerTruncateImage(),
580 ** dbSize is updated.
582 ** Variables dbOrigSize and dbFileSize are valid in states
583 ** PAGER_WRITER_LOCKED and higher. dbOrigSize is a copy of the dbSize
584 ** variable at the start of the transaction. It is used during rollback,
585 ** and to determine whether or not pages need to be journalled before
586 ** being modified.
588 ** Throughout a write-transaction, dbFileSize contains the size of
589 ** the file on disk in pages. It is set to a copy of dbSize when the
590 ** write-transaction is first opened, and updated when VFS calls are made
591 ** to write or truncate the database file on disk.
593 ** The only reason the dbFileSize variable is required is to suppress
594 ** unnecessary calls to xTruncate() after committing a transaction. If,
595 ** when a transaction is committed, the dbFileSize variable indicates
596 ** that the database file is larger than the database image (Pager.dbSize),
597 ** pager_truncate() is called. The pager_truncate() call uses xFilesize()
598 ** to measure the database file on disk, and then truncates it if required.
599 ** dbFileSize is not used when rolling back a transaction. In this case
600 ** pager_truncate() is called unconditionally (which means there may be
601 ** a call to xFilesize() that is not strictly required). In either case,
602 ** pager_truncate() may cause the file to become smaller or larger.
604 ** dbHintSize
606 ** The dbHintSize variable is used to limit the number of calls made to
607 ** the VFS xFileControl(FCNTL_SIZE_HINT) method.
609 ** dbHintSize is set to a copy of the dbSize variable when a
610 ** write-transaction is opened (at the same time as dbFileSize and
611 ** dbOrigSize). If the xFileControl(FCNTL_SIZE_HINT) method is called,
612 ** dbHintSize is increased to the number of pages that correspond to the
613 ** size-hint passed to the method call. See pager_write_pagelist() for
614 ** details.
616 ** errCode
618 ** The Pager.errCode variable is only ever used in PAGER_ERROR state. It
619 ** is set to zero in all other states. In PAGER_ERROR state, Pager.errCode
620 ** is always set to SQLITE_FULL, SQLITE_IOERR or one of the SQLITE_IOERR_XXX
621 ** sub-codes.
623 ** syncFlags, walSyncFlags
625 ** syncFlags is either SQLITE_SYNC_NORMAL (0x02) or SQLITE_SYNC_FULL (0x03).
626 ** syncFlags is used for rollback mode. walSyncFlags is used for WAL mode
627 ** and contains the flags used to sync the checkpoint operations in the
628 ** lower two bits, and sync flags used for transaction commits in the WAL
629 ** file in bits 0x04 and 0x08. In other words, to get the correct sync flags
630 ** for checkpoint operations, use (walSyncFlags&0x03) and to get the correct
631 ** sync flags for transaction commit, use ((walSyncFlags>>2)&0x03). Note
632 ** that with synchronous=NORMAL in WAL mode, transaction commit is not synced
633 ** meaning that the 0x04 and 0x08 bits are both zero.
635 struct Pager {
636 sqlite3_vfs *pVfs; /* OS functions to use for IO */
637 u8 exclusiveMode; /* Boolean. True if locking_mode==EXCLUSIVE */
638 u8 journalMode; /* One of the PAGER_JOURNALMODE_* values */
639 u8 useJournal; /* Use a rollback journal on this file */
640 u8 noSync; /* Do not sync the journal if true */
641 u8 fullSync; /* Do extra syncs of the journal for robustness */
642 u8 extraSync; /* sync directory after journal delete */
643 u8 syncFlags; /* SYNC_NORMAL or SYNC_FULL otherwise */
644 u8 walSyncFlags; /* See description above */
645 u8 tempFile; /* zFilename is a temporary or immutable file */
646 u8 noLock; /* Do not lock (except in WAL mode) */
647 u8 readOnly; /* True for a read-only database */
648 u8 memDb; /* True to inhibit all file I/O */
649 u8 memVfs; /* VFS-implemented memory database */
651 /**************************************************************************
652 ** The following block contains those class members that change during
653 ** routine operation. Class members not in this block are either fixed
654 ** when the pager is first created or else only change when there is a
655 ** significant mode change (such as changing the page_size, locking_mode,
656 ** or the journal_mode). From another view, these class members describe
657 ** the "state" of the pager, while other class members describe the
658 ** "configuration" of the pager.
660 u8 eState; /* Pager state (OPEN, READER, WRITER_LOCKED..) */
661 u8 eLock; /* Current lock held on database file */
662 u8 changeCountDone; /* Set after incrementing the change-counter */
663 u8 setSuper; /* Super-jrnl name is written into jrnl */
664 u8 doNotSpill; /* Do not spill the cache when non-zero */
665 u8 subjInMemory; /* True to use in-memory sub-journals */
666 u8 bUseFetch; /* True to use xFetch() */
667 u8 hasHeldSharedLock; /* True if a shared lock has ever been held */
668 Pgno dbSize; /* Number of pages in the database */
669 Pgno dbOrigSize; /* dbSize before the current transaction */
670 Pgno dbFileSize; /* Number of pages in the database file */
671 Pgno dbHintSize; /* Value passed to FCNTL_SIZE_HINT call */
672 int errCode; /* One of several kinds of errors */
673 int nRec; /* Pages journalled since last j-header written */
674 u32 cksumInit; /* Quasi-random value added to every checksum */
675 u32 nSubRec; /* Number of records written to sub-journal */
676 Bitvec *pInJournal; /* One bit for each page in the database file */
677 sqlite3_file *fd; /* File descriptor for database */
678 sqlite3_file *jfd; /* File descriptor for main journal */
679 sqlite3_file *sjfd; /* File descriptor for sub-journal */
680 i64 journalOff; /* Current write offset in the journal file */
681 i64 journalHdr; /* Byte offset to previous journal header */
682 sqlite3_backup *pBackup; /* Pointer to list of ongoing backup processes */
683 PagerSavepoint *aSavepoint; /* Array of active savepoints */
684 int nSavepoint; /* Number of elements in aSavepoint[] */
685 u32 iDataVersion; /* Changes whenever database content changes */
686 char dbFileVers[16]; /* Changes whenever database file changes */
688 int nMmapOut; /* Number of mmap pages currently outstanding */
689 sqlite3_int64 szMmap; /* Desired maximum mmap size */
690 PgHdr *pMmapFreelist; /* List of free mmap page headers (pDirty) */
692 ** End of the routinely-changing class members
693 ***************************************************************************/
695 u16 nExtra; /* Add this many bytes to each in-memory page */
696 i16 nReserve; /* Number of unused bytes at end of each page */
697 u32 vfsFlags; /* Flags for sqlite3_vfs.xOpen() */
698 u32 sectorSize; /* Assumed sector size during rollback */
699 Pgno mxPgno; /* Maximum allowed size of the database */
700 Pgno lckPgno; /* Page number for the locking page */
701 i64 pageSize; /* Number of bytes in a page */
702 i64 journalSizeLimit; /* Size limit for persistent journal files */
703 char *zFilename; /* Name of the database file */
704 char *zJournal; /* Name of the journal file */
705 int (*xBusyHandler)(void*); /* Function to call when busy */
706 void *pBusyHandlerArg; /* Context argument for xBusyHandler */
707 int aStat[4]; /* Total cache hits, misses, writes, spills */
708 #ifdef SQLITE_TEST
709 int nRead; /* Database pages read */
710 #endif
711 void (*xReiniter)(DbPage*); /* Call this routine when reloading pages */
712 int (*xGet)(Pager*,Pgno,DbPage**,int); /* Routine to fetch a patch */
713 /* BEGIN SQLCIPHER */
714 #ifdef SQLITE_HAS_CODEC
715 void *(*xCodec)(void*,void*,Pgno,int); /* Routine for en/decoding data */
716 void (*xCodecSizeChng)(void*,int,int); /* Notify of page size changes */
717 void (*xCodecFree)(void*); /* Destructor for the codec */
718 void *pCodec; /* First argument to xCodec... methods */
719 #endif
720 /* END SQLCIPHER */
721 char *pTmpSpace; /* Pager.pageSize bytes of space for tmp use */
722 PCache *pPCache; /* Pointer to page cache object */
723 #ifndef SQLITE_OMIT_WAL
724 Wal *pWal; /* Write-ahead log used by "journal_mode=wal" */
725 char *zWal; /* File name for write-ahead log */
726 #endif
730 ** Indexes for use with Pager.aStat[]. The Pager.aStat[] array contains
731 ** the values accessed by passing SQLITE_DBSTATUS_CACHE_HIT, CACHE_MISS
732 ** or CACHE_WRITE to sqlite3_db_status().
734 #define PAGER_STAT_HIT 0
735 #define PAGER_STAT_MISS 1
736 #define PAGER_STAT_WRITE 2
737 #define PAGER_STAT_SPILL 3
740 ** The following global variables hold counters used for
741 ** testing purposes only. These variables do not exist in
742 ** a non-testing build. These variables are not thread-safe.
744 #ifdef SQLITE_TEST
745 int sqlite3_pager_readdb_count = 0; /* Number of full pages read from DB */
746 int sqlite3_pager_writedb_count = 0; /* Number of full pages written to DB */
747 int sqlite3_pager_writej_count = 0; /* Number of pages written to journal */
748 # define PAGER_INCR(v) v++
749 #else
750 # define PAGER_INCR(v)
751 #endif
756 ** Journal files begin with the following magic string. The data
757 ** was obtained from /dev/random. It is used only as a sanity check.
759 ** Since version 2.8.0, the journal format contains additional sanity
760 ** checking information. If the power fails while the journal is being
761 ** written, semi-random garbage data might appear in the journal
762 ** file after power is restored. If an attempt is then made
763 ** to roll the journal back, the database could be corrupted. The additional
764 ** sanity checking data is an attempt to discover the garbage in the
765 ** journal and ignore it.
767 ** The sanity checking information for the new journal format consists
768 ** of a 32-bit checksum on each page of data. The checksum covers both
769 ** the page number and the pPager->pageSize bytes of data for the page.
770 ** This cksum is initialized to a 32-bit random value that appears in the
771 ** journal file right after the header. The random initializer is important,
772 ** because garbage data that appears at the end of a journal is likely
773 ** data that was once in other files that have now been deleted. If the
774 ** garbage data came from an obsolete journal file, the checksums might
775 ** be correct. But by initializing the checksum to random value which
776 ** is different for every journal, we minimize that risk.
778 static const unsigned char aJournalMagic[] = {
779 0xd9, 0xd5, 0x05, 0xf9, 0x20, 0xa1, 0x63, 0xd7,
783 ** The size of the of each page record in the journal is given by
784 ** the following macro.
786 #define JOURNAL_PG_SZ(pPager) ((pPager->pageSize) + 8)
789 ** The journal header size for this pager. This is usually the same
790 ** size as a single disk sector. See also setSectorSize().
792 #define JOURNAL_HDR_SZ(pPager) (pPager->sectorSize)
795 ** The macro MEMDB is true if we are dealing with an in-memory database.
796 ** We do this as a macro so that if the SQLITE_OMIT_MEMORYDB macro is set,
797 ** the value of MEMDB will be a constant and the compiler will optimize
798 ** out code that would never execute.
800 #ifdef SQLITE_OMIT_MEMORYDB
801 # define MEMDB 0
802 #else
803 # define MEMDB pPager->memDb
804 #endif
807 ** The macro USEFETCH is true if we are allowed to use the xFetch and xUnfetch
808 ** interfaces to access the database using memory-mapped I/O.
810 #if SQLITE_MAX_MMAP_SIZE>0
811 # define USEFETCH(x) ((x)->bUseFetch)
812 #else
813 # define USEFETCH(x) 0
814 #endif
817 ** The argument to this macro is a file descriptor (type sqlite3_file*).
818 ** Return 0 if it is not open, or non-zero (but not 1) if it is.
820 ** This is so that expressions can be written as:
822 ** if( isOpen(pPager->jfd) ){ ...
824 ** instead of
826 ** if( pPager->jfd->pMethods ){ ...
828 #define isOpen(pFd) ((pFd)->pMethods!=0)
830 #ifdef SQLITE_DIRECT_OVERFLOW_READ
832 ** Return true if page pgno can be read directly from the database file
833 ** by the b-tree layer. This is the case if:
835 ** * the database file is open,
836 ** * there are no dirty pages in the cache, and
837 ** * the desired page is not currently in the wal file.
839 int sqlite3PagerDirectReadOk(Pager *pPager, Pgno pgno){
840 if( pPager->fd->pMethods==0 ) return 0;
841 if( sqlite3PCacheIsDirty(pPager->pPCache) ) return 0;
842 /* BEGIN SQLCIPHER */
843 #ifdef SQLITE_HAS_CODEC
844 if( pPager->xCodec!=0 ) return 0;
845 #endif
846 /* END SQLCIPHER */
847 #ifndef SQLITE_OMIT_WAL
848 if( pPager->pWal ){
849 u32 iRead = 0;
850 int rc;
851 rc = sqlite3WalFindFrame(pPager->pWal, pgno, &iRead);
852 return (rc==SQLITE_OK && iRead==0);
854 #endif
855 return 1;
857 #endif
859 #ifndef SQLITE_OMIT_WAL
860 # define pagerUseWal(x) ((x)->pWal!=0)
861 #else
862 # define pagerUseWal(x) 0
863 # define pagerRollbackWal(x) 0
864 # define pagerWalFrames(v,w,x,y) 0
865 # define pagerOpenWalIfPresent(z) SQLITE_OK
866 # define pagerBeginReadTransaction(z) SQLITE_OK
867 #endif
869 #ifndef NDEBUG
871 ** Usage:
873 ** assert( assert_pager_state(pPager) );
875 ** This function runs many asserts to try to find inconsistencies in
876 ** the internal state of the Pager object.
878 static int assert_pager_state(Pager *p){
879 Pager *pPager = p;
881 /* State must be valid. */
882 assert( p->eState==PAGER_OPEN
883 || p->eState==PAGER_READER
884 || p->eState==PAGER_WRITER_LOCKED
885 || p->eState==PAGER_WRITER_CACHEMOD
886 || p->eState==PAGER_WRITER_DBMOD
887 || p->eState==PAGER_WRITER_FINISHED
888 || p->eState==PAGER_ERROR
891 /* Regardless of the current state, a temp-file connection always behaves
892 ** as if it has an exclusive lock on the database file. It never updates
893 ** the change-counter field, so the changeCountDone flag is always set.
895 assert( p->tempFile==0 || p->eLock==EXCLUSIVE_LOCK );
896 assert( p->tempFile==0 || pPager->changeCountDone );
898 /* If the useJournal flag is clear, the journal-mode must be "OFF".
899 ** And if the journal-mode is "OFF", the journal file must not be open.
901 assert( p->journalMode==PAGER_JOURNALMODE_OFF || p->useJournal );
902 assert( p->journalMode!=PAGER_JOURNALMODE_OFF || !isOpen(p->jfd) );
904 /* Check that MEMDB implies noSync. And an in-memory journal. Since
905 ** this means an in-memory pager performs no IO at all, it cannot encounter
906 ** either SQLITE_IOERR or SQLITE_FULL during rollback or while finalizing
907 ** a journal file. (although the in-memory journal implementation may
908 ** return SQLITE_IOERR_NOMEM while the journal file is being written). It
909 ** is therefore not possible for an in-memory pager to enter the ERROR
910 ** state.
912 if( MEMDB ){
913 assert( !isOpen(p->fd) );
914 assert( p->noSync );
915 assert( p->journalMode==PAGER_JOURNALMODE_OFF
916 || p->journalMode==PAGER_JOURNALMODE_MEMORY
918 assert( p->eState!=PAGER_ERROR && p->eState!=PAGER_OPEN );
919 assert( pagerUseWal(p)==0 );
922 /* If changeCountDone is set, a RESERVED lock or greater must be held
923 ** on the file.
925 assert( pPager->changeCountDone==0 || pPager->eLock>=RESERVED_LOCK );
926 assert( p->eLock!=PENDING_LOCK );
928 switch( p->eState ){
929 case PAGER_OPEN:
930 assert( !MEMDB );
931 assert( pPager->errCode==SQLITE_OK );
932 assert( sqlite3PcacheRefCount(pPager->pPCache)==0 || pPager->tempFile );
933 break;
935 case PAGER_READER:
936 assert( pPager->errCode==SQLITE_OK );
937 assert( p->eLock!=UNKNOWN_LOCK );
938 assert( p->eLock>=SHARED_LOCK );
939 break;
941 case PAGER_WRITER_LOCKED:
942 assert( p->eLock!=UNKNOWN_LOCK );
943 assert( pPager->errCode==SQLITE_OK );
944 if( !pagerUseWal(pPager) ){
945 assert( p->eLock>=RESERVED_LOCK );
947 assert( pPager->dbSize==pPager->dbOrigSize );
948 assert( pPager->dbOrigSize==pPager->dbFileSize );
949 assert( pPager->dbOrigSize==pPager->dbHintSize );
950 assert( pPager->setSuper==0 );
951 break;
953 case PAGER_WRITER_CACHEMOD:
954 assert( p->eLock!=UNKNOWN_LOCK );
955 assert( pPager->errCode==SQLITE_OK );
956 if( !pagerUseWal(pPager) ){
957 /* It is possible that if journal_mode=wal here that neither the
958 ** journal file nor the WAL file are open. This happens during
959 ** a rollback transaction that switches from journal_mode=off
960 ** to journal_mode=wal.
962 assert( p->eLock>=RESERVED_LOCK );
963 assert( isOpen(p->jfd)
964 || p->journalMode==PAGER_JOURNALMODE_OFF
965 || p->journalMode==PAGER_JOURNALMODE_WAL
968 assert( pPager->dbOrigSize==pPager->dbFileSize );
969 assert( pPager->dbOrigSize==pPager->dbHintSize );
970 break;
972 case PAGER_WRITER_DBMOD:
973 assert( p->eLock==EXCLUSIVE_LOCK );
974 assert( pPager->errCode==SQLITE_OK );
975 assert( !pagerUseWal(pPager) );
976 assert( p->eLock>=EXCLUSIVE_LOCK );
977 assert( isOpen(p->jfd)
978 || p->journalMode==PAGER_JOURNALMODE_OFF
979 || p->journalMode==PAGER_JOURNALMODE_WAL
980 || (sqlite3OsDeviceCharacteristics(p->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
982 assert( pPager->dbOrigSize<=pPager->dbHintSize );
983 break;
985 case PAGER_WRITER_FINISHED:
986 assert( p->eLock==EXCLUSIVE_LOCK );
987 assert( pPager->errCode==SQLITE_OK );
988 assert( !pagerUseWal(pPager) );
989 assert( isOpen(p->jfd)
990 || p->journalMode==PAGER_JOURNALMODE_OFF
991 || p->journalMode==PAGER_JOURNALMODE_WAL
992 || (sqlite3OsDeviceCharacteristics(p->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
994 break;
996 case PAGER_ERROR:
997 /* There must be at least one outstanding reference to the pager if
998 ** in ERROR state. Otherwise the pager should have already dropped
999 ** back to OPEN state.
1001 assert( pPager->errCode!=SQLITE_OK );
1002 assert( sqlite3PcacheRefCount(pPager->pPCache)>0 || pPager->tempFile );
1003 break;
1006 return 1;
1008 #endif /* ifndef NDEBUG */
1010 #ifdef SQLITE_DEBUG
1012 ** Return a pointer to a human readable string in a static buffer
1013 ** containing the state of the Pager object passed as an argument. This
1014 ** is intended to be used within debuggers. For example, as an alternative
1015 ** to "print *pPager" in gdb:
1017 ** (gdb) printf "%s", print_pager_state(pPager)
1019 ** This routine has external linkage in order to suppress compiler warnings
1020 ** about an unused function. It is enclosed within SQLITE_DEBUG and so does
1021 ** not appear in normal builds.
1023 char *print_pager_state(Pager *p){
1024 static char zRet[1024];
1026 sqlite3_snprintf(1024, zRet,
1027 "Filename: %s\n"
1028 "State: %s errCode=%d\n"
1029 "Lock: %s\n"
1030 "Locking mode: locking_mode=%s\n"
1031 "Journal mode: journal_mode=%s\n"
1032 "Backing store: tempFile=%d memDb=%d useJournal=%d\n"
1033 "Journal: journalOff=%lld journalHdr=%lld\n"
1034 "Size: dbsize=%d dbOrigSize=%d dbFileSize=%d\n"
1035 , p->zFilename
1036 , p->eState==PAGER_OPEN ? "OPEN" :
1037 p->eState==PAGER_READER ? "READER" :
1038 p->eState==PAGER_WRITER_LOCKED ? "WRITER_LOCKED" :
1039 p->eState==PAGER_WRITER_CACHEMOD ? "WRITER_CACHEMOD" :
1040 p->eState==PAGER_WRITER_DBMOD ? "WRITER_DBMOD" :
1041 p->eState==PAGER_WRITER_FINISHED ? "WRITER_FINISHED" :
1042 p->eState==PAGER_ERROR ? "ERROR" : "?error?"
1043 , (int)p->errCode
1044 , p->eLock==NO_LOCK ? "NO_LOCK" :
1045 p->eLock==RESERVED_LOCK ? "RESERVED" :
1046 p->eLock==EXCLUSIVE_LOCK ? "EXCLUSIVE" :
1047 p->eLock==SHARED_LOCK ? "SHARED" :
1048 p->eLock==UNKNOWN_LOCK ? "UNKNOWN" : "?error?"
1049 , p->exclusiveMode ? "exclusive" : "normal"
1050 , p->journalMode==PAGER_JOURNALMODE_MEMORY ? "memory" :
1051 p->journalMode==PAGER_JOURNALMODE_OFF ? "off" :
1052 p->journalMode==PAGER_JOURNALMODE_DELETE ? "delete" :
1053 p->journalMode==PAGER_JOURNALMODE_PERSIST ? "persist" :
1054 p->journalMode==PAGER_JOURNALMODE_TRUNCATE ? "truncate" :
1055 p->journalMode==PAGER_JOURNALMODE_WAL ? "wal" : "?error?"
1056 , (int)p->tempFile, (int)p->memDb, (int)p->useJournal
1057 , p->journalOff, p->journalHdr
1058 , (int)p->dbSize, (int)p->dbOrigSize, (int)p->dbFileSize
1061 return zRet;
1063 #endif
1065 /* Forward references to the various page getters */
1066 static int getPageNormal(Pager*,Pgno,DbPage**,int);
1067 static int getPageError(Pager*,Pgno,DbPage**,int);
1068 #if SQLITE_MAX_MMAP_SIZE>0
1069 static int getPageMMap(Pager*,Pgno,DbPage**,int);
1070 #endif
1073 ** Set the Pager.xGet method for the appropriate routine used to fetch
1074 ** content from the pager.
1076 static void setGetterMethod(Pager *pPager){
1077 if( pPager->errCode ){
1078 pPager->xGet = getPageError;
1079 #if SQLITE_MAX_MMAP_SIZE>0
1080 }else if( USEFETCH(pPager)
1081 /* BEGIN SQLCIPHER */
1082 #ifdef SQLITE_HAS_CODEC
1083 && pPager->xCodec==0
1084 #endif
1085 /* END SQLCIPHER */
1087 pPager->xGet = getPageMMap;
1088 #endif /* SQLITE_MAX_MMAP_SIZE>0 */
1089 }else{
1090 pPager->xGet = getPageNormal;
1095 ** Return true if it is necessary to write page *pPg into the sub-journal.
1096 ** A page needs to be written into the sub-journal if there exists one
1097 ** or more open savepoints for which:
1099 ** * The page-number is less than or equal to PagerSavepoint.nOrig, and
1100 ** * The bit corresponding to the page-number is not set in
1101 ** PagerSavepoint.pInSavepoint.
1103 static int subjRequiresPage(PgHdr *pPg){
1104 Pager *pPager = pPg->pPager;
1105 PagerSavepoint *p;
1106 Pgno pgno = pPg->pgno;
1107 int i;
1108 for(i=0; i<pPager->nSavepoint; i++){
1109 p = &pPager->aSavepoint[i];
1110 if( p->nOrig>=pgno && 0==sqlite3BitvecTestNotNull(p->pInSavepoint, pgno) ){
1111 for(i=i+1; i<pPager->nSavepoint; i++){
1112 pPager->aSavepoint[i].bTruncateOnRelease = 0;
1114 return 1;
1117 return 0;
1120 #ifdef SQLITE_DEBUG
1122 ** Return true if the page is already in the journal file.
1124 static int pageInJournal(Pager *pPager, PgHdr *pPg){
1125 return sqlite3BitvecTest(pPager->pInJournal, pPg->pgno);
1127 #endif
1130 ** Read a 32-bit integer from the given file descriptor. Store the integer
1131 ** that is read in *pRes. Return SQLITE_OK if everything worked, or an
1132 ** error code is something goes wrong.
1134 ** All values are stored on disk as big-endian.
1136 static int read32bits(sqlite3_file *fd, i64 offset, u32 *pRes){
1137 unsigned char ac[4];
1138 int rc = sqlite3OsRead(fd, ac, sizeof(ac), offset);
1139 if( rc==SQLITE_OK ){
1140 *pRes = sqlite3Get4byte(ac);
1142 return rc;
1146 ** Write a 32-bit integer into a string buffer in big-endian byte order.
1148 #define put32bits(A,B) sqlite3Put4byte((u8*)A,B)
1152 ** Write a 32-bit integer into the given file descriptor. Return SQLITE_OK
1153 ** on success or an error code is something goes wrong.
1155 static int write32bits(sqlite3_file *fd, i64 offset, u32 val){
1156 char ac[4];
1157 put32bits(ac, val);
1158 return sqlite3OsWrite(fd, ac, 4, offset);
1162 ** Unlock the database file to level eLock, which must be either NO_LOCK
1163 ** or SHARED_LOCK. Regardless of whether or not the call to xUnlock()
1164 ** succeeds, set the Pager.eLock variable to match the (attempted) new lock.
1166 ** Except, if Pager.eLock is set to UNKNOWN_LOCK when this function is
1167 ** called, do not modify it. See the comment above the #define of
1168 ** UNKNOWN_LOCK for an explanation of this.
1170 static int pagerUnlockDb(Pager *pPager, int eLock){
1171 int rc = SQLITE_OK;
1173 assert( !pPager->exclusiveMode || pPager->eLock==eLock );
1174 assert( eLock==NO_LOCK || eLock==SHARED_LOCK );
1175 assert( eLock!=NO_LOCK || pagerUseWal(pPager)==0 );
1176 if( isOpen(pPager->fd) ){
1177 assert( pPager->eLock>=eLock );
1178 rc = pPager->noLock ? SQLITE_OK : sqlite3OsUnlock(pPager->fd, eLock);
1179 if( pPager->eLock!=UNKNOWN_LOCK ){
1180 pPager->eLock = (u8)eLock;
1182 IOTRACE(("UNLOCK %p %d\n", pPager, eLock))
1184 pPager->changeCountDone = pPager->tempFile; /* ticket fb3b3024ea238d5c */
1185 return rc;
1189 ** Lock the database file to level eLock, which must be either SHARED_LOCK,
1190 ** RESERVED_LOCK or EXCLUSIVE_LOCK. If the caller is successful, set the
1191 ** Pager.eLock variable to the new locking state.
1193 ** Except, if Pager.eLock is set to UNKNOWN_LOCK when this function is
1194 ** called, do not modify it unless the new locking state is EXCLUSIVE_LOCK.
1195 ** See the comment above the #define of UNKNOWN_LOCK for an explanation
1196 ** of this.
1198 static int pagerLockDb(Pager *pPager, int eLock){
1199 int rc = SQLITE_OK;
1201 assert( eLock==SHARED_LOCK || eLock==RESERVED_LOCK || eLock==EXCLUSIVE_LOCK );
1202 if( pPager->eLock<eLock || pPager->eLock==UNKNOWN_LOCK ){
1203 rc = pPager->noLock ? SQLITE_OK : sqlite3OsLock(pPager->fd, eLock);
1204 if( rc==SQLITE_OK && (pPager->eLock!=UNKNOWN_LOCK||eLock==EXCLUSIVE_LOCK) ){
1205 pPager->eLock = (u8)eLock;
1206 IOTRACE(("LOCK %p %d\n", pPager, eLock))
1209 return rc;
1213 ** This function determines whether or not the atomic-write or
1214 ** atomic-batch-write optimizations can be used with this pager. The
1215 ** atomic-write optimization can be used if:
1217 ** (a) the value returned by OsDeviceCharacteristics() indicates that
1218 ** a database page may be written atomically, and
1219 ** (b) the value returned by OsSectorSize() is less than or equal
1220 ** to the page size.
1222 ** If it can be used, then the value returned is the size of the journal
1223 ** file when it contains rollback data for exactly one page.
1225 ** The atomic-batch-write optimization can be used if OsDeviceCharacteristics()
1226 ** returns a value with the SQLITE_IOCAP_BATCH_ATOMIC bit set. -1 is
1227 ** returned in this case.
1229 ** If neither optimization can be used, 0 is returned.
1231 static int jrnlBufferSize(Pager *pPager){
1232 assert( !MEMDB );
1234 #if defined(SQLITE_ENABLE_ATOMIC_WRITE) \
1235 || defined(SQLITE_ENABLE_BATCH_ATOMIC_WRITE)
1236 int dc; /* Device characteristics */
1238 assert( isOpen(pPager->fd) );
1239 dc = sqlite3OsDeviceCharacteristics(pPager->fd);
1240 #else
1241 UNUSED_PARAMETER(pPager);
1242 #endif
1244 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
1245 if( pPager->dbSize>0 && (dc&SQLITE_IOCAP_BATCH_ATOMIC) ){
1246 return -1;
1248 #endif
1250 #ifdef SQLITE_ENABLE_ATOMIC_WRITE
1252 int nSector = pPager->sectorSize;
1253 int szPage = pPager->pageSize;
1255 assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
1256 assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));
1257 if( 0==(dc&(SQLITE_IOCAP_ATOMIC|(szPage>>8)) || nSector>szPage) ){
1258 return 0;
1262 return JOURNAL_HDR_SZ(pPager) + JOURNAL_PG_SZ(pPager);
1263 #endif
1265 return 0;
1269 ** If SQLITE_CHECK_PAGES is defined then we do some sanity checking
1270 ** on the cache using a hash function. This is used for testing
1271 ** and debugging only.
1273 #ifdef SQLITE_CHECK_PAGES
1275 ** Return a 32-bit hash of the page data for pPage.
1277 static u32 pager_datahash(int nByte, unsigned char *pData){
1278 u32 hash = 0;
1279 int i;
1280 for(i=0; i<nByte; i++){
1281 hash = (hash*1039) + pData[i];
1283 return hash;
1285 static u32 pager_pagehash(PgHdr *pPage){
1286 return pager_datahash(pPage->pPager->pageSize, (unsigned char *)pPage->pData);
1288 static void pager_set_pagehash(PgHdr *pPage){
1289 pPage->pageHash = pager_pagehash(pPage);
1293 ** The CHECK_PAGE macro takes a PgHdr* as an argument. If SQLITE_CHECK_PAGES
1294 ** is defined, and NDEBUG is not defined, an assert() statement checks
1295 ** that the page is either dirty or still matches the calculated page-hash.
1297 #define CHECK_PAGE(x) checkPage(x)
1298 static void checkPage(PgHdr *pPg){
1299 Pager *pPager = pPg->pPager;
1300 assert( pPager->eState!=PAGER_ERROR );
1301 assert( (pPg->flags&PGHDR_DIRTY) || pPg->pageHash==pager_pagehash(pPg) );
1304 #else
1305 #define pager_datahash(X,Y) 0
1306 #define pager_pagehash(X) 0
1307 #define pager_set_pagehash(X)
1308 #define CHECK_PAGE(x)
1309 #endif /* SQLITE_CHECK_PAGES */
1312 ** When this is called the journal file for pager pPager must be open.
1313 ** This function attempts to read a super-journal file name from the
1314 ** end of the file and, if successful, copies it into memory supplied
1315 ** by the caller. See comments above writeSuperJournal() for the format
1316 ** used to store a super-journal file name at the end of a journal file.
1318 ** zSuper must point to a buffer of at least nSuper bytes allocated by
1319 ** the caller. This should be sqlite3_vfs.mxPathname+1 (to ensure there is
1320 ** enough space to write the super-journal name). If the super-journal
1321 ** name in the journal is longer than nSuper bytes (including a
1322 ** nul-terminator), then this is handled as if no super-journal name
1323 ** were present in the journal.
1325 ** If a super-journal file name is present at the end of the journal
1326 ** file, then it is copied into the buffer pointed to by zSuper. A
1327 ** nul-terminator byte is appended to the buffer following the
1328 ** super-journal file name.
1330 ** If it is determined that no super-journal file name is present
1331 ** zSuper[0] is set to 0 and SQLITE_OK returned.
1333 ** If an error occurs while reading from the journal file, an SQLite
1334 ** error code is returned.
1336 static int readSuperJournal(sqlite3_file *pJrnl, char *zSuper, u32 nSuper){
1337 int rc; /* Return code */
1338 u32 len; /* Length in bytes of super-journal name */
1339 i64 szJ; /* Total size in bytes of journal file pJrnl */
1340 u32 cksum; /* MJ checksum value read from journal */
1341 u32 u; /* Unsigned loop counter */
1342 unsigned char aMagic[8]; /* A buffer to hold the magic header */
1343 zSuper[0] = '\0';
1345 if( SQLITE_OK!=(rc = sqlite3OsFileSize(pJrnl, &szJ))
1346 || szJ<16
1347 || SQLITE_OK!=(rc = read32bits(pJrnl, szJ-16, &len))
1348 || len>=nSuper
1349 || len>szJ-16
1350 || len==0
1351 || SQLITE_OK!=(rc = read32bits(pJrnl, szJ-12, &cksum))
1352 || SQLITE_OK!=(rc = sqlite3OsRead(pJrnl, aMagic, 8, szJ-8))
1353 || memcmp(aMagic, aJournalMagic, 8)
1354 || SQLITE_OK!=(rc = sqlite3OsRead(pJrnl, zSuper, len, szJ-16-len))
1356 return rc;
1359 /* See if the checksum matches the super-journal name */
1360 for(u=0; u<len; u++){
1361 cksum -= zSuper[u];
1363 if( cksum ){
1364 /* If the checksum doesn't add up, then one or more of the disk sectors
1365 ** containing the super-journal filename is corrupted. This means
1366 ** definitely roll back, so just return SQLITE_OK and report a (nul)
1367 ** super-journal filename.
1369 len = 0;
1371 zSuper[len] = '\0';
1372 zSuper[len+1] = '\0';
1374 return SQLITE_OK;
1378 ** Return the offset of the sector boundary at or immediately
1379 ** following the value in pPager->journalOff, assuming a sector
1380 ** size of pPager->sectorSize bytes.
1382 ** i.e for a sector size of 512:
1384 ** Pager.journalOff Return value
1385 ** ---------------------------------------
1386 ** 0 0
1387 ** 512 512
1388 ** 100 512
1389 ** 2000 2048
1392 static i64 journalHdrOffset(Pager *pPager){
1393 i64 offset = 0;
1394 i64 c = pPager->journalOff;
1395 if( c ){
1396 offset = ((c-1)/JOURNAL_HDR_SZ(pPager) + 1) * JOURNAL_HDR_SZ(pPager);
1398 assert( offset%JOURNAL_HDR_SZ(pPager)==0 );
1399 assert( offset>=c );
1400 assert( (offset-c)<JOURNAL_HDR_SZ(pPager) );
1401 return offset;
1405 ** The journal file must be open when this function is called.
1407 ** This function is a no-op if the journal file has not been written to
1408 ** within the current transaction (i.e. if Pager.journalOff==0).
1410 ** If doTruncate is non-zero or the Pager.journalSizeLimit variable is
1411 ** set to 0, then truncate the journal file to zero bytes in size. Otherwise,
1412 ** zero the 28-byte header at the start of the journal file. In either case,
1413 ** if the pager is not in no-sync mode, sync the journal file immediately
1414 ** after writing or truncating it.
1416 ** If Pager.journalSizeLimit is set to a positive, non-zero value, and
1417 ** following the truncation or zeroing described above the size of the
1418 ** journal file in bytes is larger than this value, then truncate the
1419 ** journal file to Pager.journalSizeLimit bytes. The journal file does
1420 ** not need to be synced following this operation.
1422 ** If an IO error occurs, abandon processing and return the IO error code.
1423 ** Otherwise, return SQLITE_OK.
1425 static int zeroJournalHdr(Pager *pPager, int doTruncate){
1426 int rc = SQLITE_OK; /* Return code */
1427 assert( isOpen(pPager->jfd) );
1428 assert( !sqlite3JournalIsInMemory(pPager->jfd) );
1429 if( pPager->journalOff ){
1430 const i64 iLimit = pPager->journalSizeLimit; /* Local cache of jsl */
1432 IOTRACE(("JZEROHDR %p\n", pPager))
1433 if( doTruncate || iLimit==0 ){
1434 rc = sqlite3OsTruncate(pPager->jfd, 0);
1435 }else{
1436 static const char zeroHdr[28] = {0};
1437 rc = sqlite3OsWrite(pPager->jfd, zeroHdr, sizeof(zeroHdr), 0);
1439 if( rc==SQLITE_OK && !pPager->noSync ){
1440 rc = sqlite3OsSync(pPager->jfd, SQLITE_SYNC_DATAONLY|pPager->syncFlags);
1443 /* At this point the transaction is committed but the write lock
1444 ** is still held on the file. If there is a size limit configured for
1445 ** the persistent journal and the journal file currently consumes more
1446 ** space than that limit allows for, truncate it now. There is no need
1447 ** to sync the file following this operation.
1449 if( rc==SQLITE_OK && iLimit>0 ){
1450 i64 sz;
1451 rc = sqlite3OsFileSize(pPager->jfd, &sz);
1452 if( rc==SQLITE_OK && sz>iLimit ){
1453 rc = sqlite3OsTruncate(pPager->jfd, iLimit);
1457 return rc;
1461 ** The journal file must be open when this routine is called. A journal
1462 ** header (JOURNAL_HDR_SZ bytes) is written into the journal file at the
1463 ** current location.
1465 ** The format for the journal header is as follows:
1466 ** - 8 bytes: Magic identifying journal format.
1467 ** - 4 bytes: Number of records in journal, or -1 no-sync mode is on.
1468 ** - 4 bytes: Random number used for page hash.
1469 ** - 4 bytes: Initial database page count.
1470 ** - 4 bytes: Sector size used by the process that wrote this journal.
1471 ** - 4 bytes: Database page size.
1473 ** Followed by (JOURNAL_HDR_SZ - 28) bytes of unused space.
1475 static int writeJournalHdr(Pager *pPager){
1476 int rc = SQLITE_OK; /* Return code */
1477 char *zHeader = pPager->pTmpSpace; /* Temporary space used to build header */
1478 u32 nHeader = (u32)pPager->pageSize;/* Size of buffer pointed to by zHeader */
1479 u32 nWrite; /* Bytes of header sector written */
1480 int ii; /* Loop counter */
1482 assert( isOpen(pPager->jfd) ); /* Journal file must be open. */
1484 if( nHeader>JOURNAL_HDR_SZ(pPager) ){
1485 nHeader = JOURNAL_HDR_SZ(pPager);
1488 /* If there are active savepoints and any of them were created
1489 ** since the most recent journal header was written, update the
1490 ** PagerSavepoint.iHdrOffset fields now.
1492 for(ii=0; ii<pPager->nSavepoint; ii++){
1493 if( pPager->aSavepoint[ii].iHdrOffset==0 ){
1494 pPager->aSavepoint[ii].iHdrOffset = pPager->journalOff;
1498 pPager->journalHdr = pPager->journalOff = journalHdrOffset(pPager);
1501 ** Write the nRec Field - the number of page records that follow this
1502 ** journal header. Normally, zero is written to this value at this time.
1503 ** After the records are added to the journal (and the journal synced,
1504 ** if in full-sync mode), the zero is overwritten with the true number
1505 ** of records (see syncJournal()).
1507 ** A faster alternative is to write 0xFFFFFFFF to the nRec field. When
1508 ** reading the journal this value tells SQLite to assume that the
1509 ** rest of the journal file contains valid page records. This assumption
1510 ** is dangerous, as if a failure occurred whilst writing to the journal
1511 ** file it may contain some garbage data. There are two scenarios
1512 ** where this risk can be ignored:
1514 ** * When the pager is in no-sync mode. Corruption can follow a
1515 ** power failure in this case anyway.
1517 ** * When the SQLITE_IOCAP_SAFE_APPEND flag is set. This guarantees
1518 ** that garbage data is never appended to the journal file.
1520 assert( isOpen(pPager->fd) || pPager->noSync );
1521 if( pPager->noSync || (pPager->journalMode==PAGER_JOURNALMODE_MEMORY)
1522 || (sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_SAFE_APPEND)
1524 memcpy(zHeader, aJournalMagic, sizeof(aJournalMagic));
1525 put32bits(&zHeader[sizeof(aJournalMagic)], 0xffffffff);
1526 }else{
1527 memset(zHeader, 0, sizeof(aJournalMagic)+4);
1530 /* The random check-hash initializer */
1531 sqlite3_randomness(sizeof(pPager->cksumInit), &pPager->cksumInit);
1532 put32bits(&zHeader[sizeof(aJournalMagic)+4], pPager->cksumInit);
1533 /* The initial database size */
1534 put32bits(&zHeader[sizeof(aJournalMagic)+8], pPager->dbOrigSize);
1535 /* The assumed sector size for this process */
1536 put32bits(&zHeader[sizeof(aJournalMagic)+12], pPager->sectorSize);
1538 /* The page size */
1539 put32bits(&zHeader[sizeof(aJournalMagic)+16], pPager->pageSize);
1541 /* Initializing the tail of the buffer is not necessary. Everything
1542 ** works find if the following memset() is omitted. But initializing
1543 ** the memory prevents valgrind from complaining, so we are willing to
1544 ** take the performance hit.
1546 memset(&zHeader[sizeof(aJournalMagic)+20], 0,
1547 nHeader-(sizeof(aJournalMagic)+20));
1549 /* In theory, it is only necessary to write the 28 bytes that the
1550 ** journal header consumes to the journal file here. Then increment the
1551 ** Pager.journalOff variable by JOURNAL_HDR_SZ so that the next
1552 ** record is written to the following sector (leaving a gap in the file
1553 ** that will be implicitly filled in by the OS).
1555 ** However it has been discovered that on some systems this pattern can
1556 ** be significantly slower than contiguously writing data to the file,
1557 ** even if that means explicitly writing data to the block of
1558 ** (JOURNAL_HDR_SZ - 28) bytes that will not be used. So that is what
1559 ** is done.
1561 ** The loop is required here in case the sector-size is larger than the
1562 ** database page size. Since the zHeader buffer is only Pager.pageSize
1563 ** bytes in size, more than one call to sqlite3OsWrite() may be required
1564 ** to populate the entire journal header sector.
1566 for(nWrite=0; rc==SQLITE_OK&&nWrite<JOURNAL_HDR_SZ(pPager); nWrite+=nHeader){
1567 IOTRACE(("JHDR %p %lld %d\n", pPager, pPager->journalHdr, nHeader))
1568 rc = sqlite3OsWrite(pPager->jfd, zHeader, nHeader, pPager->journalOff);
1569 assert( pPager->journalHdr <= pPager->journalOff );
1570 pPager->journalOff += nHeader;
1573 return rc;
1577 ** The journal file must be open when this is called. A journal header file
1578 ** (JOURNAL_HDR_SZ bytes) is read from the current location in the journal
1579 ** file. The current location in the journal file is given by
1580 ** pPager->journalOff. See comments above function writeJournalHdr() for
1581 ** a description of the journal header format.
1583 ** If the header is read successfully, *pNRec is set to the number of
1584 ** page records following this header and *pDbSize is set to the size of the
1585 ** database before the transaction began, in pages. Also, pPager->cksumInit
1586 ** is set to the value read from the journal header. SQLITE_OK is returned
1587 ** in this case.
1589 ** If the journal header file appears to be corrupted, SQLITE_DONE is
1590 ** returned and *pNRec and *PDbSize are undefined. If JOURNAL_HDR_SZ bytes
1591 ** cannot be read from the journal file an error code is returned.
1593 static int readJournalHdr(
1594 Pager *pPager, /* Pager object */
1595 int isHot,
1596 i64 journalSize, /* Size of the open journal file in bytes */
1597 u32 *pNRec, /* OUT: Value read from the nRec field */
1598 u32 *pDbSize /* OUT: Value of original database size field */
1600 int rc; /* Return code */
1601 unsigned char aMagic[8]; /* A buffer to hold the magic header */
1602 i64 iHdrOff; /* Offset of journal header being read */
1604 assert( isOpen(pPager->jfd) ); /* Journal file must be open. */
1606 /* Advance Pager.journalOff to the start of the next sector. If the
1607 ** journal file is too small for there to be a header stored at this
1608 ** point, return SQLITE_DONE.
1610 pPager->journalOff = journalHdrOffset(pPager);
1611 if( pPager->journalOff+JOURNAL_HDR_SZ(pPager) > journalSize ){
1612 return SQLITE_DONE;
1614 iHdrOff = pPager->journalOff;
1616 /* Read in the first 8 bytes of the journal header. If they do not match
1617 ** the magic string found at the start of each journal header, return
1618 ** SQLITE_DONE. If an IO error occurs, return an error code. Otherwise,
1619 ** proceed.
1621 if( isHot || iHdrOff!=pPager->journalHdr ){
1622 rc = sqlite3OsRead(pPager->jfd, aMagic, sizeof(aMagic), iHdrOff);
1623 if( rc ){
1624 return rc;
1626 if( memcmp(aMagic, aJournalMagic, sizeof(aMagic))!=0 ){
1627 return SQLITE_DONE;
1631 /* Read the first three 32-bit fields of the journal header: The nRec
1632 ** field, the checksum-initializer and the database size at the start
1633 ** of the transaction. Return an error code if anything goes wrong.
1635 if( SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+8, pNRec))
1636 || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+12, &pPager->cksumInit))
1637 || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+16, pDbSize))
1639 return rc;
1642 if( pPager->journalOff==0 ){
1643 u32 iPageSize; /* Page-size field of journal header */
1644 u32 iSectorSize; /* Sector-size field of journal header */
1646 /* Read the page-size and sector-size journal header fields. */
1647 if( SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+20, &iSectorSize))
1648 || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+24, &iPageSize))
1650 return rc;
1653 /* Versions of SQLite prior to 3.5.8 set the page-size field of the
1654 ** journal header to zero. In this case, assume that the Pager.pageSize
1655 ** variable is already set to the correct page size.
1657 if( iPageSize==0 ){
1658 iPageSize = pPager->pageSize;
1661 /* Check that the values read from the page-size and sector-size fields
1662 ** are within range. To be 'in range', both values need to be a power
1663 ** of two greater than or equal to 512 or 32, and not greater than their
1664 ** respective compile time maximum limits.
1666 if( iPageSize<512 || iSectorSize<32
1667 || iPageSize>SQLITE_MAX_PAGE_SIZE || iSectorSize>MAX_SECTOR_SIZE
1668 || ((iPageSize-1)&iPageSize)!=0 || ((iSectorSize-1)&iSectorSize)!=0
1670 /* If the either the page-size or sector-size in the journal-header is
1671 ** invalid, then the process that wrote the journal-header must have
1672 ** crashed before the header was synced. In this case stop reading
1673 ** the journal file here.
1675 return SQLITE_DONE;
1678 /* Update the page-size to match the value read from the journal.
1679 ** Use a testcase() macro to make sure that malloc failure within
1680 ** PagerSetPagesize() is tested.
1682 rc = sqlite3PagerSetPagesize(pPager, &iPageSize, -1);
1683 testcase( rc!=SQLITE_OK );
1685 /* Update the assumed sector-size to match the value used by
1686 ** the process that created this journal. If this journal was
1687 ** created by a process other than this one, then this routine
1688 ** is being called from within pager_playback(). The local value
1689 ** of Pager.sectorSize is restored at the end of that routine.
1691 pPager->sectorSize = iSectorSize;
1694 pPager->journalOff += JOURNAL_HDR_SZ(pPager);
1695 return rc;
1700 ** Write the supplied super-journal name into the journal file for pager
1701 ** pPager at the current location. The super-journal name must be the last
1702 ** thing written to a journal file. If the pager is in full-sync mode, the
1703 ** journal file descriptor is advanced to the next sector boundary before
1704 ** anything is written. The format is:
1706 ** + 4 bytes: PAGER_SJ_PGNO.
1707 ** + N bytes: super-journal filename in utf-8.
1708 ** + 4 bytes: N (length of super-journal name in bytes, no nul-terminator).
1709 ** + 4 bytes: super-journal name checksum.
1710 ** + 8 bytes: aJournalMagic[].
1712 ** The super-journal page checksum is the sum of the bytes in thesuper-journal
1713 ** name, where each byte is interpreted as a signed 8-bit integer.
1715 ** If zSuper is a NULL pointer (occurs for a single database transaction),
1716 ** this call is a no-op.
1718 static int writeSuperJournal(Pager *pPager, const char *zSuper){
1719 int rc; /* Return code */
1720 int nSuper; /* Length of string zSuper */
1721 i64 iHdrOff; /* Offset of header in journal file */
1722 i64 jrnlSize; /* Size of journal file on disk */
1723 u32 cksum = 0; /* Checksum of string zSuper */
1725 assert( pPager->setSuper==0 );
1726 assert( !pagerUseWal(pPager) );
1728 if( !zSuper
1729 || pPager->journalMode==PAGER_JOURNALMODE_MEMORY
1730 || !isOpen(pPager->jfd)
1732 return SQLITE_OK;
1734 pPager->setSuper = 1;
1735 assert( pPager->journalHdr <= pPager->journalOff );
1737 /* Calculate the length in bytes and the checksum of zSuper */
1738 for(nSuper=0; zSuper[nSuper]; nSuper++){
1739 cksum += zSuper[nSuper];
1742 /* If in full-sync mode, advance to the next disk sector before writing
1743 ** the super-journal name. This is in case the previous page written to
1744 ** the journal has already been synced.
1746 if( pPager->fullSync ){
1747 pPager->journalOff = journalHdrOffset(pPager);
1749 iHdrOff = pPager->journalOff;
1751 /* Write the super-journal data to the end of the journal file. If
1752 ** an error occurs, return the error code to the caller.
1754 if( (0 != (rc = write32bits(pPager->jfd, iHdrOff, PAGER_SJ_PGNO(pPager))))
1755 || (0 != (rc = sqlite3OsWrite(pPager->jfd, zSuper, nSuper, iHdrOff+4)))
1756 || (0 != (rc = write32bits(pPager->jfd, iHdrOff+4+nSuper, nSuper)))
1757 || (0 != (rc = write32bits(pPager->jfd, iHdrOff+4+nSuper+4, cksum)))
1758 || (0 != (rc = sqlite3OsWrite(pPager->jfd, aJournalMagic, 8,
1759 iHdrOff+4+nSuper+8)))
1761 return rc;
1763 pPager->journalOff += (nSuper+20);
1765 /* If the pager is in peristent-journal mode, then the physical
1766 ** journal-file may extend past the end of the super-journal name
1767 ** and 8 bytes of magic data just written to the file. This is
1768 ** dangerous because the code to rollback a hot-journal file
1769 ** will not be able to find the super-journal name to determine
1770 ** whether or not the journal is hot.
1772 ** Easiest thing to do in this scenario is to truncate the journal
1773 ** file to the required size.
1775 if( SQLITE_OK==(rc = sqlite3OsFileSize(pPager->jfd, &jrnlSize))
1776 && jrnlSize>pPager->journalOff
1778 rc = sqlite3OsTruncate(pPager->jfd, pPager->journalOff);
1780 return rc;
1784 ** Discard the entire contents of the in-memory page-cache.
1786 static void pager_reset(Pager *pPager){
1787 pPager->iDataVersion++;
1788 sqlite3BackupRestart(pPager->pBackup);
1789 sqlite3PcacheClear(pPager->pPCache);
1793 ** Return the pPager->iDataVersion value
1795 u32 sqlite3PagerDataVersion(Pager *pPager){
1796 return pPager->iDataVersion;
1800 ** Free all structures in the Pager.aSavepoint[] array and set both
1801 ** Pager.aSavepoint and Pager.nSavepoint to zero. Close the sub-journal
1802 ** if it is open and the pager is not in exclusive mode.
1804 static void releaseAllSavepoints(Pager *pPager){
1805 int ii; /* Iterator for looping through Pager.aSavepoint */
1806 for(ii=0; ii<pPager->nSavepoint; ii++){
1807 sqlite3BitvecDestroy(pPager->aSavepoint[ii].pInSavepoint);
1809 if( !pPager->exclusiveMode || sqlite3JournalIsInMemory(pPager->sjfd) ){
1810 sqlite3OsClose(pPager->sjfd);
1812 sqlite3_free(pPager->aSavepoint);
1813 pPager->aSavepoint = 0;
1814 pPager->nSavepoint = 0;
1815 pPager->nSubRec = 0;
1819 ** Set the bit number pgno in the PagerSavepoint.pInSavepoint
1820 ** bitvecs of all open savepoints. Return SQLITE_OK if successful
1821 ** or SQLITE_NOMEM if a malloc failure occurs.
1823 static int addToSavepointBitvecs(Pager *pPager, Pgno pgno){
1824 int ii; /* Loop counter */
1825 int rc = SQLITE_OK; /* Result code */
1827 for(ii=0; ii<pPager->nSavepoint; ii++){
1828 PagerSavepoint *p = &pPager->aSavepoint[ii];
1829 if( pgno<=p->nOrig ){
1830 rc |= sqlite3BitvecSet(p->pInSavepoint, pgno);
1831 testcase( rc==SQLITE_NOMEM );
1832 assert( rc==SQLITE_OK || rc==SQLITE_NOMEM );
1835 return rc;
1839 ** This function is a no-op if the pager is in exclusive mode and not
1840 ** in the ERROR state. Otherwise, it switches the pager to PAGER_OPEN
1841 ** state.
1843 ** If the pager is not in exclusive-access mode, the database file is
1844 ** completely unlocked. If the file is unlocked and the file-system does
1845 ** not exhibit the UNDELETABLE_WHEN_OPEN property, the journal file is
1846 ** closed (if it is open).
1848 ** If the pager is in ERROR state when this function is called, the
1849 ** contents of the pager cache are discarded before switching back to
1850 ** the OPEN state. Regardless of whether the pager is in exclusive-mode
1851 ** or not, any journal file left in the file-system will be treated
1852 ** as a hot-journal and rolled back the next time a read-transaction
1853 ** is opened (by this or by any other connection).
1855 static void pager_unlock(Pager *pPager){
1857 assert( pPager->eState==PAGER_READER
1858 || pPager->eState==PAGER_OPEN
1859 || pPager->eState==PAGER_ERROR
1862 sqlite3BitvecDestroy(pPager->pInJournal);
1863 pPager->pInJournal = 0;
1864 releaseAllSavepoints(pPager);
1866 if( pagerUseWal(pPager) ){
1867 assert( !isOpen(pPager->jfd) );
1868 sqlite3WalEndReadTransaction(pPager->pWal);
1869 pPager->eState = PAGER_OPEN;
1870 }else if( !pPager->exclusiveMode ){
1871 int rc; /* Error code returned by pagerUnlockDb() */
1872 int iDc = isOpen(pPager->fd)?sqlite3OsDeviceCharacteristics(pPager->fd):0;
1874 /* If the operating system support deletion of open files, then
1875 ** close the journal file when dropping the database lock. Otherwise
1876 ** another connection with journal_mode=delete might delete the file
1877 ** out from under us.
1879 assert( (PAGER_JOURNALMODE_MEMORY & 5)!=1 );
1880 assert( (PAGER_JOURNALMODE_OFF & 5)!=1 );
1881 assert( (PAGER_JOURNALMODE_WAL & 5)!=1 );
1882 assert( (PAGER_JOURNALMODE_DELETE & 5)!=1 );
1883 assert( (PAGER_JOURNALMODE_TRUNCATE & 5)==1 );
1884 assert( (PAGER_JOURNALMODE_PERSIST & 5)==1 );
1885 if( 0==(iDc & SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN)
1886 || 1!=(pPager->journalMode & 5)
1888 sqlite3OsClose(pPager->jfd);
1891 /* If the pager is in the ERROR state and the call to unlock the database
1892 ** file fails, set the current lock to UNKNOWN_LOCK. See the comment
1893 ** above the #define for UNKNOWN_LOCK for an explanation of why this
1894 ** is necessary.
1896 rc = pagerUnlockDb(pPager, NO_LOCK);
1897 if( rc!=SQLITE_OK && pPager->eState==PAGER_ERROR ){
1898 pPager->eLock = UNKNOWN_LOCK;
1901 /* The pager state may be changed from PAGER_ERROR to PAGER_OPEN here
1902 ** without clearing the error code. This is intentional - the error
1903 ** code is cleared and the cache reset in the block below.
1905 assert( pPager->errCode || pPager->eState!=PAGER_ERROR );
1906 pPager->eState = PAGER_OPEN;
1909 /* If Pager.errCode is set, the contents of the pager cache cannot be
1910 ** trusted. Now that there are no outstanding references to the pager,
1911 ** it can safely move back to PAGER_OPEN state. This happens in both
1912 ** normal and exclusive-locking mode.
1914 assert( pPager->errCode==SQLITE_OK || !MEMDB );
1915 if( pPager->errCode ){
1916 if( pPager->tempFile==0 ){
1917 pager_reset(pPager);
1918 pPager->changeCountDone = 0;
1919 pPager->eState = PAGER_OPEN;
1920 }else{
1921 pPager->eState = (isOpen(pPager->jfd) ? PAGER_OPEN : PAGER_READER);
1923 if( USEFETCH(pPager) ) sqlite3OsUnfetch(pPager->fd, 0, 0);
1924 pPager->errCode = SQLITE_OK;
1925 setGetterMethod(pPager);
1928 pPager->journalOff = 0;
1929 pPager->journalHdr = 0;
1930 pPager->setSuper = 0;
1934 ** This function is called whenever an IOERR or FULL error that requires
1935 ** the pager to transition into the ERROR state may ahve occurred.
1936 ** The first argument is a pointer to the pager structure, the second
1937 ** the error-code about to be returned by a pager API function. The
1938 ** value returned is a copy of the second argument to this function.
1940 ** If the second argument is SQLITE_FULL, SQLITE_IOERR or one of the
1941 ** IOERR sub-codes, the pager enters the ERROR state and the error code
1942 ** is stored in Pager.errCode. While the pager remains in the ERROR state,
1943 ** all major API calls on the Pager will immediately return Pager.errCode.
1945 ** The ERROR state indicates that the contents of the pager-cache
1946 ** cannot be trusted. This state can be cleared by completely discarding
1947 ** the contents of the pager-cache. If a transaction was active when
1948 ** the persistent error occurred, then the rollback journal may need
1949 ** to be replayed to restore the contents of the database file (as if
1950 ** it were a hot-journal).
1952 static int pager_error(Pager *pPager, int rc){
1953 int rc2 = rc & 0xff;
1954 assert( rc==SQLITE_OK || !MEMDB );
1955 assert(
1956 pPager->errCode==SQLITE_FULL ||
1957 pPager->errCode==SQLITE_OK ||
1958 (pPager->errCode & 0xff)==SQLITE_IOERR
1960 if( rc2==SQLITE_FULL || rc2==SQLITE_IOERR ){
1961 pPager->errCode = rc;
1962 pPager->eState = PAGER_ERROR;
1963 setGetterMethod(pPager);
1965 return rc;
1968 static int pager_truncate(Pager *pPager, Pgno nPage);
1971 ** The write transaction open on pPager is being committed (bCommit==1)
1972 ** or rolled back (bCommit==0).
1974 ** Return TRUE if and only if all dirty pages should be flushed to disk.
1976 ** Rules:
1978 ** * For non-TEMP databases, always sync to disk. This is necessary
1979 ** for transactions to be durable.
1981 ** * Sync TEMP database only on a COMMIT (not a ROLLBACK) when the backing
1982 ** file has been created already (via a spill on pagerStress()) and
1983 ** when the number of dirty pages in memory exceeds 25% of the total
1984 ** cache size.
1986 static int pagerFlushOnCommit(Pager *pPager, int bCommit){
1987 if( pPager->tempFile==0 ) return 1;
1988 if( !bCommit ) return 0;
1989 if( !isOpen(pPager->fd) ) return 0;
1990 return (sqlite3PCachePercentDirty(pPager->pPCache)>=25);
1994 ** This routine ends a transaction. A transaction is usually ended by
1995 ** either a COMMIT or a ROLLBACK operation. This routine may be called
1996 ** after rollback of a hot-journal, or if an error occurs while opening
1997 ** the journal file or writing the very first journal-header of a
1998 ** database transaction.
2000 ** This routine is never called in PAGER_ERROR state. If it is called
2001 ** in PAGER_NONE or PAGER_SHARED state and the lock held is less
2002 ** exclusive than a RESERVED lock, it is a no-op.
2004 ** Otherwise, any active savepoints are released.
2006 ** If the journal file is open, then it is "finalized". Once a journal
2007 ** file has been finalized it is not possible to use it to roll back a
2008 ** transaction. Nor will it be considered to be a hot-journal by this
2009 ** or any other database connection. Exactly how a journal is finalized
2010 ** depends on whether or not the pager is running in exclusive mode and
2011 ** the current journal-mode (Pager.journalMode value), as follows:
2013 ** journalMode==MEMORY
2014 ** Journal file descriptor is simply closed. This destroys an
2015 ** in-memory journal.
2017 ** journalMode==TRUNCATE
2018 ** Journal file is truncated to zero bytes in size.
2020 ** journalMode==PERSIST
2021 ** The first 28 bytes of the journal file are zeroed. This invalidates
2022 ** the first journal header in the file, and hence the entire journal
2023 ** file. An invalid journal file cannot be rolled back.
2025 ** journalMode==DELETE
2026 ** The journal file is closed and deleted using sqlite3OsDelete().
2028 ** If the pager is running in exclusive mode, this method of finalizing
2029 ** the journal file is never used. Instead, if the journalMode is
2030 ** DELETE and the pager is in exclusive mode, the method described under
2031 ** journalMode==PERSIST is used instead.
2033 ** After the journal is finalized, the pager moves to PAGER_READER state.
2034 ** If running in non-exclusive rollback mode, the lock on the file is
2035 ** downgraded to a SHARED_LOCK.
2037 ** SQLITE_OK is returned if no error occurs. If an error occurs during
2038 ** any of the IO operations to finalize the journal file or unlock the
2039 ** database then the IO error code is returned to the user. If the
2040 ** operation to finalize the journal file fails, then the code still
2041 ** tries to unlock the database file if not in exclusive mode. If the
2042 ** unlock operation fails as well, then the first error code related
2043 ** to the first error encountered (the journal finalization one) is
2044 ** returned.
2046 static int pager_end_transaction(Pager *pPager, int hasSuper, int bCommit){
2047 int rc = SQLITE_OK; /* Error code from journal finalization operation */
2048 int rc2 = SQLITE_OK; /* Error code from db file unlock operation */
2050 /* Do nothing if the pager does not have an open write transaction
2051 ** or at least a RESERVED lock. This function may be called when there
2052 ** is no write-transaction active but a RESERVED or greater lock is
2053 ** held under two circumstances:
2055 ** 1. After a successful hot-journal rollback, it is called with
2056 ** eState==PAGER_NONE and eLock==EXCLUSIVE_LOCK.
2058 ** 2. If a connection with locking_mode=exclusive holding an EXCLUSIVE
2059 ** lock switches back to locking_mode=normal and then executes a
2060 ** read-transaction, this function is called with eState==PAGER_READER
2061 ** and eLock==EXCLUSIVE_LOCK when the read-transaction is closed.
2063 assert( assert_pager_state(pPager) );
2064 assert( pPager->eState!=PAGER_ERROR );
2065 if( pPager->eState<PAGER_WRITER_LOCKED && pPager->eLock<RESERVED_LOCK ){
2066 return SQLITE_OK;
2069 releaseAllSavepoints(pPager);
2070 assert( isOpen(pPager->jfd) || pPager->pInJournal==0
2071 || (sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
2073 if( isOpen(pPager->jfd) ){
2074 assert( !pagerUseWal(pPager) );
2076 /* Finalize the journal file. */
2077 if( sqlite3JournalIsInMemory(pPager->jfd) ){
2078 /* assert( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ); */
2079 sqlite3OsClose(pPager->jfd);
2080 }else if( pPager->journalMode==PAGER_JOURNALMODE_TRUNCATE ){
2081 if( pPager->journalOff==0 ){
2082 rc = SQLITE_OK;
2083 }else{
2084 rc = sqlite3OsTruncate(pPager->jfd, 0);
2085 if( rc==SQLITE_OK && pPager->fullSync ){
2086 /* Make sure the new file size is written into the inode right away.
2087 ** Otherwise the journal might resurrect following a power loss and
2088 ** cause the last transaction to roll back. See
2089 ** https://bugzilla.mozilla.org/show_bug.cgi?id=1072773
2091 rc = sqlite3OsSync(pPager->jfd, pPager->syncFlags);
2094 pPager->journalOff = 0;
2095 }else if( pPager->journalMode==PAGER_JOURNALMODE_PERSIST
2096 || (pPager->exclusiveMode && pPager->journalMode!=PAGER_JOURNALMODE_WAL)
2098 rc = zeroJournalHdr(pPager, hasSuper||pPager->tempFile);
2099 pPager->journalOff = 0;
2100 }else{
2101 /* This branch may be executed with Pager.journalMode==MEMORY if
2102 ** a hot-journal was just rolled back. In this case the journal
2103 ** file should be closed and deleted. If this connection writes to
2104 ** the database file, it will do so using an in-memory journal.
2106 int bDelete = !pPager->tempFile;
2107 assert( sqlite3JournalIsInMemory(pPager->jfd)==0 );
2108 assert( pPager->journalMode==PAGER_JOURNALMODE_DELETE
2109 || pPager->journalMode==PAGER_JOURNALMODE_MEMORY
2110 || pPager->journalMode==PAGER_JOURNALMODE_WAL
2112 sqlite3OsClose(pPager->jfd);
2113 if( bDelete ){
2114 rc = sqlite3OsDelete(pPager->pVfs, pPager->zJournal, pPager->extraSync);
2119 #ifdef SQLITE_CHECK_PAGES
2120 sqlite3PcacheIterateDirty(pPager->pPCache, pager_set_pagehash);
2121 if( pPager->dbSize==0 && sqlite3PcacheRefCount(pPager->pPCache)>0 ){
2122 PgHdr *p = sqlite3PagerLookup(pPager, 1);
2123 if( p ){
2124 p->pageHash = 0;
2125 sqlite3PagerUnrefNotNull(p);
2128 #endif
2130 sqlite3BitvecDestroy(pPager->pInJournal);
2131 pPager->pInJournal = 0;
2132 pPager->nRec = 0;
2133 if( rc==SQLITE_OK ){
2134 if( MEMDB || pagerFlushOnCommit(pPager, bCommit) ){
2135 sqlite3PcacheCleanAll(pPager->pPCache);
2136 }else{
2137 sqlite3PcacheClearWritable(pPager->pPCache);
2139 sqlite3PcacheTruncate(pPager->pPCache, pPager->dbSize);
2142 if( pagerUseWal(pPager) ){
2143 /* Drop the WAL write-lock, if any. Also, if the connection was in
2144 ** locking_mode=exclusive mode but is no longer, drop the EXCLUSIVE
2145 ** lock held on the database file.
2147 rc2 = sqlite3WalEndWriteTransaction(pPager->pWal);
2148 assert( rc2==SQLITE_OK );
2149 }else if( rc==SQLITE_OK && bCommit && pPager->dbFileSize>pPager->dbSize ){
2150 /* This branch is taken when committing a transaction in rollback-journal
2151 ** mode if the database file on disk is larger than the database image.
2152 ** At this point the journal has been finalized and the transaction
2153 ** successfully committed, but the EXCLUSIVE lock is still held on the
2154 ** file. So it is safe to truncate the database file to its minimum
2155 ** required size. */
2156 assert( pPager->eLock==EXCLUSIVE_LOCK );
2157 rc = pager_truncate(pPager, pPager->dbSize);
2160 if( rc==SQLITE_OK && bCommit ){
2161 rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_COMMIT_PHASETWO, 0);
2162 if( rc==SQLITE_NOTFOUND ) rc = SQLITE_OK;
2165 if( !pPager->exclusiveMode
2166 && (!pagerUseWal(pPager) || sqlite3WalExclusiveMode(pPager->pWal, 0))
2168 rc2 = pagerUnlockDb(pPager, SHARED_LOCK);
2170 pPager->eState = PAGER_READER;
2171 pPager->setSuper = 0;
2173 return (rc==SQLITE_OK?rc2:rc);
2177 ** Execute a rollback if a transaction is active and unlock the
2178 ** database file.
2180 ** If the pager has already entered the ERROR state, do not attempt
2181 ** the rollback at this time. Instead, pager_unlock() is called. The
2182 ** call to pager_unlock() will discard all in-memory pages, unlock
2183 ** the database file and move the pager back to OPEN state. If this
2184 ** means that there is a hot-journal left in the file-system, the next
2185 ** connection to obtain a shared lock on the pager (which may be this one)
2186 ** will roll it back.
2188 ** If the pager has not already entered the ERROR state, but an IO or
2189 ** malloc error occurs during a rollback, then this will itself cause
2190 ** the pager to enter the ERROR state. Which will be cleared by the
2191 ** call to pager_unlock(), as described above.
2193 static void pagerUnlockAndRollback(Pager *pPager){
2194 if( pPager->eState!=PAGER_ERROR && pPager->eState!=PAGER_OPEN ){
2195 assert( assert_pager_state(pPager) );
2196 if( pPager->eState>=PAGER_WRITER_LOCKED ){
2197 sqlite3BeginBenignMalloc();
2198 sqlite3PagerRollback(pPager);
2199 sqlite3EndBenignMalloc();
2200 }else if( !pPager->exclusiveMode ){
2201 assert( pPager->eState==PAGER_READER );
2202 pager_end_transaction(pPager, 0, 0);
2205 pager_unlock(pPager);
2209 ** Parameter aData must point to a buffer of pPager->pageSize bytes
2210 ** of data. Compute and return a checksum based ont the contents of the
2211 ** page of data and the current value of pPager->cksumInit.
2213 ** This is not a real checksum. It is really just the sum of the
2214 ** random initial value (pPager->cksumInit) and every 200th byte
2215 ** of the page data, starting with byte offset (pPager->pageSize%200).
2216 ** Each byte is interpreted as an 8-bit unsigned integer.
2218 ** Changing the formula used to compute this checksum results in an
2219 ** incompatible journal file format.
2221 ** If journal corruption occurs due to a power failure, the most likely
2222 ** scenario is that one end or the other of the record will be changed.
2223 ** It is much less likely that the two ends of the journal record will be
2224 ** correct and the middle be corrupt. Thus, this "checksum" scheme,
2225 ** though fast and simple, catches the mostly likely kind of corruption.
2227 static u32 pager_cksum(Pager *pPager, const u8 *aData){
2228 u32 cksum = pPager->cksumInit; /* Checksum value to return */
2229 int i = pPager->pageSize-200; /* Loop counter */
2230 while( i>0 ){
2231 cksum += aData[i];
2232 i -= 200;
2234 return cksum;
2238 ** Report the current page size and number of reserved bytes back
2239 ** to the codec.
2241 /* BEGIN SQLCIPHER */
2242 #ifdef SQLITE_HAS_CODEC
2243 static void pagerReportSize(Pager *pPager){
2244 if( pPager->xCodecSizeChng ){
2245 pPager->xCodecSizeChng(pPager->pCodec, pPager->pageSize,
2246 (int)pPager->nReserve);
2249 #else
2250 # define pagerReportSize(X) /* No-op if we do not support a codec */
2251 #endif
2252 /* END SQLCIPHER */
2254 /* BEGIN SQLCIPHER */
2255 #ifdef SQLITE_HAS_CODEC
2257 ** Make sure the number of reserved bits is the same in the destination
2258 ** pager as it is in the source. This comes up when a VACUUM changes the
2259 ** number of reserved bits to the "optimal" amount.
2261 void sqlite3PagerAlignReserve(Pager *pDest, Pager *pSrc){
2262 if( pDest->nReserve!=pSrc->nReserve ){
2263 pDest->nReserve = pSrc->nReserve;
2264 pagerReportSize(pDest);
2267 #endif
2268 /* END SQLCIPHER */
2271 ** Read a single page from either the journal file (if isMainJrnl==1) or
2272 ** from the sub-journal (if isMainJrnl==0) and playback that page.
2273 ** The page begins at offset *pOffset into the file. The *pOffset
2274 ** value is increased to the start of the next page in the journal.
2276 ** The main rollback journal uses checksums - the statement journal does
2277 ** not.
2279 ** If the page number of the page record read from the (sub-)journal file
2280 ** is greater than the current value of Pager.dbSize, then playback is
2281 ** skipped and SQLITE_OK is returned.
2283 ** If pDone is not NULL, then it is a record of pages that have already
2284 ** been played back. If the page at *pOffset has already been played back
2285 ** (if the corresponding pDone bit is set) then skip the playback.
2286 ** Make sure the pDone bit corresponding to the *pOffset page is set
2287 ** prior to returning.
2289 ** If the page record is successfully read from the (sub-)journal file
2290 ** and played back, then SQLITE_OK is returned. If an IO error occurs
2291 ** while reading the record from the (sub-)journal file or while writing
2292 ** to the database file, then the IO error code is returned. If data
2293 ** is successfully read from the (sub-)journal file but appears to be
2294 ** corrupted, SQLITE_DONE is returned. Data is considered corrupted in
2295 ** two circumstances:
2297 ** * If the record page-number is illegal (0 or PAGER_SJ_PGNO), or
2298 ** * If the record is being rolled back from the main journal file
2299 ** and the checksum field does not match the record content.
2301 ** Neither of these two scenarios are possible during a savepoint rollback.
2303 ** If this is a savepoint rollback, then memory may have to be dynamically
2304 ** allocated by this function. If this is the case and an allocation fails,
2305 ** SQLITE_NOMEM is returned.
2307 static int pager_playback_one_page(
2308 Pager *pPager, /* The pager being played back */
2309 i64 *pOffset, /* Offset of record to playback */
2310 Bitvec *pDone, /* Bitvec of pages already played back */
2311 int isMainJrnl, /* 1 -> main journal. 0 -> sub-journal. */
2312 int isSavepnt /* True for a savepoint rollback */
2314 int rc;
2315 PgHdr *pPg; /* An existing page in the cache */
2316 Pgno pgno; /* The page number of a page in journal */
2317 u32 cksum; /* Checksum used for sanity checking */
2318 char *aData; /* Temporary storage for the page */
2319 sqlite3_file *jfd; /* The file descriptor for the journal file */
2320 int isSynced; /* True if journal page is synced */
2321 /* BEGIN SQLCIPHER */
2322 #ifdef SQLITE_HAS_CODEC
2323 /* The jrnlEnc flag is true if Journal pages should be passed through
2324 ** the codec. It is false for pure in-memory journals. */
2325 const int jrnlEnc = (isMainJrnl || pPager->subjInMemory==0);
2326 #endif
2327 /* END SQLCIPHER */
2329 assert( (isMainJrnl&~1)==0 ); /* isMainJrnl is 0 or 1 */
2330 assert( (isSavepnt&~1)==0 ); /* isSavepnt is 0 or 1 */
2331 assert( isMainJrnl || pDone ); /* pDone always used on sub-journals */
2332 assert( isSavepnt || pDone==0 ); /* pDone never used on non-savepoint */
2334 aData = pPager->pTmpSpace;
2335 assert( aData ); /* Temp storage must have already been allocated */
2336 assert( pagerUseWal(pPager)==0 || (!isMainJrnl && isSavepnt) );
2338 /* Either the state is greater than PAGER_WRITER_CACHEMOD (a transaction
2339 ** or savepoint rollback done at the request of the caller) or this is
2340 ** a hot-journal rollback. If it is a hot-journal rollback, the pager
2341 ** is in state OPEN and holds an EXCLUSIVE lock. Hot-journal rollback
2342 ** only reads from the main journal, not the sub-journal.
2344 assert( pPager->eState>=PAGER_WRITER_CACHEMOD
2345 || (pPager->eState==PAGER_OPEN && pPager->eLock==EXCLUSIVE_LOCK)
2347 assert( pPager->eState>=PAGER_WRITER_CACHEMOD || isMainJrnl );
2349 /* Read the page number and page data from the journal or sub-journal
2350 ** file. Return an error code to the caller if an IO error occurs.
2352 jfd = isMainJrnl ? pPager->jfd : pPager->sjfd;
2353 rc = read32bits(jfd, *pOffset, &pgno);
2354 if( rc!=SQLITE_OK ) return rc;
2355 rc = sqlite3OsRead(jfd, (u8*)aData, pPager->pageSize, (*pOffset)+4);
2356 if( rc!=SQLITE_OK ) return rc;
2357 *pOffset += pPager->pageSize + 4 + isMainJrnl*4;
2359 /* Sanity checking on the page. This is more important that I originally
2360 ** thought. If a power failure occurs while the journal is being written,
2361 ** it could cause invalid data to be written into the journal. We need to
2362 ** detect this invalid data (with high probability) and ignore it.
2364 if( pgno==0 || pgno==PAGER_SJ_PGNO(pPager) ){
2365 assert( !isSavepnt );
2366 return SQLITE_DONE;
2368 if( pgno>(Pgno)pPager->dbSize || sqlite3BitvecTest(pDone, pgno) ){
2369 return SQLITE_OK;
2371 if( isMainJrnl ){
2372 rc = read32bits(jfd, (*pOffset)-4, &cksum);
2373 if( rc ) return rc;
2374 if( !isSavepnt && pager_cksum(pPager, (u8*)aData)!=cksum ){
2375 return SQLITE_DONE;
2379 /* If this page has already been played back before during the current
2380 ** rollback, then don't bother to play it back again.
2382 if( pDone && (rc = sqlite3BitvecSet(pDone, pgno))!=SQLITE_OK ){
2383 return rc;
2386 /* When playing back page 1, restore the nReserve setting
2388 if( pgno==1 && pPager->nReserve!=((u8*)aData)[20] ){
2389 pPager->nReserve = ((u8*)aData)[20];
2390 pagerReportSize(pPager);
2393 /* If the pager is in CACHEMOD state, then there must be a copy of this
2394 ** page in the pager cache. In this case just update the pager cache,
2395 ** not the database file. The page is left marked dirty in this case.
2397 ** An exception to the above rule: If the database is in no-sync mode
2398 ** and a page is moved during an incremental vacuum then the page may
2399 ** not be in the pager cache. Later: if a malloc() or IO error occurs
2400 ** during a Movepage() call, then the page may not be in the cache
2401 ** either. So the condition described in the above paragraph is not
2402 ** assert()able.
2404 ** If in WRITER_DBMOD, WRITER_FINISHED or OPEN state, then we update the
2405 ** pager cache if it exists and the main file. The page is then marked
2406 ** not dirty. Since this code is only executed in PAGER_OPEN state for
2407 ** a hot-journal rollback, it is guaranteed that the page-cache is empty
2408 ** if the pager is in OPEN state.
2410 ** Ticket #1171: The statement journal might contain page content that is
2411 ** different from the page content at the start of the transaction.
2412 ** This occurs when a page is changed prior to the start of a statement
2413 ** then changed again within the statement. When rolling back such a
2414 ** statement we must not write to the original database unless we know
2415 ** for certain that original page contents are synced into the main rollback
2416 ** journal. Otherwise, a power loss might leave modified data in the
2417 ** database file without an entry in the rollback journal that can
2418 ** restore the database to its original form. Two conditions must be
2419 ** met before writing to the database files. (1) the database must be
2420 ** locked. (2) we know that the original page content is fully synced
2421 ** in the main journal either because the page is not in cache or else
2422 ** the page is marked as needSync==0.
2424 ** 2008-04-14: When attempting to vacuum a corrupt database file, it
2425 ** is possible to fail a statement on a database that does not yet exist.
2426 ** Do not attempt to write if database file has never been opened.
2428 if( pagerUseWal(pPager) ){
2429 pPg = 0;
2430 }else{
2431 pPg = sqlite3PagerLookup(pPager, pgno);
2433 assert( pPg || !MEMDB );
2434 assert( pPager->eState!=PAGER_OPEN || pPg==0 || pPager->tempFile );
2435 PAGERTRACE(("PLAYBACK %d page %d hash(%08x) %s\n",
2436 PAGERID(pPager), pgno, pager_datahash(pPager->pageSize, (u8*)aData),
2437 (isMainJrnl?"main-journal":"sub-journal")
2439 if( isMainJrnl ){
2440 isSynced = pPager->noSync || (*pOffset <= pPager->journalHdr);
2441 }else{
2442 isSynced = (pPg==0 || 0==(pPg->flags & PGHDR_NEED_SYNC));
2444 if( isOpen(pPager->fd)
2445 && (pPager->eState>=PAGER_WRITER_DBMOD || pPager->eState==PAGER_OPEN)
2446 && isSynced
2448 i64 ofst = (pgno-1)*(i64)pPager->pageSize;
2449 testcase( !isSavepnt && pPg!=0 && (pPg->flags&PGHDR_NEED_SYNC)!=0 );
2450 assert( !pagerUseWal(pPager) );
2452 /* Write the data read from the journal back into the database file.
2453 ** This is usually safe even for an encrypted database - as the data
2454 ** was encrypted before it was written to the journal file. The exception
2455 ** is if the data was just read from an in-memory sub-journal. In that
2456 ** case it must be encrypted here before it is copied into the database
2457 ** file. */
2458 /* BEGIN SQLCIPHER */
2459 #ifdef SQLITE_HAS_CODEC
2460 if( !jrnlEnc ){
2461 CODEC2(pPager, aData, pgno, 7, rc=SQLITE_NOMEM_BKPT, aData);
2462 rc = sqlite3OsWrite(pPager->fd, (u8 *)aData, pPager->pageSize, ofst);
2463 CODEC1(pPager, aData, pgno, 3, rc=SQLITE_NOMEM_BKPT);
2464 }else
2465 #endif
2466 /* END SQLCIPHER */
2467 rc = sqlite3OsWrite(pPager->fd, (u8 *)aData, pPager->pageSize, ofst);
2469 if( pgno>pPager->dbFileSize ){
2470 pPager->dbFileSize = pgno;
2472 if( pPager->pBackup ){
2473 /* BEGIN SQLCIPHER */
2474 #ifdef SQLITE_HAS_CODEC
2475 if( jrnlEnc ){
2476 CODEC1(pPager, aData, pgno, 3, rc=SQLITE_NOMEM_BKPT);
2477 sqlite3BackupUpdate(pPager->pBackup, pgno, (u8*)aData);
2478 CODEC2(pPager, aData, pgno, 7, rc=SQLITE_NOMEM_BKPT,aData);
2479 }else
2480 #endif
2481 /* END SQLCIPHER */
2482 sqlite3BackupUpdate(pPager->pBackup, pgno, (u8*)aData);
2484 }else if( !isMainJrnl && pPg==0 ){
2485 /* If this is a rollback of a savepoint and data was not written to
2486 ** the database and the page is not in-memory, there is a potential
2487 ** problem. When the page is next fetched by the b-tree layer, it
2488 ** will be read from the database file, which may or may not be
2489 ** current.
2491 ** There are a couple of different ways this can happen. All are quite
2492 ** obscure. When running in synchronous mode, this can only happen
2493 ** if the page is on the free-list at the start of the transaction, then
2494 ** populated, then moved using sqlite3PagerMovepage().
2496 ** The solution is to add an in-memory page to the cache containing
2497 ** the data just read from the sub-journal. Mark the page as dirty
2498 ** and if the pager requires a journal-sync, then mark the page as
2499 ** requiring a journal-sync before it is written.
2501 assert( isSavepnt );
2502 assert( (pPager->doNotSpill & SPILLFLAG_ROLLBACK)==0 );
2503 pPager->doNotSpill |= SPILLFLAG_ROLLBACK;
2504 rc = sqlite3PagerGet(pPager, pgno, &pPg, 1);
2505 assert( (pPager->doNotSpill & SPILLFLAG_ROLLBACK)!=0 );
2506 pPager->doNotSpill &= ~SPILLFLAG_ROLLBACK;
2507 if( rc!=SQLITE_OK ) return rc;
2508 sqlite3PcacheMakeDirty(pPg);
2510 if( pPg ){
2511 /* No page should ever be explicitly rolled back that is in use, except
2512 ** for page 1 which is held in use in order to keep the lock on the
2513 ** database active. However such a page may be rolled back as a result
2514 ** of an internal error resulting in an automatic call to
2515 ** sqlite3PagerRollback().
2517 void *pData;
2518 pData = pPg->pData;
2519 memcpy(pData, (u8*)aData, pPager->pageSize);
2520 pPager->xReiniter(pPg);
2521 /* It used to be that sqlite3PcacheMakeClean(pPg) was called here. But
2522 ** that call was dangerous and had no detectable benefit since the cache
2523 ** is normally cleaned by sqlite3PcacheCleanAll() after rollback and so
2524 ** has been removed. */
2525 pager_set_pagehash(pPg);
2527 /* If this was page 1, then restore the value of Pager.dbFileVers.
2528 ** Do this before any decoding. */
2529 if( pgno==1 ){
2530 memcpy(&pPager->dbFileVers, &((u8*)pData)[24],sizeof(pPager->dbFileVers));
2533 /* Decode the page just read from disk */
2534 /* BEGIN SQLCIPHER */
2535 #if SQLITE_HAS_CODEC
2536 if( jrnlEnc ){ CODEC1(pPager, pData, pPg->pgno, 3, rc=SQLITE_NOMEM_BKPT); }
2537 #endif
2538 /* END SQLCIPHER */
2539 sqlite3PcacheRelease(pPg);
2541 return rc;
2545 ** Parameter zSuper is the name of a super-journal file. A single journal
2546 ** file that referred to the super-journal file has just been rolled back.
2547 ** This routine checks if it is possible to delete the super-journal file,
2548 ** and does so if it is.
2550 ** Argument zSuper may point to Pager.pTmpSpace. So that buffer is not
2551 ** available for use within this function.
2553 ** When a super-journal file is created, it is populated with the names
2554 ** of all of its child journals, one after another, formatted as utf-8
2555 ** encoded text. The end of each child journal file is marked with a
2556 ** nul-terminator byte (0x00). i.e. the entire contents of a super-journal
2557 ** file for a transaction involving two databases might be:
2559 ** "/home/bill/a.db-journal\x00/home/bill/b.db-journal\x00"
2561 ** A super-journal file may only be deleted once all of its child
2562 ** journals have been rolled back.
2564 ** This function reads the contents of the super-journal file into
2565 ** memory and loops through each of the child journal names. For
2566 ** each child journal, it checks if:
2568 ** * if the child journal exists, and if so
2569 ** * if the child journal contains a reference to super-journal
2570 ** file zSuper
2572 ** If a child journal can be found that matches both of the criteria
2573 ** above, this function returns without doing anything. Otherwise, if
2574 ** no such child journal can be found, file zSuper is deleted from
2575 ** the file-system using sqlite3OsDelete().
2577 ** If an IO error within this function, an error code is returned. This
2578 ** function allocates memory by calling sqlite3Malloc(). If an allocation
2579 ** fails, SQLITE_NOMEM is returned. Otherwise, if no IO or malloc errors
2580 ** occur, SQLITE_OK is returned.
2582 ** TODO: This function allocates a single block of memory to load
2583 ** the entire contents of the super-journal file. This could be
2584 ** a couple of kilobytes or so - potentially larger than the page
2585 ** size.
2587 static int pager_delsuper(Pager *pPager, const char *zSuper){
2588 sqlite3_vfs *pVfs = pPager->pVfs;
2589 int rc; /* Return code */
2590 sqlite3_file *pSuper; /* Malloc'd super-journal file descriptor */
2591 sqlite3_file *pJournal; /* Malloc'd child-journal file descriptor */
2592 char *zSuperJournal = 0; /* Contents of super-journal file */
2593 i64 nSuperJournal; /* Size of super-journal file */
2594 char *zJournal; /* Pointer to one journal within MJ file */
2595 char *zSuperPtr; /* Space to hold super-journal filename */
2596 char *zFree = 0; /* Free this buffer */
2597 int nSuperPtr; /* Amount of space allocated to zSuperPtr[] */
2599 /* Allocate space for both the pJournal and pSuper file descriptors.
2600 ** If successful, open the super-journal file for reading.
2602 pSuper = (sqlite3_file *)sqlite3MallocZero(pVfs->szOsFile * 2);
2603 if( !pSuper ){
2604 rc = SQLITE_NOMEM_BKPT;
2605 pJournal = 0;
2606 }else{
2607 const int flags = (SQLITE_OPEN_READONLY|SQLITE_OPEN_SUPER_JOURNAL);
2608 rc = sqlite3OsOpen(pVfs, zSuper, pSuper, flags, 0);
2609 pJournal = (sqlite3_file *)(((u8 *)pSuper) + pVfs->szOsFile);
2611 if( rc!=SQLITE_OK ) goto delsuper_out;
2613 /* Load the entire super-journal file into space obtained from
2614 ** sqlite3_malloc() and pointed to by zSuperJournal. Also obtain
2615 ** sufficient space (in zSuperPtr) to hold the names of super-journal
2616 ** files extracted from regular rollback-journals.
2618 rc = sqlite3OsFileSize(pSuper, &nSuperJournal);
2619 if( rc!=SQLITE_OK ) goto delsuper_out;
2620 nSuperPtr = pVfs->mxPathname+1;
2621 zFree = sqlite3Malloc(4 + nSuperJournal + nSuperPtr + 2);
2622 if( !zFree ){
2623 rc = SQLITE_NOMEM_BKPT;
2624 goto delsuper_out;
2626 zFree[0] = zFree[1] = zFree[2] = zFree[3] = 0;
2627 zSuperJournal = &zFree[4];
2628 zSuperPtr = &zSuperJournal[nSuperJournal+2];
2629 rc = sqlite3OsRead(pSuper, zSuperJournal, (int)nSuperJournal, 0);
2630 if( rc!=SQLITE_OK ) goto delsuper_out;
2631 zSuperJournal[nSuperJournal] = 0;
2632 zSuperJournal[nSuperJournal+1] = 0;
2634 zJournal = zSuperJournal;
2635 while( (zJournal-zSuperJournal)<nSuperJournal ){
2636 int exists;
2637 rc = sqlite3OsAccess(pVfs, zJournal, SQLITE_ACCESS_EXISTS, &exists);
2638 if( rc!=SQLITE_OK ){
2639 goto delsuper_out;
2641 if( exists ){
2642 /* One of the journals pointed to by the super-journal exists.
2643 ** Open it and check if it points at the super-journal. If
2644 ** so, return without deleting the super-journal file.
2645 ** NB: zJournal is really a MAIN_JOURNAL. But call it a
2646 ** SUPER_JOURNAL here so that the VFS will not send the zJournal
2647 ** name into sqlite3_database_file_object().
2649 int c;
2650 int flags = (SQLITE_OPEN_READONLY|SQLITE_OPEN_SUPER_JOURNAL);
2651 rc = sqlite3OsOpen(pVfs, zJournal, pJournal, flags, 0);
2652 if( rc!=SQLITE_OK ){
2653 goto delsuper_out;
2656 rc = readSuperJournal(pJournal, zSuperPtr, nSuperPtr);
2657 sqlite3OsClose(pJournal);
2658 if( rc!=SQLITE_OK ){
2659 goto delsuper_out;
2662 c = zSuperPtr[0]!=0 && strcmp(zSuperPtr, zSuper)==0;
2663 if( c ){
2664 /* We have a match. Do not delete the super-journal file. */
2665 goto delsuper_out;
2668 zJournal += (sqlite3Strlen30(zJournal)+1);
2671 sqlite3OsClose(pSuper);
2672 rc = sqlite3OsDelete(pVfs, zSuper, 0);
2674 delsuper_out:
2675 sqlite3_free(zFree);
2676 if( pSuper ){
2677 sqlite3OsClose(pSuper);
2678 assert( !isOpen(pJournal) );
2679 sqlite3_free(pSuper);
2681 return rc;
2686 ** This function is used to change the actual size of the database
2687 ** file in the file-system. This only happens when committing a transaction,
2688 ** or rolling back a transaction (including rolling back a hot-journal).
2690 ** If the main database file is not open, or the pager is not in either
2691 ** DBMOD or OPEN state, this function is a no-op. Otherwise, the size
2692 ** of the file is changed to nPage pages (nPage*pPager->pageSize bytes).
2693 ** If the file on disk is currently larger than nPage pages, then use the VFS
2694 ** xTruncate() method to truncate it.
2696 ** Or, it might be the case that the file on disk is smaller than
2697 ** nPage pages. Some operating system implementations can get confused if
2698 ** you try to truncate a file to some size that is larger than it
2699 ** currently is, so detect this case and write a single zero byte to
2700 ** the end of the new file instead.
2702 ** If successful, return SQLITE_OK. If an IO error occurs while modifying
2703 ** the database file, return the error code to the caller.
2705 static int pager_truncate(Pager *pPager, Pgno nPage){
2706 int rc = SQLITE_OK;
2707 assert( pPager->eState!=PAGER_ERROR );
2708 assert( pPager->eState!=PAGER_READER );
2710 if( isOpen(pPager->fd)
2711 && (pPager->eState>=PAGER_WRITER_DBMOD || pPager->eState==PAGER_OPEN)
2713 i64 currentSize, newSize;
2714 int szPage = pPager->pageSize;
2715 assert( pPager->eLock==EXCLUSIVE_LOCK );
2716 /* TODO: Is it safe to use Pager.dbFileSize here? */
2717 rc = sqlite3OsFileSize(pPager->fd, &currentSize);
2718 newSize = szPage*(i64)nPage;
2719 if( rc==SQLITE_OK && currentSize!=newSize ){
2720 if( currentSize>newSize ){
2721 rc = sqlite3OsTruncate(pPager->fd, newSize);
2722 }else if( (currentSize+szPage)<=newSize ){
2723 char *pTmp = pPager->pTmpSpace;
2724 memset(pTmp, 0, szPage);
2725 testcase( (newSize-szPage) == currentSize );
2726 testcase( (newSize-szPage) > currentSize );
2727 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &newSize);
2728 rc = sqlite3OsWrite(pPager->fd, pTmp, szPage, newSize-szPage);
2730 if( rc==SQLITE_OK ){
2731 pPager->dbFileSize = nPage;
2735 return rc;
2739 ** Return a sanitized version of the sector-size of OS file pFile. The
2740 ** return value is guaranteed to lie between 32 and MAX_SECTOR_SIZE.
2742 int sqlite3SectorSize(sqlite3_file *pFile){
2743 int iRet = sqlite3OsSectorSize(pFile);
2744 if( iRet<32 ){
2745 iRet = 512;
2746 }else if( iRet>MAX_SECTOR_SIZE ){
2747 assert( MAX_SECTOR_SIZE>=512 );
2748 iRet = MAX_SECTOR_SIZE;
2750 return iRet;
2754 ** Set the value of the Pager.sectorSize variable for the given
2755 ** pager based on the value returned by the xSectorSize method
2756 ** of the open database file. The sector size will be used
2757 ** to determine the size and alignment of journal header and
2758 ** super-journal pointers within created journal files.
2760 ** For temporary files the effective sector size is always 512 bytes.
2762 ** Otherwise, for non-temporary files, the effective sector size is
2763 ** the value returned by the xSectorSize() method rounded up to 32 if
2764 ** it is less than 32, or rounded down to MAX_SECTOR_SIZE if it
2765 ** is greater than MAX_SECTOR_SIZE.
2767 ** If the file has the SQLITE_IOCAP_POWERSAFE_OVERWRITE property, then set
2768 ** the effective sector size to its minimum value (512). The purpose of
2769 ** pPager->sectorSize is to define the "blast radius" of bytes that
2770 ** might change if a crash occurs while writing to a single byte in
2771 ** that range. But with POWERSAFE_OVERWRITE, the blast radius is zero
2772 ** (that is what POWERSAFE_OVERWRITE means), so we minimize the sector
2773 ** size. For backwards compatibility of the rollback journal file format,
2774 ** we cannot reduce the effective sector size below 512.
2776 static void setSectorSize(Pager *pPager){
2777 assert( isOpen(pPager->fd) || pPager->tempFile );
2779 if( pPager->tempFile
2780 || (sqlite3OsDeviceCharacteristics(pPager->fd) &
2781 SQLITE_IOCAP_POWERSAFE_OVERWRITE)!=0
2783 /* Sector size doesn't matter for temporary files. Also, the file
2784 ** may not have been opened yet, in which case the OsSectorSize()
2785 ** call will segfault. */
2786 pPager->sectorSize = 512;
2787 }else{
2788 pPager->sectorSize = sqlite3SectorSize(pPager->fd);
2793 ** Playback the journal and thus restore the database file to
2794 ** the state it was in before we started making changes.
2796 ** The journal file format is as follows:
2798 ** (1) 8 byte prefix. A copy of aJournalMagic[].
2799 ** (2) 4 byte big-endian integer which is the number of valid page records
2800 ** in the journal. If this value is 0xffffffff, then compute the
2801 ** number of page records from the journal size.
2802 ** (3) 4 byte big-endian integer which is the initial value for the
2803 ** sanity checksum.
2804 ** (4) 4 byte integer which is the number of pages to truncate the
2805 ** database to during a rollback.
2806 ** (5) 4 byte big-endian integer which is the sector size. The header
2807 ** is this many bytes in size.
2808 ** (6) 4 byte big-endian integer which is the page size.
2809 ** (7) zero padding out to the next sector size.
2810 ** (8) Zero or more pages instances, each as follows:
2811 ** + 4 byte page number.
2812 ** + pPager->pageSize bytes of data.
2813 ** + 4 byte checksum
2815 ** When we speak of the journal header, we mean the first 7 items above.
2816 ** Each entry in the journal is an instance of the 8th item.
2818 ** Call the value from the second bullet "nRec". nRec is the number of
2819 ** valid page entries in the journal. In most cases, you can compute the
2820 ** value of nRec from the size of the journal file. But if a power
2821 ** failure occurred while the journal was being written, it could be the
2822 ** case that the size of the journal file had already been increased but
2823 ** the extra entries had not yet made it safely to disk. In such a case,
2824 ** the value of nRec computed from the file size would be too large. For
2825 ** that reason, we always use the nRec value in the header.
2827 ** If the nRec value is 0xffffffff it means that nRec should be computed
2828 ** from the file size. This value is used when the user selects the
2829 ** no-sync option for the journal. A power failure could lead to corruption
2830 ** in this case. But for things like temporary table (which will be
2831 ** deleted when the power is restored) we don't care.
2833 ** If the file opened as the journal file is not a well-formed
2834 ** journal file then all pages up to the first corrupted page are rolled
2835 ** back (or no pages if the journal header is corrupted). The journal file
2836 ** is then deleted and SQLITE_OK returned, just as if no corruption had
2837 ** been encountered.
2839 ** If an I/O or malloc() error occurs, the journal-file is not deleted
2840 ** and an error code is returned.
2842 ** The isHot parameter indicates that we are trying to rollback a journal
2843 ** that might be a hot journal. Or, it could be that the journal is
2844 ** preserved because of JOURNALMODE_PERSIST or JOURNALMODE_TRUNCATE.
2845 ** If the journal really is hot, reset the pager cache prior rolling
2846 ** back any content. If the journal is merely persistent, no reset is
2847 ** needed.
2849 static int pager_playback(Pager *pPager, int isHot){
2850 sqlite3_vfs *pVfs = pPager->pVfs;
2851 i64 szJ; /* Size of the journal file in bytes */
2852 u32 nRec; /* Number of Records in the journal */
2853 u32 u; /* Unsigned loop counter */
2854 Pgno mxPg = 0; /* Size of the original file in pages */
2855 int rc; /* Result code of a subroutine */
2856 int res = 1; /* Value returned by sqlite3OsAccess() */
2857 char *zSuper = 0; /* Name of super-journal file if any */
2858 int needPagerReset; /* True to reset page prior to first page rollback */
2859 int nPlayback = 0; /* Total number of pages restored from journal */
2860 u32 savedPageSize = pPager->pageSize;
2862 /* Figure out how many records are in the journal. Abort early if
2863 ** the journal is empty.
2865 assert( isOpen(pPager->jfd) );
2866 rc = sqlite3OsFileSize(pPager->jfd, &szJ);
2867 if( rc!=SQLITE_OK ){
2868 goto end_playback;
2871 /* Read the super-journal name from the journal, if it is present.
2872 ** If a super-journal file name is specified, but the file is not
2873 ** present on disk, then the journal is not hot and does not need to be
2874 ** played back.
2876 ** TODO: Technically the following is an error because it assumes that
2877 ** buffer Pager.pTmpSpace is (mxPathname+1) bytes or larger. i.e. that
2878 ** (pPager->pageSize >= pPager->pVfs->mxPathname+1). Using os_unix.c,
2879 ** mxPathname is 512, which is the same as the minimum allowable value
2880 ** for pageSize.
2882 zSuper = pPager->pTmpSpace;
2883 rc = readSuperJournal(pPager->jfd, zSuper, pPager->pVfs->mxPathname+1);
2884 if( rc==SQLITE_OK && zSuper[0] ){
2885 rc = sqlite3OsAccess(pVfs, zSuper, SQLITE_ACCESS_EXISTS, &res);
2887 zSuper = 0;
2888 if( rc!=SQLITE_OK || !res ){
2889 goto end_playback;
2891 pPager->journalOff = 0;
2892 needPagerReset = isHot;
2894 /* This loop terminates either when a readJournalHdr() or
2895 ** pager_playback_one_page() call returns SQLITE_DONE or an IO error
2896 ** occurs.
2898 while( 1 ){
2899 /* Read the next journal header from the journal file. If there are
2900 ** not enough bytes left in the journal file for a complete header, or
2901 ** it is corrupted, then a process must have failed while writing it.
2902 ** This indicates nothing more needs to be rolled back.
2904 rc = readJournalHdr(pPager, isHot, szJ, &nRec, &mxPg);
2905 if( rc!=SQLITE_OK ){
2906 if( rc==SQLITE_DONE ){
2907 rc = SQLITE_OK;
2909 goto end_playback;
2912 /* If nRec is 0xffffffff, then this journal was created by a process
2913 ** working in no-sync mode. This means that the rest of the journal
2914 ** file consists of pages, there are no more journal headers. Compute
2915 ** the value of nRec based on this assumption.
2917 if( nRec==0xffffffff ){
2918 assert( pPager->journalOff==JOURNAL_HDR_SZ(pPager) );
2919 nRec = (int)((szJ - JOURNAL_HDR_SZ(pPager))/JOURNAL_PG_SZ(pPager));
2922 /* If nRec is 0 and this rollback is of a transaction created by this
2923 ** process and if this is the final header in the journal, then it means
2924 ** that this part of the journal was being filled but has not yet been
2925 ** synced to disk. Compute the number of pages based on the remaining
2926 ** size of the file.
2928 ** The third term of the test was added to fix ticket #2565.
2929 ** When rolling back a hot journal, nRec==0 always means that the next
2930 ** chunk of the journal contains zero pages to be rolled back. But
2931 ** when doing a ROLLBACK and the nRec==0 chunk is the last chunk in
2932 ** the journal, it means that the journal might contain additional
2933 ** pages that need to be rolled back and that the number of pages
2934 ** should be computed based on the journal file size.
2936 if( nRec==0 && !isHot &&
2937 pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff ){
2938 nRec = (int)((szJ - pPager->journalOff) / JOURNAL_PG_SZ(pPager));
2941 /* If this is the first header read from the journal, truncate the
2942 ** database file back to its original size.
2944 if( pPager->journalOff==JOURNAL_HDR_SZ(pPager) ){
2945 rc = pager_truncate(pPager, mxPg);
2946 if( rc!=SQLITE_OK ){
2947 goto end_playback;
2949 pPager->dbSize = mxPg;
2950 if( pPager->mxPgno<mxPg ){
2951 pPager->mxPgno = mxPg;
2955 /* Copy original pages out of the journal and back into the
2956 ** database file and/or page cache.
2958 for(u=0; u<nRec; u++){
2959 if( needPagerReset ){
2960 pager_reset(pPager);
2961 needPagerReset = 0;
2963 rc = pager_playback_one_page(pPager,&pPager->journalOff,0,1,0);
2964 if( rc==SQLITE_OK ){
2965 nPlayback++;
2966 }else{
2967 if( rc==SQLITE_DONE ){
2968 pPager->journalOff = szJ;
2969 break;
2970 }else if( rc==SQLITE_IOERR_SHORT_READ ){
2971 /* If the journal has been truncated, simply stop reading and
2972 ** processing the journal. This might happen if the journal was
2973 ** not completely written and synced prior to a crash. In that
2974 ** case, the database should have never been written in the
2975 ** first place so it is OK to simply abandon the rollback. */
2976 rc = SQLITE_OK;
2977 goto end_playback;
2978 }else{
2979 /* If we are unable to rollback, quit and return the error
2980 ** code. This will cause the pager to enter the error state
2981 ** so that no further harm will be done. Perhaps the next
2982 ** process to come along will be able to rollback the database.
2984 goto end_playback;
2989 /*NOTREACHED*/
2990 assert( 0 );
2992 end_playback:
2993 if( rc==SQLITE_OK ){
2994 rc = sqlite3PagerSetPagesize(pPager, &savedPageSize, -1);
2996 /* Following a rollback, the database file should be back in its original
2997 ** state prior to the start of the transaction, so invoke the
2998 ** SQLITE_FCNTL_DB_UNCHANGED file-control method to disable the
2999 ** assertion that the transaction counter was modified.
3001 #ifdef SQLITE_DEBUG
3002 sqlite3OsFileControlHint(pPager->fd,SQLITE_FCNTL_DB_UNCHANGED,0);
3003 #endif
3005 /* If this playback is happening automatically as a result of an IO or
3006 ** malloc error that occurred after the change-counter was updated but
3007 ** before the transaction was committed, then the change-counter
3008 ** modification may just have been reverted. If this happens in exclusive
3009 ** mode, then subsequent transactions performed by the connection will not
3010 ** update the change-counter at all. This may lead to cache inconsistency
3011 ** problems for other processes at some point in the future. So, just
3012 ** in case this has happened, clear the changeCountDone flag now.
3014 pPager->changeCountDone = pPager->tempFile;
3016 if( rc==SQLITE_OK ){
3017 /* Leave 4 bytes of space before the super-journal filename in memory.
3018 ** This is because it may end up being passed to sqlite3OsOpen(), in
3019 ** which case it requires 4 0x00 bytes in memory immediately before
3020 ** the filename. */
3021 zSuper = &pPager->pTmpSpace[4];
3022 rc = readSuperJournal(pPager->jfd, zSuper, pPager->pVfs->mxPathname+1);
3023 testcase( rc!=SQLITE_OK );
3025 if( rc==SQLITE_OK
3026 && (pPager->eState>=PAGER_WRITER_DBMOD || pPager->eState==PAGER_OPEN)
3028 rc = sqlite3PagerSync(pPager, 0);
3030 if( rc==SQLITE_OK ){
3031 rc = pager_end_transaction(pPager, zSuper[0]!='\0', 0);
3032 testcase( rc!=SQLITE_OK );
3034 if( rc==SQLITE_OK && zSuper[0] && res ){
3035 /* If there was a super-journal and this routine will return success,
3036 ** see if it is possible to delete the super-journal.
3038 assert( zSuper==&pPager->pTmpSpace[4] );
3039 memset(&zSuper[-4], 0, 4);
3040 rc = pager_delsuper(pPager, zSuper);
3041 testcase( rc!=SQLITE_OK );
3043 if( isHot && nPlayback ){
3044 sqlite3_log(SQLITE_NOTICE_RECOVER_ROLLBACK, "recovered %d pages from %s",
3045 nPlayback, pPager->zJournal);
3048 /* The Pager.sectorSize variable may have been updated while rolling
3049 ** back a journal created by a process with a different sector size
3050 ** value. Reset it to the correct value for this process.
3052 setSectorSize(pPager);
3053 return rc;
3058 ** Read the content for page pPg out of the database file (or out of
3059 ** the WAL if that is where the most recent copy if found) into
3060 ** pPg->pData. A shared lock or greater must be held on the database
3061 ** file before this function is called.
3063 ** If page 1 is read, then the value of Pager.dbFileVers[] is set to
3064 ** the value read from the database file.
3066 ** If an IO error occurs, then the IO error is returned to the caller.
3067 ** Otherwise, SQLITE_OK is returned.
3069 static int readDbPage(PgHdr *pPg){
3070 Pager *pPager = pPg->pPager; /* Pager object associated with page pPg */
3071 int rc = SQLITE_OK; /* Return code */
3073 #ifndef SQLITE_OMIT_WAL
3074 u32 iFrame = 0; /* Frame of WAL containing pgno */
3076 assert( pPager->eState>=PAGER_READER && !MEMDB );
3077 assert( isOpen(pPager->fd) );
3079 if( pagerUseWal(pPager) ){
3080 rc = sqlite3WalFindFrame(pPager->pWal, pPg->pgno, &iFrame);
3081 if( rc ) return rc;
3083 if( iFrame ){
3084 rc = sqlite3WalReadFrame(pPager->pWal, iFrame,pPager->pageSize,pPg->pData);
3085 }else
3086 #endif
3088 i64 iOffset = (pPg->pgno-1)*(i64)pPager->pageSize;
3089 rc = sqlite3OsRead(pPager->fd, pPg->pData, pPager->pageSize, iOffset);
3090 if( rc==SQLITE_IOERR_SHORT_READ ){
3091 rc = SQLITE_OK;
3095 if( pPg->pgno==1 ){
3096 if( rc ){
3097 /* If the read is unsuccessful, set the dbFileVers[] to something
3098 ** that will never be a valid file version. dbFileVers[] is a copy
3099 ** of bytes 24..39 of the database. Bytes 28..31 should always be
3100 ** zero or the size of the database in page. Bytes 32..35 and 35..39
3101 ** should be page numbers which are never 0xffffffff. So filling
3102 ** pPager->dbFileVers[] with all 0xff bytes should suffice.
3104 ** For an encrypted database, the situation is more complex: bytes
3105 ** 24..39 of the database are white noise. But the probability of
3106 ** white noise equaling 16 bytes of 0xff is vanishingly small so
3107 ** we should still be ok.
3109 memset(pPager->dbFileVers, 0xff, sizeof(pPager->dbFileVers));
3110 }else{
3111 u8 *dbFileVers = &((u8*)pPg->pData)[24];
3112 memcpy(&pPager->dbFileVers, dbFileVers, sizeof(pPager->dbFileVers));
3115 CODEC1(pPager, pPg->pData, pPg->pgno, 3, rc = SQLITE_NOMEM_BKPT);
3117 PAGER_INCR(sqlite3_pager_readdb_count);
3118 PAGER_INCR(pPager->nRead);
3119 IOTRACE(("PGIN %p %d\n", pPager, pPg->pgno));
3120 PAGERTRACE(("FETCH %d page %d hash(%08x)\n",
3121 PAGERID(pPager), pPg->pgno, pager_pagehash(pPg)));
3123 return rc;
3127 ** Update the value of the change-counter at offsets 24 and 92 in
3128 ** the header and the sqlite version number at offset 96.
3130 ** This is an unconditional update. See also the pager_incr_changecounter()
3131 ** routine which only updates the change-counter if the update is actually
3132 ** needed, as determined by the pPager->changeCountDone state variable.
3134 static void pager_write_changecounter(PgHdr *pPg){
3135 u32 change_counter;
3136 if( NEVER(pPg==0) ) return;
3138 /* Increment the value just read and write it back to byte 24. */
3139 change_counter = sqlite3Get4byte((u8*)pPg->pPager->dbFileVers)+1;
3140 put32bits(((char*)pPg->pData)+24, change_counter);
3142 /* Also store the SQLite version number in bytes 96..99 and in
3143 ** bytes 92..95 store the change counter for which the version number
3144 ** is valid. */
3145 put32bits(((char*)pPg->pData)+92, change_counter);
3146 put32bits(((char*)pPg->pData)+96, SQLITE_VERSION_NUMBER);
3149 #ifndef SQLITE_OMIT_WAL
3151 ** This function is invoked once for each page that has already been
3152 ** written into the log file when a WAL transaction is rolled back.
3153 ** Parameter iPg is the page number of said page. The pCtx argument
3154 ** is actually a pointer to the Pager structure.
3156 ** If page iPg is present in the cache, and has no outstanding references,
3157 ** it is discarded. Otherwise, if there are one or more outstanding
3158 ** references, the page content is reloaded from the database. If the
3159 ** attempt to reload content from the database is required and fails,
3160 ** return an SQLite error code. Otherwise, SQLITE_OK.
3162 static int pagerUndoCallback(void *pCtx, Pgno iPg){
3163 int rc = SQLITE_OK;
3164 Pager *pPager = (Pager *)pCtx;
3165 PgHdr *pPg;
3167 assert( pagerUseWal(pPager) );
3168 pPg = sqlite3PagerLookup(pPager, iPg);
3169 if( pPg ){
3170 if( sqlite3PcachePageRefcount(pPg)==1 ){
3171 sqlite3PcacheDrop(pPg);
3172 }else{
3173 rc = readDbPage(pPg);
3174 if( rc==SQLITE_OK ){
3175 pPager->xReiniter(pPg);
3177 sqlite3PagerUnrefNotNull(pPg);
3181 /* Normally, if a transaction is rolled back, any backup processes are
3182 ** updated as data is copied out of the rollback journal and into the
3183 ** database. This is not generally possible with a WAL database, as
3184 ** rollback involves simply truncating the log file. Therefore, if one
3185 ** or more frames have already been written to the log (and therefore
3186 ** also copied into the backup databases) as part of this transaction,
3187 ** the backups must be restarted.
3189 sqlite3BackupRestart(pPager->pBackup);
3191 return rc;
3195 ** This function is called to rollback a transaction on a WAL database.
3197 static int pagerRollbackWal(Pager *pPager){
3198 int rc; /* Return Code */
3199 PgHdr *pList; /* List of dirty pages to revert */
3201 /* For all pages in the cache that are currently dirty or have already
3202 ** been written (but not committed) to the log file, do one of the
3203 ** following:
3205 ** + Discard the cached page (if refcount==0), or
3206 ** + Reload page content from the database (if refcount>0).
3208 pPager->dbSize = pPager->dbOrigSize;
3209 rc = sqlite3WalUndo(pPager->pWal, pagerUndoCallback, (void *)pPager);
3210 pList = sqlite3PcacheDirtyList(pPager->pPCache);
3211 while( pList && rc==SQLITE_OK ){
3212 PgHdr *pNext = pList->pDirty;
3213 rc = pagerUndoCallback((void *)pPager, pList->pgno);
3214 pList = pNext;
3217 return rc;
3221 ** This function is a wrapper around sqlite3WalFrames(). As well as logging
3222 ** the contents of the list of pages headed by pList (connected by pDirty),
3223 ** this function notifies any active backup processes that the pages have
3224 ** changed.
3226 ** The list of pages passed into this routine is always sorted by page number.
3227 ** Hence, if page 1 appears anywhere on the list, it will be the first page.
3229 static int pagerWalFrames(
3230 Pager *pPager, /* Pager object */
3231 PgHdr *pList, /* List of frames to log */
3232 Pgno nTruncate, /* Database size after this commit */
3233 int isCommit /* True if this is a commit */
3235 int rc; /* Return code */
3236 int nList; /* Number of pages in pList */
3237 PgHdr *p; /* For looping over pages */
3239 assert( pPager->pWal );
3240 assert( pList );
3241 #ifdef SQLITE_DEBUG
3242 /* Verify that the page list is in accending order */
3243 for(p=pList; p && p->pDirty; p=p->pDirty){
3244 assert( p->pgno < p->pDirty->pgno );
3246 #endif
3248 assert( pList->pDirty==0 || isCommit );
3249 if( isCommit ){
3250 /* If a WAL transaction is being committed, there is no point in writing
3251 ** any pages with page numbers greater than nTruncate into the WAL file.
3252 ** They will never be read by any client. So remove them from the pDirty
3253 ** list here. */
3254 PgHdr **ppNext = &pList;
3255 nList = 0;
3256 for(p=pList; (*ppNext = p)!=0; p=p->pDirty){
3257 if( p->pgno<=nTruncate ){
3258 ppNext = &p->pDirty;
3259 nList++;
3262 assert( pList );
3263 }else{
3264 nList = 1;
3266 pPager->aStat[PAGER_STAT_WRITE] += nList;
3268 if( pList->pgno==1 ) pager_write_changecounter(pList);
3269 rc = sqlite3WalFrames(pPager->pWal,
3270 pPager->pageSize, pList, nTruncate, isCommit, pPager->walSyncFlags
3272 if( rc==SQLITE_OK && pPager->pBackup ){
3273 for(p=pList; p; p=p->pDirty){
3274 sqlite3BackupUpdate(pPager->pBackup, p->pgno, (u8 *)p->pData);
3278 #ifdef SQLITE_CHECK_PAGES
3279 pList = sqlite3PcacheDirtyList(pPager->pPCache);
3280 for(p=pList; p; p=p->pDirty){
3281 pager_set_pagehash(p);
3283 #endif
3285 return rc;
3289 ** Begin a read transaction on the WAL.
3291 ** This routine used to be called "pagerOpenSnapshot()" because it essentially
3292 ** makes a snapshot of the database at the current point in time and preserves
3293 ** that snapshot for use by the reader in spite of concurrently changes by
3294 ** other writers or checkpointers.
3296 static int pagerBeginReadTransaction(Pager *pPager){
3297 int rc; /* Return code */
3298 int changed = 0; /* True if cache must be reset */
3300 assert( pagerUseWal(pPager) );
3301 assert( pPager->eState==PAGER_OPEN || pPager->eState==PAGER_READER );
3303 /* sqlite3WalEndReadTransaction() was not called for the previous
3304 ** transaction in locking_mode=EXCLUSIVE. So call it now. If we
3305 ** are in locking_mode=NORMAL and EndRead() was previously called,
3306 ** the duplicate call is harmless.
3308 sqlite3WalEndReadTransaction(pPager->pWal);
3310 rc = sqlite3WalBeginReadTransaction(pPager->pWal, &changed);
3311 if( rc!=SQLITE_OK || changed ){
3312 pager_reset(pPager);
3313 if( USEFETCH(pPager) ) sqlite3OsUnfetch(pPager->fd, 0, 0);
3316 return rc;
3318 #endif
3321 ** This function is called as part of the transition from PAGER_OPEN
3322 ** to PAGER_READER state to determine the size of the database file
3323 ** in pages (assuming the page size currently stored in Pager.pageSize).
3325 ** If no error occurs, SQLITE_OK is returned and the size of the database
3326 ** in pages is stored in *pnPage. Otherwise, an error code (perhaps
3327 ** SQLITE_IOERR_FSTAT) is returned and *pnPage is left unmodified.
3329 static int pagerPagecount(Pager *pPager, Pgno *pnPage){
3330 Pgno nPage; /* Value to return via *pnPage */
3332 /* Query the WAL sub-system for the database size. The WalDbsize()
3333 ** function returns zero if the WAL is not open (i.e. Pager.pWal==0), or
3334 ** if the database size is not available. The database size is not
3335 ** available from the WAL sub-system if the log file is empty or
3336 ** contains no valid committed transactions.
3338 assert( pPager->eState==PAGER_OPEN );
3339 assert( pPager->eLock>=SHARED_LOCK );
3340 assert( isOpen(pPager->fd) );
3341 assert( pPager->tempFile==0 );
3342 nPage = sqlite3WalDbsize(pPager->pWal);
3344 /* If the number of pages in the database is not available from the
3345 ** WAL sub-system, determine the page count based on the size of
3346 ** the database file. If the size of the database file is not an
3347 ** integer multiple of the page-size, round up the result.
3349 if( nPage==0 && ALWAYS(isOpen(pPager->fd)) ){
3350 i64 n = 0; /* Size of db file in bytes */
3351 int rc = sqlite3OsFileSize(pPager->fd, &n);
3352 if( rc!=SQLITE_OK ){
3353 return rc;
3355 nPage = (Pgno)((n+pPager->pageSize-1) / pPager->pageSize);
3358 /* If the current number of pages in the file is greater than the
3359 ** configured maximum pager number, increase the allowed limit so
3360 ** that the file can be read.
3362 if( nPage>pPager->mxPgno ){
3363 pPager->mxPgno = (Pgno)nPage;
3366 *pnPage = nPage;
3367 return SQLITE_OK;
3370 #ifndef SQLITE_OMIT_WAL
3372 ** Check if the *-wal file that corresponds to the database opened by pPager
3373 ** exists if the database is not empy, or verify that the *-wal file does
3374 ** not exist (by deleting it) if the database file is empty.
3376 ** If the database is not empty and the *-wal file exists, open the pager
3377 ** in WAL mode. If the database is empty or if no *-wal file exists and
3378 ** if no error occurs, make sure Pager.journalMode is not set to
3379 ** PAGER_JOURNALMODE_WAL.
3381 ** Return SQLITE_OK or an error code.
3383 ** The caller must hold a SHARED lock on the database file to call this
3384 ** function. Because an EXCLUSIVE lock on the db file is required to delete
3385 ** a WAL on a none-empty database, this ensures there is no race condition
3386 ** between the xAccess() below and an xDelete() being executed by some
3387 ** other connection.
3389 static int pagerOpenWalIfPresent(Pager *pPager){
3390 int rc = SQLITE_OK;
3391 assert( pPager->eState==PAGER_OPEN );
3392 assert( pPager->eLock>=SHARED_LOCK );
3394 if( !pPager->tempFile ){
3395 int isWal; /* True if WAL file exists */
3396 rc = sqlite3OsAccess(
3397 pPager->pVfs, pPager->zWal, SQLITE_ACCESS_EXISTS, &isWal
3399 if( rc==SQLITE_OK ){
3400 if( isWal ){
3401 Pgno nPage; /* Size of the database file */
3403 rc = pagerPagecount(pPager, &nPage);
3404 if( rc ) return rc;
3405 if( nPage==0 ){
3406 rc = sqlite3OsDelete(pPager->pVfs, pPager->zWal, 0);
3407 }else{
3408 testcase( sqlite3PcachePagecount(pPager->pPCache)==0 );
3409 rc = sqlite3PagerOpenWal(pPager, 0);
3411 }else if( pPager->journalMode==PAGER_JOURNALMODE_WAL ){
3412 pPager->journalMode = PAGER_JOURNALMODE_DELETE;
3416 return rc;
3418 #endif
3421 ** Playback savepoint pSavepoint. Or, if pSavepoint==NULL, then playback
3422 ** the entire super-journal file. The case pSavepoint==NULL occurs when
3423 ** a ROLLBACK TO command is invoked on a SAVEPOINT that is a transaction
3424 ** savepoint.
3426 ** When pSavepoint is not NULL (meaning a non-transaction savepoint is
3427 ** being rolled back), then the rollback consists of up to three stages,
3428 ** performed in the order specified:
3430 ** * Pages are played back from the main journal starting at byte
3431 ** offset PagerSavepoint.iOffset and continuing to
3432 ** PagerSavepoint.iHdrOffset, or to the end of the main journal
3433 ** file if PagerSavepoint.iHdrOffset is zero.
3435 ** * If PagerSavepoint.iHdrOffset is not zero, then pages are played
3436 ** back starting from the journal header immediately following
3437 ** PagerSavepoint.iHdrOffset to the end of the main journal file.
3439 ** * Pages are then played back from the sub-journal file, starting
3440 ** with the PagerSavepoint.iSubRec and continuing to the end of
3441 ** the journal file.
3443 ** Throughout the rollback process, each time a page is rolled back, the
3444 ** corresponding bit is set in a bitvec structure (variable pDone in the
3445 ** implementation below). This is used to ensure that a page is only
3446 ** rolled back the first time it is encountered in either journal.
3448 ** If pSavepoint is NULL, then pages are only played back from the main
3449 ** journal file. There is no need for a bitvec in this case.
3451 ** In either case, before playback commences the Pager.dbSize variable
3452 ** is reset to the value that it held at the start of the savepoint
3453 ** (or transaction). No page with a page-number greater than this value
3454 ** is played back. If one is encountered it is simply skipped.
3456 static int pagerPlaybackSavepoint(Pager *pPager, PagerSavepoint *pSavepoint){
3457 i64 szJ; /* Effective size of the main journal */
3458 i64 iHdrOff; /* End of first segment of main-journal records */
3459 int rc = SQLITE_OK; /* Return code */
3460 Bitvec *pDone = 0; /* Bitvec to ensure pages played back only once */
3462 assert( pPager->eState!=PAGER_ERROR );
3463 assert( pPager->eState>=PAGER_WRITER_LOCKED );
3465 /* Allocate a bitvec to use to store the set of pages rolled back */
3466 if( pSavepoint ){
3467 pDone = sqlite3BitvecCreate(pSavepoint->nOrig);
3468 if( !pDone ){
3469 return SQLITE_NOMEM_BKPT;
3473 /* Set the database size back to the value it was before the savepoint
3474 ** being reverted was opened.
3476 pPager->dbSize = pSavepoint ? pSavepoint->nOrig : pPager->dbOrigSize;
3477 pPager->changeCountDone = pPager->tempFile;
3479 if( !pSavepoint && pagerUseWal(pPager) ){
3480 return pagerRollbackWal(pPager);
3483 /* Use pPager->journalOff as the effective size of the main rollback
3484 ** journal. The actual file might be larger than this in
3485 ** PAGER_JOURNALMODE_TRUNCATE or PAGER_JOURNALMODE_PERSIST. But anything
3486 ** past pPager->journalOff is off-limits to us.
3488 szJ = pPager->journalOff;
3489 assert( pagerUseWal(pPager)==0 || szJ==0 );
3491 /* Begin by rolling back records from the main journal starting at
3492 ** PagerSavepoint.iOffset and continuing to the next journal header.
3493 ** There might be records in the main journal that have a page number
3494 ** greater than the current database size (pPager->dbSize) but those
3495 ** will be skipped automatically. Pages are added to pDone as they
3496 ** are played back.
3498 if( pSavepoint && !pagerUseWal(pPager) ){
3499 iHdrOff = pSavepoint->iHdrOffset ? pSavepoint->iHdrOffset : szJ;
3500 pPager->journalOff = pSavepoint->iOffset;
3501 while( rc==SQLITE_OK && pPager->journalOff<iHdrOff ){
3502 rc = pager_playback_one_page(pPager, &pPager->journalOff, pDone, 1, 1);
3504 assert( rc!=SQLITE_DONE );
3505 }else{
3506 pPager->journalOff = 0;
3509 /* Continue rolling back records out of the main journal starting at
3510 ** the first journal header seen and continuing until the effective end
3511 ** of the main journal file. Continue to skip out-of-range pages and
3512 ** continue adding pages rolled back to pDone.
3514 while( rc==SQLITE_OK && pPager->journalOff<szJ ){
3515 u32 ii; /* Loop counter */
3516 u32 nJRec = 0; /* Number of Journal Records */
3517 u32 dummy;
3518 rc = readJournalHdr(pPager, 0, szJ, &nJRec, &dummy);
3519 assert( rc!=SQLITE_DONE );
3522 ** The "pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff"
3523 ** test is related to ticket #2565. See the discussion in the
3524 ** pager_playback() function for additional information.
3526 if( nJRec==0
3527 && pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff
3529 nJRec = (u32)((szJ - pPager->journalOff)/JOURNAL_PG_SZ(pPager));
3531 for(ii=0; rc==SQLITE_OK && ii<nJRec && pPager->journalOff<szJ; ii++){
3532 rc = pager_playback_one_page(pPager, &pPager->journalOff, pDone, 1, 1);
3534 assert( rc!=SQLITE_DONE );
3536 assert( rc!=SQLITE_OK || pPager->journalOff>=szJ );
3538 /* Finally, rollback pages from the sub-journal. Page that were
3539 ** previously rolled back out of the main journal (and are hence in pDone)
3540 ** will be skipped. Out-of-range pages are also skipped.
3542 if( pSavepoint ){
3543 u32 ii; /* Loop counter */
3544 i64 offset = (i64)pSavepoint->iSubRec*(4+pPager->pageSize);
3546 if( pagerUseWal(pPager) ){
3547 rc = sqlite3WalSavepointUndo(pPager->pWal, pSavepoint->aWalData);
3549 for(ii=pSavepoint->iSubRec; rc==SQLITE_OK && ii<pPager->nSubRec; ii++){
3550 assert( offset==(i64)ii*(4+pPager->pageSize) );
3551 rc = pager_playback_one_page(pPager, &offset, pDone, 0, 1);
3553 assert( rc!=SQLITE_DONE );
3556 sqlite3BitvecDestroy(pDone);
3557 if( rc==SQLITE_OK ){
3558 pPager->journalOff = szJ;
3561 return rc;
3565 ** Change the maximum number of in-memory pages that are allowed
3566 ** before attempting to recycle clean and unused pages.
3568 void sqlite3PagerSetCachesize(Pager *pPager, int mxPage){
3569 sqlite3PcacheSetCachesize(pPager->pPCache, mxPage);
3573 ** Change the maximum number of in-memory pages that are allowed
3574 ** before attempting to spill pages to journal.
3576 int sqlite3PagerSetSpillsize(Pager *pPager, int mxPage){
3577 return sqlite3PcacheSetSpillsize(pPager->pPCache, mxPage);
3581 ** Invoke SQLITE_FCNTL_MMAP_SIZE based on the current value of szMmap.
3583 static void pagerFixMaplimit(Pager *pPager){
3584 #if SQLITE_MAX_MMAP_SIZE>0
3585 sqlite3_file *fd = pPager->fd;
3586 if( isOpen(fd) && fd->pMethods->iVersion>=3 ){
3587 sqlite3_int64 sz;
3588 sz = pPager->szMmap;
3589 pPager->bUseFetch = (sz>0);
3590 setGetterMethod(pPager);
3591 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_MMAP_SIZE, &sz);
3593 #endif
3597 ** Change the maximum size of any memory mapping made of the database file.
3599 void sqlite3PagerSetMmapLimit(Pager *pPager, sqlite3_int64 szMmap){
3600 pPager->szMmap = szMmap;
3601 pagerFixMaplimit(pPager);
3605 ** Free as much memory as possible from the pager.
3607 void sqlite3PagerShrink(Pager *pPager){
3608 sqlite3PcacheShrink(pPager->pPCache);
3612 ** Adjust settings of the pager to those specified in the pgFlags parameter.
3614 ** The "level" in pgFlags & PAGER_SYNCHRONOUS_MASK sets the robustness
3615 ** of the database to damage due to OS crashes or power failures by
3616 ** changing the number of syncs()s when writing the journals.
3617 ** There are four levels:
3619 ** OFF sqlite3OsSync() is never called. This is the default
3620 ** for temporary and transient files.
3622 ** NORMAL The journal is synced once before writes begin on the
3623 ** database. This is normally adequate protection, but
3624 ** it is theoretically possible, though very unlikely,
3625 ** that an inopertune power failure could leave the journal
3626 ** in a state which would cause damage to the database
3627 ** when it is rolled back.
3629 ** FULL The journal is synced twice before writes begin on the
3630 ** database (with some additional information - the nRec field
3631 ** of the journal header - being written in between the two
3632 ** syncs). If we assume that writing a
3633 ** single disk sector is atomic, then this mode provides
3634 ** assurance that the journal will not be corrupted to the
3635 ** point of causing damage to the database during rollback.
3637 ** EXTRA This is like FULL except that is also syncs the directory
3638 ** that contains the rollback journal after the rollback
3639 ** journal is unlinked.
3641 ** The above is for a rollback-journal mode. For WAL mode, OFF continues
3642 ** to mean that no syncs ever occur. NORMAL means that the WAL is synced
3643 ** prior to the start of checkpoint and that the database file is synced
3644 ** at the conclusion of the checkpoint if the entire content of the WAL
3645 ** was written back into the database. But no sync operations occur for
3646 ** an ordinary commit in NORMAL mode with WAL. FULL means that the WAL
3647 ** file is synced following each commit operation, in addition to the
3648 ** syncs associated with NORMAL. There is no difference between FULL
3649 ** and EXTRA for WAL mode.
3651 ** Do not confuse synchronous=FULL with SQLITE_SYNC_FULL. The
3652 ** SQLITE_SYNC_FULL macro means to use the MacOSX-style full-fsync
3653 ** using fcntl(F_FULLFSYNC). SQLITE_SYNC_NORMAL means to do an
3654 ** ordinary fsync() call. There is no difference between SQLITE_SYNC_FULL
3655 ** and SQLITE_SYNC_NORMAL on platforms other than MacOSX. But the
3656 ** synchronous=FULL versus synchronous=NORMAL setting determines when
3657 ** the xSync primitive is called and is relevant to all platforms.
3659 ** Numeric values associated with these states are OFF==1, NORMAL=2,
3660 ** and FULL=3.
3662 #ifndef SQLITE_OMIT_PAGER_PRAGMAS
3663 void sqlite3PagerSetFlags(
3664 Pager *pPager, /* The pager to set safety level for */
3665 unsigned pgFlags /* Various flags */
3667 unsigned level = pgFlags & PAGER_SYNCHRONOUS_MASK;
3668 if( pPager->tempFile ){
3669 pPager->noSync = 1;
3670 pPager->fullSync = 0;
3671 pPager->extraSync = 0;
3672 }else{
3673 pPager->noSync = level==PAGER_SYNCHRONOUS_OFF ?1:0;
3674 pPager->fullSync = level>=PAGER_SYNCHRONOUS_FULL ?1:0;
3675 pPager->extraSync = level==PAGER_SYNCHRONOUS_EXTRA ?1:0;
3677 if( pPager->noSync ){
3678 pPager->syncFlags = 0;
3679 }else if( pgFlags & PAGER_FULLFSYNC ){
3680 pPager->syncFlags = SQLITE_SYNC_FULL;
3681 }else{
3682 pPager->syncFlags = SQLITE_SYNC_NORMAL;
3684 pPager->walSyncFlags = (pPager->syncFlags<<2);
3685 if( pPager->fullSync ){
3686 pPager->walSyncFlags |= pPager->syncFlags;
3688 if( (pgFlags & PAGER_CKPT_FULLFSYNC) && !pPager->noSync ){
3689 pPager->walSyncFlags |= (SQLITE_SYNC_FULL<<2);
3691 if( pgFlags & PAGER_CACHESPILL ){
3692 pPager->doNotSpill &= ~SPILLFLAG_OFF;
3693 }else{
3694 pPager->doNotSpill |= SPILLFLAG_OFF;
3697 #endif
3700 ** The following global variable is incremented whenever the library
3701 ** attempts to open a temporary file. This information is used for
3702 ** testing and analysis only.
3704 #ifdef SQLITE_TEST
3705 int sqlite3_opentemp_count = 0;
3706 #endif
3709 ** Open a temporary file.
3711 ** Write the file descriptor into *pFile. Return SQLITE_OK on success
3712 ** or some other error code if we fail. The OS will automatically
3713 ** delete the temporary file when it is closed.
3715 ** The flags passed to the VFS layer xOpen() call are those specified
3716 ** by parameter vfsFlags ORed with the following:
3718 ** SQLITE_OPEN_READWRITE
3719 ** SQLITE_OPEN_CREATE
3720 ** SQLITE_OPEN_EXCLUSIVE
3721 ** SQLITE_OPEN_DELETEONCLOSE
3723 static int pagerOpentemp(
3724 Pager *pPager, /* The pager object */
3725 sqlite3_file *pFile, /* Write the file descriptor here */
3726 int vfsFlags /* Flags passed through to the VFS */
3728 int rc; /* Return code */
3730 #ifdef SQLITE_TEST
3731 sqlite3_opentemp_count++; /* Used for testing and analysis only */
3732 #endif
3734 vfsFlags |= SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE |
3735 SQLITE_OPEN_EXCLUSIVE | SQLITE_OPEN_DELETEONCLOSE;
3736 rc = sqlite3OsOpen(pPager->pVfs, 0, pFile, vfsFlags, 0);
3737 assert( rc!=SQLITE_OK || isOpen(pFile) );
3738 return rc;
3742 ** Set the busy handler function.
3744 ** The pager invokes the busy-handler if sqlite3OsLock() returns
3745 ** SQLITE_BUSY when trying to upgrade from no-lock to a SHARED lock,
3746 ** or when trying to upgrade from a RESERVED lock to an EXCLUSIVE
3747 ** lock. It does *not* invoke the busy handler when upgrading from
3748 ** SHARED to RESERVED, or when upgrading from SHARED to EXCLUSIVE
3749 ** (which occurs during hot-journal rollback). Summary:
3751 ** Transition | Invokes xBusyHandler
3752 ** --------------------------------------------------------
3753 ** NO_LOCK -> SHARED_LOCK | Yes
3754 ** SHARED_LOCK -> RESERVED_LOCK | No
3755 ** SHARED_LOCK -> EXCLUSIVE_LOCK | No
3756 ** RESERVED_LOCK -> EXCLUSIVE_LOCK | Yes
3758 ** If the busy-handler callback returns non-zero, the lock is
3759 ** retried. If it returns zero, then the SQLITE_BUSY error is
3760 ** returned to the caller of the pager API function.
3762 void sqlite3PagerSetBusyHandler(
3763 Pager *pPager, /* Pager object */
3764 int (*xBusyHandler)(void *), /* Pointer to busy-handler function */
3765 void *pBusyHandlerArg /* Argument to pass to xBusyHandler */
3767 void **ap;
3768 pPager->xBusyHandler = xBusyHandler;
3769 pPager->pBusyHandlerArg = pBusyHandlerArg;
3770 ap = (void **)&pPager->xBusyHandler;
3771 assert( ((int(*)(void *))(ap[0]))==xBusyHandler );
3772 assert( ap[1]==pBusyHandlerArg );
3773 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_BUSYHANDLER, (void *)ap);
3777 ** Change the page size used by the Pager object. The new page size
3778 ** is passed in *pPageSize.
3780 ** If the pager is in the error state when this function is called, it
3781 ** is a no-op. The value returned is the error state error code (i.e.
3782 ** one of SQLITE_IOERR, an SQLITE_IOERR_xxx sub-code or SQLITE_FULL).
3784 ** Otherwise, if all of the following are true:
3786 ** * the new page size (value of *pPageSize) is valid (a power
3787 ** of two between 512 and SQLITE_MAX_PAGE_SIZE, inclusive), and
3789 ** * there are no outstanding page references, and
3791 ** * the database is either not an in-memory database or it is
3792 ** an in-memory database that currently consists of zero pages.
3794 ** then the pager object page size is set to *pPageSize.
3796 ** If the page size is changed, then this function uses sqlite3PagerMalloc()
3797 ** to obtain a new Pager.pTmpSpace buffer. If this allocation attempt
3798 ** fails, SQLITE_NOMEM is returned and the page size remains unchanged.
3799 ** In all other cases, SQLITE_OK is returned.
3801 ** If the page size is not changed, either because one of the enumerated
3802 ** conditions above is not true, the pager was in error state when this
3803 ** function was called, or because the memory allocation attempt failed,
3804 ** then *pPageSize is set to the old, retained page size before returning.
3806 int sqlite3PagerSetPagesize(Pager *pPager, u32 *pPageSize, int nReserve){
3807 int rc = SQLITE_OK;
3809 /* It is not possible to do a full assert_pager_state() here, as this
3810 ** function may be called from within PagerOpen(), before the state
3811 ** of the Pager object is internally consistent.
3813 ** At one point this function returned an error if the pager was in
3814 ** PAGER_ERROR state. But since PAGER_ERROR state guarantees that
3815 ** there is at least one outstanding page reference, this function
3816 ** is a no-op for that case anyhow.
3819 u32 pageSize = *pPageSize;
3820 assert( pageSize==0 || (pageSize>=512 && pageSize<=SQLITE_MAX_PAGE_SIZE) );
3821 if( (pPager->memDb==0 || pPager->dbSize==0)
3822 && sqlite3PcacheRefCount(pPager->pPCache)==0
3823 && pageSize && pageSize!=(u32)pPager->pageSize
3825 char *pNew = NULL; /* New temp space */
3826 i64 nByte = 0;
3828 if( pPager->eState>PAGER_OPEN && isOpen(pPager->fd) ){
3829 rc = sqlite3OsFileSize(pPager->fd, &nByte);
3831 if( rc==SQLITE_OK ){
3832 /* 8 bytes of zeroed overrun space is sufficient so that the b-tree
3833 * cell header parser will never run off the end of the allocation */
3834 pNew = (char *)sqlite3PageMalloc(pageSize+8);
3835 if( !pNew ){
3836 rc = SQLITE_NOMEM_BKPT;
3837 }else{
3838 memset(pNew+pageSize, 0, 8);
3842 if( rc==SQLITE_OK ){
3843 pager_reset(pPager);
3844 rc = sqlite3PcacheSetPageSize(pPager->pPCache, pageSize);
3846 if( rc==SQLITE_OK ){
3847 sqlite3PageFree(pPager->pTmpSpace);
3848 pPager->pTmpSpace = pNew;
3849 pPager->dbSize = (Pgno)((nByte+pageSize-1)/pageSize);
3850 pPager->pageSize = pageSize;
3851 pPager->lckPgno = (Pgno)(PENDING_BYTE/pageSize) + 1;
3852 }else{
3853 sqlite3PageFree(pNew);
3857 *pPageSize = pPager->pageSize;
3858 if( rc==SQLITE_OK ){
3859 if( nReserve<0 ) nReserve = pPager->nReserve;
3860 assert( nReserve>=0 && nReserve<1000 );
3861 pPager->nReserve = (i16)nReserve;
3862 pagerReportSize(pPager);
3863 pagerFixMaplimit(pPager);
3865 return rc;
3869 ** Return a pointer to the "temporary page" buffer held internally
3870 ** by the pager. This is a buffer that is big enough to hold the
3871 ** entire content of a database page. This buffer is used internally
3872 ** during rollback and will be overwritten whenever a rollback
3873 ** occurs. But other modules are free to use it too, as long as
3874 ** no rollbacks are happening.
3876 void *sqlite3PagerTempSpace(Pager *pPager){
3877 return pPager->pTmpSpace;
3881 ** Attempt to set the maximum database page count if mxPage is positive.
3882 ** Make no changes if mxPage is zero or negative. And never reduce the
3883 ** maximum page count below the current size of the database.
3885 ** Regardless of mxPage, return the current maximum page count.
3887 Pgno sqlite3PagerMaxPageCount(Pager *pPager, Pgno mxPage){
3888 if( mxPage>0 ){
3889 pPager->mxPgno = mxPage;
3891 assert( pPager->eState!=PAGER_OPEN ); /* Called only by OP_MaxPgcnt */
3892 /* assert( pPager->mxPgno>=pPager->dbSize ); */
3893 /* OP_MaxPgcnt ensures that the parameter passed to this function is not
3894 ** less than the total number of valid pages in the database. But this
3895 ** may be less than Pager.dbSize, and so the assert() above is not valid */
3896 return pPager->mxPgno;
3900 ** The following set of routines are used to disable the simulated
3901 ** I/O error mechanism. These routines are used to avoid simulated
3902 ** errors in places where we do not care about errors.
3904 ** Unless -DSQLITE_TEST=1 is used, these routines are all no-ops
3905 ** and generate no code.
3907 #ifdef SQLITE_TEST
3908 extern int sqlite3_io_error_pending;
3909 extern int sqlite3_io_error_hit;
3910 static int saved_cnt;
3911 void disable_simulated_io_errors(void){
3912 saved_cnt = sqlite3_io_error_pending;
3913 sqlite3_io_error_pending = -1;
3915 void enable_simulated_io_errors(void){
3916 sqlite3_io_error_pending = saved_cnt;
3918 #else
3919 # define disable_simulated_io_errors()
3920 # define enable_simulated_io_errors()
3921 #endif
3924 ** Read the first N bytes from the beginning of the file into memory
3925 ** that pDest points to.
3927 ** If the pager was opened on a transient file (zFilename==""), or
3928 ** opened on a file less than N bytes in size, the output buffer is
3929 ** zeroed and SQLITE_OK returned. The rationale for this is that this
3930 ** function is used to read database headers, and a new transient or
3931 ** zero sized database has a header than consists entirely of zeroes.
3933 ** If any IO error apart from SQLITE_IOERR_SHORT_READ is encountered,
3934 ** the error code is returned to the caller and the contents of the
3935 ** output buffer undefined.
3937 int sqlite3PagerReadFileheader(Pager *pPager, int N, unsigned char *pDest){
3938 int rc = SQLITE_OK;
3939 memset(pDest, 0, N);
3940 assert( isOpen(pPager->fd) || pPager->tempFile );
3942 /* This routine is only called by btree immediately after creating
3943 ** the Pager object. There has not been an opportunity to transition
3944 ** to WAL mode yet.
3946 assert( !pagerUseWal(pPager) );
3948 if( isOpen(pPager->fd) ){
3949 IOTRACE(("DBHDR %p 0 %d\n", pPager, N))
3950 rc = sqlite3OsRead(pPager->fd, pDest, N, 0);
3951 if( rc==SQLITE_IOERR_SHORT_READ ){
3952 rc = SQLITE_OK;
3955 return rc;
3959 ** This function may only be called when a read-transaction is open on
3960 ** the pager. It returns the total number of pages in the database.
3962 ** However, if the file is between 1 and <page-size> bytes in size, then
3963 ** this is considered a 1 page file.
3965 void sqlite3PagerPagecount(Pager *pPager, int *pnPage){
3966 assert( pPager->eState>=PAGER_READER );
3967 assert( pPager->eState!=PAGER_WRITER_FINISHED );
3968 *pnPage = (int)pPager->dbSize;
3973 ** Try to obtain a lock of type locktype on the database file. If
3974 ** a similar or greater lock is already held, this function is a no-op
3975 ** (returning SQLITE_OK immediately).
3977 ** Otherwise, attempt to obtain the lock using sqlite3OsLock(). Invoke
3978 ** the busy callback if the lock is currently not available. Repeat
3979 ** until the busy callback returns false or until the attempt to
3980 ** obtain the lock succeeds.
3982 ** Return SQLITE_OK on success and an error code if we cannot obtain
3983 ** the lock. If the lock is obtained successfully, set the Pager.state
3984 ** variable to locktype before returning.
3986 static int pager_wait_on_lock(Pager *pPager, int locktype){
3987 int rc; /* Return code */
3989 /* Check that this is either a no-op (because the requested lock is
3990 ** already held), or one of the transitions that the busy-handler
3991 ** may be invoked during, according to the comment above
3992 ** sqlite3PagerSetBusyhandler().
3994 assert( (pPager->eLock>=locktype)
3995 || (pPager->eLock==NO_LOCK && locktype==SHARED_LOCK)
3996 || (pPager->eLock==RESERVED_LOCK && locktype==EXCLUSIVE_LOCK)
3999 do {
4000 rc = pagerLockDb(pPager, locktype);
4001 }while( rc==SQLITE_BUSY && pPager->xBusyHandler(pPager->pBusyHandlerArg) );
4002 return rc;
4006 ** Function assertTruncateConstraint(pPager) checks that one of the
4007 ** following is true for all dirty pages currently in the page-cache:
4009 ** a) The page number is less than or equal to the size of the
4010 ** current database image, in pages, OR
4012 ** b) if the page content were written at this time, it would not
4013 ** be necessary to write the current content out to the sub-journal.
4015 ** If the condition asserted by this function were not true, and the
4016 ** dirty page were to be discarded from the cache via the pagerStress()
4017 ** routine, pagerStress() would not write the current page content to
4018 ** the database file. If a savepoint transaction were rolled back after
4019 ** this happened, the correct behavior would be to restore the current
4020 ** content of the page. However, since this content is not present in either
4021 ** the database file or the portion of the rollback journal and
4022 ** sub-journal rolled back the content could not be restored and the
4023 ** database image would become corrupt. It is therefore fortunate that
4024 ** this circumstance cannot arise.
4026 #if defined(SQLITE_DEBUG)
4027 static void assertTruncateConstraintCb(PgHdr *pPg){
4028 Pager *pPager = pPg->pPager;
4029 assert( pPg->flags&PGHDR_DIRTY );
4030 if( pPg->pgno>pPager->dbSize ){ /* if (a) is false */
4031 Pgno pgno = pPg->pgno;
4032 int i;
4033 for(i=0; i<pPg->pPager->nSavepoint; i++){
4034 PagerSavepoint *p = &pPager->aSavepoint[i];
4035 assert( p->nOrig<pgno || sqlite3BitvecTestNotNull(p->pInSavepoint,pgno) );
4039 static void assertTruncateConstraint(Pager *pPager){
4040 sqlite3PcacheIterateDirty(pPager->pPCache, assertTruncateConstraintCb);
4042 #else
4043 # define assertTruncateConstraint(pPager)
4044 #endif
4047 ** Truncate the in-memory database file image to nPage pages. This
4048 ** function does not actually modify the database file on disk. It
4049 ** just sets the internal state of the pager object so that the
4050 ** truncation will be done when the current transaction is committed.
4052 ** This function is only called right before committing a transaction.
4053 ** Once this function has been called, the transaction must either be
4054 ** rolled back or committed. It is not safe to call this function and
4055 ** then continue writing to the database.
4057 void sqlite3PagerTruncateImage(Pager *pPager, Pgno nPage){
4058 assert( pPager->dbSize>=nPage || CORRUPT_DB );
4059 assert( pPager->eState>=PAGER_WRITER_CACHEMOD );
4060 pPager->dbSize = nPage;
4062 /* At one point the code here called assertTruncateConstraint() to
4063 ** ensure that all pages being truncated away by this operation are,
4064 ** if one or more savepoints are open, present in the savepoint
4065 ** journal so that they can be restored if the savepoint is rolled
4066 ** back. This is no longer necessary as this function is now only
4067 ** called right before committing a transaction. So although the
4068 ** Pager object may still have open savepoints (Pager.nSavepoint!=0),
4069 ** they cannot be rolled back. So the assertTruncateConstraint() call
4070 ** is no longer correct. */
4075 ** This function is called before attempting a hot-journal rollback. It
4076 ** syncs the journal file to disk, then sets pPager->journalHdr to the
4077 ** size of the journal file so that the pager_playback() routine knows
4078 ** that the entire journal file has been synced.
4080 ** Syncing a hot-journal to disk before attempting to roll it back ensures
4081 ** that if a power-failure occurs during the rollback, the process that
4082 ** attempts rollback following system recovery sees the same journal
4083 ** content as this process.
4085 ** If everything goes as planned, SQLITE_OK is returned. Otherwise,
4086 ** an SQLite error code.
4088 static int pagerSyncHotJournal(Pager *pPager){
4089 int rc = SQLITE_OK;
4090 if( !pPager->noSync ){
4091 rc = sqlite3OsSync(pPager->jfd, SQLITE_SYNC_NORMAL);
4093 if( rc==SQLITE_OK ){
4094 rc = sqlite3OsFileSize(pPager->jfd, &pPager->journalHdr);
4096 return rc;
4099 #if SQLITE_MAX_MMAP_SIZE>0
4101 ** Obtain a reference to a memory mapped page object for page number pgno.
4102 ** The new object will use the pointer pData, obtained from xFetch().
4103 ** If successful, set *ppPage to point to the new page reference
4104 ** and return SQLITE_OK. Otherwise, return an SQLite error code and set
4105 ** *ppPage to zero.
4107 ** Page references obtained by calling this function should be released
4108 ** by calling pagerReleaseMapPage().
4110 static int pagerAcquireMapPage(
4111 Pager *pPager, /* Pager object */
4112 Pgno pgno, /* Page number */
4113 void *pData, /* xFetch()'d data for this page */
4114 PgHdr **ppPage /* OUT: Acquired page object */
4116 PgHdr *p; /* Memory mapped page to return */
4118 if( pPager->pMmapFreelist ){
4119 *ppPage = p = pPager->pMmapFreelist;
4120 pPager->pMmapFreelist = p->pDirty;
4121 p->pDirty = 0;
4122 assert( pPager->nExtra>=8 );
4123 memset(p->pExtra, 0, 8);
4124 }else{
4125 *ppPage = p = (PgHdr *)sqlite3MallocZero(sizeof(PgHdr) + pPager->nExtra);
4126 if( p==0 ){
4127 sqlite3OsUnfetch(pPager->fd, (i64)(pgno-1) * pPager->pageSize, pData);
4128 return SQLITE_NOMEM_BKPT;
4130 p->pExtra = (void *)&p[1];
4131 p->flags = PGHDR_MMAP;
4132 p->nRef = 1;
4133 p->pPager = pPager;
4136 assert( p->pExtra==(void *)&p[1] );
4137 assert( p->pPage==0 );
4138 assert( p->flags==PGHDR_MMAP );
4139 assert( p->pPager==pPager );
4140 assert( p->nRef==1 );
4142 p->pgno = pgno;
4143 p->pData = pData;
4144 pPager->nMmapOut++;
4146 return SQLITE_OK;
4148 #endif
4151 ** Release a reference to page pPg. pPg must have been returned by an
4152 ** earlier call to pagerAcquireMapPage().
4154 static void pagerReleaseMapPage(PgHdr *pPg){
4155 Pager *pPager = pPg->pPager;
4156 pPager->nMmapOut--;
4157 pPg->pDirty = pPager->pMmapFreelist;
4158 pPager->pMmapFreelist = pPg;
4160 assert( pPager->fd->pMethods->iVersion>=3 );
4161 sqlite3OsUnfetch(pPager->fd, (i64)(pPg->pgno-1)*pPager->pageSize, pPg->pData);
4165 ** Free all PgHdr objects stored in the Pager.pMmapFreelist list.
4167 static void pagerFreeMapHdrs(Pager *pPager){
4168 PgHdr *p;
4169 PgHdr *pNext;
4170 for(p=pPager->pMmapFreelist; p; p=pNext){
4171 pNext = p->pDirty;
4172 sqlite3_free(p);
4176 /* Verify that the database file has not be deleted or renamed out from
4177 ** under the pager. Return SQLITE_OK if the database is still where it ought
4178 ** to be on disk. Return non-zero (SQLITE_READONLY_DBMOVED or some other error
4179 ** code from sqlite3OsAccess()) if the database has gone missing.
4181 static int databaseIsUnmoved(Pager *pPager){
4182 int bHasMoved = 0;
4183 int rc;
4185 if( pPager->tempFile ) return SQLITE_OK;
4186 if( pPager->dbSize==0 ) return SQLITE_OK;
4187 assert( pPager->zFilename && pPager->zFilename[0] );
4188 rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_HAS_MOVED, &bHasMoved);
4189 if( rc==SQLITE_NOTFOUND ){
4190 /* If the HAS_MOVED file-control is unimplemented, assume that the file
4191 ** has not been moved. That is the historical behavior of SQLite: prior to
4192 ** version 3.8.3, it never checked */
4193 rc = SQLITE_OK;
4194 }else if( rc==SQLITE_OK && bHasMoved ){
4195 rc = SQLITE_READONLY_DBMOVED;
4197 return rc;
4202 ** Shutdown the page cache. Free all memory and close all files.
4204 ** If a transaction was in progress when this routine is called, that
4205 ** transaction is rolled back. All outstanding pages are invalidated
4206 ** and their memory is freed. Any attempt to use a page associated
4207 ** with this page cache after this function returns will likely
4208 ** result in a coredump.
4210 ** This function always succeeds. If a transaction is active an attempt
4211 ** is made to roll it back. If an error occurs during the rollback
4212 ** a hot journal may be left in the filesystem but no error is returned
4213 ** to the caller.
4215 int sqlite3PagerClose(Pager *pPager, sqlite3 *db){
4216 u8 *pTmp = (u8*)pPager->pTmpSpace;
4217 assert( db || pagerUseWal(pPager)==0 );
4218 assert( assert_pager_state(pPager) );
4219 disable_simulated_io_errors();
4220 sqlite3BeginBenignMalloc();
4221 pagerFreeMapHdrs(pPager);
4222 /* pPager->errCode = 0; */
4223 pPager->exclusiveMode = 0;
4224 #ifndef SQLITE_OMIT_WAL
4226 u8 *a = 0;
4227 assert( db || pPager->pWal==0 );
4228 if( db && 0==(db->flags & SQLITE_NoCkptOnClose)
4229 && SQLITE_OK==databaseIsUnmoved(pPager)
4231 a = pTmp;
4233 sqlite3WalClose(pPager->pWal, db, pPager->walSyncFlags, pPager->pageSize,a);
4234 pPager->pWal = 0;
4236 #endif
4237 pager_reset(pPager);
4238 if( MEMDB ){
4239 pager_unlock(pPager);
4240 }else{
4241 /* If it is open, sync the journal file before calling UnlockAndRollback.
4242 ** If this is not done, then an unsynced portion of the open journal
4243 ** file may be played back into the database. If a power failure occurs
4244 ** while this is happening, the database could become corrupt.
4246 ** If an error occurs while trying to sync the journal, shift the pager
4247 ** into the ERROR state. This causes UnlockAndRollback to unlock the
4248 ** database and close the journal file without attempting to roll it
4249 ** back or finalize it. The next database user will have to do hot-journal
4250 ** rollback before accessing the database file.
4252 if( isOpen(pPager->jfd) ){
4253 pager_error(pPager, pagerSyncHotJournal(pPager));
4255 pagerUnlockAndRollback(pPager);
4257 sqlite3EndBenignMalloc();
4258 enable_simulated_io_errors();
4259 PAGERTRACE(("CLOSE %d\n", PAGERID(pPager)));
4260 IOTRACE(("CLOSE %p\n", pPager))
4261 sqlite3OsClose(pPager->jfd);
4262 sqlite3OsClose(pPager->fd);
4263 sqlite3PageFree(pTmp);
4264 sqlite3PcacheClose(pPager->pPCache);
4266 /* BEGIN SQLCIPHER */
4267 #ifdef SQLITE_HAS_CODEC
4268 if( pPager->xCodecFree ) pPager->xCodecFree(pPager->pCodec);
4269 #endif
4270 /* END SQLCIPHER */
4272 assert( !pPager->aSavepoint && !pPager->pInJournal );
4273 assert( !isOpen(pPager->jfd) && !isOpen(pPager->sjfd) );
4275 sqlite3_free(pPager);
4276 return SQLITE_OK;
4279 #if !defined(NDEBUG) || defined(SQLITE_TEST)
4281 ** Return the page number for page pPg.
4283 Pgno sqlite3PagerPagenumber(DbPage *pPg){
4284 return pPg->pgno;
4286 #endif
4289 ** Increment the reference count for page pPg.
4291 void sqlite3PagerRef(DbPage *pPg){
4292 sqlite3PcacheRef(pPg);
4296 ** Sync the journal. In other words, make sure all the pages that have
4297 ** been written to the journal have actually reached the surface of the
4298 ** disk and can be restored in the event of a hot-journal rollback.
4300 ** If the Pager.noSync flag is set, then this function is a no-op.
4301 ** Otherwise, the actions required depend on the journal-mode and the
4302 ** device characteristics of the file-system, as follows:
4304 ** * If the journal file is an in-memory journal file, no action need
4305 ** be taken.
4307 ** * Otherwise, if the device does not support the SAFE_APPEND property,
4308 ** then the nRec field of the most recently written journal header
4309 ** is updated to contain the number of journal records that have
4310 ** been written following it. If the pager is operating in full-sync
4311 ** mode, then the journal file is synced before this field is updated.
4313 ** * If the device does not support the SEQUENTIAL property, then
4314 ** journal file is synced.
4316 ** Or, in pseudo-code:
4318 ** if( NOT <in-memory journal> ){
4319 ** if( NOT SAFE_APPEND ){
4320 ** if( <full-sync mode> ) xSync(<journal file>);
4321 ** <update nRec field>
4322 ** }
4323 ** if( NOT SEQUENTIAL ) xSync(<journal file>);
4324 ** }
4326 ** If successful, this routine clears the PGHDR_NEED_SYNC flag of every
4327 ** page currently held in memory before returning SQLITE_OK. If an IO
4328 ** error is encountered, then the IO error code is returned to the caller.
4330 static int syncJournal(Pager *pPager, int newHdr){
4331 int rc; /* Return code */
4333 assert( pPager->eState==PAGER_WRITER_CACHEMOD
4334 || pPager->eState==PAGER_WRITER_DBMOD
4336 assert( assert_pager_state(pPager) );
4337 assert( !pagerUseWal(pPager) );
4339 rc = sqlite3PagerExclusiveLock(pPager);
4340 if( rc!=SQLITE_OK ) return rc;
4342 if( !pPager->noSync ){
4343 assert( !pPager->tempFile );
4344 if( isOpen(pPager->jfd) && pPager->journalMode!=PAGER_JOURNALMODE_MEMORY ){
4345 const int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
4346 assert( isOpen(pPager->jfd) );
4348 if( 0==(iDc&SQLITE_IOCAP_SAFE_APPEND) ){
4349 /* This block deals with an obscure problem. If the last connection
4350 ** that wrote to this database was operating in persistent-journal
4351 ** mode, then the journal file may at this point actually be larger
4352 ** than Pager.journalOff bytes. If the next thing in the journal
4353 ** file happens to be a journal-header (written as part of the
4354 ** previous connection's transaction), and a crash or power-failure
4355 ** occurs after nRec is updated but before this connection writes
4356 ** anything else to the journal file (or commits/rolls back its
4357 ** transaction), then SQLite may become confused when doing the
4358 ** hot-journal rollback following recovery. It may roll back all
4359 ** of this connections data, then proceed to rolling back the old,
4360 ** out-of-date data that follows it. Database corruption.
4362 ** To work around this, if the journal file does appear to contain
4363 ** a valid header following Pager.journalOff, then write a 0x00
4364 ** byte to the start of it to prevent it from being recognized.
4366 ** Variable iNextHdrOffset is set to the offset at which this
4367 ** problematic header will occur, if it exists. aMagic is used
4368 ** as a temporary buffer to inspect the first couple of bytes of
4369 ** the potential journal header.
4371 i64 iNextHdrOffset;
4372 u8 aMagic[8];
4373 u8 zHeader[sizeof(aJournalMagic)+4];
4375 memcpy(zHeader, aJournalMagic, sizeof(aJournalMagic));
4376 put32bits(&zHeader[sizeof(aJournalMagic)], pPager->nRec);
4378 iNextHdrOffset = journalHdrOffset(pPager);
4379 rc = sqlite3OsRead(pPager->jfd, aMagic, 8, iNextHdrOffset);
4380 if( rc==SQLITE_OK && 0==memcmp(aMagic, aJournalMagic, 8) ){
4381 static const u8 zerobyte = 0;
4382 rc = sqlite3OsWrite(pPager->jfd, &zerobyte, 1, iNextHdrOffset);
4384 if( rc!=SQLITE_OK && rc!=SQLITE_IOERR_SHORT_READ ){
4385 return rc;
4388 /* Write the nRec value into the journal file header. If in
4389 ** full-synchronous mode, sync the journal first. This ensures that
4390 ** all data has really hit the disk before nRec is updated to mark
4391 ** it as a candidate for rollback.
4393 ** This is not required if the persistent media supports the
4394 ** SAFE_APPEND property. Because in this case it is not possible
4395 ** for garbage data to be appended to the file, the nRec field
4396 ** is populated with 0xFFFFFFFF when the journal header is written
4397 ** and never needs to be updated.
4399 if( pPager->fullSync && 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
4400 PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager)));
4401 IOTRACE(("JSYNC %p\n", pPager))
4402 rc = sqlite3OsSync(pPager->jfd, pPager->syncFlags);
4403 if( rc!=SQLITE_OK ) return rc;
4405 IOTRACE(("JHDR %p %lld\n", pPager, pPager->journalHdr));
4406 rc = sqlite3OsWrite(
4407 pPager->jfd, zHeader, sizeof(zHeader), pPager->journalHdr
4409 if( rc!=SQLITE_OK ) return rc;
4411 if( 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
4412 PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager)));
4413 IOTRACE(("JSYNC %p\n", pPager))
4414 rc = sqlite3OsSync(pPager->jfd, pPager->syncFlags|
4415 (pPager->syncFlags==SQLITE_SYNC_FULL?SQLITE_SYNC_DATAONLY:0)
4417 if( rc!=SQLITE_OK ) return rc;
4420 pPager->journalHdr = pPager->journalOff;
4421 if( newHdr && 0==(iDc&SQLITE_IOCAP_SAFE_APPEND) ){
4422 pPager->nRec = 0;
4423 rc = writeJournalHdr(pPager);
4424 if( rc!=SQLITE_OK ) return rc;
4426 }else{
4427 pPager->journalHdr = pPager->journalOff;
4431 /* Unless the pager is in noSync mode, the journal file was just
4432 ** successfully synced. Either way, clear the PGHDR_NEED_SYNC flag on
4433 ** all pages.
4435 sqlite3PcacheClearSyncFlags(pPager->pPCache);
4436 pPager->eState = PAGER_WRITER_DBMOD;
4437 assert( assert_pager_state(pPager) );
4438 return SQLITE_OK;
4442 ** The argument is the first in a linked list of dirty pages connected
4443 ** by the PgHdr.pDirty pointer. This function writes each one of the
4444 ** in-memory pages in the list to the database file. The argument may
4445 ** be NULL, representing an empty list. In this case this function is
4446 ** a no-op.
4448 ** The pager must hold at least a RESERVED lock when this function
4449 ** is called. Before writing anything to the database file, this lock
4450 ** is upgraded to an EXCLUSIVE lock. If the lock cannot be obtained,
4451 ** SQLITE_BUSY is returned and no data is written to the database file.
4453 ** If the pager is a temp-file pager and the actual file-system file
4454 ** is not yet open, it is created and opened before any data is
4455 ** written out.
4457 ** Once the lock has been upgraded and, if necessary, the file opened,
4458 ** the pages are written out to the database file in list order. Writing
4459 ** a page is skipped if it meets either of the following criteria:
4461 ** * The page number is greater than Pager.dbSize, or
4462 ** * The PGHDR_DONT_WRITE flag is set on the page.
4464 ** If writing out a page causes the database file to grow, Pager.dbFileSize
4465 ** is updated accordingly. If page 1 is written out, then the value cached
4466 ** in Pager.dbFileVers[] is updated to match the new value stored in
4467 ** the database file.
4469 ** If everything is successful, SQLITE_OK is returned. If an IO error
4470 ** occurs, an IO error code is returned. Or, if the EXCLUSIVE lock cannot
4471 ** be obtained, SQLITE_BUSY is returned.
4473 static int pager_write_pagelist(Pager *pPager, PgHdr *pList){
4474 int rc = SQLITE_OK; /* Return code */
4476 /* This function is only called for rollback pagers in WRITER_DBMOD state. */
4477 assert( !pagerUseWal(pPager) );
4478 assert( pPager->tempFile || pPager->eState==PAGER_WRITER_DBMOD );
4479 assert( pPager->eLock==EXCLUSIVE_LOCK );
4480 assert( isOpen(pPager->fd) || pList->pDirty==0 );
4482 /* If the file is a temp-file has not yet been opened, open it now. It
4483 ** is not possible for rc to be other than SQLITE_OK if this branch
4484 ** is taken, as pager_wait_on_lock() is a no-op for temp-files.
4486 if( !isOpen(pPager->fd) ){
4487 assert( pPager->tempFile && rc==SQLITE_OK );
4488 rc = pagerOpentemp(pPager, pPager->fd, pPager->vfsFlags);
4491 /* Before the first write, give the VFS a hint of what the final
4492 ** file size will be.
4494 assert( rc!=SQLITE_OK || isOpen(pPager->fd) );
4495 if( rc==SQLITE_OK
4496 && pPager->dbHintSize<pPager->dbSize
4497 && (pList->pDirty || pList->pgno>pPager->dbHintSize)
4499 sqlite3_int64 szFile = pPager->pageSize * (sqlite3_int64)pPager->dbSize;
4500 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &szFile);
4501 pPager->dbHintSize = pPager->dbSize;
4504 while( rc==SQLITE_OK && pList ){
4505 Pgno pgno = pList->pgno;
4507 /* If there are dirty pages in the page cache with page numbers greater
4508 ** than Pager.dbSize, this means sqlite3PagerTruncateImage() was called to
4509 ** make the file smaller (presumably by auto-vacuum code). Do not write
4510 ** any such pages to the file.
4512 ** Also, do not write out any page that has the PGHDR_DONT_WRITE flag
4513 ** set (set by sqlite3PagerDontWrite()).
4515 if( pgno<=pPager->dbSize && 0==(pList->flags&PGHDR_DONT_WRITE) ){
4516 i64 offset = (pgno-1)*(i64)pPager->pageSize; /* Offset to write */
4517 char *pData; /* Data to write */
4519 assert( (pList->flags&PGHDR_NEED_SYNC)==0 );
4520 if( pList->pgno==1 ) pager_write_changecounter(pList);
4522 /* Encode the database */
4523 CODEC2(pPager, pList->pData, pgno, 6, return SQLITE_NOMEM_BKPT, pData);
4525 /* Write out the page data. */
4526 rc = sqlite3OsWrite(pPager->fd, pData, pPager->pageSize, offset);
4528 /* If page 1 was just written, update Pager.dbFileVers to match
4529 ** the value now stored in the database file. If writing this
4530 ** page caused the database file to grow, update dbFileSize.
4532 if( pgno==1 ){
4533 memcpy(&pPager->dbFileVers, &pData[24], sizeof(pPager->dbFileVers));
4535 if( pgno>pPager->dbFileSize ){
4536 pPager->dbFileSize = pgno;
4538 pPager->aStat[PAGER_STAT_WRITE]++;
4540 /* Update any backup objects copying the contents of this pager. */
4541 sqlite3BackupUpdate(pPager->pBackup, pgno, (u8*)pList->pData);
4543 PAGERTRACE(("STORE %d page %d hash(%08x)\n",
4544 PAGERID(pPager), pgno, pager_pagehash(pList)));
4545 IOTRACE(("PGOUT %p %d\n", pPager, pgno));
4546 PAGER_INCR(sqlite3_pager_writedb_count);
4547 }else{
4548 PAGERTRACE(("NOSTORE %d page %d\n", PAGERID(pPager), pgno));
4550 pager_set_pagehash(pList);
4551 pList = pList->pDirty;
4554 return rc;
4558 ** Ensure that the sub-journal file is open. If it is already open, this
4559 ** function is a no-op.
4561 ** SQLITE_OK is returned if everything goes according to plan. An
4562 ** SQLITE_IOERR_XXX error code is returned if a call to sqlite3OsOpen()
4563 ** fails.
4565 static int openSubJournal(Pager *pPager){
4566 int rc = SQLITE_OK;
4567 if( !isOpen(pPager->sjfd) ){
4568 const int flags = SQLITE_OPEN_SUBJOURNAL | SQLITE_OPEN_READWRITE
4569 | SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE
4570 | SQLITE_OPEN_DELETEONCLOSE;
4571 int nStmtSpill = sqlite3Config.nStmtSpill;
4572 if( pPager->journalMode==PAGER_JOURNALMODE_MEMORY || pPager->subjInMemory ){
4573 nStmtSpill = -1;
4575 rc = sqlite3JournalOpen(pPager->pVfs, 0, pPager->sjfd, flags, nStmtSpill);
4577 return rc;
4581 ** Append a record of the current state of page pPg to the sub-journal.
4583 ** If successful, set the bit corresponding to pPg->pgno in the bitvecs
4584 ** for all open savepoints before returning.
4586 ** This function returns SQLITE_OK if everything is successful, an IO
4587 ** error code if the attempt to write to the sub-journal fails, or
4588 ** SQLITE_NOMEM if a malloc fails while setting a bit in a savepoint
4589 ** bitvec.
4591 static int subjournalPage(PgHdr *pPg){
4592 int rc = SQLITE_OK;
4593 Pager *pPager = pPg->pPager;
4594 if( pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
4596 /* Open the sub-journal, if it has not already been opened */
4597 assert( pPager->useJournal );
4598 assert( isOpen(pPager->jfd) || pagerUseWal(pPager) );
4599 assert( isOpen(pPager->sjfd) || pPager->nSubRec==0 );
4600 assert( pagerUseWal(pPager)
4601 || pageInJournal(pPager, pPg)
4602 || pPg->pgno>pPager->dbOrigSize
4604 rc = openSubJournal(pPager);
4606 /* If the sub-journal was opened successfully (or was already open),
4607 ** write the journal record into the file. */
4608 if( rc==SQLITE_OK ){
4609 void *pData = pPg->pData;
4610 i64 offset = (i64)pPager->nSubRec*(4+pPager->pageSize);
4611 char *pData2;
4613 /* BEGIN SQLCIPHER */
4614 #if SQLITE_HAS_CODEC
4615 if( !pPager->subjInMemory ){
4616 CODEC2(pPager, pData, pPg->pgno, 7, return SQLITE_NOMEM_BKPT, pData2);
4617 }else
4618 #endif
4619 /* END SQLCIPHER */
4620 pData2 = pData;
4621 PAGERTRACE(("STMT-JOURNAL %d page %d\n", PAGERID(pPager), pPg->pgno));
4622 rc = write32bits(pPager->sjfd, offset, pPg->pgno);
4623 if( rc==SQLITE_OK ){
4624 rc = sqlite3OsWrite(pPager->sjfd, pData2, pPager->pageSize, offset+4);
4628 if( rc==SQLITE_OK ){
4629 pPager->nSubRec++;
4630 assert( pPager->nSavepoint>0 );
4631 rc = addToSavepointBitvecs(pPager, pPg->pgno);
4633 return rc;
4635 static int subjournalPageIfRequired(PgHdr *pPg){
4636 if( subjRequiresPage(pPg) ){
4637 return subjournalPage(pPg);
4638 }else{
4639 return SQLITE_OK;
4644 ** This function is called by the pcache layer when it has reached some
4645 ** soft memory limit. The first argument is a pointer to a Pager object
4646 ** (cast as a void*). The pager is always 'purgeable' (not an in-memory
4647 ** database). The second argument is a reference to a page that is
4648 ** currently dirty but has no outstanding references. The page
4649 ** is always associated with the Pager object passed as the first
4650 ** argument.
4652 ** The job of this function is to make pPg clean by writing its contents
4653 ** out to the database file, if possible. This may involve syncing the
4654 ** journal file.
4656 ** If successful, sqlite3PcacheMakeClean() is called on the page and
4657 ** SQLITE_OK returned. If an IO error occurs while trying to make the
4658 ** page clean, the IO error code is returned. If the page cannot be
4659 ** made clean for some other reason, but no error occurs, then SQLITE_OK
4660 ** is returned by sqlite3PcacheMakeClean() is not called.
4662 static int pagerStress(void *p, PgHdr *pPg){
4663 Pager *pPager = (Pager *)p;
4664 int rc = SQLITE_OK;
4666 assert( pPg->pPager==pPager );
4667 assert( pPg->flags&PGHDR_DIRTY );
4669 /* The doNotSpill NOSYNC bit is set during times when doing a sync of
4670 ** journal (and adding a new header) is not allowed. This occurs
4671 ** during calls to sqlite3PagerWrite() while trying to journal multiple
4672 ** pages belonging to the same sector.
4674 ** The doNotSpill ROLLBACK and OFF bits inhibits all cache spilling
4675 ** regardless of whether or not a sync is required. This is set during
4676 ** a rollback or by user request, respectively.
4678 ** Spilling is also prohibited when in an error state since that could
4679 ** lead to database corruption. In the current implementation it
4680 ** is impossible for sqlite3PcacheFetch() to be called with createFlag==3
4681 ** while in the error state, hence it is impossible for this routine to
4682 ** be called in the error state. Nevertheless, we include a NEVER()
4683 ** test for the error state as a safeguard against future changes.
4685 if( NEVER(pPager->errCode) ) return SQLITE_OK;
4686 testcase( pPager->doNotSpill & SPILLFLAG_ROLLBACK );
4687 testcase( pPager->doNotSpill & SPILLFLAG_OFF );
4688 testcase( pPager->doNotSpill & SPILLFLAG_NOSYNC );
4689 if( pPager->doNotSpill
4690 && ((pPager->doNotSpill & (SPILLFLAG_ROLLBACK|SPILLFLAG_OFF))!=0
4691 || (pPg->flags & PGHDR_NEED_SYNC)!=0)
4693 return SQLITE_OK;
4696 pPager->aStat[PAGER_STAT_SPILL]++;
4697 pPg->pDirty = 0;
4698 if( pagerUseWal(pPager) ){
4699 /* Write a single frame for this page to the log. */
4700 rc = subjournalPageIfRequired(pPg);
4701 if( rc==SQLITE_OK ){
4702 rc = pagerWalFrames(pPager, pPg, 0, 0);
4704 }else{
4706 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
4707 if( pPager->tempFile==0 ){
4708 rc = sqlite3JournalCreate(pPager->jfd);
4709 if( rc!=SQLITE_OK ) return pager_error(pPager, rc);
4711 #endif
4713 /* Sync the journal file if required. */
4714 if( pPg->flags&PGHDR_NEED_SYNC
4715 || pPager->eState==PAGER_WRITER_CACHEMOD
4717 rc = syncJournal(pPager, 1);
4720 /* Write the contents of the page out to the database file. */
4721 if( rc==SQLITE_OK ){
4722 assert( (pPg->flags&PGHDR_NEED_SYNC)==0 );
4723 rc = pager_write_pagelist(pPager, pPg);
4727 /* Mark the page as clean. */
4728 if( rc==SQLITE_OK ){
4729 PAGERTRACE(("STRESS %d page %d\n", PAGERID(pPager), pPg->pgno));
4730 sqlite3PcacheMakeClean(pPg);
4733 return pager_error(pPager, rc);
4737 ** Flush all unreferenced dirty pages to disk.
4739 int sqlite3PagerFlush(Pager *pPager){
4740 int rc = pPager->errCode;
4741 if( !MEMDB ){
4742 PgHdr *pList = sqlite3PcacheDirtyList(pPager->pPCache);
4743 assert( assert_pager_state(pPager) );
4744 while( rc==SQLITE_OK && pList ){
4745 PgHdr *pNext = pList->pDirty;
4746 if( pList->nRef==0 ){
4747 rc = pagerStress((void*)pPager, pList);
4749 pList = pNext;
4753 return rc;
4757 ** Allocate and initialize a new Pager object and put a pointer to it
4758 ** in *ppPager. The pager should eventually be freed by passing it
4759 ** to sqlite3PagerClose().
4761 ** The zFilename argument is the path to the database file to open.
4762 ** If zFilename is NULL then a randomly-named temporary file is created
4763 ** and used as the file to be cached. Temporary files are be deleted
4764 ** automatically when they are closed. If zFilename is ":memory:" then
4765 ** all information is held in cache. It is never written to disk.
4766 ** This can be used to implement an in-memory database.
4768 ** The nExtra parameter specifies the number of bytes of space allocated
4769 ** along with each page reference. This space is available to the user
4770 ** via the sqlite3PagerGetExtra() API. When a new page is allocated, the
4771 ** first 8 bytes of this space are zeroed but the remainder is uninitialized.
4772 ** (The extra space is used by btree as the MemPage object.)
4774 ** The flags argument is used to specify properties that affect the
4775 ** operation of the pager. It should be passed some bitwise combination
4776 ** of the PAGER_* flags.
4778 ** The vfsFlags parameter is a bitmask to pass to the flags parameter
4779 ** of the xOpen() method of the supplied VFS when opening files.
4781 ** If the pager object is allocated and the specified file opened
4782 ** successfully, SQLITE_OK is returned and *ppPager set to point to
4783 ** the new pager object. If an error occurs, *ppPager is set to NULL
4784 ** and error code returned. This function may return SQLITE_NOMEM
4785 ** (sqlite3Malloc() is used to allocate memory), SQLITE_CANTOPEN or
4786 ** various SQLITE_IO_XXX errors.
4788 int sqlite3PagerOpen(
4789 sqlite3_vfs *pVfs, /* The virtual file system to use */
4790 Pager **ppPager, /* OUT: Return the Pager structure here */
4791 const char *zFilename, /* Name of the database file to open */
4792 int nExtra, /* Extra bytes append to each in-memory page */
4793 int flags, /* flags controlling this file */
4794 int vfsFlags, /* flags passed through to sqlite3_vfs.xOpen() */
4795 void (*xReinit)(DbPage*) /* Function to reinitialize pages */
4797 u8 *pPtr;
4798 Pager *pPager = 0; /* Pager object to allocate and return */
4799 int rc = SQLITE_OK; /* Return code */
4800 int tempFile = 0; /* True for temp files (incl. in-memory files) */
4801 int memDb = 0; /* True if this is an in-memory file */
4802 #ifndef SQLITE_OMIT_DESERIALIZE
4803 int memJM = 0; /* Memory journal mode */
4804 #else
4805 # define memJM 0
4806 #endif
4807 int readOnly = 0; /* True if this is a read-only file */
4808 int journalFileSize; /* Bytes to allocate for each journal fd */
4809 char *zPathname = 0; /* Full path to database file */
4810 int nPathname = 0; /* Number of bytes in zPathname */
4811 int useJournal = (flags & PAGER_OMIT_JOURNAL)==0; /* False to omit journal */
4812 int pcacheSize = sqlite3PcacheSize(); /* Bytes to allocate for PCache */
4813 u32 szPageDflt = SQLITE_DEFAULT_PAGE_SIZE; /* Default page size */
4814 const char *zUri = 0; /* URI args to copy */
4815 int nUriByte = 1; /* Number of bytes of URI args at *zUri */
4816 int nUri = 0; /* Number of URI parameters */
4818 /* Figure out how much space is required for each journal file-handle
4819 ** (there are two of them, the main journal and the sub-journal). */
4820 journalFileSize = ROUND8(sqlite3JournalSize(pVfs));
4822 /* Set the output variable to NULL in case an error occurs. */
4823 *ppPager = 0;
4825 #ifndef SQLITE_OMIT_MEMORYDB
4826 if( flags & PAGER_MEMORY ){
4827 memDb = 1;
4828 if( zFilename && zFilename[0] ){
4829 zPathname = sqlite3DbStrDup(0, zFilename);
4830 if( zPathname==0 ) return SQLITE_NOMEM_BKPT;
4831 nPathname = sqlite3Strlen30(zPathname);
4832 zFilename = 0;
4835 #endif
4837 /* Compute and store the full pathname in an allocated buffer pointed
4838 ** to by zPathname, length nPathname. Or, if this is a temporary file,
4839 ** leave both nPathname and zPathname set to 0.
4841 if( zFilename && zFilename[0] ){
4842 const char *z;
4843 nPathname = pVfs->mxPathname+1;
4844 zPathname = sqlite3DbMallocRaw(0, nPathname*2);
4845 if( zPathname==0 ){
4846 return SQLITE_NOMEM_BKPT;
4848 zPathname[0] = 0; /* Make sure initialized even if FullPathname() fails */
4849 rc = sqlite3OsFullPathname(pVfs, zFilename, nPathname, zPathname);
4850 if( rc!=SQLITE_OK ){
4851 if( rc==SQLITE_OK_SYMLINK ){
4852 if( vfsFlags & SQLITE_OPEN_NOFOLLOW ){
4853 rc = SQLITE_CANTOPEN_SYMLINK;
4854 }else{
4855 rc = SQLITE_OK;
4859 nPathname = sqlite3Strlen30(zPathname);
4860 z = zUri = &zFilename[sqlite3Strlen30(zFilename)+1];
4861 while( *z ){
4862 z += strlen(z)+1;
4863 z += strlen(z)+1;
4864 nUri++;
4866 nUriByte = (int)(&z[1] - zUri);
4867 assert( nUriByte>=1 );
4868 if( rc==SQLITE_OK && nPathname+8>pVfs->mxPathname ){
4869 /* This branch is taken when the journal path required by
4870 ** the database being opened will be more than pVfs->mxPathname
4871 ** bytes in length. This means the database cannot be opened,
4872 ** as it will not be possible to open the journal file or even
4873 ** check for a hot-journal before reading.
4875 rc = SQLITE_CANTOPEN_BKPT;
4877 if( rc!=SQLITE_OK ){
4878 sqlite3DbFree(0, zPathname);
4879 return rc;
4883 /* Allocate memory for the Pager structure, PCache object, the
4884 ** three file descriptors, the database file name and the journal
4885 ** file name. The layout in memory is as follows:
4887 ** Pager object (sizeof(Pager) bytes)
4888 ** PCache object (sqlite3PcacheSize() bytes)
4889 ** Database file handle (pVfs->szOsFile bytes)
4890 ** Sub-journal file handle (journalFileSize bytes)
4891 ** Main journal file handle (journalFileSize bytes)
4892 ** Ptr back to the Pager (sizeof(Pager*) bytes)
4893 ** \0\0\0\0 database prefix (4 bytes)
4894 ** Database file name (nPathname+1 bytes)
4895 ** URI query parameters (nUriByte bytes)
4896 ** Journal filename (nPathname+8+1 bytes)
4897 ** WAL filename (nPathname+4+1 bytes)
4898 ** \0\0\0 terminator (3 bytes)
4900 ** Some 3rd-party software, over which we have no control, depends on
4901 ** the specific order of the filenames and the \0 separators between them
4902 ** so that it can (for example) find the database filename given the WAL
4903 ** filename without using the sqlite3_filename_database() API. This is a
4904 ** misuse of SQLite and a bug in the 3rd-party software, but the 3rd-party
4905 ** software is in widespread use, so we try to avoid changing the filename
4906 ** order and formatting if possible. In particular, the details of the
4907 ** filename format expected by 3rd-party software should be as follows:
4909 ** - Main Database Path
4910 ** - \0
4911 ** - Multiple URI components consisting of:
4912 ** - Key
4913 ** - \0
4914 ** - Value
4915 ** - \0
4916 ** - \0
4917 ** - Journal Path
4918 ** - \0
4919 ** - WAL Path (zWALName)
4920 ** - \0
4922 ** The sqlite3_create_filename() interface and the databaseFilename() utility
4923 ** that is used by sqlite3_filename_database() and kin also depend on the
4924 ** specific formatting and order of the various filenames, so if the format
4925 ** changes here, be sure to change it there as well.
4927 pPtr = (u8 *)sqlite3MallocZero(
4928 ROUND8(sizeof(*pPager)) + /* Pager structure */
4929 ROUND8(pcacheSize) + /* PCache object */
4930 ROUND8(pVfs->szOsFile) + /* The main db file */
4931 journalFileSize * 2 + /* The two journal files */
4932 sizeof(pPager) + /* Space to hold a pointer */
4933 4 + /* Database prefix */
4934 nPathname + 1 + /* database filename */
4935 nUriByte + /* query parameters */
4936 nPathname + 8 + 1 + /* Journal filename */
4937 #ifndef SQLITE_OMIT_WAL
4938 nPathname + 4 + 1 + /* WAL filename */
4939 #endif
4940 3 /* Terminator */
4942 assert( EIGHT_BYTE_ALIGNMENT(SQLITE_INT_TO_PTR(journalFileSize)) );
4943 if( !pPtr ){
4944 sqlite3DbFree(0, zPathname);
4945 return SQLITE_NOMEM_BKPT;
4947 pPager = (Pager*)pPtr; pPtr += ROUND8(sizeof(*pPager));
4948 pPager->pPCache = (PCache*)pPtr; pPtr += ROUND8(pcacheSize);
4949 pPager->fd = (sqlite3_file*)pPtr; pPtr += ROUND8(pVfs->szOsFile);
4950 pPager->sjfd = (sqlite3_file*)pPtr; pPtr += journalFileSize;
4951 pPager->jfd = (sqlite3_file*)pPtr; pPtr += journalFileSize;
4952 assert( EIGHT_BYTE_ALIGNMENT(pPager->jfd) );
4953 memcpy(pPtr, &pPager, sizeof(pPager)); pPtr += sizeof(pPager);
4955 /* Fill in the Pager.zFilename and pPager.zQueryParam fields */
4956 pPtr += 4; /* Skip zero prefix */
4957 pPager->zFilename = (char*)pPtr;
4958 if( nPathname>0 ){
4959 memcpy(pPtr, zPathname, nPathname); pPtr += nPathname + 1;
4960 if( zUri ){
4961 memcpy(pPtr, zUri, nUriByte); pPtr += nUriByte;
4962 }else{
4963 pPtr++;
4968 /* Fill in Pager.zJournal */
4969 if( nPathname>0 ){
4970 pPager->zJournal = (char*)pPtr;
4971 memcpy(pPtr, zPathname, nPathname); pPtr += nPathname;
4972 memcpy(pPtr, "-journal",8); pPtr += 8 + 1;
4973 #ifdef SQLITE_ENABLE_8_3_NAMES
4974 sqlite3FileSuffix3(zFilename,pPager->zJournal);
4975 pPtr = (u8*)(pPager->zJournal + sqlite3Strlen30(pPager->zJournal)+1);
4976 #endif
4977 }else{
4978 pPager->zJournal = 0;
4981 #ifndef SQLITE_OMIT_WAL
4982 /* Fill in Pager.zWal */
4983 if( nPathname>0 ){
4984 pPager->zWal = (char*)pPtr;
4985 memcpy(pPtr, zPathname, nPathname); pPtr += nPathname;
4986 memcpy(pPtr, "-wal", 4); pPtr += 4 + 1;
4987 #ifdef SQLITE_ENABLE_8_3_NAMES
4988 sqlite3FileSuffix3(zFilename, pPager->zWal);
4989 pPtr = (u8*)(pPager->zWal + sqlite3Strlen30(pPager->zWal)+1);
4990 #endif
4991 }else{
4992 pPager->zWal = 0;
4994 #endif
4995 (void)pPtr; /* Suppress warning about unused pPtr value */
4997 if( nPathname ) sqlite3DbFree(0, zPathname);
4998 pPager->pVfs = pVfs;
4999 pPager->vfsFlags = vfsFlags;
5001 /* Open the pager file.
5003 if( zFilename && zFilename[0] ){
5004 int fout = 0; /* VFS flags returned by xOpen() */
5005 rc = sqlite3OsOpen(pVfs, pPager->zFilename, pPager->fd, vfsFlags, &fout);
5006 assert( !memDb );
5007 #ifndef SQLITE_OMIT_DESERIALIZE
5008 pPager->memVfs = memJM = (fout&SQLITE_OPEN_MEMORY)!=0;
5009 #endif
5010 readOnly = (fout&SQLITE_OPEN_READONLY)!=0;
5012 /* If the file was successfully opened for read/write access,
5013 ** choose a default page size in case we have to create the
5014 ** database file. The default page size is the maximum of:
5016 ** + SQLITE_DEFAULT_PAGE_SIZE,
5017 ** + The value returned by sqlite3OsSectorSize()
5018 ** + The largest page size that can be written atomically.
5020 if( rc==SQLITE_OK ){
5021 int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
5022 if( !readOnly ){
5023 setSectorSize(pPager);
5024 assert(SQLITE_DEFAULT_PAGE_SIZE<=SQLITE_MAX_DEFAULT_PAGE_SIZE);
5025 if( szPageDflt<pPager->sectorSize ){
5026 if( pPager->sectorSize>SQLITE_MAX_DEFAULT_PAGE_SIZE ){
5027 szPageDflt = SQLITE_MAX_DEFAULT_PAGE_SIZE;
5028 }else{
5029 szPageDflt = (u32)pPager->sectorSize;
5032 #ifdef SQLITE_ENABLE_ATOMIC_WRITE
5034 int ii;
5035 assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
5036 assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));
5037 assert(SQLITE_MAX_DEFAULT_PAGE_SIZE<=65536);
5038 for(ii=szPageDflt; ii<=SQLITE_MAX_DEFAULT_PAGE_SIZE; ii=ii*2){
5039 if( iDc&(SQLITE_IOCAP_ATOMIC|(ii>>8)) ){
5040 szPageDflt = ii;
5044 #endif
5046 pPager->noLock = sqlite3_uri_boolean(pPager->zFilename, "nolock", 0);
5047 if( (iDc & SQLITE_IOCAP_IMMUTABLE)!=0
5048 || sqlite3_uri_boolean(pPager->zFilename, "immutable", 0) ){
5049 vfsFlags |= SQLITE_OPEN_READONLY;
5050 goto act_like_temp_file;
5053 }else{
5054 /* If a temporary file is requested, it is not opened immediately.
5055 ** In this case we accept the default page size and delay actually
5056 ** opening the file until the first call to OsWrite().
5058 ** This branch is also run for an in-memory database. An in-memory
5059 ** database is the same as a temp-file that is never written out to
5060 ** disk and uses an in-memory rollback journal.
5062 ** This branch also runs for files marked as immutable.
5064 act_like_temp_file:
5065 tempFile = 1;
5066 pPager->eState = PAGER_READER; /* Pretend we already have a lock */
5067 pPager->eLock = EXCLUSIVE_LOCK; /* Pretend we are in EXCLUSIVE mode */
5068 pPager->noLock = 1; /* Do no locking */
5069 readOnly = (vfsFlags&SQLITE_OPEN_READONLY);
5072 /* The following call to PagerSetPagesize() serves to set the value of
5073 ** Pager.pageSize and to allocate the Pager.pTmpSpace buffer.
5075 if( rc==SQLITE_OK ){
5076 assert( pPager->memDb==0 );
5077 rc = sqlite3PagerSetPagesize(pPager, &szPageDflt, -1);
5078 testcase( rc!=SQLITE_OK );
5081 /* Initialize the PCache object. */
5082 if( rc==SQLITE_OK ){
5083 nExtra = ROUND8(nExtra);
5084 assert( nExtra>=8 && nExtra<1000 );
5085 rc = sqlite3PcacheOpen(szPageDflt, nExtra, !memDb,
5086 !memDb?pagerStress:0, (void *)pPager, pPager->pPCache);
5089 /* If an error occurred above, free the Pager structure and close the file.
5091 if( rc!=SQLITE_OK ){
5092 sqlite3OsClose(pPager->fd);
5093 sqlite3PageFree(pPager->pTmpSpace);
5094 sqlite3_free(pPager);
5095 return rc;
5098 PAGERTRACE(("OPEN %d %s\n", FILEHANDLEID(pPager->fd), pPager->zFilename));
5099 IOTRACE(("OPEN %p %s\n", pPager, pPager->zFilename))
5101 pPager->useJournal = (u8)useJournal;
5102 /* pPager->stmtOpen = 0; */
5103 /* pPager->stmtInUse = 0; */
5104 /* pPager->nRef = 0; */
5105 /* pPager->stmtSize = 0; */
5106 /* pPager->stmtJSize = 0; */
5107 /* pPager->nPage = 0; */
5108 pPager->mxPgno = SQLITE_MAX_PAGE_COUNT;
5109 /* pPager->state = PAGER_UNLOCK; */
5110 /* pPager->errMask = 0; */
5111 pPager->tempFile = (u8)tempFile;
5112 assert( tempFile==PAGER_LOCKINGMODE_NORMAL
5113 || tempFile==PAGER_LOCKINGMODE_EXCLUSIVE );
5114 assert( PAGER_LOCKINGMODE_EXCLUSIVE==1 );
5115 pPager->exclusiveMode = (u8)tempFile;
5116 pPager->changeCountDone = pPager->tempFile;
5117 pPager->memDb = (u8)memDb;
5118 pPager->readOnly = (u8)readOnly;
5119 assert( useJournal || pPager->tempFile );
5120 pPager->noSync = pPager->tempFile;
5121 if( pPager->noSync ){
5122 assert( pPager->fullSync==0 );
5123 assert( pPager->extraSync==0 );
5124 assert( pPager->syncFlags==0 );
5125 assert( pPager->walSyncFlags==0 );
5126 }else{
5127 pPager->fullSync = 1;
5128 pPager->extraSync = 0;
5129 pPager->syncFlags = SQLITE_SYNC_NORMAL;
5130 pPager->walSyncFlags = SQLITE_SYNC_NORMAL | (SQLITE_SYNC_NORMAL<<2);
5132 /* pPager->pFirst = 0; */
5133 /* pPager->pFirstSynced = 0; */
5134 /* pPager->pLast = 0; */
5135 pPager->nExtra = (u16)nExtra;
5136 pPager->journalSizeLimit = SQLITE_DEFAULT_JOURNAL_SIZE_LIMIT;
5137 assert( isOpen(pPager->fd) || tempFile );
5138 setSectorSize(pPager);
5139 if( !useJournal ){
5140 pPager->journalMode = PAGER_JOURNALMODE_OFF;
5141 }else if( memDb || memJM ){
5142 pPager->journalMode = PAGER_JOURNALMODE_MEMORY;
5144 /* pPager->xBusyHandler = 0; */
5145 /* pPager->pBusyHandlerArg = 0; */
5146 pPager->xReiniter = xReinit;
5147 setGetterMethod(pPager);
5148 /* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */
5149 /* pPager->szMmap = SQLITE_DEFAULT_MMAP_SIZE // will be set by btree.c */
5151 *ppPager = pPager;
5152 return SQLITE_OK;
5156 ** Return the sqlite3_file for the main database given the name
5157 ** of the corresonding WAL or Journal name as passed into
5158 ** xOpen.
5160 sqlite3_file *sqlite3_database_file_object(const char *zName){
5161 Pager *pPager;
5162 while( zName[-1]!=0 || zName[-2]!=0 || zName[-3]!=0 || zName[-4]!=0 ){
5163 zName--;
5165 pPager = *(Pager**)(zName - 4 - sizeof(Pager*));
5166 return pPager->fd;
5171 ** This function is called after transitioning from PAGER_UNLOCK to
5172 ** PAGER_SHARED state. It tests if there is a hot journal present in
5173 ** the file-system for the given pager. A hot journal is one that
5174 ** needs to be played back. According to this function, a hot-journal
5175 ** file exists if the following criteria are met:
5177 ** * The journal file exists in the file system, and
5178 ** * No process holds a RESERVED or greater lock on the database file, and
5179 ** * The database file itself is greater than 0 bytes in size, and
5180 ** * The first byte of the journal file exists and is not 0x00.
5182 ** If the current size of the database file is 0 but a journal file
5183 ** exists, that is probably an old journal left over from a prior
5184 ** database with the same name. In this case the journal file is
5185 ** just deleted using OsDelete, *pExists is set to 0 and SQLITE_OK
5186 ** is returned.
5188 ** This routine does not check if there is a super-journal filename
5189 ** at the end of the file. If there is, and that super-journal file
5190 ** does not exist, then the journal file is not really hot. In this
5191 ** case this routine will return a false-positive. The pager_playback()
5192 ** routine will discover that the journal file is not really hot and
5193 ** will not roll it back.
5195 ** If a hot-journal file is found to exist, *pExists is set to 1 and
5196 ** SQLITE_OK returned. If no hot-journal file is present, *pExists is
5197 ** set to 0 and SQLITE_OK returned. If an IO error occurs while trying
5198 ** to determine whether or not a hot-journal file exists, the IO error
5199 ** code is returned and the value of *pExists is undefined.
5201 static int hasHotJournal(Pager *pPager, int *pExists){
5202 sqlite3_vfs * const pVfs = pPager->pVfs;
5203 int rc = SQLITE_OK; /* Return code */
5204 int exists = 1; /* True if a journal file is present */
5205 int jrnlOpen = !!isOpen(pPager->jfd);
5207 assert( pPager->useJournal );
5208 assert( isOpen(pPager->fd) );
5209 assert( pPager->eState==PAGER_OPEN );
5211 assert( jrnlOpen==0 || ( sqlite3OsDeviceCharacteristics(pPager->jfd) &
5212 SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN
5215 *pExists = 0;
5216 if( !jrnlOpen ){
5217 rc = sqlite3OsAccess(pVfs, pPager->zJournal, SQLITE_ACCESS_EXISTS, &exists);
5219 if( rc==SQLITE_OK && exists ){
5220 int locked = 0; /* True if some process holds a RESERVED lock */
5222 /* Race condition here: Another process might have been holding the
5223 ** the RESERVED lock and have a journal open at the sqlite3OsAccess()
5224 ** call above, but then delete the journal and drop the lock before
5225 ** we get to the following sqlite3OsCheckReservedLock() call. If that
5226 ** is the case, this routine might think there is a hot journal when
5227 ** in fact there is none. This results in a false-positive which will
5228 ** be dealt with by the playback routine. Ticket #3883.
5230 rc = sqlite3OsCheckReservedLock(pPager->fd, &locked);
5231 if( rc==SQLITE_OK && !locked ){
5232 Pgno nPage; /* Number of pages in database file */
5234 assert( pPager->tempFile==0 );
5235 rc = pagerPagecount(pPager, &nPage);
5236 if( rc==SQLITE_OK ){
5237 /* If the database is zero pages in size, that means that either (1) the
5238 ** journal is a remnant from a prior database with the same name where
5239 ** the database file but not the journal was deleted, or (2) the initial
5240 ** transaction that populates a new database is being rolled back.
5241 ** In either case, the journal file can be deleted. However, take care
5242 ** not to delete the journal file if it is already open due to
5243 ** journal_mode=PERSIST.
5245 if( nPage==0 && !jrnlOpen ){
5246 sqlite3BeginBenignMalloc();
5247 if( pagerLockDb(pPager, RESERVED_LOCK)==SQLITE_OK ){
5248 sqlite3OsDelete(pVfs, pPager->zJournal, 0);
5249 if( !pPager->exclusiveMode ) pagerUnlockDb(pPager, SHARED_LOCK);
5251 sqlite3EndBenignMalloc();
5252 }else{
5253 /* The journal file exists and no other connection has a reserved
5254 ** or greater lock on the database file. Now check that there is
5255 ** at least one non-zero bytes at the start of the journal file.
5256 ** If there is, then we consider this journal to be hot. If not,
5257 ** it can be ignored.
5259 if( !jrnlOpen ){
5260 int f = SQLITE_OPEN_READONLY|SQLITE_OPEN_MAIN_JOURNAL;
5261 rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, f, &f);
5263 if( rc==SQLITE_OK ){
5264 u8 first = 0;
5265 rc = sqlite3OsRead(pPager->jfd, (void *)&first, 1, 0);
5266 if( rc==SQLITE_IOERR_SHORT_READ ){
5267 rc = SQLITE_OK;
5269 if( !jrnlOpen ){
5270 sqlite3OsClose(pPager->jfd);
5272 *pExists = (first!=0);
5273 }else if( rc==SQLITE_CANTOPEN ){
5274 /* If we cannot open the rollback journal file in order to see if
5275 ** it has a zero header, that might be due to an I/O error, or
5276 ** it might be due to the race condition described above and in
5277 ** ticket #3883. Either way, assume that the journal is hot.
5278 ** This might be a false positive. But if it is, then the
5279 ** automatic journal playback and recovery mechanism will deal
5280 ** with it under an EXCLUSIVE lock where we do not need to
5281 ** worry so much with race conditions.
5283 *pExists = 1;
5284 rc = SQLITE_OK;
5291 return rc;
5295 ** This function is called to obtain a shared lock on the database file.
5296 ** It is illegal to call sqlite3PagerGet() until after this function
5297 ** has been successfully called. If a shared-lock is already held when
5298 ** this function is called, it is a no-op.
5300 ** The following operations are also performed by this function.
5302 ** 1) If the pager is currently in PAGER_OPEN state (no lock held
5303 ** on the database file), then an attempt is made to obtain a
5304 ** SHARED lock on the database file. Immediately after obtaining
5305 ** the SHARED lock, the file-system is checked for a hot-journal,
5306 ** which is played back if present. Following any hot-journal
5307 ** rollback, the contents of the cache are validated by checking
5308 ** the 'change-counter' field of the database file header and
5309 ** discarded if they are found to be invalid.
5311 ** 2) If the pager is running in exclusive-mode, and there are currently
5312 ** no outstanding references to any pages, and is in the error state,
5313 ** then an attempt is made to clear the error state by discarding
5314 ** the contents of the page cache and rolling back any open journal
5315 ** file.
5317 ** If everything is successful, SQLITE_OK is returned. If an IO error
5318 ** occurs while locking the database, checking for a hot-journal file or
5319 ** rolling back a journal file, the IO error code is returned.
5321 int sqlite3PagerSharedLock(Pager *pPager){
5322 int rc = SQLITE_OK; /* Return code */
5324 /* This routine is only called from b-tree and only when there are no
5325 ** outstanding pages. This implies that the pager state should either
5326 ** be OPEN or READER. READER is only possible if the pager is or was in
5327 ** exclusive access mode. */
5328 assert( sqlite3PcacheRefCount(pPager->pPCache)==0 );
5329 assert( assert_pager_state(pPager) );
5330 assert( pPager->eState==PAGER_OPEN || pPager->eState==PAGER_READER );
5331 assert( pPager->errCode==SQLITE_OK );
5333 if( !pagerUseWal(pPager) && pPager->eState==PAGER_OPEN ){
5334 int bHotJournal = 1; /* True if there exists a hot journal-file */
5336 assert( !MEMDB );
5337 assert( pPager->tempFile==0 || pPager->eLock==EXCLUSIVE_LOCK );
5339 rc = pager_wait_on_lock(pPager, SHARED_LOCK);
5340 if( rc!=SQLITE_OK ){
5341 assert( pPager->eLock==NO_LOCK || pPager->eLock==UNKNOWN_LOCK );
5342 goto failed;
5345 /* If a journal file exists, and there is no RESERVED lock on the
5346 ** database file, then it either needs to be played back or deleted.
5348 if( pPager->eLock<=SHARED_LOCK ){
5349 rc = hasHotJournal(pPager, &bHotJournal);
5351 if( rc!=SQLITE_OK ){
5352 goto failed;
5354 if( bHotJournal ){
5355 if( pPager->readOnly ){
5356 rc = SQLITE_READONLY_ROLLBACK;
5357 goto failed;
5360 /* Get an EXCLUSIVE lock on the database file. At this point it is
5361 ** important that a RESERVED lock is not obtained on the way to the
5362 ** EXCLUSIVE lock. If it were, another process might open the
5363 ** database file, detect the RESERVED lock, and conclude that the
5364 ** database is safe to read while this process is still rolling the
5365 ** hot-journal back.
5367 ** Because the intermediate RESERVED lock is not requested, any
5368 ** other process attempting to access the database file will get to
5369 ** this point in the code and fail to obtain its own EXCLUSIVE lock
5370 ** on the database file.
5372 ** Unless the pager is in locking_mode=exclusive mode, the lock is
5373 ** downgraded to SHARED_LOCK before this function returns.
5375 rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
5376 if( rc!=SQLITE_OK ){
5377 goto failed;
5380 /* If it is not already open and the file exists on disk, open the
5381 ** journal for read/write access. Write access is required because
5382 ** in exclusive-access mode the file descriptor will be kept open
5383 ** and possibly used for a transaction later on. Also, write-access
5384 ** is usually required to finalize the journal in journal_mode=persist
5385 ** mode (and also for journal_mode=truncate on some systems).
5387 ** If the journal does not exist, it usually means that some
5388 ** other connection managed to get in and roll it back before
5389 ** this connection obtained the exclusive lock above. Or, it
5390 ** may mean that the pager was in the error-state when this
5391 ** function was called and the journal file does not exist.
5393 if( !isOpen(pPager->jfd) && pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
5394 sqlite3_vfs * const pVfs = pPager->pVfs;
5395 int bExists; /* True if journal file exists */
5396 rc = sqlite3OsAccess(
5397 pVfs, pPager->zJournal, SQLITE_ACCESS_EXISTS, &bExists);
5398 if( rc==SQLITE_OK && bExists ){
5399 int fout = 0;
5400 int f = SQLITE_OPEN_READWRITE|SQLITE_OPEN_MAIN_JOURNAL;
5401 assert( !pPager->tempFile );
5402 rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, f, &fout);
5403 assert( rc!=SQLITE_OK || isOpen(pPager->jfd) );
5404 if( rc==SQLITE_OK && fout&SQLITE_OPEN_READONLY ){
5405 rc = SQLITE_CANTOPEN_BKPT;
5406 sqlite3OsClose(pPager->jfd);
5411 /* Playback and delete the journal. Drop the database write
5412 ** lock and reacquire the read lock. Purge the cache before
5413 ** playing back the hot-journal so that we don't end up with
5414 ** an inconsistent cache. Sync the hot journal before playing
5415 ** it back since the process that crashed and left the hot journal
5416 ** probably did not sync it and we are required to always sync
5417 ** the journal before playing it back.
5419 if( isOpen(pPager->jfd) ){
5420 assert( rc==SQLITE_OK );
5421 rc = pagerSyncHotJournal(pPager);
5422 if( rc==SQLITE_OK ){
5423 rc = pager_playback(pPager, !pPager->tempFile);
5424 pPager->eState = PAGER_OPEN;
5426 }else if( !pPager->exclusiveMode ){
5427 pagerUnlockDb(pPager, SHARED_LOCK);
5430 if( rc!=SQLITE_OK ){
5431 /* This branch is taken if an error occurs while trying to open
5432 ** or roll back a hot-journal while holding an EXCLUSIVE lock. The
5433 ** pager_unlock() routine will be called before returning to unlock
5434 ** the file. If the unlock attempt fails, then Pager.eLock must be
5435 ** set to UNKNOWN_LOCK (see the comment above the #define for
5436 ** UNKNOWN_LOCK above for an explanation).
5438 ** In order to get pager_unlock() to do this, set Pager.eState to
5439 ** PAGER_ERROR now. This is not actually counted as a transition
5440 ** to ERROR state in the state diagram at the top of this file,
5441 ** since we know that the same call to pager_unlock() will very
5442 ** shortly transition the pager object to the OPEN state. Calling
5443 ** assert_pager_state() would fail now, as it should not be possible
5444 ** to be in ERROR state when there are zero outstanding page
5445 ** references.
5447 pager_error(pPager, rc);
5448 goto failed;
5451 assert( pPager->eState==PAGER_OPEN );
5452 assert( (pPager->eLock==SHARED_LOCK)
5453 || (pPager->exclusiveMode && pPager->eLock>SHARED_LOCK)
5457 if( !pPager->tempFile && pPager->hasHeldSharedLock ){
5458 /* The shared-lock has just been acquired then check to
5459 ** see if the database has been modified. If the database has changed,
5460 ** flush the cache. The hasHeldSharedLock flag prevents this from
5461 ** occurring on the very first access to a file, in order to save a
5462 ** single unnecessary sqlite3OsRead() call at the start-up.
5464 ** Database changes are detected by looking at 15 bytes beginning
5465 ** at offset 24 into the file. The first 4 of these 16 bytes are
5466 ** a 32-bit counter that is incremented with each change. The
5467 ** other bytes change randomly with each file change when
5468 ** a codec is in use.
5470 ** There is a vanishingly small chance that a change will not be
5471 ** detected. The chance of an undetected change is so small that
5472 ** it can be neglected.
5474 char dbFileVers[sizeof(pPager->dbFileVers)];
5476 IOTRACE(("CKVERS %p %d\n", pPager, sizeof(dbFileVers)));
5477 rc = sqlite3OsRead(pPager->fd, &dbFileVers, sizeof(dbFileVers), 24);
5478 if( rc!=SQLITE_OK ){
5479 if( rc!=SQLITE_IOERR_SHORT_READ ){
5480 goto failed;
5482 memset(dbFileVers, 0, sizeof(dbFileVers));
5485 if( memcmp(pPager->dbFileVers, dbFileVers, sizeof(dbFileVers))!=0 ){
5486 pager_reset(pPager);
5488 /* Unmap the database file. It is possible that external processes
5489 ** may have truncated the database file and then extended it back
5490 ** to its original size while this process was not holding a lock.
5491 ** In this case there may exist a Pager.pMap mapping that appears
5492 ** to be the right size but is not actually valid. Avoid this
5493 ** possibility by unmapping the db here. */
5494 if( USEFETCH(pPager) ){
5495 sqlite3OsUnfetch(pPager->fd, 0, 0);
5500 /* If there is a WAL file in the file-system, open this database in WAL
5501 ** mode. Otherwise, the following function call is a no-op.
5503 rc = pagerOpenWalIfPresent(pPager);
5504 #ifndef SQLITE_OMIT_WAL
5505 assert( pPager->pWal==0 || rc==SQLITE_OK );
5506 #endif
5509 if( pagerUseWal(pPager) ){
5510 assert( rc==SQLITE_OK );
5511 rc = pagerBeginReadTransaction(pPager);
5514 if( pPager->tempFile==0 && pPager->eState==PAGER_OPEN && rc==SQLITE_OK ){
5515 rc = pagerPagecount(pPager, &pPager->dbSize);
5518 failed:
5519 if( rc!=SQLITE_OK ){
5520 assert( !MEMDB );
5521 pager_unlock(pPager);
5522 assert( pPager->eState==PAGER_OPEN );
5523 }else{
5524 pPager->eState = PAGER_READER;
5525 pPager->hasHeldSharedLock = 1;
5527 return rc;
5531 ** If the reference count has reached zero, rollback any active
5532 ** transaction and unlock the pager.
5534 ** Except, in locking_mode=EXCLUSIVE when there is nothing to in
5535 ** the rollback journal, the unlock is not performed and there is
5536 ** nothing to rollback, so this routine is a no-op.
5538 static void pagerUnlockIfUnused(Pager *pPager){
5539 if( sqlite3PcacheRefCount(pPager->pPCache)==0 ){
5540 assert( pPager->nMmapOut==0 ); /* because page1 is never memory mapped */
5541 pagerUnlockAndRollback(pPager);
5546 ** The page getter methods each try to acquire a reference to a
5547 ** page with page number pgno. If the requested reference is
5548 ** successfully obtained, it is copied to *ppPage and SQLITE_OK returned.
5550 ** There are different implementations of the getter method depending
5551 ** on the current state of the pager.
5553 ** getPageNormal() -- The normal getter
5554 ** getPageError() -- Used if the pager is in an error state
5555 ** getPageMmap() -- Used if memory-mapped I/O is enabled
5557 ** If the requested page is already in the cache, it is returned.
5558 ** Otherwise, a new page object is allocated and populated with data
5559 ** read from the database file. In some cases, the pcache module may
5560 ** choose not to allocate a new page object and may reuse an existing
5561 ** object with no outstanding references.
5563 ** The extra data appended to a page is always initialized to zeros the
5564 ** first time a page is loaded into memory. If the page requested is
5565 ** already in the cache when this function is called, then the extra
5566 ** data is left as it was when the page object was last used.
5568 ** If the database image is smaller than the requested page or if
5569 ** the flags parameter contains the PAGER_GET_NOCONTENT bit and the
5570 ** requested page is not already stored in the cache, then no
5571 ** actual disk read occurs. In this case the memory image of the
5572 ** page is initialized to all zeros.
5574 ** If PAGER_GET_NOCONTENT is true, it means that we do not care about
5575 ** the contents of the page. This occurs in two scenarios:
5577 ** a) When reading a free-list leaf page from the database, and
5579 ** b) When a savepoint is being rolled back and we need to load
5580 ** a new page into the cache to be filled with the data read
5581 ** from the savepoint journal.
5583 ** If PAGER_GET_NOCONTENT is true, then the data returned is zeroed instead
5584 ** of being read from the database. Additionally, the bits corresponding
5585 ** to pgno in Pager.pInJournal (bitvec of pages already written to the
5586 ** journal file) and the PagerSavepoint.pInSavepoint bitvecs of any open
5587 ** savepoints are set. This means if the page is made writable at any
5588 ** point in the future, using a call to sqlite3PagerWrite(), its contents
5589 ** will not be journaled. This saves IO.
5591 ** The acquisition might fail for several reasons. In all cases,
5592 ** an appropriate error code is returned and *ppPage is set to NULL.
5594 ** See also sqlite3PagerLookup(). Both this routine and Lookup() attempt
5595 ** to find a page in the in-memory cache first. If the page is not already
5596 ** in memory, this routine goes to disk to read it in whereas Lookup()
5597 ** just returns 0. This routine acquires a read-lock the first time it
5598 ** has to go to disk, and could also playback an old journal if necessary.
5599 ** Since Lookup() never goes to disk, it never has to deal with locks
5600 ** or journal files.
5602 static int getPageNormal(
5603 Pager *pPager, /* The pager open on the database file */
5604 Pgno pgno, /* Page number to fetch */
5605 DbPage **ppPage, /* Write a pointer to the page here */
5606 int flags /* PAGER_GET_XXX flags */
5608 int rc = SQLITE_OK;
5609 PgHdr *pPg;
5610 u8 noContent; /* True if PAGER_GET_NOCONTENT is set */
5611 sqlite3_pcache_page *pBase;
5613 assert( pPager->errCode==SQLITE_OK );
5614 assert( pPager->eState>=PAGER_READER );
5615 assert( assert_pager_state(pPager) );
5616 assert( pPager->hasHeldSharedLock==1 );
5618 if( pgno==0 ) return SQLITE_CORRUPT_BKPT;
5619 pBase = sqlite3PcacheFetch(pPager->pPCache, pgno, 3);
5620 if( pBase==0 ){
5621 pPg = 0;
5622 rc = sqlite3PcacheFetchStress(pPager->pPCache, pgno, &pBase);
5623 if( rc!=SQLITE_OK ) goto pager_acquire_err;
5624 if( pBase==0 ){
5625 rc = SQLITE_NOMEM_BKPT;
5626 goto pager_acquire_err;
5629 pPg = *ppPage = sqlite3PcacheFetchFinish(pPager->pPCache, pgno, pBase);
5630 assert( pPg==(*ppPage) );
5631 assert( pPg->pgno==pgno );
5632 assert( pPg->pPager==pPager || pPg->pPager==0 );
5634 noContent = (flags & PAGER_GET_NOCONTENT)!=0;
5635 if( pPg->pPager && !noContent ){
5636 /* In this case the pcache already contains an initialized copy of
5637 ** the page. Return without further ado. */
5638 assert( pgno!=PAGER_SJ_PGNO(pPager) );
5639 pPager->aStat[PAGER_STAT_HIT]++;
5640 return SQLITE_OK;
5642 }else{
5643 /* The pager cache has created a new page. Its content needs to
5644 ** be initialized. But first some error checks:
5646 ** (*) obsolete. Was: maximum page number is 2^31
5647 ** (2) Never try to fetch the locking page
5649 if( pgno==PAGER_SJ_PGNO(pPager) ){
5650 rc = SQLITE_CORRUPT_BKPT;
5651 goto pager_acquire_err;
5654 pPg->pPager = pPager;
5656 assert( !isOpen(pPager->fd) || !MEMDB );
5657 if( !isOpen(pPager->fd) || pPager->dbSize<pgno || noContent ){
5658 if( pgno>pPager->mxPgno ){
5659 rc = SQLITE_FULL;
5660 goto pager_acquire_err;
5662 if( noContent ){
5663 /* Failure to set the bits in the InJournal bit-vectors is benign.
5664 ** It merely means that we might do some extra work to journal a
5665 ** page that does not need to be journaled. Nevertheless, be sure
5666 ** to test the case where a malloc error occurs while trying to set
5667 ** a bit in a bit vector.
5669 sqlite3BeginBenignMalloc();
5670 if( pgno<=pPager->dbOrigSize ){
5671 TESTONLY( rc = ) sqlite3BitvecSet(pPager->pInJournal, pgno);
5672 testcase( rc==SQLITE_NOMEM );
5674 TESTONLY( rc = ) addToSavepointBitvecs(pPager, pgno);
5675 testcase( rc==SQLITE_NOMEM );
5676 sqlite3EndBenignMalloc();
5678 memset(pPg->pData, 0, pPager->pageSize);
5679 IOTRACE(("ZERO %p %d\n", pPager, pgno));
5680 }else{
5681 assert( pPg->pPager==pPager );
5682 pPager->aStat[PAGER_STAT_MISS]++;
5683 rc = readDbPage(pPg);
5684 if( rc!=SQLITE_OK ){
5685 goto pager_acquire_err;
5688 pager_set_pagehash(pPg);
5690 return SQLITE_OK;
5692 pager_acquire_err:
5693 assert( rc!=SQLITE_OK );
5694 if( pPg ){
5695 sqlite3PcacheDrop(pPg);
5697 pagerUnlockIfUnused(pPager);
5698 *ppPage = 0;
5699 return rc;
5702 #if SQLITE_MAX_MMAP_SIZE>0
5703 /* The page getter for when memory-mapped I/O is enabled */
5704 static int getPageMMap(
5705 Pager *pPager, /* The pager open on the database file */
5706 Pgno pgno, /* Page number to fetch */
5707 DbPage **ppPage, /* Write a pointer to the page here */
5708 int flags /* PAGER_GET_XXX flags */
5710 int rc = SQLITE_OK;
5711 PgHdr *pPg = 0;
5712 u32 iFrame = 0; /* Frame to read from WAL file */
5714 /* It is acceptable to use a read-only (mmap) page for any page except
5715 ** page 1 if there is no write-transaction open or the ACQUIRE_READONLY
5716 ** flag was specified by the caller. And so long as the db is not a
5717 ** temporary or in-memory database. */
5718 const int bMmapOk = (pgno>1
5719 && (pPager->eState==PAGER_READER || (flags & PAGER_GET_READONLY))
5722 assert( USEFETCH(pPager) );
5723 /* BEGIN SQLCIPHER */
5724 #ifdef SQLITE_HAS_CODEC
5725 assert( pPager->xCodec==0 );
5726 #endif
5727 /* END SQLCIPHER */
5729 /* Optimization note: Adding the "pgno<=1" term before "pgno==0" here
5730 ** allows the compiler optimizer to reuse the results of the "pgno>1"
5731 ** test in the previous statement, and avoid testing pgno==0 in the
5732 ** common case where pgno is large. */
5733 if( pgno<=1 && pgno==0 ){
5734 return SQLITE_CORRUPT_BKPT;
5736 assert( pPager->eState>=PAGER_READER );
5737 assert( assert_pager_state(pPager) );
5738 assert( pPager->hasHeldSharedLock==1 );
5739 assert( pPager->errCode==SQLITE_OK );
5741 if( bMmapOk && pagerUseWal(pPager) ){
5742 rc = sqlite3WalFindFrame(pPager->pWal, pgno, &iFrame);
5743 if( rc!=SQLITE_OK ){
5744 *ppPage = 0;
5745 return rc;
5748 if( bMmapOk && iFrame==0 ){
5749 void *pData = 0;
5750 rc = sqlite3OsFetch(pPager->fd,
5751 (i64)(pgno-1) * pPager->pageSize, pPager->pageSize, &pData
5753 if( rc==SQLITE_OK && pData ){
5754 if( pPager->eState>PAGER_READER || pPager->tempFile ){
5755 pPg = sqlite3PagerLookup(pPager, pgno);
5757 if( pPg==0 ){
5758 rc = pagerAcquireMapPage(pPager, pgno, pData, &pPg);
5759 }else{
5760 sqlite3OsUnfetch(pPager->fd, (i64)(pgno-1)*pPager->pageSize, pData);
5762 if( pPg ){
5763 assert( rc==SQLITE_OK );
5764 *ppPage = pPg;
5765 return SQLITE_OK;
5768 if( rc!=SQLITE_OK ){
5769 *ppPage = 0;
5770 return rc;
5773 return getPageNormal(pPager, pgno, ppPage, flags);
5775 #endif /* SQLITE_MAX_MMAP_SIZE>0 */
5777 /* The page getter method for when the pager is an error state */
5778 static int getPageError(
5779 Pager *pPager, /* The pager open on the database file */
5780 Pgno pgno, /* Page number to fetch */
5781 DbPage **ppPage, /* Write a pointer to the page here */
5782 int flags /* PAGER_GET_XXX flags */
5784 UNUSED_PARAMETER(pgno);
5785 UNUSED_PARAMETER(flags);
5786 assert( pPager->errCode!=SQLITE_OK );
5787 *ppPage = 0;
5788 return pPager->errCode;
5792 /* Dispatch all page fetch requests to the appropriate getter method.
5794 int sqlite3PagerGet(
5795 Pager *pPager, /* The pager open on the database file */
5796 Pgno pgno, /* Page number to fetch */
5797 DbPage **ppPage, /* Write a pointer to the page here */
5798 int flags /* PAGER_GET_XXX flags */
5800 /* printf("PAGE %u\n", pgno); fflush(stdout); */
5801 return pPager->xGet(pPager, pgno, ppPage, flags);
5805 ** Acquire a page if it is already in the in-memory cache. Do
5806 ** not read the page from disk. Return a pointer to the page,
5807 ** or 0 if the page is not in cache.
5809 ** See also sqlite3PagerGet(). The difference between this routine
5810 ** and sqlite3PagerGet() is that _get() will go to the disk and read
5811 ** in the page if the page is not already in cache. This routine
5812 ** returns NULL if the page is not in cache or if a disk I/O error
5813 ** has ever happened.
5815 DbPage *sqlite3PagerLookup(Pager *pPager, Pgno pgno){
5816 sqlite3_pcache_page *pPage;
5817 assert( pPager!=0 );
5818 assert( pgno!=0 );
5819 assert( pPager->pPCache!=0 );
5820 pPage = sqlite3PcacheFetch(pPager->pPCache, pgno, 0);
5821 assert( pPage==0 || pPager->hasHeldSharedLock );
5822 if( pPage==0 ) return 0;
5823 return sqlite3PcacheFetchFinish(pPager->pPCache, pgno, pPage);
5827 ** Release a page reference.
5829 ** The sqlite3PagerUnref() and sqlite3PagerUnrefNotNull() may only be
5830 ** used if we know that the page being released is not the last page.
5831 ** The btree layer always holds page1 open until the end, so these first
5832 ** to routines can be used to release any page other than BtShared.pPage1.
5834 ** Use sqlite3PagerUnrefPageOne() to release page1. This latter routine
5835 ** checks the total number of outstanding pages and if the number of
5836 ** pages reaches zero it drops the database lock.
5838 void sqlite3PagerUnrefNotNull(DbPage *pPg){
5839 TESTONLY( Pager *pPager = pPg->pPager; )
5840 assert( pPg!=0 );
5841 if( pPg->flags & PGHDR_MMAP ){
5842 assert( pPg->pgno!=1 ); /* Page1 is never memory mapped */
5843 pagerReleaseMapPage(pPg);
5844 }else{
5845 sqlite3PcacheRelease(pPg);
5847 /* Do not use this routine to release the last reference to page1 */
5848 assert( sqlite3PcacheRefCount(pPager->pPCache)>0 );
5850 void sqlite3PagerUnref(DbPage *pPg){
5851 if( pPg ) sqlite3PagerUnrefNotNull(pPg);
5853 void sqlite3PagerUnrefPageOne(DbPage *pPg){
5854 Pager *pPager;
5855 assert( pPg!=0 );
5856 assert( pPg->pgno==1 );
5857 assert( (pPg->flags & PGHDR_MMAP)==0 ); /* Page1 is never memory mapped */
5858 pPager = pPg->pPager;
5859 sqlite3PcacheRelease(pPg);
5860 pagerUnlockIfUnused(pPager);
5864 ** This function is called at the start of every write transaction.
5865 ** There must already be a RESERVED or EXCLUSIVE lock on the database
5866 ** file when this routine is called.
5868 ** Open the journal file for pager pPager and write a journal header
5869 ** to the start of it. If there are active savepoints, open the sub-journal
5870 ** as well. This function is only used when the journal file is being
5871 ** opened to write a rollback log for a transaction. It is not used
5872 ** when opening a hot journal file to roll it back.
5874 ** If the journal file is already open (as it may be in exclusive mode),
5875 ** then this function just writes a journal header to the start of the
5876 ** already open file.
5878 ** Whether or not the journal file is opened by this function, the
5879 ** Pager.pInJournal bitvec structure is allocated.
5881 ** Return SQLITE_OK if everything is successful. Otherwise, return
5882 ** SQLITE_NOMEM if the attempt to allocate Pager.pInJournal fails, or
5883 ** an IO error code if opening or writing the journal file fails.
5885 static int pager_open_journal(Pager *pPager){
5886 int rc = SQLITE_OK; /* Return code */
5887 sqlite3_vfs * const pVfs = pPager->pVfs; /* Local cache of vfs pointer */
5889 assert( pPager->eState==PAGER_WRITER_LOCKED );
5890 assert( assert_pager_state(pPager) );
5891 assert( pPager->pInJournal==0 );
5893 /* If already in the error state, this function is a no-op. But on
5894 ** the other hand, this routine is never called if we are already in
5895 ** an error state. */
5896 if( NEVER(pPager->errCode) ) return pPager->errCode;
5898 if( !pagerUseWal(pPager) && pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
5899 pPager->pInJournal = sqlite3BitvecCreate(pPager->dbSize);
5900 if( pPager->pInJournal==0 ){
5901 return SQLITE_NOMEM_BKPT;
5904 /* Open the journal file if it is not already open. */
5905 if( !isOpen(pPager->jfd) ){
5906 if( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ){
5907 sqlite3MemJournalOpen(pPager->jfd);
5908 }else{
5909 int flags = SQLITE_OPEN_READWRITE|SQLITE_OPEN_CREATE;
5910 int nSpill;
5912 if( pPager->tempFile ){
5913 flags |= (SQLITE_OPEN_DELETEONCLOSE|SQLITE_OPEN_TEMP_JOURNAL);
5914 nSpill = sqlite3Config.nStmtSpill;
5915 }else{
5916 flags |= SQLITE_OPEN_MAIN_JOURNAL;
5917 nSpill = jrnlBufferSize(pPager);
5920 /* Verify that the database still has the same name as it did when
5921 ** it was originally opened. */
5922 rc = databaseIsUnmoved(pPager);
5923 if( rc==SQLITE_OK ){
5924 rc = sqlite3JournalOpen (
5925 pVfs, pPager->zJournal, pPager->jfd, flags, nSpill
5929 assert( rc!=SQLITE_OK || isOpen(pPager->jfd) );
5933 /* Write the first journal header to the journal file and open
5934 ** the sub-journal if necessary.
5936 if( rc==SQLITE_OK ){
5937 /* TODO: Check if all of these are really required. */
5938 pPager->nRec = 0;
5939 pPager->journalOff = 0;
5940 pPager->setSuper = 0;
5941 pPager->journalHdr = 0;
5942 rc = writeJournalHdr(pPager);
5946 if( rc!=SQLITE_OK ){
5947 sqlite3BitvecDestroy(pPager->pInJournal);
5948 pPager->pInJournal = 0;
5949 }else{
5950 assert( pPager->eState==PAGER_WRITER_LOCKED );
5951 pPager->eState = PAGER_WRITER_CACHEMOD;
5954 return rc;
5958 ** Begin a write-transaction on the specified pager object. If a
5959 ** write-transaction has already been opened, this function is a no-op.
5961 ** If the exFlag argument is false, then acquire at least a RESERVED
5962 ** lock on the database file. If exFlag is true, then acquire at least
5963 ** an EXCLUSIVE lock. If such a lock is already held, no locking
5964 ** functions need be called.
5966 ** If the subjInMemory argument is non-zero, then any sub-journal opened
5967 ** within this transaction will be opened as an in-memory file. This
5968 ** has no effect if the sub-journal is already opened (as it may be when
5969 ** running in exclusive mode) or if the transaction does not require a
5970 ** sub-journal. If the subjInMemory argument is zero, then any required
5971 ** sub-journal is implemented in-memory if pPager is an in-memory database,
5972 ** or using a temporary file otherwise.
5974 int sqlite3PagerBegin(Pager *pPager, int exFlag, int subjInMemory){
5975 int rc = SQLITE_OK;
5977 if( pPager->errCode ) return pPager->errCode;
5978 assert( pPager->eState>=PAGER_READER && pPager->eState<PAGER_ERROR );
5979 pPager->subjInMemory = (u8)subjInMemory;
5981 if( pPager->eState==PAGER_READER ){
5982 assert( pPager->pInJournal==0 );
5984 if( pagerUseWal(pPager) ){
5985 /* If the pager is configured to use locking_mode=exclusive, and an
5986 ** exclusive lock on the database is not already held, obtain it now.
5988 if( pPager->exclusiveMode && sqlite3WalExclusiveMode(pPager->pWal, -1) ){
5989 rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
5990 if( rc!=SQLITE_OK ){
5991 return rc;
5993 (void)sqlite3WalExclusiveMode(pPager->pWal, 1);
5996 /* Grab the write lock on the log file. If successful, upgrade to
5997 ** PAGER_RESERVED state. Otherwise, return an error code to the caller.
5998 ** The busy-handler is not invoked if another connection already
5999 ** holds the write-lock. If possible, the upper layer will call it.
6001 rc = sqlite3WalBeginWriteTransaction(pPager->pWal);
6002 }else{
6003 /* Obtain a RESERVED lock on the database file. If the exFlag parameter
6004 ** is true, then immediately upgrade this to an EXCLUSIVE lock. The
6005 ** busy-handler callback can be used when upgrading to the EXCLUSIVE
6006 ** lock, but not when obtaining the RESERVED lock.
6008 rc = pagerLockDb(pPager, RESERVED_LOCK);
6009 if( rc==SQLITE_OK && exFlag ){
6010 rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
6014 if( rc==SQLITE_OK ){
6015 /* Change to WRITER_LOCKED state.
6017 ** WAL mode sets Pager.eState to PAGER_WRITER_LOCKED or CACHEMOD
6018 ** when it has an open transaction, but never to DBMOD or FINISHED.
6019 ** This is because in those states the code to roll back savepoint
6020 ** transactions may copy data from the sub-journal into the database
6021 ** file as well as into the page cache. Which would be incorrect in
6022 ** WAL mode.
6024 pPager->eState = PAGER_WRITER_LOCKED;
6025 pPager->dbHintSize = pPager->dbSize;
6026 pPager->dbFileSize = pPager->dbSize;
6027 pPager->dbOrigSize = pPager->dbSize;
6028 pPager->journalOff = 0;
6031 assert( rc==SQLITE_OK || pPager->eState==PAGER_READER );
6032 assert( rc!=SQLITE_OK || pPager->eState==PAGER_WRITER_LOCKED );
6033 assert( assert_pager_state(pPager) );
6036 PAGERTRACE(("TRANSACTION %d\n", PAGERID(pPager)));
6037 return rc;
6041 ** Write page pPg onto the end of the rollback journal.
6043 static SQLITE_NOINLINE int pagerAddPageToRollbackJournal(PgHdr *pPg){
6044 Pager *pPager = pPg->pPager;
6045 int rc;
6046 u32 cksum;
6047 char *pData2;
6048 i64 iOff = pPager->journalOff;
6050 /* We should never write to the journal file the page that
6051 ** contains the database locks. The following assert verifies
6052 ** that we do not. */
6053 assert( pPg->pgno!=PAGER_SJ_PGNO(pPager) );
6055 assert( pPager->journalHdr<=pPager->journalOff );
6056 CODEC2(pPager, pPg->pData, pPg->pgno, 7, return SQLITE_NOMEM_BKPT, pData2);
6057 cksum = pager_cksum(pPager, (u8*)pData2);
6059 /* Even if an IO or diskfull error occurs while journalling the
6060 ** page in the block above, set the need-sync flag for the page.
6061 ** Otherwise, when the transaction is rolled back, the logic in
6062 ** playback_one_page() will think that the page needs to be restored
6063 ** in the database file. And if an IO error occurs while doing so,
6064 ** then corruption may follow.
6066 pPg->flags |= PGHDR_NEED_SYNC;
6068 rc = write32bits(pPager->jfd, iOff, pPg->pgno);
6069 if( rc!=SQLITE_OK ) return rc;
6070 rc = sqlite3OsWrite(pPager->jfd, pData2, pPager->pageSize, iOff+4);
6071 if( rc!=SQLITE_OK ) return rc;
6072 rc = write32bits(pPager->jfd, iOff+pPager->pageSize+4, cksum);
6073 if( rc!=SQLITE_OK ) return rc;
6075 IOTRACE(("JOUT %p %d %lld %d\n", pPager, pPg->pgno,
6076 pPager->journalOff, pPager->pageSize));
6077 PAGER_INCR(sqlite3_pager_writej_count);
6078 PAGERTRACE(("JOURNAL %d page %d needSync=%d hash(%08x)\n",
6079 PAGERID(pPager), pPg->pgno,
6080 ((pPg->flags&PGHDR_NEED_SYNC)?1:0), pager_pagehash(pPg)));
6082 pPager->journalOff += 8 + pPager->pageSize;
6083 pPager->nRec++;
6084 assert( pPager->pInJournal!=0 );
6085 rc = sqlite3BitvecSet(pPager->pInJournal, pPg->pgno);
6086 testcase( rc==SQLITE_NOMEM );
6087 assert( rc==SQLITE_OK || rc==SQLITE_NOMEM );
6088 rc |= addToSavepointBitvecs(pPager, pPg->pgno);
6089 assert( rc==SQLITE_OK || rc==SQLITE_NOMEM );
6090 return rc;
6094 ** Mark a single data page as writeable. The page is written into the
6095 ** main journal or sub-journal as required. If the page is written into
6096 ** one of the journals, the corresponding bit is set in the
6097 ** Pager.pInJournal bitvec and the PagerSavepoint.pInSavepoint bitvecs
6098 ** of any open savepoints as appropriate.
6100 static int pager_write(PgHdr *pPg){
6101 Pager *pPager = pPg->pPager;
6102 int rc = SQLITE_OK;
6104 /* This routine is not called unless a write-transaction has already
6105 ** been started. The journal file may or may not be open at this point.
6106 ** It is never called in the ERROR state.
6108 assert( pPager->eState==PAGER_WRITER_LOCKED
6109 || pPager->eState==PAGER_WRITER_CACHEMOD
6110 || pPager->eState==PAGER_WRITER_DBMOD
6112 assert( assert_pager_state(pPager) );
6113 assert( pPager->errCode==0 );
6114 assert( pPager->readOnly==0 );
6115 CHECK_PAGE(pPg);
6117 /* The journal file needs to be opened. Higher level routines have already
6118 ** obtained the necessary locks to begin the write-transaction, but the
6119 ** rollback journal might not yet be open. Open it now if this is the case.
6121 ** This is done before calling sqlite3PcacheMakeDirty() on the page.
6122 ** Otherwise, if it were done after calling sqlite3PcacheMakeDirty(), then
6123 ** an error might occur and the pager would end up in WRITER_LOCKED state
6124 ** with pages marked as dirty in the cache.
6126 if( pPager->eState==PAGER_WRITER_LOCKED ){
6127 rc = pager_open_journal(pPager);
6128 if( rc!=SQLITE_OK ) return rc;
6130 assert( pPager->eState>=PAGER_WRITER_CACHEMOD );
6131 assert( assert_pager_state(pPager) );
6133 /* Mark the page that is about to be modified as dirty. */
6134 sqlite3PcacheMakeDirty(pPg);
6136 /* If a rollback journal is in use, them make sure the page that is about
6137 ** to change is in the rollback journal, or if the page is a new page off
6138 ** then end of the file, make sure it is marked as PGHDR_NEED_SYNC.
6140 assert( (pPager->pInJournal!=0) == isOpen(pPager->jfd) );
6141 if( pPager->pInJournal!=0
6142 && sqlite3BitvecTestNotNull(pPager->pInJournal, pPg->pgno)==0
6144 assert( pagerUseWal(pPager)==0 );
6145 if( pPg->pgno<=pPager->dbOrigSize ){
6146 rc = pagerAddPageToRollbackJournal(pPg);
6147 if( rc!=SQLITE_OK ){
6148 return rc;
6150 }else{
6151 if( pPager->eState!=PAGER_WRITER_DBMOD ){
6152 pPg->flags |= PGHDR_NEED_SYNC;
6154 PAGERTRACE(("APPEND %d page %d needSync=%d\n",
6155 PAGERID(pPager), pPg->pgno,
6156 ((pPg->flags&PGHDR_NEED_SYNC)?1:0)));
6160 /* The PGHDR_DIRTY bit is set above when the page was added to the dirty-list
6161 ** and before writing the page into the rollback journal. Wait until now,
6162 ** after the page has been successfully journalled, before setting the
6163 ** PGHDR_WRITEABLE bit that indicates that the page can be safely modified.
6165 pPg->flags |= PGHDR_WRITEABLE;
6167 /* If the statement journal is open and the page is not in it,
6168 ** then write the page into the statement journal.
6170 if( pPager->nSavepoint>0 ){
6171 rc = subjournalPageIfRequired(pPg);
6174 /* Update the database size and return. */
6175 if( pPager->dbSize<pPg->pgno ){
6176 pPager->dbSize = pPg->pgno;
6178 return rc;
6182 ** This is a variant of sqlite3PagerWrite() that runs when the sector size
6183 ** is larger than the page size. SQLite makes the (reasonable) assumption that
6184 ** all bytes of a sector are written together by hardware. Hence, all bytes of
6185 ** a sector need to be journalled in case of a power loss in the middle of
6186 ** a write.
6188 ** Usually, the sector size is less than or equal to the page size, in which
6189 ** case pages can be individually written. This routine only runs in the
6190 ** exceptional case where the page size is smaller than the sector size.
6192 static SQLITE_NOINLINE int pagerWriteLargeSector(PgHdr *pPg){
6193 int rc = SQLITE_OK; /* Return code */
6194 Pgno nPageCount; /* Total number of pages in database file */
6195 Pgno pg1; /* First page of the sector pPg is located on. */
6196 int nPage = 0; /* Number of pages starting at pg1 to journal */
6197 int ii; /* Loop counter */
6198 int needSync = 0; /* True if any page has PGHDR_NEED_SYNC */
6199 Pager *pPager = pPg->pPager; /* The pager that owns pPg */
6200 Pgno nPagePerSector = (pPager->sectorSize/pPager->pageSize);
6202 /* Set the doNotSpill NOSYNC bit to 1. This is because we cannot allow
6203 ** a journal header to be written between the pages journaled by
6204 ** this function.
6206 assert( !MEMDB );
6207 assert( (pPager->doNotSpill & SPILLFLAG_NOSYNC)==0 );
6208 pPager->doNotSpill |= SPILLFLAG_NOSYNC;
6210 /* This trick assumes that both the page-size and sector-size are
6211 ** an integer power of 2. It sets variable pg1 to the identifier
6212 ** of the first page of the sector pPg is located on.
6214 pg1 = ((pPg->pgno-1) & ~(nPagePerSector-1)) + 1;
6216 nPageCount = pPager->dbSize;
6217 if( pPg->pgno>nPageCount ){
6218 nPage = (pPg->pgno - pg1)+1;
6219 }else if( (pg1+nPagePerSector-1)>nPageCount ){
6220 nPage = nPageCount+1-pg1;
6221 }else{
6222 nPage = nPagePerSector;
6224 assert(nPage>0);
6225 assert(pg1<=pPg->pgno);
6226 assert((pg1+nPage)>pPg->pgno);
6228 for(ii=0; ii<nPage && rc==SQLITE_OK; ii++){
6229 Pgno pg = pg1+ii;
6230 PgHdr *pPage;
6231 if( pg==pPg->pgno || !sqlite3BitvecTest(pPager->pInJournal, pg) ){
6232 if( pg!=PAGER_SJ_PGNO(pPager) ){
6233 rc = sqlite3PagerGet(pPager, pg, &pPage, 0);
6234 if( rc==SQLITE_OK ){
6235 rc = pager_write(pPage);
6236 if( pPage->flags&PGHDR_NEED_SYNC ){
6237 needSync = 1;
6239 sqlite3PagerUnrefNotNull(pPage);
6242 }else if( (pPage = sqlite3PagerLookup(pPager, pg))!=0 ){
6243 if( pPage->flags&PGHDR_NEED_SYNC ){
6244 needSync = 1;
6246 sqlite3PagerUnrefNotNull(pPage);
6250 /* If the PGHDR_NEED_SYNC flag is set for any of the nPage pages
6251 ** starting at pg1, then it needs to be set for all of them. Because
6252 ** writing to any of these nPage pages may damage the others, the
6253 ** journal file must contain sync()ed copies of all of them
6254 ** before any of them can be written out to the database file.
6256 if( rc==SQLITE_OK && needSync ){
6257 assert( !MEMDB );
6258 for(ii=0; ii<nPage; ii++){
6259 PgHdr *pPage = sqlite3PagerLookup(pPager, pg1+ii);
6260 if( pPage ){
6261 pPage->flags |= PGHDR_NEED_SYNC;
6262 sqlite3PagerUnrefNotNull(pPage);
6267 assert( (pPager->doNotSpill & SPILLFLAG_NOSYNC)!=0 );
6268 pPager->doNotSpill &= ~SPILLFLAG_NOSYNC;
6269 return rc;
6273 ** Mark a data page as writeable. This routine must be called before
6274 ** making changes to a page. The caller must check the return value
6275 ** of this function and be careful not to change any page data unless
6276 ** this routine returns SQLITE_OK.
6278 ** The difference between this function and pager_write() is that this
6279 ** function also deals with the special case where 2 or more pages
6280 ** fit on a single disk sector. In this case all co-resident pages
6281 ** must have been written to the journal file before returning.
6283 ** If an error occurs, SQLITE_NOMEM or an IO error code is returned
6284 ** as appropriate. Otherwise, SQLITE_OK.
6286 int sqlite3PagerWrite(PgHdr *pPg){
6287 Pager *pPager = pPg->pPager;
6288 assert( (pPg->flags & PGHDR_MMAP)==0 );
6289 assert( pPager->eState>=PAGER_WRITER_LOCKED );
6290 assert( assert_pager_state(pPager) );
6291 if( (pPg->flags & PGHDR_WRITEABLE)!=0 && pPager->dbSize>=pPg->pgno ){
6292 if( pPager->nSavepoint ) return subjournalPageIfRequired(pPg);
6293 return SQLITE_OK;
6294 }else if( pPager->errCode ){
6295 return pPager->errCode;
6296 }else if( pPager->sectorSize > (u32)pPager->pageSize ){
6297 assert( pPager->tempFile==0 );
6298 return pagerWriteLargeSector(pPg);
6299 }else{
6300 return pager_write(pPg);
6305 ** Return TRUE if the page given in the argument was previously passed
6306 ** to sqlite3PagerWrite(). In other words, return TRUE if it is ok
6307 ** to change the content of the page.
6309 #ifndef NDEBUG
6310 int sqlite3PagerIswriteable(DbPage *pPg){
6311 return pPg->flags & PGHDR_WRITEABLE;
6313 #endif
6316 ** A call to this routine tells the pager that it is not necessary to
6317 ** write the information on page pPg back to the disk, even though
6318 ** that page might be marked as dirty. This happens, for example, when
6319 ** the page has been added as a leaf of the freelist and so its
6320 ** content no longer matters.
6322 ** The overlying software layer calls this routine when all of the data
6323 ** on the given page is unused. The pager marks the page as clean so
6324 ** that it does not get written to disk.
6326 ** Tests show that this optimization can quadruple the speed of large
6327 ** DELETE operations.
6329 ** This optimization cannot be used with a temp-file, as the page may
6330 ** have been dirty at the start of the transaction. In that case, if
6331 ** memory pressure forces page pPg out of the cache, the data does need
6332 ** to be written out to disk so that it may be read back in if the
6333 ** current transaction is rolled back.
6335 void sqlite3PagerDontWrite(PgHdr *pPg){
6336 Pager *pPager = pPg->pPager;
6337 if( !pPager->tempFile && (pPg->flags&PGHDR_DIRTY) && pPager->nSavepoint==0 ){
6338 PAGERTRACE(("DONT_WRITE page %d of %d\n", pPg->pgno, PAGERID(pPager)));
6339 IOTRACE(("CLEAN %p %d\n", pPager, pPg->pgno))
6340 pPg->flags |= PGHDR_DONT_WRITE;
6341 pPg->flags &= ~PGHDR_WRITEABLE;
6342 testcase( pPg->flags & PGHDR_NEED_SYNC );
6343 pager_set_pagehash(pPg);
6348 ** This routine is called to increment the value of the database file
6349 ** change-counter, stored as a 4-byte big-endian integer starting at
6350 ** byte offset 24 of the pager file. The secondary change counter at
6351 ** 92 is also updated, as is the SQLite version number at offset 96.
6353 ** But this only happens if the pPager->changeCountDone flag is false.
6354 ** To avoid excess churning of page 1, the update only happens once.
6355 ** See also the pager_write_changecounter() routine that does an
6356 ** unconditional update of the change counters.
6358 ** If the isDirectMode flag is zero, then this is done by calling
6359 ** sqlite3PagerWrite() on page 1, then modifying the contents of the
6360 ** page data. In this case the file will be updated when the current
6361 ** transaction is committed.
6363 ** The isDirectMode flag may only be non-zero if the library was compiled
6364 ** with the SQLITE_ENABLE_ATOMIC_WRITE macro defined. In this case,
6365 ** if isDirect is non-zero, then the database file is updated directly
6366 ** by writing an updated version of page 1 using a call to the
6367 ** sqlite3OsWrite() function.
6369 static int pager_incr_changecounter(Pager *pPager, int isDirectMode){
6370 int rc = SQLITE_OK;
6372 assert( pPager->eState==PAGER_WRITER_CACHEMOD
6373 || pPager->eState==PAGER_WRITER_DBMOD
6375 assert( assert_pager_state(pPager) );
6377 /* Declare and initialize constant integer 'isDirect'. If the
6378 ** atomic-write optimization is enabled in this build, then isDirect
6379 ** is initialized to the value passed as the isDirectMode parameter
6380 ** to this function. Otherwise, it is always set to zero.
6382 ** The idea is that if the atomic-write optimization is not
6383 ** enabled at compile time, the compiler can omit the tests of
6384 ** 'isDirect' below, as well as the block enclosed in the
6385 ** "if( isDirect )" condition.
6387 #ifndef SQLITE_ENABLE_ATOMIC_WRITE
6388 # define DIRECT_MODE 0
6389 assert( isDirectMode==0 );
6390 UNUSED_PARAMETER(isDirectMode);
6391 #else
6392 # define DIRECT_MODE isDirectMode
6393 #endif
6395 if( !pPager->changeCountDone && ALWAYS(pPager->dbSize>0) ){
6396 PgHdr *pPgHdr; /* Reference to page 1 */
6398 assert( !pPager->tempFile && isOpen(pPager->fd) );
6400 /* Open page 1 of the file for writing. */
6401 rc = sqlite3PagerGet(pPager, 1, &pPgHdr, 0);
6402 assert( pPgHdr==0 || rc==SQLITE_OK );
6404 /* If page one was fetched successfully, and this function is not
6405 ** operating in direct-mode, make page 1 writable. When not in
6406 ** direct mode, page 1 is always held in cache and hence the PagerGet()
6407 ** above is always successful - hence the ALWAYS on rc==SQLITE_OK.
6409 if( !DIRECT_MODE && ALWAYS(rc==SQLITE_OK) ){
6410 rc = sqlite3PagerWrite(pPgHdr);
6413 if( rc==SQLITE_OK ){
6414 /* Actually do the update of the change counter */
6415 pager_write_changecounter(pPgHdr);
6417 /* If running in direct mode, write the contents of page 1 to the file. */
6418 if( DIRECT_MODE ){
6419 const void *zBuf;
6420 assert( pPager->dbFileSize>0 );
6421 CODEC2(pPager, pPgHdr->pData, 1, 6, rc=SQLITE_NOMEM_BKPT, zBuf);
6422 if( rc==SQLITE_OK ){
6423 rc = sqlite3OsWrite(pPager->fd, zBuf, pPager->pageSize, 0);
6424 pPager->aStat[PAGER_STAT_WRITE]++;
6426 if( rc==SQLITE_OK ){
6427 /* Update the pager's copy of the change-counter. Otherwise, the
6428 ** next time a read transaction is opened the cache will be
6429 ** flushed (as the change-counter values will not match). */
6430 const void *pCopy = (const void *)&((const char *)zBuf)[24];
6431 memcpy(&pPager->dbFileVers, pCopy, sizeof(pPager->dbFileVers));
6432 pPager->changeCountDone = 1;
6434 }else{
6435 pPager->changeCountDone = 1;
6439 /* Release the page reference. */
6440 sqlite3PagerUnref(pPgHdr);
6442 return rc;
6446 ** Sync the database file to disk. This is a no-op for in-memory databases
6447 ** or pages with the Pager.noSync flag set.
6449 ** If successful, or if called on a pager for which it is a no-op, this
6450 ** function returns SQLITE_OK. Otherwise, an IO error code is returned.
6452 int sqlite3PagerSync(Pager *pPager, const char *zSuper){
6453 int rc = SQLITE_OK;
6454 void *pArg = (void*)zSuper;
6455 rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SYNC, pArg);
6456 if( rc==SQLITE_NOTFOUND ) rc = SQLITE_OK;
6457 if( rc==SQLITE_OK && !pPager->noSync ){
6458 assert( !MEMDB );
6459 rc = sqlite3OsSync(pPager->fd, pPager->syncFlags);
6461 return rc;
6465 ** This function may only be called while a write-transaction is active in
6466 ** rollback. If the connection is in WAL mode, this call is a no-op.
6467 ** Otherwise, if the connection does not already have an EXCLUSIVE lock on
6468 ** the database file, an attempt is made to obtain one.
6470 ** If the EXCLUSIVE lock is already held or the attempt to obtain it is
6471 ** successful, or the connection is in WAL mode, SQLITE_OK is returned.
6472 ** Otherwise, either SQLITE_BUSY or an SQLITE_IOERR_XXX error code is
6473 ** returned.
6475 int sqlite3PagerExclusiveLock(Pager *pPager){
6476 int rc = pPager->errCode;
6477 assert( assert_pager_state(pPager) );
6478 if( rc==SQLITE_OK ){
6479 assert( pPager->eState==PAGER_WRITER_CACHEMOD
6480 || pPager->eState==PAGER_WRITER_DBMOD
6481 || pPager->eState==PAGER_WRITER_LOCKED
6483 assert( assert_pager_state(pPager) );
6484 if( 0==pagerUseWal(pPager) ){
6485 rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
6488 return rc;
6492 ** Sync the database file for the pager pPager. zSuper points to the name
6493 ** of a super-journal file that should be written into the individual
6494 ** journal file. zSuper may be NULL, which is interpreted as no
6495 ** super-journal (a single database transaction).
6497 ** This routine ensures that:
6499 ** * The database file change-counter is updated,
6500 ** * the journal is synced (unless the atomic-write optimization is used),
6501 ** * all dirty pages are written to the database file,
6502 ** * the database file is truncated (if required), and
6503 ** * the database file synced.
6505 ** The only thing that remains to commit the transaction is to finalize
6506 ** (delete, truncate or zero the first part of) the journal file (or
6507 ** delete the super-journal file if specified).
6509 ** Note that if zSuper==NULL, this does not overwrite a previous value
6510 ** passed to an sqlite3PagerCommitPhaseOne() call.
6512 ** If the final parameter - noSync - is true, then the database file itself
6513 ** is not synced. The caller must call sqlite3PagerSync() directly to
6514 ** sync the database file before calling CommitPhaseTwo() to delete the
6515 ** journal file in this case.
6517 int sqlite3PagerCommitPhaseOne(
6518 Pager *pPager, /* Pager object */
6519 const char *zSuper, /* If not NULL, the super-journal name */
6520 int noSync /* True to omit the xSync on the db file */
6522 int rc = SQLITE_OK; /* Return code */
6524 assert( pPager->eState==PAGER_WRITER_LOCKED
6525 || pPager->eState==PAGER_WRITER_CACHEMOD
6526 || pPager->eState==PAGER_WRITER_DBMOD
6527 || pPager->eState==PAGER_ERROR
6529 assert( assert_pager_state(pPager) );
6531 /* If a prior error occurred, report that error again. */
6532 if( NEVER(pPager->errCode) ) return pPager->errCode;
6534 /* Provide the ability to easily simulate an I/O error during testing */
6535 if( sqlite3FaultSim(400) ) return SQLITE_IOERR;
6537 PAGERTRACE(("DATABASE SYNC: File=%s zSuper=%s nSize=%d\n",
6538 pPager->zFilename, zSuper, pPager->dbSize));
6540 /* If no database changes have been made, return early. */
6541 if( pPager->eState<PAGER_WRITER_CACHEMOD ) return SQLITE_OK;
6543 assert( MEMDB==0 || pPager->tempFile );
6544 assert( isOpen(pPager->fd) || pPager->tempFile );
6545 if( 0==pagerFlushOnCommit(pPager, 1) ){
6546 /* If this is an in-memory db, or no pages have been written to, or this
6547 ** function has already been called, it is mostly a no-op. However, any
6548 ** backup in progress needs to be restarted. */
6549 sqlite3BackupRestart(pPager->pBackup);
6550 }else{
6551 PgHdr *pList;
6552 if( pagerUseWal(pPager) ){
6553 PgHdr *pPageOne = 0;
6554 pList = sqlite3PcacheDirtyList(pPager->pPCache);
6555 if( pList==0 ){
6556 /* Must have at least one page for the WAL commit flag.
6557 ** Ticket [2d1a5c67dfc2363e44f29d9bbd57f] 2011-05-18 */
6558 rc = sqlite3PagerGet(pPager, 1, &pPageOne, 0);
6559 pList = pPageOne;
6560 pList->pDirty = 0;
6562 assert( rc==SQLITE_OK );
6563 if( ALWAYS(pList) ){
6564 rc = pagerWalFrames(pPager, pList, pPager->dbSize, 1);
6566 sqlite3PagerUnref(pPageOne);
6567 if( rc==SQLITE_OK ){
6568 sqlite3PcacheCleanAll(pPager->pPCache);
6570 }else{
6571 /* The bBatch boolean is true if the batch-atomic-write commit method
6572 ** should be used. No rollback journal is created if batch-atomic-write
6573 ** is enabled.
6575 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6576 sqlite3_file *fd = pPager->fd;
6577 int bBatch = zSuper==0 /* An SQLITE_IOCAP_BATCH_ATOMIC commit */
6578 && (sqlite3OsDeviceCharacteristics(fd) & SQLITE_IOCAP_BATCH_ATOMIC)
6579 && !pPager->noSync
6580 && sqlite3JournalIsInMemory(pPager->jfd);
6581 #else
6582 # define bBatch 0
6583 #endif
6585 #ifdef SQLITE_ENABLE_ATOMIC_WRITE
6586 /* The following block updates the change-counter. Exactly how it
6587 ** does this depends on whether or not the atomic-update optimization
6588 ** was enabled at compile time, and if this transaction meets the
6589 ** runtime criteria to use the operation:
6591 ** * The file-system supports the atomic-write property for
6592 ** blocks of size page-size, and
6593 ** * This commit is not part of a multi-file transaction, and
6594 ** * Exactly one page has been modified and store in the journal file.
6596 ** If the optimization was not enabled at compile time, then the
6597 ** pager_incr_changecounter() function is called to update the change
6598 ** counter in 'indirect-mode'. If the optimization is compiled in but
6599 ** is not applicable to this transaction, call sqlite3JournalCreate()
6600 ** to make sure the journal file has actually been created, then call
6601 ** pager_incr_changecounter() to update the change-counter in indirect
6602 ** mode.
6604 ** Otherwise, if the optimization is both enabled and applicable,
6605 ** then call pager_incr_changecounter() to update the change-counter
6606 ** in 'direct' mode. In this case the journal file will never be
6607 ** created for this transaction.
6609 if( bBatch==0 ){
6610 PgHdr *pPg;
6611 assert( isOpen(pPager->jfd)
6612 || pPager->journalMode==PAGER_JOURNALMODE_OFF
6613 || pPager->journalMode==PAGER_JOURNALMODE_WAL
6615 if( !zSuper && isOpen(pPager->jfd)
6616 && pPager->journalOff==jrnlBufferSize(pPager)
6617 && pPager->dbSize>=pPager->dbOrigSize
6618 && (!(pPg = sqlite3PcacheDirtyList(pPager->pPCache)) || 0==pPg->pDirty)
6620 /* Update the db file change counter via the direct-write method. The
6621 ** following call will modify the in-memory representation of page 1
6622 ** to include the updated change counter and then write page 1
6623 ** directly to the database file. Because of the atomic-write
6624 ** property of the host file-system, this is safe.
6626 rc = pager_incr_changecounter(pPager, 1);
6627 }else{
6628 rc = sqlite3JournalCreate(pPager->jfd);
6629 if( rc==SQLITE_OK ){
6630 rc = pager_incr_changecounter(pPager, 0);
6634 #else /* SQLITE_ENABLE_ATOMIC_WRITE */
6635 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6636 if( zSuper ){
6637 rc = sqlite3JournalCreate(pPager->jfd);
6638 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6639 assert( bBatch==0 );
6641 #endif
6642 rc = pager_incr_changecounter(pPager, 0);
6643 #endif /* !SQLITE_ENABLE_ATOMIC_WRITE */
6644 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6646 /* Write the super-journal name into the journal file. If a
6647 ** super-journal file name has already been written to the journal file,
6648 ** or if zSuper is NULL (no super-journal), then this call is a no-op.
6650 rc = writeSuperJournal(pPager, zSuper);
6651 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6653 /* Sync the journal file and write all dirty pages to the database.
6654 ** If the atomic-update optimization is being used, this sync will not
6655 ** create the journal file or perform any real IO.
6657 ** Because the change-counter page was just modified, unless the
6658 ** atomic-update optimization is used it is almost certain that the
6659 ** journal requires a sync here. However, in locking_mode=exclusive
6660 ** on a system under memory pressure it is just possible that this is
6661 ** not the case. In this case it is likely enough that the redundant
6662 ** xSync() call will be changed to a no-op by the OS anyhow.
6664 rc = syncJournal(pPager, 0);
6665 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6667 pList = sqlite3PcacheDirtyList(pPager->pPCache);
6668 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6669 if( bBatch ){
6670 rc = sqlite3OsFileControl(fd, SQLITE_FCNTL_BEGIN_ATOMIC_WRITE, 0);
6671 if( rc==SQLITE_OK ){
6672 rc = pager_write_pagelist(pPager, pList);
6673 if( rc==SQLITE_OK ){
6674 rc = sqlite3OsFileControl(fd, SQLITE_FCNTL_COMMIT_ATOMIC_WRITE, 0);
6676 if( rc!=SQLITE_OK ){
6677 sqlite3OsFileControlHint(fd, SQLITE_FCNTL_ROLLBACK_ATOMIC_WRITE, 0);
6681 if( (rc&0xFF)==SQLITE_IOERR && rc!=SQLITE_IOERR_NOMEM ){
6682 rc = sqlite3JournalCreate(pPager->jfd);
6683 if( rc!=SQLITE_OK ){
6684 sqlite3OsClose(pPager->jfd);
6685 goto commit_phase_one_exit;
6687 bBatch = 0;
6688 }else{
6689 sqlite3OsClose(pPager->jfd);
6692 #endif /* SQLITE_ENABLE_BATCH_ATOMIC_WRITE */
6694 if( bBatch==0 ){
6695 rc = pager_write_pagelist(pPager, pList);
6697 if( rc!=SQLITE_OK ){
6698 assert( rc!=SQLITE_IOERR_BLOCKED );
6699 goto commit_phase_one_exit;
6701 sqlite3PcacheCleanAll(pPager->pPCache);
6703 /* If the file on disk is smaller than the database image, use
6704 ** pager_truncate to grow the file here. This can happen if the database
6705 ** image was extended as part of the current transaction and then the
6706 ** last page in the db image moved to the free-list. In this case the
6707 ** last page is never written out to disk, leaving the database file
6708 ** undersized. Fix this now if it is the case. */
6709 if( pPager->dbSize>pPager->dbFileSize ){
6710 Pgno nNew = pPager->dbSize - (pPager->dbSize==PAGER_SJ_PGNO(pPager));
6711 assert( pPager->eState==PAGER_WRITER_DBMOD );
6712 rc = pager_truncate(pPager, nNew);
6713 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6716 /* Finally, sync the database file. */
6717 if( !noSync ){
6718 rc = sqlite3PagerSync(pPager, zSuper);
6720 IOTRACE(("DBSYNC %p\n", pPager))
6724 commit_phase_one_exit:
6725 if( rc==SQLITE_OK && !pagerUseWal(pPager) ){
6726 pPager->eState = PAGER_WRITER_FINISHED;
6728 return rc;
6733 ** When this function is called, the database file has been completely
6734 ** updated to reflect the changes made by the current transaction and
6735 ** synced to disk. The journal file still exists in the file-system
6736 ** though, and if a failure occurs at this point it will eventually
6737 ** be used as a hot-journal and the current transaction rolled back.
6739 ** This function finalizes the journal file, either by deleting,
6740 ** truncating or partially zeroing it, so that it cannot be used
6741 ** for hot-journal rollback. Once this is done the transaction is
6742 ** irrevocably committed.
6744 ** If an error occurs, an IO error code is returned and the pager
6745 ** moves into the error state. Otherwise, SQLITE_OK is returned.
6747 int sqlite3PagerCommitPhaseTwo(Pager *pPager){
6748 int rc = SQLITE_OK; /* Return code */
6750 /* This routine should not be called if a prior error has occurred.
6751 ** But if (due to a coding error elsewhere in the system) it does get
6752 ** called, just return the same error code without doing anything. */
6753 if( NEVER(pPager->errCode) ) return pPager->errCode;
6754 pPager->iDataVersion++;
6756 assert( pPager->eState==PAGER_WRITER_LOCKED
6757 || pPager->eState==PAGER_WRITER_FINISHED
6758 || (pagerUseWal(pPager) && pPager->eState==PAGER_WRITER_CACHEMOD)
6760 assert( assert_pager_state(pPager) );
6762 /* An optimization. If the database was not actually modified during
6763 ** this transaction, the pager is running in exclusive-mode and is
6764 ** using persistent journals, then this function is a no-op.
6766 ** The start of the journal file currently contains a single journal
6767 ** header with the nRec field set to 0. If such a journal is used as
6768 ** a hot-journal during hot-journal rollback, 0 changes will be made
6769 ** to the database file. So there is no need to zero the journal
6770 ** header. Since the pager is in exclusive mode, there is no need
6771 ** to drop any locks either.
6773 if( pPager->eState==PAGER_WRITER_LOCKED
6774 && pPager->exclusiveMode
6775 && pPager->journalMode==PAGER_JOURNALMODE_PERSIST
6777 assert( pPager->journalOff==JOURNAL_HDR_SZ(pPager) || !pPager->journalOff );
6778 pPager->eState = PAGER_READER;
6779 return SQLITE_OK;
6782 PAGERTRACE(("COMMIT %d\n", PAGERID(pPager)));
6783 rc = pager_end_transaction(pPager, pPager->setSuper, 1);
6784 return pager_error(pPager, rc);
6788 ** If a write transaction is open, then all changes made within the
6789 ** transaction are reverted and the current write-transaction is closed.
6790 ** The pager falls back to PAGER_READER state if successful, or PAGER_ERROR
6791 ** state if an error occurs.
6793 ** If the pager is already in PAGER_ERROR state when this function is called,
6794 ** it returns Pager.errCode immediately. No work is performed in this case.
6796 ** Otherwise, in rollback mode, this function performs two functions:
6798 ** 1) It rolls back the journal file, restoring all database file and
6799 ** in-memory cache pages to the state they were in when the transaction
6800 ** was opened, and
6802 ** 2) It finalizes the journal file, so that it is not used for hot
6803 ** rollback at any point in the future.
6805 ** Finalization of the journal file (task 2) is only performed if the
6806 ** rollback is successful.
6808 ** In WAL mode, all cache-entries containing data modified within the
6809 ** current transaction are either expelled from the cache or reverted to
6810 ** their pre-transaction state by re-reading data from the database or
6811 ** WAL files. The WAL transaction is then closed.
6813 int sqlite3PagerRollback(Pager *pPager){
6814 int rc = SQLITE_OK; /* Return code */
6815 PAGERTRACE(("ROLLBACK %d\n", PAGERID(pPager)));
6817 /* PagerRollback() is a no-op if called in READER or OPEN state. If
6818 ** the pager is already in the ERROR state, the rollback is not
6819 ** attempted here. Instead, the error code is returned to the caller.
6821 assert( assert_pager_state(pPager) );
6822 if( pPager->eState==PAGER_ERROR ) return pPager->errCode;
6823 if( pPager->eState<=PAGER_READER ) return SQLITE_OK;
6825 if( pagerUseWal(pPager) ){
6826 int rc2;
6827 rc = sqlite3PagerSavepoint(pPager, SAVEPOINT_ROLLBACK, -1);
6828 rc2 = pager_end_transaction(pPager, pPager->setSuper, 0);
6829 if( rc==SQLITE_OK ) rc = rc2;
6830 }else if( !isOpen(pPager->jfd) || pPager->eState==PAGER_WRITER_LOCKED ){
6831 int eState = pPager->eState;
6832 rc = pager_end_transaction(pPager, 0, 0);
6833 if( !MEMDB && eState>PAGER_WRITER_LOCKED ){
6834 /* This can happen using journal_mode=off. Move the pager to the error
6835 ** state to indicate that the contents of the cache may not be trusted.
6836 ** Any active readers will get SQLITE_ABORT.
6838 pPager->errCode = SQLITE_ABORT;
6839 pPager->eState = PAGER_ERROR;
6840 setGetterMethod(pPager);
6841 return rc;
6843 }else{
6844 rc = pager_playback(pPager, 0);
6847 assert( pPager->eState==PAGER_READER || rc!=SQLITE_OK );
6848 assert( rc==SQLITE_OK || rc==SQLITE_FULL || rc==SQLITE_CORRUPT
6849 || rc==SQLITE_NOMEM || (rc&0xFF)==SQLITE_IOERR
6850 || rc==SQLITE_CANTOPEN
6853 /* If an error occurs during a ROLLBACK, we can no longer trust the pager
6854 ** cache. So call pager_error() on the way out to make any error persistent.
6856 return pager_error(pPager, rc);
6860 ** Return TRUE if the database file is opened read-only. Return FALSE
6861 ** if the database is (in theory) writable.
6863 u8 sqlite3PagerIsreadonly(Pager *pPager){
6864 return pPager->readOnly;
6867 #ifdef SQLITE_DEBUG
6869 ** Return the sum of the reference counts for all pages held by pPager.
6871 int sqlite3PagerRefcount(Pager *pPager){
6872 return sqlite3PcacheRefCount(pPager->pPCache);
6874 #endif
6877 ** Return the approximate number of bytes of memory currently
6878 ** used by the pager and its associated cache.
6880 int sqlite3PagerMemUsed(Pager *pPager){
6881 int perPageSize = pPager->pageSize + pPager->nExtra
6882 + (int)(sizeof(PgHdr) + 5*sizeof(void*));
6883 return perPageSize*sqlite3PcachePagecount(pPager->pPCache)
6884 + sqlite3MallocSize(pPager)
6885 + pPager->pageSize;
6889 ** Return the number of references to the specified page.
6891 int sqlite3PagerPageRefcount(DbPage *pPage){
6892 return sqlite3PcachePageRefcount(pPage);
6895 #ifdef SQLITE_TEST
6897 ** This routine is used for testing and analysis only.
6899 int *sqlite3PagerStats(Pager *pPager){
6900 static int a[11];
6901 a[0] = sqlite3PcacheRefCount(pPager->pPCache);
6902 a[1] = sqlite3PcachePagecount(pPager->pPCache);
6903 a[2] = sqlite3PcacheGetCachesize(pPager->pPCache);
6904 a[3] = pPager->eState==PAGER_OPEN ? -1 : (int) pPager->dbSize;
6905 a[4] = pPager->eState;
6906 a[5] = pPager->errCode;
6907 a[6] = pPager->aStat[PAGER_STAT_HIT];
6908 a[7] = pPager->aStat[PAGER_STAT_MISS];
6909 a[8] = 0; /* Used to be pPager->nOvfl */
6910 a[9] = pPager->nRead;
6911 a[10] = pPager->aStat[PAGER_STAT_WRITE];
6912 return a;
6914 #endif
6917 ** Parameter eStat must be one of SQLITE_DBSTATUS_CACHE_HIT, _MISS, _WRITE,
6918 ** or _WRITE+1. The SQLITE_DBSTATUS_CACHE_WRITE+1 case is a translation
6919 ** of SQLITE_DBSTATUS_CACHE_SPILL. The _SPILL case is not contiguous because
6920 ** it was added later.
6922 ** Before returning, *pnVal is incremented by the
6923 ** current cache hit or miss count, according to the value of eStat. If the
6924 ** reset parameter is non-zero, the cache hit or miss count is zeroed before
6925 ** returning.
6927 void sqlite3PagerCacheStat(Pager *pPager, int eStat, int reset, int *pnVal){
6929 assert( eStat==SQLITE_DBSTATUS_CACHE_HIT
6930 || eStat==SQLITE_DBSTATUS_CACHE_MISS
6931 || eStat==SQLITE_DBSTATUS_CACHE_WRITE
6932 || eStat==SQLITE_DBSTATUS_CACHE_WRITE+1
6935 assert( SQLITE_DBSTATUS_CACHE_HIT+1==SQLITE_DBSTATUS_CACHE_MISS );
6936 assert( SQLITE_DBSTATUS_CACHE_HIT+2==SQLITE_DBSTATUS_CACHE_WRITE );
6937 assert( PAGER_STAT_HIT==0 && PAGER_STAT_MISS==1
6938 && PAGER_STAT_WRITE==2 && PAGER_STAT_SPILL==3 );
6940 eStat -= SQLITE_DBSTATUS_CACHE_HIT;
6941 *pnVal += pPager->aStat[eStat];
6942 if( reset ){
6943 pPager->aStat[eStat] = 0;
6948 ** Return true if this is an in-memory or temp-file backed pager.
6950 int sqlite3PagerIsMemdb(Pager *pPager){
6951 return pPager->tempFile || pPager->memVfs;
6955 ** Check that there are at least nSavepoint savepoints open. If there are
6956 ** currently less than nSavepoints open, then open one or more savepoints
6957 ** to make up the difference. If the number of savepoints is already
6958 ** equal to nSavepoint, then this function is a no-op.
6960 ** If a memory allocation fails, SQLITE_NOMEM is returned. If an error
6961 ** occurs while opening the sub-journal file, then an IO error code is
6962 ** returned. Otherwise, SQLITE_OK.
6964 static SQLITE_NOINLINE int pagerOpenSavepoint(Pager *pPager, int nSavepoint){
6965 int rc = SQLITE_OK; /* Return code */
6966 int nCurrent = pPager->nSavepoint; /* Current number of savepoints */
6967 int ii; /* Iterator variable */
6968 PagerSavepoint *aNew; /* New Pager.aSavepoint array */
6970 assert( pPager->eState>=PAGER_WRITER_LOCKED );
6971 assert( assert_pager_state(pPager) );
6972 assert( nSavepoint>nCurrent && pPager->useJournal );
6974 /* Grow the Pager.aSavepoint array using realloc(). Return SQLITE_NOMEM
6975 ** if the allocation fails. Otherwise, zero the new portion in case a
6976 ** malloc failure occurs while populating it in the for(...) loop below.
6978 aNew = (PagerSavepoint *)sqlite3Realloc(
6979 pPager->aSavepoint, sizeof(PagerSavepoint)*nSavepoint
6981 if( !aNew ){
6982 return SQLITE_NOMEM_BKPT;
6984 memset(&aNew[nCurrent], 0, (nSavepoint-nCurrent) * sizeof(PagerSavepoint));
6985 pPager->aSavepoint = aNew;
6987 /* Populate the PagerSavepoint structures just allocated. */
6988 for(ii=nCurrent; ii<nSavepoint; ii++){
6989 aNew[ii].nOrig = pPager->dbSize;
6990 if( isOpen(pPager->jfd) && pPager->journalOff>0 ){
6991 aNew[ii].iOffset = pPager->journalOff;
6992 }else{
6993 aNew[ii].iOffset = JOURNAL_HDR_SZ(pPager);
6995 aNew[ii].iSubRec = pPager->nSubRec;
6996 aNew[ii].pInSavepoint = sqlite3BitvecCreate(pPager->dbSize);
6997 aNew[ii].bTruncateOnRelease = 1;
6998 if( !aNew[ii].pInSavepoint ){
6999 return SQLITE_NOMEM_BKPT;
7001 if( pagerUseWal(pPager) ){
7002 sqlite3WalSavepoint(pPager->pWal, aNew[ii].aWalData);
7004 pPager->nSavepoint = ii+1;
7006 assert( pPager->nSavepoint==nSavepoint );
7007 assertTruncateConstraint(pPager);
7008 return rc;
7010 int sqlite3PagerOpenSavepoint(Pager *pPager, int nSavepoint){
7011 assert( pPager->eState>=PAGER_WRITER_LOCKED );
7012 assert( assert_pager_state(pPager) );
7014 if( nSavepoint>pPager->nSavepoint && pPager->useJournal ){
7015 return pagerOpenSavepoint(pPager, nSavepoint);
7016 }else{
7017 return SQLITE_OK;
7023 ** This function is called to rollback or release (commit) a savepoint.
7024 ** The savepoint to release or rollback need not be the most recently
7025 ** created savepoint.
7027 ** Parameter op is always either SAVEPOINT_ROLLBACK or SAVEPOINT_RELEASE.
7028 ** If it is SAVEPOINT_RELEASE, then release and destroy the savepoint with
7029 ** index iSavepoint. If it is SAVEPOINT_ROLLBACK, then rollback all changes
7030 ** that have occurred since the specified savepoint was created.
7032 ** The savepoint to rollback or release is identified by parameter
7033 ** iSavepoint. A value of 0 means to operate on the outermost savepoint
7034 ** (the first created). A value of (Pager.nSavepoint-1) means operate
7035 ** on the most recently created savepoint. If iSavepoint is greater than
7036 ** (Pager.nSavepoint-1), then this function is a no-op.
7038 ** If a negative value is passed to this function, then the current
7039 ** transaction is rolled back. This is different to calling
7040 ** sqlite3PagerRollback() because this function does not terminate
7041 ** the transaction or unlock the database, it just restores the
7042 ** contents of the database to its original state.
7044 ** In any case, all savepoints with an index greater than iSavepoint
7045 ** are destroyed. If this is a release operation (op==SAVEPOINT_RELEASE),
7046 ** then savepoint iSavepoint is also destroyed.
7048 ** This function may return SQLITE_NOMEM if a memory allocation fails,
7049 ** or an IO error code if an IO error occurs while rolling back a
7050 ** savepoint. If no errors occur, SQLITE_OK is returned.
7052 int sqlite3PagerSavepoint(Pager *pPager, int op, int iSavepoint){
7053 int rc = pPager->errCode;
7055 #ifdef SQLITE_ENABLE_ZIPVFS
7056 if( op==SAVEPOINT_RELEASE ) rc = SQLITE_OK;
7057 #endif
7059 assert( op==SAVEPOINT_RELEASE || op==SAVEPOINT_ROLLBACK );
7060 assert( iSavepoint>=0 || op==SAVEPOINT_ROLLBACK );
7062 if( rc==SQLITE_OK && iSavepoint<pPager->nSavepoint ){
7063 int ii; /* Iterator variable */
7064 int nNew; /* Number of remaining savepoints after this op. */
7066 /* Figure out how many savepoints will still be active after this
7067 ** operation. Store this value in nNew. Then free resources associated
7068 ** with any savepoints that are destroyed by this operation.
7070 nNew = iSavepoint + (( op==SAVEPOINT_RELEASE ) ? 0 : 1);
7071 for(ii=nNew; ii<pPager->nSavepoint; ii++){
7072 sqlite3BitvecDestroy(pPager->aSavepoint[ii].pInSavepoint);
7074 pPager->nSavepoint = nNew;
7076 /* Truncate the sub-journal so that it only includes the parts
7077 ** that are still in use. */
7078 if( op==SAVEPOINT_RELEASE ){
7079 PagerSavepoint *pRel = &pPager->aSavepoint[nNew];
7080 if( pRel->bTruncateOnRelease && isOpen(pPager->sjfd) ){
7081 /* Only truncate if it is an in-memory sub-journal. */
7082 if( sqlite3JournalIsInMemory(pPager->sjfd) ){
7083 i64 sz = (pPager->pageSize+4)*(i64)pRel->iSubRec;
7084 rc = sqlite3OsTruncate(pPager->sjfd, sz);
7085 assert( rc==SQLITE_OK );
7087 pPager->nSubRec = pRel->iSubRec;
7090 /* Else this is a rollback operation, playback the specified savepoint.
7091 ** If this is a temp-file, it is possible that the journal file has
7092 ** not yet been opened. In this case there have been no changes to
7093 ** the database file, so the playback operation can be skipped.
7095 else if( pagerUseWal(pPager) || isOpen(pPager->jfd) ){
7096 PagerSavepoint *pSavepoint = (nNew==0)?0:&pPager->aSavepoint[nNew-1];
7097 rc = pagerPlaybackSavepoint(pPager, pSavepoint);
7098 assert(rc!=SQLITE_DONE);
7101 #ifdef SQLITE_ENABLE_ZIPVFS
7102 /* If the cache has been modified but the savepoint cannot be rolled
7103 ** back journal_mode=off, put the pager in the error state. This way,
7104 ** if the VFS used by this pager includes ZipVFS, the entire transaction
7105 ** can be rolled back at the ZipVFS level. */
7106 else if(
7107 pPager->journalMode==PAGER_JOURNALMODE_OFF
7108 && pPager->eState>=PAGER_WRITER_CACHEMOD
7110 pPager->errCode = SQLITE_ABORT;
7111 pPager->eState = PAGER_ERROR;
7112 setGetterMethod(pPager);
7114 #endif
7117 return rc;
7121 ** Return the full pathname of the database file.
7123 ** Except, if the pager is in-memory only, then return an empty string if
7124 ** nullIfMemDb is true. This routine is called with nullIfMemDb==1 when
7125 ** used to report the filename to the user, for compatibility with legacy
7126 ** behavior. But when the Btree needs to know the filename for matching to
7127 ** shared cache, it uses nullIfMemDb==0 so that in-memory databases can
7128 ** participate in shared-cache.
7130 ** The return value to this routine is always safe to use with
7131 ** sqlite3_uri_parameter() and sqlite3_filename_database() and friends.
7133 const char *sqlite3PagerFilename(const Pager *pPager, int nullIfMemDb){
7134 static const char zFake[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
7135 return (nullIfMemDb && pPager->memDb) ? &zFake[4] : pPager->zFilename;
7139 ** Return the VFS structure for the pager.
7141 sqlite3_vfs *sqlite3PagerVfs(Pager *pPager){
7142 return pPager->pVfs;
7146 ** Return the file handle for the database file associated
7147 ** with the pager. This might return NULL if the file has
7148 ** not yet been opened.
7150 sqlite3_file *sqlite3PagerFile(Pager *pPager){
7151 return pPager->fd;
7155 ** Return the file handle for the journal file (if it exists).
7156 ** This will be either the rollback journal or the WAL file.
7158 sqlite3_file *sqlite3PagerJrnlFile(Pager *pPager){
7159 #if SQLITE_OMIT_WAL
7160 return pPager->jfd;
7161 #else
7162 return pPager->pWal ? sqlite3WalFile(pPager->pWal) : pPager->jfd;
7163 #endif
7167 ** Return the full pathname of the journal file.
7169 const char *sqlite3PagerJournalname(Pager *pPager){
7170 return pPager->zJournal;
7173 /* BEGIN SQLCIPHER */
7174 #ifdef SQLITE_HAS_CODEC
7176 ** Set or retrieve the codec for this pager
7178 void sqlcipherPagerSetCodec(
7179 Pager *pPager,
7180 void *(*xCodec)(void*,void*,Pgno,int),
7181 void (*xCodecSizeChng)(void*,int,int),
7182 void (*xCodecFree)(void*),
7183 void *pCodec
7185 if( pPager->xCodecFree ){
7186 pPager->xCodecFree(pPager->pCodec);
7187 }else{
7188 pager_reset(pPager);
7190 pPager->xCodec = pPager->memDb ? 0 : xCodec;
7191 pPager->xCodecSizeChng = xCodecSizeChng;
7192 pPager->xCodecFree = xCodecFree;
7193 pPager->pCodec = pCodec;
7194 setGetterMethod(pPager);
7195 pagerReportSize(pPager);
7197 void *sqlcipherPagerGetCodec(Pager *pPager){
7198 return pPager->pCodec;
7202 ** This function is called by the wal module when writing page content
7203 ** into the log file.
7205 ** This function returns a pointer to a buffer containing the encrypted
7206 ** page content. If a malloc fails, this function may return NULL.
7208 void *sqlcipherPagerCodec(PgHdr *pPg){
7209 void *aData = 0;
7210 CODEC2(pPg->pPager, pPg->pData, pPg->pgno, 6, return 0, aData);
7211 return aData;
7213 #endif /* SQLITE_HAS_CODEC */
7214 /* END SQLCIPHER */
7216 #ifndef SQLITE_OMIT_AUTOVACUUM
7218 ** Move the page pPg to location pgno in the file.
7220 ** There must be no references to the page previously located at
7221 ** pgno (which we call pPgOld) though that page is allowed to be
7222 ** in cache. If the page previously located at pgno is not already
7223 ** in the rollback journal, it is not put there by by this routine.
7225 ** References to the page pPg remain valid. Updating any
7226 ** meta-data associated with pPg (i.e. data stored in the nExtra bytes
7227 ** allocated along with the page) is the responsibility of the caller.
7229 ** A transaction must be active when this routine is called. It used to be
7230 ** required that a statement transaction was not active, but this restriction
7231 ** has been removed (CREATE INDEX needs to move a page when a statement
7232 ** transaction is active).
7234 ** If the fourth argument, isCommit, is non-zero, then this page is being
7235 ** moved as part of a database reorganization just before the transaction
7236 ** is being committed. In this case, it is guaranteed that the database page
7237 ** pPg refers to will not be written to again within this transaction.
7239 ** This function may return SQLITE_NOMEM or an IO error code if an error
7240 ** occurs. Otherwise, it returns SQLITE_OK.
7242 int sqlite3PagerMovepage(Pager *pPager, DbPage *pPg, Pgno pgno, int isCommit){
7243 PgHdr *pPgOld; /* The page being overwritten. */
7244 Pgno needSyncPgno = 0; /* Old value of pPg->pgno, if sync is required */
7245 int rc; /* Return code */
7246 Pgno origPgno; /* The original page number */
7248 assert( pPg->nRef>0 );
7249 assert( pPager->eState==PAGER_WRITER_CACHEMOD
7250 || pPager->eState==PAGER_WRITER_DBMOD
7252 assert( assert_pager_state(pPager) );
7254 /* In order to be able to rollback, an in-memory database must journal
7255 ** the page we are moving from.
7257 assert( pPager->tempFile || !MEMDB );
7258 if( pPager->tempFile ){
7259 rc = sqlite3PagerWrite(pPg);
7260 if( rc ) return rc;
7263 /* If the page being moved is dirty and has not been saved by the latest
7264 ** savepoint, then save the current contents of the page into the
7265 ** sub-journal now. This is required to handle the following scenario:
7267 ** BEGIN;
7268 ** <journal page X, then modify it in memory>
7269 ** SAVEPOINT one;
7270 ** <Move page X to location Y>
7271 ** ROLLBACK TO one;
7273 ** If page X were not written to the sub-journal here, it would not
7274 ** be possible to restore its contents when the "ROLLBACK TO one"
7275 ** statement were is processed.
7277 ** subjournalPage() may need to allocate space to store pPg->pgno into
7278 ** one or more savepoint bitvecs. This is the reason this function
7279 ** may return SQLITE_NOMEM.
7281 if( (pPg->flags & PGHDR_DIRTY)!=0
7282 && SQLITE_OK!=(rc = subjournalPageIfRequired(pPg))
7284 return rc;
7287 PAGERTRACE(("MOVE %d page %d (needSync=%d) moves to %d\n",
7288 PAGERID(pPager), pPg->pgno, (pPg->flags&PGHDR_NEED_SYNC)?1:0, pgno));
7289 IOTRACE(("MOVE %p %d %d\n", pPager, pPg->pgno, pgno))
7291 /* If the journal needs to be sync()ed before page pPg->pgno can
7292 ** be written to, store pPg->pgno in local variable needSyncPgno.
7294 ** If the isCommit flag is set, there is no need to remember that
7295 ** the journal needs to be sync()ed before database page pPg->pgno
7296 ** can be written to. The caller has already promised not to write to it.
7298 if( (pPg->flags&PGHDR_NEED_SYNC) && !isCommit ){
7299 needSyncPgno = pPg->pgno;
7300 assert( pPager->journalMode==PAGER_JOURNALMODE_OFF ||
7301 pageInJournal(pPager, pPg) || pPg->pgno>pPager->dbOrigSize );
7302 assert( pPg->flags&PGHDR_DIRTY );
7305 /* If the cache contains a page with page-number pgno, remove it
7306 ** from its hash chain. Also, if the PGHDR_NEED_SYNC flag was set for
7307 ** page pgno before the 'move' operation, it needs to be retained
7308 ** for the page moved there.
7310 pPg->flags &= ~PGHDR_NEED_SYNC;
7311 pPgOld = sqlite3PagerLookup(pPager, pgno);
7312 assert( !pPgOld || pPgOld->nRef==1 || CORRUPT_DB );
7313 if( pPgOld ){
7314 if( NEVER(pPgOld->nRef>1) ){
7315 sqlite3PagerUnrefNotNull(pPgOld);
7316 return SQLITE_CORRUPT_BKPT;
7318 pPg->flags |= (pPgOld->flags&PGHDR_NEED_SYNC);
7319 if( pPager->tempFile ){
7320 /* Do not discard pages from an in-memory database since we might
7321 ** need to rollback later. Just move the page out of the way. */
7322 sqlite3PcacheMove(pPgOld, pPager->dbSize+1);
7323 }else{
7324 sqlite3PcacheDrop(pPgOld);
7328 origPgno = pPg->pgno;
7329 sqlite3PcacheMove(pPg, pgno);
7330 sqlite3PcacheMakeDirty(pPg);
7332 /* For an in-memory database, make sure the original page continues
7333 ** to exist, in case the transaction needs to roll back. Use pPgOld
7334 ** as the original page since it has already been allocated.
7336 if( pPager->tempFile && pPgOld ){
7337 sqlite3PcacheMove(pPgOld, origPgno);
7338 sqlite3PagerUnrefNotNull(pPgOld);
7341 if( needSyncPgno ){
7342 /* If needSyncPgno is non-zero, then the journal file needs to be
7343 ** sync()ed before any data is written to database file page needSyncPgno.
7344 ** Currently, no such page exists in the page-cache and the
7345 ** "is journaled" bitvec flag has been set. This needs to be remedied by
7346 ** loading the page into the pager-cache and setting the PGHDR_NEED_SYNC
7347 ** flag.
7349 ** If the attempt to load the page into the page-cache fails, (due
7350 ** to a malloc() or IO failure), clear the bit in the pInJournal[]
7351 ** array. Otherwise, if the page is loaded and written again in
7352 ** this transaction, it may be written to the database file before
7353 ** it is synced into the journal file. This way, it may end up in
7354 ** the journal file twice, but that is not a problem.
7356 PgHdr *pPgHdr;
7357 rc = sqlite3PagerGet(pPager, needSyncPgno, &pPgHdr, 0);
7358 if( rc!=SQLITE_OK ){
7359 if( needSyncPgno<=pPager->dbOrigSize ){
7360 assert( pPager->pTmpSpace!=0 );
7361 sqlite3BitvecClear(pPager->pInJournal, needSyncPgno, pPager->pTmpSpace);
7363 return rc;
7365 pPgHdr->flags |= PGHDR_NEED_SYNC;
7366 sqlite3PcacheMakeDirty(pPgHdr);
7367 sqlite3PagerUnrefNotNull(pPgHdr);
7370 return SQLITE_OK;
7372 #endif
7375 ** The page handle passed as the first argument refers to a dirty page
7376 ** with a page number other than iNew. This function changes the page's
7377 ** page number to iNew and sets the value of the PgHdr.flags field to
7378 ** the value passed as the third parameter.
7380 void sqlite3PagerRekey(DbPage *pPg, Pgno iNew, u16 flags){
7381 assert( pPg->pgno!=iNew );
7382 pPg->flags = flags;
7383 sqlite3PcacheMove(pPg, iNew);
7387 ** Return a pointer to the data for the specified page.
7389 void *sqlite3PagerGetData(DbPage *pPg){
7390 assert( pPg->nRef>0 || pPg->pPager->memDb );
7391 return pPg->pData;
7395 ** Return a pointer to the Pager.nExtra bytes of "extra" space
7396 ** allocated along with the specified page.
7398 void *sqlite3PagerGetExtra(DbPage *pPg){
7399 return pPg->pExtra;
7403 ** Get/set the locking-mode for this pager. Parameter eMode must be one
7404 ** of PAGER_LOCKINGMODE_QUERY, PAGER_LOCKINGMODE_NORMAL or
7405 ** PAGER_LOCKINGMODE_EXCLUSIVE. If the parameter is not _QUERY, then
7406 ** the locking-mode is set to the value specified.
7408 ** The returned value is either PAGER_LOCKINGMODE_NORMAL or
7409 ** PAGER_LOCKINGMODE_EXCLUSIVE, indicating the current (possibly updated)
7410 ** locking-mode.
7412 int sqlite3PagerLockingMode(Pager *pPager, int eMode){
7413 assert( eMode==PAGER_LOCKINGMODE_QUERY
7414 || eMode==PAGER_LOCKINGMODE_NORMAL
7415 || eMode==PAGER_LOCKINGMODE_EXCLUSIVE );
7416 assert( PAGER_LOCKINGMODE_QUERY<0 );
7417 assert( PAGER_LOCKINGMODE_NORMAL>=0 && PAGER_LOCKINGMODE_EXCLUSIVE>=0 );
7418 assert( pPager->exclusiveMode || 0==sqlite3WalHeapMemory(pPager->pWal) );
7419 if( eMode>=0 && !pPager->tempFile && !sqlite3WalHeapMemory(pPager->pWal) ){
7420 pPager->exclusiveMode = (u8)eMode;
7422 return (int)pPager->exclusiveMode;
7426 ** Set the journal-mode for this pager. Parameter eMode must be one of:
7428 ** PAGER_JOURNALMODE_DELETE
7429 ** PAGER_JOURNALMODE_TRUNCATE
7430 ** PAGER_JOURNALMODE_PERSIST
7431 ** PAGER_JOURNALMODE_OFF
7432 ** PAGER_JOURNALMODE_MEMORY
7433 ** PAGER_JOURNALMODE_WAL
7435 ** The journalmode is set to the value specified if the change is allowed.
7436 ** The change may be disallowed for the following reasons:
7438 ** * An in-memory database can only have its journal_mode set to _OFF
7439 ** or _MEMORY.
7441 ** * Temporary databases cannot have _WAL journalmode.
7443 ** The returned indicate the current (possibly updated) journal-mode.
7445 int sqlite3PagerSetJournalMode(Pager *pPager, int eMode){
7446 u8 eOld = pPager->journalMode; /* Prior journalmode */
7448 /* The eMode parameter is always valid */
7449 assert( eMode==PAGER_JOURNALMODE_DELETE /* 0 */
7450 || eMode==PAGER_JOURNALMODE_PERSIST /* 1 */
7451 || eMode==PAGER_JOURNALMODE_OFF /* 2 */
7452 || eMode==PAGER_JOURNALMODE_TRUNCATE /* 3 */
7453 || eMode==PAGER_JOURNALMODE_MEMORY /* 4 */
7454 || eMode==PAGER_JOURNALMODE_WAL /* 5 */ );
7456 /* This routine is only called from the OP_JournalMode opcode, and
7457 ** the logic there will never allow a temporary file to be changed
7458 ** to WAL mode.
7460 assert( pPager->tempFile==0 || eMode!=PAGER_JOURNALMODE_WAL );
7462 /* Do allow the journalmode of an in-memory database to be set to
7463 ** anything other than MEMORY or OFF
7465 if( MEMDB ){
7466 assert( eOld==PAGER_JOURNALMODE_MEMORY || eOld==PAGER_JOURNALMODE_OFF );
7467 if( eMode!=PAGER_JOURNALMODE_MEMORY && eMode!=PAGER_JOURNALMODE_OFF ){
7468 eMode = eOld;
7472 if( eMode!=eOld ){
7474 /* Change the journal mode. */
7475 assert( pPager->eState!=PAGER_ERROR );
7476 pPager->journalMode = (u8)eMode;
7478 /* When transistioning from TRUNCATE or PERSIST to any other journal
7479 ** mode except WAL, unless the pager is in locking_mode=exclusive mode,
7480 ** delete the journal file.
7482 assert( (PAGER_JOURNALMODE_TRUNCATE & 5)==1 );
7483 assert( (PAGER_JOURNALMODE_PERSIST & 5)==1 );
7484 assert( (PAGER_JOURNALMODE_DELETE & 5)==0 );
7485 assert( (PAGER_JOURNALMODE_MEMORY & 5)==4 );
7486 assert( (PAGER_JOURNALMODE_OFF & 5)==0 );
7487 assert( (PAGER_JOURNALMODE_WAL & 5)==5 );
7489 assert( isOpen(pPager->fd) || pPager->exclusiveMode );
7490 if( !pPager->exclusiveMode && (eOld & 5)==1 && (eMode & 1)==0 ){
7491 /* In this case we would like to delete the journal file. If it is
7492 ** not possible, then that is not a problem. Deleting the journal file
7493 ** here is an optimization only.
7495 ** Before deleting the journal file, obtain a RESERVED lock on the
7496 ** database file. This ensures that the journal file is not deleted
7497 ** while it is in use by some other client.
7499 sqlite3OsClose(pPager->jfd);
7500 if( pPager->eLock>=RESERVED_LOCK ){
7501 sqlite3OsDelete(pPager->pVfs, pPager->zJournal, 0);
7502 }else{
7503 int rc = SQLITE_OK;
7504 int state = pPager->eState;
7505 assert( state==PAGER_OPEN || state==PAGER_READER );
7506 if( state==PAGER_OPEN ){
7507 rc = sqlite3PagerSharedLock(pPager);
7509 if( pPager->eState==PAGER_READER ){
7510 assert( rc==SQLITE_OK );
7511 rc = pagerLockDb(pPager, RESERVED_LOCK);
7513 if( rc==SQLITE_OK ){
7514 sqlite3OsDelete(pPager->pVfs, pPager->zJournal, 0);
7516 if( rc==SQLITE_OK && state==PAGER_READER ){
7517 pagerUnlockDb(pPager, SHARED_LOCK);
7518 }else if( state==PAGER_OPEN ){
7519 pager_unlock(pPager);
7521 assert( state==pPager->eState );
7523 }else if( eMode==PAGER_JOURNALMODE_OFF ){
7524 sqlite3OsClose(pPager->jfd);
7528 /* Return the new journal mode */
7529 return (int)pPager->journalMode;
7533 ** Return the current journal mode.
7535 int sqlite3PagerGetJournalMode(Pager *pPager){
7536 return (int)pPager->journalMode;
7540 ** Return TRUE if the pager is in a state where it is OK to change the
7541 ** journalmode. Journalmode changes can only happen when the database
7542 ** is unmodified.
7544 int sqlite3PagerOkToChangeJournalMode(Pager *pPager){
7545 assert( assert_pager_state(pPager) );
7546 if( pPager->eState>=PAGER_WRITER_CACHEMOD ) return 0;
7547 if( NEVER(isOpen(pPager->jfd) && pPager->journalOff>0) ) return 0;
7548 return 1;
7552 ** Get/set the size-limit used for persistent journal files.
7554 ** Setting the size limit to -1 means no limit is enforced.
7555 ** An attempt to set a limit smaller than -1 is a no-op.
7557 i64 sqlite3PagerJournalSizeLimit(Pager *pPager, i64 iLimit){
7558 if( iLimit>=-1 ){
7559 pPager->journalSizeLimit = iLimit;
7560 sqlite3WalLimit(pPager->pWal, iLimit);
7562 return pPager->journalSizeLimit;
7566 ** Return a pointer to the pPager->pBackup variable. The backup module
7567 ** in backup.c maintains the content of this variable. This module
7568 ** uses it opaquely as an argument to sqlite3BackupRestart() and
7569 ** sqlite3BackupUpdate() only.
7571 sqlite3_backup **sqlite3PagerBackupPtr(Pager *pPager){
7572 return &pPager->pBackup;
7575 #ifndef SQLITE_OMIT_VACUUM
7577 ** Unless this is an in-memory or temporary database, clear the pager cache.
7579 void sqlite3PagerClearCache(Pager *pPager){
7580 assert( MEMDB==0 || pPager->tempFile );
7581 if( pPager->tempFile==0 ) pager_reset(pPager);
7583 #endif
7586 #ifndef SQLITE_OMIT_WAL
7588 ** This function is called when the user invokes "PRAGMA wal_checkpoint",
7589 ** "PRAGMA wal_blocking_checkpoint" or calls the sqlite3_wal_checkpoint()
7590 ** or wal_blocking_checkpoint() API functions.
7592 ** Parameter eMode is one of SQLITE_CHECKPOINT_PASSIVE, FULL or RESTART.
7594 int sqlite3PagerCheckpoint(
7595 Pager *pPager, /* Checkpoint on this pager */
7596 sqlite3 *db, /* Db handle used to check for interrupts */
7597 int eMode, /* Type of checkpoint */
7598 int *pnLog, /* OUT: Final number of frames in log */
7599 int *pnCkpt /* OUT: Final number of checkpointed frames */
7601 int rc = SQLITE_OK;
7602 if( pPager->pWal==0 && pPager->journalMode==PAGER_JOURNALMODE_WAL ){
7603 /* This only happens when a database file is zero bytes in size opened and
7604 ** then "PRAGMA journal_mode=WAL" is run and then sqlite3_wal_checkpoint()
7605 ** is invoked without any intervening transactions. We need to start
7606 ** a transaction to initialize pWal. The PRAGMA table_list statement is
7607 ** used for this since it starts transactions on every database file,
7608 ** including all ATTACHed databases. This seems expensive for a single
7609 ** sqlite3_wal_checkpoint() call, but it happens very rarely.
7610 ** https://sqlite.org/forum/forumpost/fd0f19d229156939
7612 sqlite3_exec(db, "PRAGMA table_list",0,0,0);
7614 if( pPager->pWal ){
7615 rc = sqlite3WalCheckpoint(pPager->pWal, db, eMode,
7616 (eMode==SQLITE_CHECKPOINT_PASSIVE ? 0 : pPager->xBusyHandler),
7617 pPager->pBusyHandlerArg,
7618 pPager->walSyncFlags, pPager->pageSize, (u8 *)pPager->pTmpSpace,
7619 pnLog, pnCkpt
7622 return rc;
7625 int sqlite3PagerWalCallback(Pager *pPager){
7626 return sqlite3WalCallback(pPager->pWal);
7630 ** Return true if the underlying VFS for the given pager supports the
7631 ** primitives necessary for write-ahead logging.
7633 int sqlite3PagerWalSupported(Pager *pPager){
7634 const sqlite3_io_methods *pMethods = pPager->fd->pMethods;
7635 if( pPager->noLock ) return 0;
7636 return pPager->exclusiveMode || (pMethods->iVersion>=2 && pMethods->xShmMap);
7640 ** Attempt to take an exclusive lock on the database file. If a PENDING lock
7641 ** is obtained instead, immediately release it.
7643 static int pagerExclusiveLock(Pager *pPager){
7644 int rc; /* Return code */
7646 assert( pPager->eLock==SHARED_LOCK || pPager->eLock==EXCLUSIVE_LOCK );
7647 rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
7648 if( rc!=SQLITE_OK ){
7649 /* If the attempt to grab the exclusive lock failed, release the
7650 ** pending lock that may have been obtained instead. */
7651 pagerUnlockDb(pPager, SHARED_LOCK);
7654 return rc;
7658 ** Call sqlite3WalOpen() to open the WAL handle. If the pager is in
7659 ** exclusive-locking mode when this function is called, take an EXCLUSIVE
7660 ** lock on the database file and use heap-memory to store the wal-index
7661 ** in. Otherwise, use the normal shared-memory.
7663 static int pagerOpenWal(Pager *pPager){
7664 int rc = SQLITE_OK;
7666 assert( pPager->pWal==0 && pPager->tempFile==0 );
7667 assert( pPager->eLock==SHARED_LOCK || pPager->eLock==EXCLUSIVE_LOCK );
7669 /* If the pager is already in exclusive-mode, the WAL module will use
7670 ** heap-memory for the wal-index instead of the VFS shared-memory
7671 ** implementation. Take the exclusive lock now, before opening the WAL
7672 ** file, to make sure this is safe.
7674 if( pPager->exclusiveMode ){
7675 rc = pagerExclusiveLock(pPager);
7678 /* Open the connection to the log file. If this operation fails,
7679 ** (e.g. due to malloc() failure), return an error code.
7681 if( rc==SQLITE_OK ){
7682 rc = sqlite3WalOpen(pPager->pVfs,
7683 pPager->fd, pPager->zWal, pPager->exclusiveMode,
7684 pPager->journalSizeLimit, &pPager->pWal
7687 pagerFixMaplimit(pPager);
7689 return rc;
7694 ** The caller must be holding a SHARED lock on the database file to call
7695 ** this function.
7697 ** If the pager passed as the first argument is open on a real database
7698 ** file (not a temp file or an in-memory database), and the WAL file
7699 ** is not already open, make an attempt to open it now. If successful,
7700 ** return SQLITE_OK. If an error occurs or the VFS used by the pager does
7701 ** not support the xShmXXX() methods, return an error code. *pbOpen is
7702 ** not modified in either case.
7704 ** If the pager is open on a temp-file (or in-memory database), or if
7705 ** the WAL file is already open, set *pbOpen to 1 and return SQLITE_OK
7706 ** without doing anything.
7708 int sqlite3PagerOpenWal(
7709 Pager *pPager, /* Pager object */
7710 int *pbOpen /* OUT: Set to true if call is a no-op */
7712 int rc = SQLITE_OK; /* Return code */
7714 assert( assert_pager_state(pPager) );
7715 assert( pPager->eState==PAGER_OPEN || pbOpen );
7716 assert( pPager->eState==PAGER_READER || !pbOpen );
7717 assert( pbOpen==0 || *pbOpen==0 );
7718 assert( pbOpen!=0 || (!pPager->tempFile && !pPager->pWal) );
7720 if( !pPager->tempFile && !pPager->pWal ){
7721 if( !sqlite3PagerWalSupported(pPager) ) return SQLITE_CANTOPEN;
7723 /* Close any rollback journal previously open */
7724 sqlite3OsClose(pPager->jfd);
7726 rc = pagerOpenWal(pPager);
7727 if( rc==SQLITE_OK ){
7728 pPager->journalMode = PAGER_JOURNALMODE_WAL;
7729 pPager->eState = PAGER_OPEN;
7731 }else{
7732 *pbOpen = 1;
7735 return rc;
7739 ** This function is called to close the connection to the log file prior
7740 ** to switching from WAL to rollback mode.
7742 ** Before closing the log file, this function attempts to take an
7743 ** EXCLUSIVE lock on the database file. If this cannot be obtained, an
7744 ** error (SQLITE_BUSY) is returned and the log connection is not closed.
7745 ** If successful, the EXCLUSIVE lock is not released before returning.
7747 int sqlite3PagerCloseWal(Pager *pPager, sqlite3 *db){
7748 int rc = SQLITE_OK;
7750 assert( pPager->journalMode==PAGER_JOURNALMODE_WAL );
7752 /* If the log file is not already open, but does exist in the file-system,
7753 ** it may need to be checkpointed before the connection can switch to
7754 ** rollback mode. Open it now so this can happen.
7756 if( !pPager->pWal ){
7757 int logexists = 0;
7758 rc = pagerLockDb(pPager, SHARED_LOCK);
7759 if( rc==SQLITE_OK ){
7760 rc = sqlite3OsAccess(
7761 pPager->pVfs, pPager->zWal, SQLITE_ACCESS_EXISTS, &logexists
7764 if( rc==SQLITE_OK && logexists ){
7765 rc = pagerOpenWal(pPager);
7769 /* Checkpoint and close the log. Because an EXCLUSIVE lock is held on
7770 ** the database file, the log and log-summary files will be deleted.
7772 if( rc==SQLITE_OK && pPager->pWal ){
7773 rc = pagerExclusiveLock(pPager);
7774 if( rc==SQLITE_OK ){
7775 rc = sqlite3WalClose(pPager->pWal, db, pPager->walSyncFlags,
7776 pPager->pageSize, (u8*)pPager->pTmpSpace);
7777 pPager->pWal = 0;
7778 pagerFixMaplimit(pPager);
7779 if( rc && !pPager->exclusiveMode ) pagerUnlockDb(pPager, SHARED_LOCK);
7782 return rc;
7785 #ifdef SQLITE_ENABLE_SETLK_TIMEOUT
7787 ** If pager pPager is a wal-mode database not in exclusive locking mode,
7788 ** invoke the sqlite3WalWriteLock() function on the associated Wal object
7789 ** with the same db and bLock parameters as were passed to this function.
7790 ** Return an SQLite error code if an error occurs, or SQLITE_OK otherwise.
7792 int sqlite3PagerWalWriteLock(Pager *pPager, int bLock){
7793 int rc = SQLITE_OK;
7794 if( pagerUseWal(pPager) && pPager->exclusiveMode==0 ){
7795 rc = sqlite3WalWriteLock(pPager->pWal, bLock);
7797 return rc;
7801 ** Set the database handle used by the wal layer to determine if
7802 ** blocking locks are required.
7804 void sqlite3PagerWalDb(Pager *pPager, sqlite3 *db){
7805 if( pagerUseWal(pPager) ){
7806 sqlite3WalDb(pPager->pWal, db);
7809 #endif
7811 #ifdef SQLITE_ENABLE_SNAPSHOT
7813 ** If this is a WAL database, obtain a snapshot handle for the snapshot
7814 ** currently open. Otherwise, return an error.
7816 int sqlite3PagerSnapshotGet(Pager *pPager, sqlite3_snapshot **ppSnapshot){
7817 int rc = SQLITE_ERROR;
7818 if( pPager->pWal ){
7819 rc = sqlite3WalSnapshotGet(pPager->pWal, ppSnapshot);
7821 return rc;
7825 ** If this is a WAL database, store a pointer to pSnapshot. Next time a
7826 ** read transaction is opened, attempt to read from the snapshot it
7827 ** identifies. If this is not a WAL database, return an error.
7829 int sqlite3PagerSnapshotOpen(
7830 Pager *pPager,
7831 sqlite3_snapshot *pSnapshot
7833 int rc = SQLITE_OK;
7834 if( pPager->pWal ){
7835 sqlite3WalSnapshotOpen(pPager->pWal, pSnapshot);
7836 }else{
7837 rc = SQLITE_ERROR;
7839 return rc;
7843 ** If this is a WAL database, call sqlite3WalSnapshotRecover(). If this
7844 ** is not a WAL database, return an error.
7846 int sqlite3PagerSnapshotRecover(Pager *pPager){
7847 int rc;
7848 if( pPager->pWal ){
7849 rc = sqlite3WalSnapshotRecover(pPager->pWal);
7850 }else{
7851 rc = SQLITE_ERROR;
7853 return rc;
7857 ** The caller currently has a read transaction open on the database.
7858 ** If this is not a WAL database, SQLITE_ERROR is returned. Otherwise,
7859 ** this function takes a SHARED lock on the CHECKPOINTER slot and then
7860 ** checks if the snapshot passed as the second argument is still
7861 ** available. If so, SQLITE_OK is returned.
7863 ** If the snapshot is not available, SQLITE_ERROR is returned. Or, if
7864 ** the CHECKPOINTER lock cannot be obtained, SQLITE_BUSY. If any error
7865 ** occurs (any value other than SQLITE_OK is returned), the CHECKPOINTER
7866 ** lock is released before returning.
7868 int sqlite3PagerSnapshotCheck(Pager *pPager, sqlite3_snapshot *pSnapshot){
7869 int rc;
7870 if( pPager->pWal ){
7871 rc = sqlite3WalSnapshotCheck(pPager->pWal, pSnapshot);
7872 }else{
7873 rc = SQLITE_ERROR;
7875 return rc;
7879 ** Release a lock obtained by an earlier successful call to
7880 ** sqlite3PagerSnapshotCheck().
7882 void sqlite3PagerSnapshotUnlock(Pager *pPager){
7883 assert( pPager->pWal );
7884 sqlite3WalSnapshotUnlock(pPager->pWal);
7887 #endif /* SQLITE_ENABLE_SNAPSHOT */
7888 #endif /* !SQLITE_OMIT_WAL */
7890 #ifdef SQLITE_ENABLE_ZIPVFS
7892 ** A read-lock must be held on the pager when this function is called. If
7893 ** the pager is in WAL mode and the WAL file currently contains one or more
7894 ** frames, return the size in bytes of the page images stored within the
7895 ** WAL frames. Otherwise, if this is not a WAL database or the WAL file
7896 ** is empty, return 0.
7898 int sqlite3PagerWalFramesize(Pager *pPager){
7899 assert( pPager->eState>=PAGER_READER );
7900 return sqlite3WalFramesize(pPager->pWal);
7902 #endif
7904 #endif /* SQLITE_OMIT_DISKIO */
7906 /* BEGIN SQLCIPHER */
7907 #ifdef SQLITE_HAS_CODEC
7909 int sqlite3pager_is_sj_pgno(Pager *pPager, Pgno pgno) {
7910 return (PAGER_SJ_PGNO(pPager) == pgno) ? 1 : 0;
7913 void sqlite3pager_error(Pager *pPager, int error) {
7914 pPager->errCode = error;
7915 pPager->eState = PAGER_ERROR;
7916 setGetterMethod(pPager);
7919 void sqlite3pager_reset(Pager *pPager){
7920 pager_reset(pPager);
7923 #endif
7924 /* END SQLCIPHER */